File: io.rst

package info (click to toggle)
pandas 1.5.3%2Bdfsg-2
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 56,516 kB
  • sloc: python: 382,477; ansic: 8,695; sh: 119; xml: 102; makefile: 97
file content (6601 lines) | stat: -rw-r--r-- 216,273 bytes parent folder | download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
5953
5954
5955
5956
5957
5958
5959
5960
5961
5962
5963
5964
5965
5966
5967
5968
5969
5970
5971
5972
5973
5974
5975
5976
5977
5978
5979
5980
5981
5982
5983
5984
5985
5986
5987
5988
5989
5990
5991
5992
5993
5994
5995
5996
5997
5998
5999
6000
6001
6002
6003
6004
6005
6006
6007
6008
6009
6010
6011
6012
6013
6014
6015
6016
6017
6018
6019
6020
6021
6022
6023
6024
6025
6026
6027
6028
6029
6030
6031
6032
6033
6034
6035
6036
6037
6038
6039
6040
6041
6042
6043
6044
6045
6046
6047
6048
6049
6050
6051
6052
6053
6054
6055
6056
6057
6058
6059
6060
6061
6062
6063
6064
6065
6066
6067
6068
6069
6070
6071
6072
6073
6074
6075
6076
6077
6078
6079
6080
6081
6082
6083
6084
6085
6086
6087
6088
6089
6090
6091
6092
6093
6094
6095
6096
6097
6098
6099
6100
6101
6102
6103
6104
6105
6106
6107
6108
6109
6110
6111
6112
6113
6114
6115
6116
6117
6118
6119
6120
6121
6122
6123
6124
6125
6126
6127
6128
6129
6130
6131
6132
6133
6134
6135
6136
6137
6138
6139
6140
6141
6142
6143
6144
6145
6146
6147
6148
6149
6150
6151
6152
6153
6154
6155
6156
6157
6158
6159
6160
6161
6162
6163
6164
6165
6166
6167
6168
6169
6170
6171
6172
6173
6174
6175
6176
6177
6178
6179
6180
6181
6182
6183
6184
6185
6186
6187
6188
6189
6190
6191
6192
6193
6194
6195
6196
6197
6198
6199
6200
6201
6202
6203
6204
6205
6206
6207
6208
6209
6210
6211
6212
6213
6214
6215
6216
6217
6218
6219
6220
6221
6222
6223
6224
6225
6226
6227
6228
6229
6230
6231
6232
6233
6234
6235
6236
6237
6238
6239
6240
6241
6242
6243
6244
6245
6246
6247
6248
6249
6250
6251
6252
6253
6254
6255
6256
6257
6258
6259
6260
6261
6262
6263
6264
6265
6266
6267
6268
6269
6270
6271
6272
6273
6274
6275
6276
6277
6278
6279
6280
6281
6282
6283
6284
6285
6286
6287
6288
6289
6290
6291
6292
6293
6294
6295
6296
6297
6298
6299
6300
6301
6302
6303
6304
6305
6306
6307
6308
6309
6310
6311
6312
6313
6314
6315
6316
6317
6318
6319
6320
6321
6322
6323
6324
6325
6326
6327
6328
6329
6330
6331
6332
6333
6334
6335
6336
6337
6338
6339
6340
6341
6342
6343
6344
6345
6346
6347
6348
6349
6350
6351
6352
6353
6354
6355
6356
6357
6358
6359
6360
6361
6362
6363
6364
6365
6366
6367
6368
6369
6370
6371
6372
6373
6374
6375
6376
6377
6378
6379
6380
6381
6382
6383
6384
6385
6386
6387
6388
6389
6390
6391
6392
6393
6394
6395
6396
6397
6398
6399
6400
6401
6402
6403
6404
6405
6406
6407
6408
6409
6410
6411
6412
6413
6414
6415
6416
6417
6418
6419
6420
6421
6422
6423
6424
6425
6426
6427
6428
6429
6430
6431
6432
6433
6434
6435
6436
6437
6438
6439
6440
6441
6442
6443
6444
6445
6446
6447
6448
6449
6450
6451
6452
6453
6454
6455
6456
6457
6458
6459
6460
6461
6462
6463
6464
6465
6466
6467
6468
6469
6470
6471
6472
6473
6474
6475
6476
6477
6478
6479
6480
6481
6482
6483
6484
6485
6486
6487
6488
6489
6490
6491
6492
6493
6494
6495
6496
6497
6498
6499
6500
6501
6502
6503
6504
6505
6506
6507
6508
6509
6510
6511
6512
6513
6514
6515
6516
6517
6518
6519
6520
6521
6522
6523
6524
6525
6526
6527
6528
6529
6530
6531
6532
6533
6534
6535
6536
6537
6538
6539
6540
6541
6542
6543
6544
6545
6546
6547
6548
6549
6550
6551
6552
6553
6554
6555
6556
6557
6558
6559
6560
6561
6562
6563
6564
6565
6566
6567
6568
6569
6570
6571
6572
6573
6574
6575
6576
6577
6578
6579
6580
6581
6582
6583
6584
6585
6586
6587
6588
6589
6590
6591
6592
6593
6594
6595
6596
6597
6598
6599
6600
6601
.. _io:

.. currentmodule:: pandas


===============================
IO tools (text, CSV, HDF5, ...)
===============================

The pandas I/O API is a set of top level ``reader`` functions accessed like
:func:`pandas.read_csv` that generally return a pandas object. The corresponding
``writer`` functions are object methods that are accessed like
:meth:`DataFrame.to_csv`. Below is a table containing available ``readers`` and
``writers``.

.. csv-table::
    :header: "Format Type", "Data Description", "Reader", "Writer"
    :widths: 30, 100, 60, 60
    :delim: ;

    text;`CSV <https://en.wikipedia.org/wiki/Comma-separated_values>`__;:ref:`read_csv<io.read_csv_table>`;:ref:`to_csv<io.store_in_csv>`
    text;Fixed-Width Text File;:ref:`read_fwf<io.fwf_reader>`
    text;`JSON <https://www.json.org/>`__;:ref:`read_json<io.json_reader>`;:ref:`to_json<io.json_writer>`
    text;`HTML <https://en.wikipedia.org/wiki/HTML>`__;:ref:`read_html<io.read_html>`;:ref:`to_html<io.html>`
    text;`LaTeX <https://en.wikipedia.org/wiki/LaTeX>`__;;:ref:`Styler.to_latex<io.latex>`
    text;`XML <https://www.w3.org/standards/xml/core>`__;:ref:`read_xml<io.read_xml>`;:ref:`to_xml<io.xml>`
    text; Local clipboard;:ref:`read_clipboard<io.clipboard>`;:ref:`to_clipboard<io.clipboard>`
    binary;`MS Excel <https://en.wikipedia.org/wiki/Microsoft_Excel>`__;:ref:`read_excel<io.excel_reader>`;:ref:`to_excel<io.excel_writer>`
    binary;`OpenDocument <http://opendocumentformat.org>`__;:ref:`read_excel<io.ods>`;
    binary;`HDF5 Format <https://support.hdfgroup.org/HDF5/whatishdf5.html>`__;:ref:`read_hdf<io.hdf5>`;:ref:`to_hdf<io.hdf5>`
    binary;`Feather Format <https://github.com/wesm/feather>`__;:ref:`read_feather<io.feather>`;:ref:`to_feather<io.feather>`
    binary;`Parquet Format <https://parquet.apache.org/>`__;:ref:`read_parquet<io.parquet>`;:ref:`to_parquet<io.parquet>`
    binary;`ORC Format <https://orc.apache.org/>`__;:ref:`read_orc<io.orc>`;:ref:`to_orc<io.orc>`
    binary;`Stata <https://en.wikipedia.org/wiki/Stata>`__;:ref:`read_stata<io.stata_reader>`;:ref:`to_stata<io.stata_writer>`
    binary;`SAS <https://en.wikipedia.org/wiki/SAS_(software)>`__;:ref:`read_sas<io.sas_reader>`;
    binary;`SPSS <https://en.wikipedia.org/wiki/SPSS>`__;:ref:`read_spss<io.spss_reader>`;
    binary;`Python Pickle Format <https://docs.python.org/3/library/pickle.html>`__;:ref:`read_pickle<io.pickle>`;:ref:`to_pickle<io.pickle>`
    SQL;`SQL <https://en.wikipedia.org/wiki/SQL>`__;:ref:`read_sql<io.sql>`;:ref:`to_sql<io.sql>`
    SQL;`Google BigQuery <https://en.wikipedia.org/wiki/BigQuery>`__;:ref:`read_gbq<io.bigquery>`;:ref:`to_gbq<io.bigquery>`

:ref:`Here <io.perf>` is an informal performance comparison for some of these IO methods.

.. note::
   For examples that use the ``StringIO`` class, make sure you import it
   with ``from io import StringIO`` for Python 3.

.. _io.read_csv_table:

CSV & text files
----------------

The workhorse function for reading text files (a.k.a. flat files) is
:func:`read_csv`. See the :ref:`cookbook<cookbook.csv>` for some advanced strategies.

Parsing options
'''''''''''''''

:func:`read_csv` accepts the following common arguments:

Basic
+++++

filepath_or_buffer : various
  Either a path to a file (a :class:`python:str`, :class:`python:pathlib.Path`,
  or :class:`py:py._path.local.LocalPath`), URL (including http, ftp, and S3
  locations), or any object with a ``read()`` method (such as an open file or
  :class:`~python:io.StringIO`).
sep : str, defaults to ``','`` for :func:`read_csv`, ``\t`` for :func:`read_table`
  Delimiter to use. If sep is ``None``, the C engine cannot automatically detect
  the separator, but the Python parsing engine can, meaning the latter will be
  used and automatically detect the separator by Python's builtin sniffer tool,
  :class:`python:csv.Sniffer`. In addition, separators longer than 1 character and
  different from ``'\s+'`` will be interpreted as regular expressions and
  will also force the use of the Python parsing engine. Note that regex
  delimiters are prone to ignoring quoted data. Regex example: ``'\\r\\t'``.
delimiter : str, default ``None``
  Alternative argument name for sep.
delim_whitespace : boolean, default False
  Specifies whether or not whitespace (e.g. ``' '`` or ``'\t'``)
  will be used as the delimiter. Equivalent to setting ``sep='\s+'``.
  If this option is set to ``True``, nothing should be passed in for the
  ``delimiter`` parameter.

Column and index locations and names
++++++++++++++++++++++++++++++++++++

header : int or list of ints, default ``'infer'``
  Row number(s) to use as the column names, and the start of the
  data. Default behavior is to infer the column names: if no names are
  passed the behavior is identical to ``header=0`` and column names
  are inferred from the first line of the file, if column names are
  passed explicitly then the behavior is identical to
  ``header=None``. Explicitly pass ``header=0`` to be able to replace
  existing names.

  The header can be a list of ints that specify row locations
  for a MultiIndex on the columns e.g. ``[0,1,3]``. Intervening rows
  that are not specified will be skipped (e.g. 2 in this example is
  skipped). Note that this parameter ignores commented lines and empty
  lines if ``skip_blank_lines=True``, so header=0 denotes the first
  line of data rather than the first line of the file.
names : array-like, default ``None``
  List of column names to use. If file contains no header row, then you should
  explicitly pass ``header=None``. Duplicates in this list are not allowed.
index_col : int, str, sequence of int / str, or False, optional, default ``None``
  Column(s) to use as the row labels of the ``DataFrame``, either given as
  string name or column index. If a sequence of int / str is given, a
  MultiIndex is used.

  .. note::
     ``index_col=False`` can be used to force pandas to *not* use the first
     column as the index, e.g. when you have a malformed file with delimiters at
     the end of each line.

  The default value of ``None`` instructs pandas to guess. If the number of
  fields in the column header row is equal to the number of fields in the body
  of the data file, then a default index is used.  If it is larger, then
  the first columns are used as index so that the remaining number of fields in
  the body are equal to the number of fields in the header.

  The first row after the header is used to determine the number of columns,
  which will go into the index. If the subsequent rows contain less columns
  than the first row, they are filled with ``NaN``.

  This can be avoided through ``usecols``. This ensures that the columns are
  taken as is and the trailing data are ignored.
usecols : list-like or callable, default ``None``
  Return a subset of the columns. If list-like, all elements must either
  be positional (i.e. integer indices into the document columns) or strings
  that correspond to column names provided either by the user in ``names`` or
  inferred from the document header row(s). If ``names`` are given, the document
  header row(s) are not taken into account. For example, a valid list-like
  ``usecols`` parameter would be ``[0, 1, 2]`` or ``['foo', 'bar', 'baz']``.

  Element order is ignored, so ``usecols=[0, 1]`` is the same as ``[1, 0]``. To
  instantiate a DataFrame from ``data`` with element order preserved use
  ``pd.read_csv(data, usecols=['foo', 'bar'])[['foo', 'bar']]`` for columns
  in ``['foo', 'bar']`` order or
  ``pd.read_csv(data, usecols=['foo', 'bar'])[['bar', 'foo']]`` for
  ``['bar', 'foo']`` order.

  If callable, the callable function will be evaluated against the column names,
  returning names where the callable function evaluates to True:

  .. ipython:: python

     import pandas as pd
     from io import StringIO

     data = "col1,col2,col3\na,b,1\na,b,2\nc,d,3"
     pd.read_csv(StringIO(data))
     pd.read_csv(StringIO(data), usecols=lambda x: x.upper() in ["COL1", "COL3"])

  Using this parameter results in much faster parsing time and lower memory usage
  when using the c engine. The Python engine loads the data first before deciding
  which columns to drop.
squeeze : boolean, default ``False``
  If the parsed data only contains one column then return a ``Series``.

  .. deprecated:: 1.4.0
     Append ``.squeeze("columns")`` to the call to ``{func_name}`` to squeeze
     the data.
prefix : str, default ``None``
  Prefix to add to column numbers when no header, e.g. 'X' for X0, X1, ...

  .. deprecated:: 1.4.0
     Use a list comprehension on the DataFrame's columns after calling ``read_csv``.

  .. ipython:: python

     data = "col1,col2,col3\na,b,1"

     df = pd.read_csv(StringIO(data))
     df.columns = [f"pre_{col}" for col in df.columns]
     df

mangle_dupe_cols : boolean, default ``True``
  Duplicate columns will be specified as 'X', 'X.1'...'X.N', rather than 'X'...'X'.
  Passing in ``False`` will cause data to be overwritten if there are duplicate
  names in the columns.

  .. deprecated:: 1.5.0
     The argument was never implemented, and a new argument where the
     renaming pattern can be specified will be added instead.

General parsing configuration
+++++++++++++++++++++++++++++

dtype : Type name or dict of column -> type, default ``None``
  Data type for data or columns. E.g. ``{'a': np.float64, 'b': np.int32, 'c': 'Int64'}``
  Use ``str`` or ``object`` together with suitable ``na_values`` settings to preserve
  and not interpret dtype. If converters are specified, they will be applied INSTEAD
  of dtype conversion.

  .. versionadded:: 1.5.0

     Support for defaultdict was added. Specify a defaultdict as input where
     the default determines the dtype of the columns which are not explicitly
     listed.
engine : {``'c'``, ``'python'``, ``'pyarrow'``}
  Parser engine to use. The C and pyarrow engines are faster, while the python engine
  is currently more feature-complete. Multithreading is currently only supported by
  the pyarrow engine.

  .. versionadded:: 1.4.0

     The "pyarrow" engine was added as an *experimental* engine, and some features
     are unsupported, or may not work correctly, with this engine.
converters : dict, default ``None``
  Dict of functions for converting values in certain columns. Keys can either be
  integers or column labels.
true_values : list, default ``None``
  Values to consider as ``True``.
false_values : list, default ``None``
  Values to consider as ``False``.
skipinitialspace : boolean, default ``False``
  Skip spaces after delimiter.
skiprows : list-like or integer, default ``None``
  Line numbers to skip (0-indexed) or number of lines to skip (int) at the start
  of the file.

  If callable, the callable function will be evaluated against the row
  indices, returning True if the row should be skipped and False otherwise:

  .. ipython:: python

     data = "col1,col2,col3\na,b,1\na,b,2\nc,d,3"
     pd.read_csv(StringIO(data))
     pd.read_csv(StringIO(data), skiprows=lambda x: x % 2 != 0)

skipfooter : int, default ``0``
  Number of lines at bottom of file to skip (unsupported with engine='c').

nrows : int, default ``None``
  Number of rows of file to read. Useful for reading pieces of large files.
low_memory : boolean, default ``True``
  Internally process the file in chunks, resulting in lower memory use
  while parsing, but possibly mixed type inference.  To ensure no mixed
  types either set ``False``, or specify the type with the ``dtype`` parameter.
  Note that the entire file is read into a single ``DataFrame`` regardless,
  use the ``chunksize`` or ``iterator`` parameter to return the data in chunks.
  (Only valid with C parser)
memory_map : boolean, default False
  If a filepath is provided for ``filepath_or_buffer``, map the file object
  directly onto memory and access the data directly from there. Using this
  option can improve performance because there is no longer any I/O overhead.

NA and missing data handling
++++++++++++++++++++++++++++

na_values : scalar, str, list-like, or dict, default ``None``
  Additional strings to recognize as NA/NaN. If dict passed, specific per-column
  NA values. See :ref:`na values const <io.navaluesconst>` below
  for a list of the values interpreted as NaN by default.

keep_default_na : boolean, default ``True``
  Whether or not to include the default NaN values when parsing the data.
  Depending on whether ``na_values`` is passed in, the behavior is as follows:

  * If ``keep_default_na`` is ``True``, and ``na_values`` are specified, ``na_values``
    is appended to the default NaN values used for parsing.
  * If ``keep_default_na`` is ``True``, and ``na_values`` are not specified, only
    the default NaN values are used for parsing.
  * If ``keep_default_na`` is ``False``, and ``na_values`` are specified, only
    the NaN values specified ``na_values`` are used for parsing.
  * If ``keep_default_na`` is ``False``, and ``na_values`` are not specified, no
    strings will be parsed as NaN.

  Note that if ``na_filter`` is passed in as ``False``, the ``keep_default_na`` and
  ``na_values`` parameters will be ignored.
na_filter : boolean, default ``True``
  Detect missing value markers (empty strings and the value of na_values). In
  data without any NAs, passing ``na_filter=False`` can improve the performance
  of reading a large file.
verbose : boolean, default ``False``
  Indicate number of NA values placed in non-numeric columns.
skip_blank_lines : boolean, default ``True``
  If ``True``, skip over blank lines rather than interpreting as NaN values.

.. _io.read_csv_table.datetime:

Datetime handling
+++++++++++++++++

parse_dates : boolean or list of ints or names or list of lists or dict, default ``False``.
  * If ``True`` -> try parsing the index.
  * If ``[1, 2, 3]`` ->  try parsing columns 1, 2, 3 each as a separate date
    column.
  * If ``[[1, 3]]`` -> combine columns 1 and 3 and parse as a single date
    column.
  * If ``{'foo': [1, 3]}`` -> parse columns 1, 3 as date and call result 'foo'.

  .. note::
     A fast-path exists for iso8601-formatted dates.
infer_datetime_format : boolean, default ``False``
  If ``True`` and parse_dates is enabled for a column, attempt to infer the
  datetime format to speed up the processing.
keep_date_col : boolean, default ``False``
  If ``True`` and parse_dates specifies combining multiple columns then keep the
  original columns.
date_parser : function, default ``None``
  Function to use for converting a sequence of string columns to an array of
  datetime instances. The default uses ``dateutil.parser.parser`` to do the
  conversion. pandas will try to call date_parser in three different ways,
  advancing to the next if an exception occurs: 1) Pass one or more arrays (as
  defined by parse_dates) as arguments; 2) concatenate (row-wise) the string
  values from the columns defined by parse_dates into a single array and pass
  that; and 3) call date_parser once for each row using one or more strings
  (corresponding to the columns defined by parse_dates) as arguments.
dayfirst : boolean, default ``False``
  DD/MM format dates, international and European format.
cache_dates : boolean, default True
  If True, use a cache of unique, converted dates to apply the datetime
  conversion. May produce significant speed-up when parsing duplicate
  date strings, especially ones with timezone offsets.

  .. versionadded:: 0.25.0

Iteration
+++++++++

iterator : boolean, default ``False``
  Return ``TextFileReader`` object for iteration or getting chunks with
  ``get_chunk()``.
chunksize : int, default ``None``
  Return ``TextFileReader`` object for iteration. See :ref:`iterating and chunking
  <io.chunking>` below.

Quoting, compression, and file format
+++++++++++++++++++++++++++++++++++++

compression : {``'infer'``, ``'gzip'``, ``'bz2'``, ``'zip'``, ``'xz'``, ``'zstd'``, ``None``, ``dict``}, default ``'infer'``
  For on-the-fly decompression of on-disk data. If 'infer', then use gzip,
  bz2, zip, xz, or zstandard if ``filepath_or_buffer`` is path-like ending in '.gz', '.bz2',
  '.zip', '.xz', '.zst', respectively, and no decompression otherwise. If using 'zip',
  the ZIP file must contain only one data file to be read in.
  Set to ``None`` for no decompression. Can also be a dict with key ``'method'``
  set to one of {``'zip'``, ``'gzip'``, ``'bz2'``, ``'zstd'``} and other key-value pairs are
  forwarded to ``zipfile.ZipFile``, ``gzip.GzipFile``, ``bz2.BZ2File``, or ``zstandard.ZstdDecompressor``.
  As an example, the following could be passed for faster compression and to
  create a reproducible gzip archive:
  ``compression={'method': 'gzip', 'compresslevel': 1, 'mtime': 1}``.

  .. versionchanged:: 1.1.0 dict option extended to support ``gzip`` and ``bz2``.
  .. versionchanged:: 1.2.0 Previous versions forwarded dict entries for 'gzip' to ``gzip.open``.
thousands : str, default ``None``
  Thousands separator.
decimal : str, default ``'.'``
  Character to recognize as decimal point. E.g. use ``','`` for European data.
float_precision : string, default None
  Specifies which converter the C engine should use for floating-point values.
  The options are ``None`` for the ordinary converter, ``high`` for the
  high-precision converter, and ``round_trip`` for the round-trip converter.
lineterminator : str (length 1), default ``None``
  Character to break file into lines. Only valid with C parser.
quotechar : str (length 1)
  The character used to denote the start and end of a quoted item. Quoted items
  can include the delimiter and it will be ignored.
quoting : int or ``csv.QUOTE_*`` instance, default ``0``
  Control field quoting behavior per ``csv.QUOTE_*`` constants. Use one of
  ``QUOTE_MINIMAL`` (0), ``QUOTE_ALL`` (1), ``QUOTE_NONNUMERIC`` (2) or
  ``QUOTE_NONE`` (3).
doublequote : boolean, default ``True``
   When ``quotechar`` is specified and ``quoting`` is not ``QUOTE_NONE``,
   indicate whether or not to interpret two consecutive ``quotechar`` elements
   **inside** a field as a single ``quotechar`` element.
escapechar : str (length 1), default ``None``
  One-character string used to escape delimiter when quoting is ``QUOTE_NONE``.
comment : str, default ``None``
  Indicates remainder of line should not be parsed. If found at the beginning of
  a line, the line will be ignored altogether. This parameter must be a single
  character. Like empty lines (as long as ``skip_blank_lines=True``), fully
  commented lines are ignored by the parameter ``header`` but not by ``skiprows``.
  For example, if ``comment='#'``, parsing '#empty\\na,b,c\\n1,2,3' with
  ``header=0`` will result in 'a,b,c' being treated as the header.
encoding : str, default ``None``
  Encoding to use for UTF when reading/writing (e.g. ``'utf-8'``). `List of
  Python standard encodings
  <https://docs.python.org/3/library/codecs.html#standard-encodings>`_.
dialect : str or :class:`python:csv.Dialect` instance, default ``None``
  If provided, this parameter will override values (default or not) for the
  following parameters: ``delimiter``, ``doublequote``, ``escapechar``,
  ``skipinitialspace``, ``quotechar``, and ``quoting``. If it is necessary to
  override values, a ParserWarning will be issued. See :class:`python:csv.Dialect`
  documentation for more details.

Error handling
++++++++++++++

error_bad_lines : boolean, optional, default ``None``
  Lines with too many fields (e.g. a csv line with too many commas) will by
  default cause an exception to be raised, and no ``DataFrame`` will be
  returned. If ``False``, then these "bad lines" will dropped from the
  ``DataFrame`` that is returned. See :ref:`bad lines <io.bad_lines>`
  below.

  .. deprecated:: 1.3.0
     The ``on_bad_lines`` parameter should be used instead to specify behavior upon
     encountering a bad line instead.
warn_bad_lines : boolean, optional, default ``None``
  If error_bad_lines is ``False``, and warn_bad_lines is ``True``, a warning for
  each "bad line" will be output.

  .. deprecated:: 1.3.0
     The ``on_bad_lines`` parameter should be used instead to specify behavior upon
     encountering a bad line instead.
on_bad_lines : {{'error', 'warn', 'skip'}}, default 'error'
    Specifies what to do upon encountering a bad line (a line with too many fields).
    Allowed values are :

        - 'error', raise an ParserError when a bad line is encountered.
        - 'warn', print a warning when a bad line is encountered and skip that line.
        - 'skip', skip bad lines without raising or warning when they are encountered.

    .. versionadded:: 1.3.0

.. _io.dtypes:

Specifying column data types
''''''''''''''''''''''''''''

You can indicate the data type for the whole ``DataFrame`` or individual
columns:

.. ipython:: python

    import numpy as np

    data = "a,b,c,d\n1,2,3,4\n5,6,7,8\n9,10,11"
    print(data)

    df = pd.read_csv(StringIO(data), dtype=object)
    df
    df["a"][0]
    df = pd.read_csv(StringIO(data), dtype={"b": object, "c": np.float64, "d": "Int64"})
    df.dtypes

Fortunately, pandas offers more than one way to ensure that your column(s)
contain only one ``dtype``. If you're unfamiliar with these concepts, you can
see :ref:`here<basics.dtypes>` to learn more about dtypes, and
:ref:`here<basics.object_conversion>` to learn more about ``object`` conversion in
pandas.


For instance, you can use the ``converters`` argument
of :func:`~pandas.read_csv`:

.. ipython:: python

    data = "col_1\n1\n2\n'A'\n4.22"
    df = pd.read_csv(StringIO(data), converters={"col_1": str})
    df
    df["col_1"].apply(type).value_counts()

Or you can use the :func:`~pandas.to_numeric` function to coerce the
dtypes after reading in the data,

.. ipython:: python

    df2 = pd.read_csv(StringIO(data))
    df2["col_1"] = pd.to_numeric(df2["col_1"], errors="coerce")
    df2
    df2["col_1"].apply(type).value_counts()

which will convert all valid parsing to floats, leaving the invalid parsing
as ``NaN``.

Ultimately, how you deal with reading in columns containing mixed dtypes
depends on your specific needs. In the case above, if you wanted to ``NaN`` out
the data anomalies, then :func:`~pandas.to_numeric` is probably your best option.
However, if you wanted for all the data to be coerced, no matter the type, then
using the ``converters`` argument of :func:`~pandas.read_csv` would certainly be
worth trying.

.. note::
   In some cases, reading in abnormal data with columns containing mixed dtypes
   will result in an inconsistent dataset. If you rely on pandas to infer the
   dtypes of your columns, the parsing engine will go and infer the dtypes for
   different chunks of the data, rather than the whole dataset at once. Consequently,
   you can end up with column(s) with mixed dtypes. For example,

   .. ipython:: python
        :okwarning:

        col_1 = list(range(500000)) + ["a", "b"] + list(range(500000))
        df = pd.DataFrame({"col_1": col_1})
        df.to_csv("foo.csv")
        mixed_df = pd.read_csv("foo.csv")
        mixed_df["col_1"].apply(type).value_counts()
        mixed_df["col_1"].dtype

   will result with ``mixed_df`` containing an ``int`` dtype for certain chunks
   of the column, and ``str`` for others due to the mixed dtypes from the
   data that was read in. It is important to note that the overall column will be
   marked with a ``dtype`` of ``object``, which is used for columns with mixed dtypes.

.. ipython:: python
   :suppress:

   import os

   os.remove("foo.csv")

.. _io.categorical:

Specifying categorical dtype
''''''''''''''''''''''''''''

``Categorical`` columns can be parsed directly by specifying ``dtype='category'`` or
``dtype=CategoricalDtype(categories, ordered)``.

.. ipython:: python

   data = "col1,col2,col3\na,b,1\na,b,2\nc,d,3"

   pd.read_csv(StringIO(data))
   pd.read_csv(StringIO(data)).dtypes
   pd.read_csv(StringIO(data), dtype="category").dtypes

Individual columns can be parsed as a ``Categorical`` using a dict
specification:

.. ipython:: python

   pd.read_csv(StringIO(data), dtype={"col1": "category"}).dtypes

Specifying ``dtype='category'`` will result in an unordered ``Categorical``
whose ``categories`` are the unique values observed in the data. For more
control on the categories and order, create a
:class:`~pandas.api.types.CategoricalDtype` ahead of time, and pass that for
that column's ``dtype``.

.. ipython:: python

   from pandas.api.types import CategoricalDtype

   dtype = CategoricalDtype(["d", "c", "b", "a"], ordered=True)
   pd.read_csv(StringIO(data), dtype={"col1": dtype}).dtypes

When using ``dtype=CategoricalDtype``, "unexpected" values outside of
``dtype.categories`` are treated as missing values.

.. ipython:: python

   dtype = CategoricalDtype(["a", "b", "d"])  # No 'c'
   pd.read_csv(StringIO(data), dtype={"col1": dtype}).col1

This matches the behavior of :meth:`Categorical.set_categories`.

.. note::

   With ``dtype='category'``, the resulting categories will always be parsed
   as strings (object dtype). If the categories are numeric they can be
   converted using the :func:`to_numeric` function, or as appropriate, another
   converter such as :func:`to_datetime`.

   When ``dtype`` is a ``CategoricalDtype`` with homogeneous ``categories`` (
   all numeric, all datetimes, etc.), the conversion is done automatically.

   .. ipython:: python

      df = pd.read_csv(StringIO(data), dtype="category")
      df.dtypes
      df["col3"]
      new_categories = pd.to_numeric(df["col3"].cat.categories)
      df["col3"] = df["col3"].cat.rename_categories(new_categories)
      df["col3"]


Naming and using columns
''''''''''''''''''''''''

.. _io.headers:

Handling column names
+++++++++++++++++++++

A file may or may not have a header row. pandas assumes the first row should be
used as the column names:

.. ipython:: python

    data = "a,b,c\n1,2,3\n4,5,6\n7,8,9"
    print(data)
    pd.read_csv(StringIO(data))

By specifying the ``names`` argument in conjunction with ``header`` you can
indicate other names to use and whether or not to throw away the header row (if
any):

.. ipython:: python

    print(data)
    pd.read_csv(StringIO(data), names=["foo", "bar", "baz"], header=0)
    pd.read_csv(StringIO(data), names=["foo", "bar", "baz"], header=None)

If the header is in a row other than the first, pass the row number to
``header``. This will skip the preceding rows:

.. ipython:: python

    data = "skip this skip it\na,b,c\n1,2,3\n4,5,6\n7,8,9"
    pd.read_csv(StringIO(data), header=1)

.. note::

  Default behavior is to infer the column names: if no names are
  passed the behavior is identical to ``header=0`` and column names
  are inferred from the first non-blank line of the file, if column
  names are passed explicitly then the behavior is identical to
  ``header=None``.

.. _io.dupe_names:

Duplicate names parsing
'''''''''''''''''''''''

  .. deprecated:: 1.5.0
     ``mangle_dupe_cols`` was never implemented, and a new argument where the
     renaming pattern can be specified will be added instead.

If the file or header contains duplicate names, pandas will by default
distinguish between them so as to prevent overwriting data:

.. ipython:: python

   data = "a,b,a\n0,1,2\n3,4,5"
   pd.read_csv(StringIO(data))

There is no more duplicate data because ``mangle_dupe_cols=True`` by default,
which modifies a series of duplicate columns 'X', ..., 'X' to become
'X', 'X.1', ..., 'X.N'.

.. _io.usecols:

Filtering columns (``usecols``)
+++++++++++++++++++++++++++++++

The ``usecols`` argument allows you to select any subset of the columns in a
file, either using the column names, position numbers or a callable:

.. ipython:: python

    data = "a,b,c,d\n1,2,3,foo\n4,5,6,bar\n7,8,9,baz"
    pd.read_csv(StringIO(data))
    pd.read_csv(StringIO(data), usecols=["b", "d"])
    pd.read_csv(StringIO(data), usecols=[0, 2, 3])
    pd.read_csv(StringIO(data), usecols=lambda x: x.upper() in ["A", "C"])

The ``usecols`` argument can also be used to specify which columns not to
use in the final result:

.. ipython:: python

   pd.read_csv(StringIO(data), usecols=lambda x: x not in ["a", "c"])

In this case, the callable is specifying that we exclude the "a" and "c"
columns from the output.

Comments and empty lines
''''''''''''''''''''''''

.. _io.skiplines:

Ignoring line comments and empty lines
++++++++++++++++++++++++++++++++++++++

If the ``comment`` parameter is specified, then completely commented lines will
be ignored. By default, completely blank lines will be ignored as well.

.. ipython:: python

   data = "\na,b,c\n  \n# commented line\n1,2,3\n\n4,5,6"
   print(data)
   pd.read_csv(StringIO(data), comment="#")

If ``skip_blank_lines=False``, then ``read_csv`` will not ignore blank lines:

.. ipython:: python

   data = "a,b,c\n\n1,2,3\n\n\n4,5,6"
   pd.read_csv(StringIO(data), skip_blank_lines=False)

.. warning::

   The presence of ignored lines might create ambiguities involving line numbers;
   the parameter ``header`` uses row numbers (ignoring commented/empty
   lines), while ``skiprows`` uses line numbers (including commented/empty lines):

   .. ipython:: python

      data = "#comment\na,b,c\nA,B,C\n1,2,3"
      pd.read_csv(StringIO(data), comment="#", header=1)
      data = "A,B,C\n#comment\na,b,c\n1,2,3"
      pd.read_csv(StringIO(data), comment="#", skiprows=2)

   If both ``header`` and ``skiprows`` are specified, ``header`` will be
   relative to the end of ``skiprows``. For example:

.. ipython:: python

   data = (
       "# empty\n"
       "# second empty line\n"
       "# third emptyline\n"
       "X,Y,Z\n"
       "1,2,3\n"
       "A,B,C\n"
       "1,2.,4.\n"
       "5.,NaN,10.0\n"
   )
   print(data)
   pd.read_csv(StringIO(data), comment="#", skiprows=4, header=1)

.. _io.comments:

Comments
++++++++

Sometimes comments or meta data may be included in a file:

.. ipython:: python
   :suppress:

   data = (
       "ID,level,category\n"
       "Patient1,123000,x # really unpleasant\n"
       "Patient2,23000,y # wouldn't take his medicine\n"
       "Patient3,1234018,z # awesome"
   )

   with open("tmp.csv", "w") as fh:
       fh.write(data)

.. ipython:: python

   print(open("tmp.csv").read())

By default, the parser includes the comments in the output:

.. ipython:: python

   df = pd.read_csv("tmp.csv")
   df

We can suppress the comments using the ``comment`` keyword:

.. ipython:: python

   df = pd.read_csv("tmp.csv", comment="#")
   df

.. ipython:: python
   :suppress:

   os.remove("tmp.csv")

.. _io.unicode:

Dealing with Unicode data
'''''''''''''''''''''''''

The ``encoding`` argument should be used for encoded unicode data, which will
result in byte strings being decoded to unicode in the result:

.. ipython:: python

   from io import BytesIO

   data = b"word,length\n" b"Tr\xc3\xa4umen,7\n" b"Gr\xc3\xbc\xc3\x9fe,5"
   data = data.decode("utf8").encode("latin-1")
   df = pd.read_csv(BytesIO(data), encoding="latin-1")
   df
   df["word"][1]

Some formats which encode all characters as multiple bytes, like UTF-16, won't
parse correctly at all without specifying the encoding. `Full list of Python
standard encodings
<https://docs.python.org/3/library/codecs.html#standard-encodings>`_.

.. _io.index_col:

Index columns and trailing delimiters
'''''''''''''''''''''''''''''''''''''

If a file has one more column of data than the number of column names, the
first column will be used as the ``DataFrame``'s row names:

.. ipython:: python

    data = "a,b,c\n4,apple,bat,5.7\n8,orange,cow,10"
    pd.read_csv(StringIO(data))

.. ipython:: python

    data = "index,a,b,c\n4,apple,bat,5.7\n8,orange,cow,10"
    pd.read_csv(StringIO(data), index_col=0)

Ordinarily, you can achieve this behavior using the ``index_col`` option.

There are some exception cases when a file has been prepared with delimiters at
the end of each data line, confusing the parser. To explicitly disable the
index column inference and discard the last column, pass ``index_col=False``:

.. ipython:: python

    data = "a,b,c\n4,apple,bat,\n8,orange,cow,"
    print(data)
    pd.read_csv(StringIO(data))
    pd.read_csv(StringIO(data), index_col=False)

If a subset of data is being parsed using the ``usecols`` option, the
``index_col`` specification is based on that subset, not the original data.

.. ipython:: python

    data = "a,b,c\n4,apple,bat,\n8,orange,cow,"
    print(data)
    pd.read_csv(StringIO(data), usecols=["b", "c"])
    pd.read_csv(StringIO(data), usecols=["b", "c"], index_col=0)

.. _io.parse_dates:

Date Handling
'''''''''''''

Specifying date columns
+++++++++++++++++++++++

To better facilitate working with datetime data, :func:`read_csv`
uses the keyword arguments ``parse_dates`` and ``date_parser``
to allow users to specify a variety of columns and date/time formats to turn the
input text data into ``datetime`` objects.

The simplest case is to just pass in ``parse_dates=True``:

.. ipython:: python

   with open("foo.csv", mode="w") as f:
       f.write("date,A,B,C\n20090101,a,1,2\n20090102,b,3,4\n20090103,c,4,5")

   # Use a column as an index, and parse it as dates.
   df = pd.read_csv("foo.csv", index_col=0, parse_dates=True)
   df

   # These are Python datetime objects
   df.index

It is often the case that we may want to store date and time data separately,
or store various date fields separately. the ``parse_dates`` keyword can be
used to specify a combination of columns to parse the dates and/or times from.

You can specify a list of column lists to ``parse_dates``, the resulting date
columns will be prepended to the output (so as to not affect the existing column
order) and the new column names will be the concatenation of the component
column names:

.. ipython:: python

   data = (
       "KORD,19990127, 19:00:00, 18:56:00, 0.8100\n"
       "KORD,19990127, 20:00:00, 19:56:00, 0.0100\n"
       "KORD,19990127, 21:00:00, 20:56:00, -0.5900\n"
       "KORD,19990127, 21:00:00, 21:18:00, -0.9900\n"
       "KORD,19990127, 22:00:00, 21:56:00, -0.5900\n"
       "KORD,19990127, 23:00:00, 22:56:00, -0.5900"
   )

   with open("tmp.csv", "w") as fh:
       fh.write(data)

    df = pd.read_csv("tmp.csv", header=None, parse_dates=[[1, 2], [1, 3]])
    df

By default the parser removes the component date columns, but you can choose
to retain them via the ``keep_date_col`` keyword:

.. ipython:: python

   df = pd.read_csv(
       "tmp.csv", header=None, parse_dates=[[1, 2], [1, 3]], keep_date_col=True
   )
   df

Note that if you wish to combine multiple columns into a single date column, a
nested list must be used. In other words, ``parse_dates=[1, 2]`` indicates that
the second and third columns should each be parsed as separate date columns
while ``parse_dates=[[1, 2]]`` means the two columns should be parsed into a
single column.

You can also use a dict to specify custom name columns:

.. ipython:: python

   date_spec = {"nominal": [1, 2], "actual": [1, 3]}
   df = pd.read_csv("tmp.csv", header=None, parse_dates=date_spec)
   df

It is important to remember that if multiple text columns are to be parsed into
a single date column, then a new column is prepended to the data. The ``index_col``
specification is based off of this new set of columns rather than the original
data columns:


.. ipython:: python

   date_spec = {"nominal": [1, 2], "actual": [1, 3]}
   df = pd.read_csv(
       "tmp.csv", header=None, parse_dates=date_spec, index_col=0
   )  # index is the nominal column
   df

.. note::
   If a column or index contains an unparsable date, the entire column or
   index will be returned unaltered as an object data type. For non-standard
   datetime parsing, use :func:`to_datetime` after ``pd.read_csv``.


.. note::
   read_csv has a fast_path for parsing datetime strings in iso8601 format,
   e.g "2000-01-01T00:01:02+00:00" and similar variations. If you can arrange
   for your data to store datetimes in this format, load times will be
   significantly faster, ~20x has been observed.


Date parsing functions
++++++++++++++++++++++

Finally, the parser allows you to specify a custom ``date_parser`` function to
take full advantage of the flexibility of the date parsing API:

.. ipython:: python

   df = pd.read_csv(
       "tmp.csv", header=None, parse_dates=date_spec, date_parser=pd.to_datetime
   )
   df

pandas will try to call the ``date_parser`` function in three different ways. If
an exception is raised, the next one is tried:

1. ``date_parser`` is first called with one or more arrays as arguments,
   as defined using ``parse_dates`` (e.g., ``date_parser(['2013', '2013'], ['1', '2'])``).

2. If #1 fails, ``date_parser`` is called with all the columns
   concatenated row-wise into a single array (e.g., ``date_parser(['2013 1', '2013 2'])``).

Note that performance-wise, you should try these methods of parsing dates in order:

1. Try to infer the format using ``infer_datetime_format=True`` (see section below).

2. If you know the format, use ``pd.to_datetime()``:
   ``date_parser=lambda x: pd.to_datetime(x, format=...)``.

3. If you have a really non-standard format, use a custom ``date_parser`` function.
   For optimal performance, this should be vectorized, i.e., it should accept arrays
   as arguments.


.. ipython:: python
   :suppress:

   os.remove("tmp.csv")


.. _io.csv.mixed_timezones:

Parsing a CSV with mixed timezones
++++++++++++++++++++++++++++++++++

pandas cannot natively represent a column or index with mixed timezones. If your CSV
file contains columns with a mixture of timezones, the default result will be
an object-dtype column with strings, even with ``parse_dates``.


.. ipython:: python

   content = """\
   a
   2000-01-01T00:00:00+05:00
   2000-01-01T00:00:00+06:00"""
   df = pd.read_csv(StringIO(content), parse_dates=["a"])
   df["a"]

To parse the mixed-timezone values as a datetime column, pass a partially-applied
:func:`to_datetime` with ``utc=True`` as the ``date_parser``.

.. ipython:: python

   df = pd.read_csv(
       StringIO(content),
       parse_dates=["a"],
       date_parser=lambda col: pd.to_datetime(col, utc=True),
   )
   df["a"]


.. _io.dayfirst:


Inferring datetime format
+++++++++++++++++++++++++

If you have ``parse_dates`` enabled for some or all of your columns, and your
datetime strings are all formatted the same way, you may get a large speed
up by setting ``infer_datetime_format=True``.  If set, pandas will attempt
to guess the format of your datetime strings, and then use a faster means
of parsing the strings.  5-10x parsing speeds have been observed.  pandas
will fallback to the usual parsing if either the format cannot be guessed
or the format that was guessed cannot properly parse the entire column
of strings.  So in general, ``infer_datetime_format`` should not have any
negative consequences if enabled.

Here are some examples of datetime strings that can be guessed (All
representing December 30th, 2011 at 00:00:00):

* "20111230"
* "2011/12/30"
* "20111230 00:00:00"
* "12/30/2011 00:00:00"
* "30/Dec/2011 00:00:00"
* "30/December/2011 00:00:00"

Note that ``infer_datetime_format`` is sensitive to ``dayfirst``.  With
``dayfirst=True``, it will guess "01/12/2011" to be December 1st. With
``dayfirst=False`` (default) it will guess "01/12/2011" to be January 12th.

.. ipython:: python

   # Try to infer the format for the index column
   df = pd.read_csv(
       "foo.csv",
       index_col=0,
       parse_dates=True,
       infer_datetime_format=True,
   )
   df

.. ipython:: python
   :suppress:

   os.remove("foo.csv")

International date formats
++++++++++++++++++++++++++

While US date formats tend to be MM/DD/YYYY, many international formats use
DD/MM/YYYY instead. For convenience, a ``dayfirst`` keyword is provided:

.. ipython:: python

   data = "date,value,cat\n1/6/2000,5,a\n2/6/2000,10,b\n3/6/2000,15,c"
   print(data)
   with open("tmp.csv", "w") as fh:
       fh.write(data)

   pd.read_csv("tmp.csv", parse_dates=[0])
   pd.read_csv("tmp.csv", dayfirst=True, parse_dates=[0])

.. ipython:: python
   :suppress:

   os.remove("tmp.csv")

Writing CSVs to binary file objects
+++++++++++++++++++++++++++++++++++

.. versionadded:: 1.2.0

``df.to_csv(..., mode="wb")`` allows writing a CSV to a file object
opened binary mode. In most cases, it is not necessary to specify
``mode`` as Pandas will auto-detect whether the file object is
opened in text or binary mode.

.. ipython:: python

   import io

   data = pd.DataFrame([0, 1, 2])
   buffer = io.BytesIO()
   data.to_csv(buffer, encoding="utf-8", compression="gzip")

.. _io.float_precision:

Specifying method for floating-point conversion
'''''''''''''''''''''''''''''''''''''''''''''''

The parameter ``float_precision`` can be specified in order to use
a specific floating-point converter during parsing with the C engine.
The options are the ordinary converter, the high-precision converter, and
the round-trip converter (which is guaranteed to round-trip values after
writing to a file). For example:

.. ipython:: python

   val = "0.3066101993807095471566981359501369297504425048828125"
   data = "a,b,c\n1,2,{0}".format(val)
   abs(
       pd.read_csv(
           StringIO(data),
           engine="c",
           float_precision=None,
       )["c"][0] - float(val)
   )
   abs(
       pd.read_csv(
           StringIO(data),
           engine="c",
           float_precision="high",
       )["c"][0] - float(val)
   )
   abs(
       pd.read_csv(StringIO(data), engine="c", float_precision="round_trip")["c"][0]
       - float(val)
   )


.. _io.thousands:

Thousand separators
'''''''''''''''''''

For large numbers that have been written with a thousands separator, you can
set the ``thousands`` keyword to a string of length 1 so that integers will be parsed
correctly:

By default, numbers with a thousands separator will be parsed as strings:

.. ipython:: python

   data = (
       "ID|level|category\n"
       "Patient1|123,000|x\n"
       "Patient2|23,000|y\n"
       "Patient3|1,234,018|z"
   )

   with open("tmp.csv", "w") as fh:
       fh.write(data)

    df = pd.read_csv("tmp.csv", sep="|")
    df

    df.level.dtype

The ``thousands`` keyword allows integers to be parsed correctly:

.. ipython:: python

    df = pd.read_csv("tmp.csv", sep="|", thousands=",")
    df

    df.level.dtype

.. ipython:: python
   :suppress:

   os.remove("tmp.csv")

.. _io.na_values:

NA values
'''''''''

To control which values are parsed as missing values (which are signified by
``NaN``), specify a string in ``na_values``. If you specify a list of strings,
then all values in it are considered to be missing values. If you specify a
number (a ``float``, like ``5.0`` or an ``integer`` like ``5``), the
corresponding equivalent values will also imply a missing value (in this case
effectively ``[5.0, 5]`` are recognized as ``NaN``).

To completely override the default values that are recognized as missing, specify ``keep_default_na=False``.

.. _io.navaluesconst:

The default ``NaN`` recognized values are ``['-1.#IND', '1.#QNAN', '1.#IND', '-1.#QNAN', '#N/A N/A', '#N/A', 'N/A',
'n/a', 'NA', '<NA>', '#NA', 'NULL', 'null', 'NaN', '-NaN', 'nan', '-nan', '']``.

Let us consider some examples:

.. code-block:: python

   pd.read_csv("path_to_file.csv", na_values=[5])

In the example above ``5`` and ``5.0`` will be recognized as ``NaN``, in
addition to the defaults. A string will first be interpreted as a numerical
``5``, then as a ``NaN``.

.. code-block:: python

   pd.read_csv("path_to_file.csv", keep_default_na=False, na_values=[""])

Above, only an empty field will be recognized as ``NaN``.

.. code-block:: python

   pd.read_csv("path_to_file.csv", keep_default_na=False, na_values=["NA", "0"])

Above, both ``NA`` and ``0`` as strings are ``NaN``.

.. code-block:: python

   pd.read_csv("path_to_file.csv", na_values=["Nope"])

The default values, in addition to the string ``"Nope"`` are recognized as
``NaN``.

.. _io.infinity:

Infinity
''''''''

``inf`` like values will be parsed as ``np.inf`` (positive infinity), and ``-inf`` as ``-np.inf`` (negative infinity).
These will ignore the case of the value, meaning ``Inf``, will also be parsed as ``np.inf``.


Returning Series
''''''''''''''''

Using the ``squeeze`` keyword, the parser will return output with a single column
as a ``Series``:

.. deprecated:: 1.4.0
   Users should append ``.squeeze("columns")`` to the DataFrame returned by
   ``read_csv`` instead.

.. ipython:: python
   :okwarning:

   data = "level\nPatient1,123000\nPatient2,23000\nPatient3,1234018"

   with open("tmp.csv", "w") as fh:
       fh.write(data)

   print(open("tmp.csv").read())

   output = pd.read_csv("tmp.csv", squeeze=True)
   output

   type(output)

.. ipython:: python
   :suppress:

   os.remove("tmp.csv")

.. _io.boolean:

Boolean values
''''''''''''''

The common values ``True``, ``False``, ``TRUE``, and ``FALSE`` are all
recognized as boolean. Occasionally you might want to recognize other values
as being boolean. To do this, use the ``true_values`` and ``false_values``
options as follows:

.. ipython:: python

    data = "a,b,c\n1,Yes,2\n3,No,4"
    print(data)
    pd.read_csv(StringIO(data))
    pd.read_csv(StringIO(data), true_values=["Yes"], false_values=["No"])

.. _io.bad_lines:

Handling "bad" lines
''''''''''''''''''''

Some files may have malformed lines with too few fields or too many. Lines with
too few fields will have NA values filled in the trailing fields. Lines with
too many fields will raise an error by default:

.. ipython:: python
    :okexcept:

    data = "a,b,c\n1,2,3\n4,5,6,7\n8,9,10"
    pd.read_csv(StringIO(data))

You can elect to skip bad lines:

.. code-block:: ipython

    In [29]: pd.read_csv(StringIO(data), on_bad_lines="warn")
    Skipping line 3: expected 3 fields, saw 4

    Out[29]:
       a  b   c
    0  1  2   3
    1  8  9  10

Or pass a callable function to handle the bad line if ``engine="python"``.
The bad line will be a list of strings that was split by the ``sep``:

.. code-block:: ipython

    In [29]: external_list = []

    In [30]: def bad_lines_func(line):
        ...:     external_list.append(line)
        ...:     return line[-3:]

    In [31]: pd.read_csv(StringIO(data), on_bad_lines=bad_lines_func, engine="python")
    Out[31]:
       a  b   c
    0  1  2   3
    1  5  6   7
    2  8  9  10

    In [32]: external_list
    Out[32]: [4, 5, 6, 7]

    .. versionadded:: 1.4.0


You can also use the ``usecols`` parameter to eliminate extraneous column
data that appear in some lines but not others:

.. code-block:: ipython

   In [33]: pd.read_csv(StringIO(data), usecols=[0, 1, 2])

    Out[33]:
       a  b   c
    0  1  2   3
    1  4  5   6
    2  8  9  10

In case you want to keep all data including the lines with too many fields, you can
specify a sufficient number of ``names``. This ensures that lines with not enough
fields are filled with ``NaN``.

.. code-block:: ipython

   In [34]: pd.read_csv(StringIO(data), names=['a', 'b', 'c', 'd'])

   Out[34]:
       a  b   c  d
    0  1  2   3  NaN
    1  4  5   6  7
    2  8  9  10  NaN

.. _io.dialect:

Dialect
'''''''

The ``dialect`` keyword gives greater flexibility in specifying the file format.
By default it uses the Excel dialect but you can specify either the dialect name
or a :class:`python:csv.Dialect` instance.

Suppose you had data with unenclosed quotes:

.. ipython:: python

   data = "label1,label2,label3\n" 'index1,"a,c,e\n' "index2,b,d,f"
   print(data)

By default, ``read_csv`` uses the Excel dialect and treats the double quote as
the quote character, which causes it to fail when it finds a newline before it
finds the closing double quote.

We can get around this using ``dialect``:

.. ipython:: python
   :okwarning:

   import csv

   dia = csv.excel()
   dia.quoting = csv.QUOTE_NONE
   pd.read_csv(StringIO(data), dialect=dia)

All of the dialect options can be specified separately by keyword arguments:

.. ipython:: python

    data = "a,b,c~1,2,3~4,5,6"
    pd.read_csv(StringIO(data), lineterminator="~")

Another common dialect option is ``skipinitialspace``, to skip any whitespace
after a delimiter:

.. ipython:: python

   data = "a, b, c\n1, 2, 3\n4, 5, 6"
   print(data)
   pd.read_csv(StringIO(data), skipinitialspace=True)

The parsers make every attempt to "do the right thing" and not be fragile. Type
inference is a pretty big deal. If a column can be coerced to integer dtype
without altering the contents, the parser will do so. Any non-numeric
columns will come through as object dtype as with the rest of pandas objects.

.. _io.quoting:

Quoting and Escape Characters
'''''''''''''''''''''''''''''

Quotes (and other escape characters) in embedded fields can be handled in any
number of ways. One way is to use backslashes; to properly parse this data, you
should pass the ``escapechar`` option:

.. ipython:: python

   data = 'a,b\n"hello, \\"Bob\\", nice to see you",5'
   print(data)
   pd.read_csv(StringIO(data), escapechar="\\")

.. _io.fwf_reader:
.. _io.fwf:

Files with fixed width columns
''''''''''''''''''''''''''''''

While :func:`read_csv` reads delimited data, the :func:`read_fwf` function works
with data files that have known and fixed column widths. The function parameters
to ``read_fwf`` are largely the same as ``read_csv`` with two extra parameters, and
a different usage of the ``delimiter`` parameter:

* ``colspecs``: A list of pairs (tuples) giving the extents of the
  fixed-width fields of each line as half-open intervals (i.e.,  [from, to[ ).
  String value 'infer' can be used to instruct the parser to try detecting
  the column specifications from the first 100 rows of the data. Default
  behavior, if not specified, is to infer.
* ``widths``: A list of field widths which can be used instead of 'colspecs'
  if the intervals are contiguous.
* ``delimiter``: Characters to consider as filler characters in the fixed-width file.
  Can be used to specify the filler character of the fields
  if it is not spaces (e.g., '~').

Consider a typical fixed-width data file:

.. ipython:: python

   data1 = (
       "id8141    360.242940   149.910199   11950.7\n"
       "id1594    444.953632   166.985655   11788.4\n"
       "id1849    364.136849   183.628767   11806.2\n"
       "id1230    413.836124   184.375703   11916.8\n"
       "id1948    502.953953   173.237159   12468.3"
   )
   with open("bar.csv", "w") as f:
       f.write(data1)

In order to parse this file into a ``DataFrame``, we simply need to supply the
column specifications to the ``read_fwf`` function along with the file name:

.. ipython:: python

   # Column specifications are a list of half-intervals
   colspecs = [(0, 6), (8, 20), (21, 33), (34, 43)]
   df = pd.read_fwf("bar.csv", colspecs=colspecs, header=None, index_col=0)
   df

Note how the parser automatically picks column names X.<column number> when
``header=None`` argument is specified. Alternatively, you can supply just the
column widths for contiguous columns:

.. ipython:: python

   # Widths are a list of integers
   widths = [6, 14, 13, 10]
   df = pd.read_fwf("bar.csv", widths=widths, header=None)
   df

The parser will take care of extra white spaces around the columns
so it's ok to have extra separation between the columns in the file.

By default, ``read_fwf`` will try to infer the file's ``colspecs`` by using the
first 100 rows of the file. It can do it only in cases when the columns are
aligned and correctly separated by the provided ``delimiter`` (default delimiter
is whitespace).

.. ipython:: python

   df = pd.read_fwf("bar.csv", header=None, index_col=0)
   df

``read_fwf`` supports the ``dtype`` parameter for specifying the types of
parsed columns to be different from the inferred type.

.. ipython:: python

   pd.read_fwf("bar.csv", header=None, index_col=0).dtypes
   pd.read_fwf("bar.csv", header=None, dtype={2: "object"}).dtypes

.. ipython:: python
   :suppress:

   os.remove("bar.csv")


Indexes
'''''''

Files with an "implicit" index column
+++++++++++++++++++++++++++++++++++++

Consider a file with one less entry in the header than the number of data
column:

.. ipython:: python

   data = "A,B,C\n20090101,a,1,2\n20090102,b,3,4\n20090103,c,4,5"
   print(data)
   with open("foo.csv", "w") as f:
       f.write(data)

In this special case, ``read_csv`` assumes that the first column is to be used
as the index of the ``DataFrame``:

.. ipython:: python

   pd.read_csv("foo.csv")

Note that the dates weren't automatically parsed. In that case you would need
to do as before:

.. ipython:: python

   df = pd.read_csv("foo.csv", parse_dates=True)
   df.index

.. ipython:: python
   :suppress:

   os.remove("foo.csv")


Reading an index with a ``MultiIndex``
++++++++++++++++++++++++++++++++++++++

.. _io.csv_multiindex:

Suppose you have data indexed by two columns:

.. ipython:: python

   data = 'year,indiv,zit,xit\n1977,"A",1.2,.6\n1977,"B",1.5,.5'
   print(data)
   with open("mindex_ex.csv", mode="w") as f:
       f.write(data)

The ``index_col`` argument to ``read_csv`` can take a list of
column numbers to turn multiple columns into a ``MultiIndex`` for the index of the
returned object:

.. ipython:: python

   df = pd.read_csv("mindex_ex.csv", index_col=[0, 1])
   df
   df.loc[1977]

.. ipython:: python
   :suppress:

   os.remove("mindex_ex.csv")

.. _io.multi_index_columns:

Reading columns with a ``MultiIndex``
+++++++++++++++++++++++++++++++++++++

By specifying list of row locations for the ``header`` argument, you
can read in a ``MultiIndex`` for the columns. Specifying non-consecutive
rows will skip the intervening rows.

.. ipython:: python

   from pandas._testing import makeCustomDataframe as mkdf

   df = mkdf(5, 3, r_idx_nlevels=2, c_idx_nlevels=4)
   df.to_csv("mi.csv")
   print(open("mi.csv").read())
   pd.read_csv("mi.csv", header=[0, 1, 2, 3], index_col=[0, 1])

``read_csv`` is also able to interpret a more common format
of multi-columns indices.

.. ipython:: python

   data = ",a,a,a,b,c,c\n,q,r,s,t,u,v\none,1,2,3,4,5,6\ntwo,7,8,9,10,11,12"
   print(data)
   with open("mi2.csv", "w") as fh:
       fh.write(data)

   pd.read_csv("mi2.csv", header=[0, 1], index_col=0)

.. note::
   If an ``index_col`` is not specified (e.g. you don't have an index, or wrote it
   with ``df.to_csv(..., index=False)``, then any ``names`` on the columns index will
   be *lost*.

.. ipython:: python
   :suppress:

   os.remove("mi.csv")
   os.remove("mi2.csv")

.. _io.sniff:

Automatically "sniffing" the delimiter
''''''''''''''''''''''''''''''''''''''

``read_csv`` is capable of inferring delimited (not necessarily
comma-separated) files, as pandas uses the :class:`python:csv.Sniffer`
class of the csv module. For this, you have to specify ``sep=None``.

.. ipython:: python

   df = pd.DataFrame(np.random.randn(10, 4))
   df.to_csv("tmp.csv", sep="|")
   df.to_csv("tmp2.csv", sep=":")
   pd.read_csv("tmp2.csv", sep=None, engine="python")

.. ipython:: python
   :suppress:

   os.remove("tmp2.csv")

.. _io.multiple_files:

Reading multiple files to create a single DataFrame
'''''''''''''''''''''''''''''''''''''''''''''''''''

It's best to use :func:`~pandas.concat` to combine multiple files.
See the :ref:`cookbook<cookbook.csv.multiple_files>` for an example.

.. _io.chunking:

Iterating through files chunk by chunk
''''''''''''''''''''''''''''''''''''''

Suppose you wish to iterate through a (potentially very large) file lazily
rather than reading the entire file into memory, such as the following:


.. ipython:: python

   df = pd.DataFrame(np.random.randn(10, 4))
   df.to_csv("tmp.csv", sep="|")
   table = pd.read_csv("tmp.csv", sep="|")
   table


By specifying a ``chunksize`` to ``read_csv``, the return
value will be an iterable object of type ``TextFileReader``:

.. ipython:: python

   with pd.read_csv("tmp.csv", sep="|", chunksize=4) as reader:
       reader
       for chunk in reader:
           print(chunk)

.. versionchanged:: 1.2

  ``read_csv/json/sas`` return a context-manager when iterating through a file.

Specifying ``iterator=True`` will also return the ``TextFileReader`` object:

.. ipython:: python

   with pd.read_csv("tmp.csv", sep="|", iterator=True) as reader:
       reader.get_chunk(5)

.. ipython:: python
   :suppress:

   os.remove("tmp.csv")

Specifying the parser engine
''''''''''''''''''''''''''''

Pandas currently supports three engines, the C engine, the python engine, and an experimental
pyarrow engine (requires the ``pyarrow`` package). In general, the pyarrow engine is fastest
on larger workloads and is equivalent in speed to the C engine on most other workloads.
The python engine tends to be slower than the pyarrow and C engines on most workloads. However,
the pyarrow engine is much less robust than the C engine, which lacks a few features compared to the
Python engine.

Where possible, pandas uses the C parser (specified as ``engine='c'``), but it may fall
back to Python if C-unsupported options are specified.

Currently, options unsupported by the C and pyarrow engines include:

* ``sep`` other than a single character (e.g. regex separators)
* ``skipfooter``
* ``sep=None`` with ``delim_whitespace=False``

Specifying any of the above options will produce a ``ParserWarning`` unless the
python engine is selected explicitly using ``engine='python'``.

Options that are unsupported by the pyarrow engine which are not covered by the list above include:

* ``float_precision``
* ``chunksize``
* ``comment``
* ``nrows``
* ``thousands``
* ``memory_map``
* ``dialect``
* ``warn_bad_lines``
* ``error_bad_lines``
* ``on_bad_lines``
* ``delim_whitespace``
* ``quoting``
* ``lineterminator``
* ``converters``
* ``decimal``
* ``iterator``
* ``dayfirst``
* ``infer_datetime_format``
* ``verbose``
* ``skipinitialspace``
* ``low_memory``

Specifying these options with ``engine='pyarrow'`` will raise a ``ValueError``.

.. _io.remote:

Reading/writing remote files
''''''''''''''''''''''''''''

You can pass in a URL to read or write remote files to many of pandas' IO
functions - the following example shows reading a CSV file:

.. code-block:: python

   df = pd.read_csv("https://download.bls.gov/pub/time.series/cu/cu.item", sep="\t")

.. versionadded:: 1.3.0

A custom header can be sent alongside HTTP(s) requests by passing a dictionary
of header key value mappings to the ``storage_options`` keyword argument as shown below:

.. code-block:: python

   headers = {"User-Agent": "pandas"}
   df = pd.read_csv(
       "https://download.bls.gov/pub/time.series/cu/cu.item",
       sep="\t",
       storage_options=headers
   )

All URLs which are not local files or HTTP(s) are handled by
`fsspec`_, if installed, and its various filesystem implementations
(including Amazon S3, Google Cloud, SSH, FTP, webHDFS...).
Some of these implementations will require additional packages to be
installed, for example
S3 URLs require the `s3fs
<https://pypi.org/project/s3fs/>`_ library:

.. code-block:: python

   df = pd.read_json("s3://pandas-test/adatafile.json")

When dealing with remote storage systems, you might need
extra configuration with environment variables or config files in
special locations. For example, to access data in your S3 bucket,
you will need to define credentials in one of the several ways listed in
the `S3Fs documentation
<https://s3fs.readthedocs.io/en/latest/#credentials>`_. The same is true
for several of the storage backends, and you should follow the links
at `fsimpl1`_ for implementations built into ``fsspec`` and `fsimpl2`_
for those not included in the main ``fsspec``
distribution.

You can also pass parameters directly to the backend driver. For example,
if you do *not* have S3 credentials, you can still access public data by
specifying an anonymous connection, such as

.. versionadded:: 1.2.0

.. code-block:: python

   pd.read_csv(
       "s3://ncei-wcsd-archive/data/processed/SH1305/18kHz/SaKe2013"
       "-D20130523-T080854_to_SaKe2013-D20130523-T085643.csv",
       storage_options={"anon": True},
   )

``fsspec`` also allows complex URLs, for accessing data in compressed
archives, local caching of files, and more. To locally cache the above
example, you would modify the call to

.. code-block:: python

   pd.read_csv(
       "simplecache::s3://ncei-wcsd-archive/data/processed/SH1305/18kHz/"
       "SaKe2013-D20130523-T080854_to_SaKe2013-D20130523-T085643.csv",
       storage_options={"s3": {"anon": True}},
   )

where we specify that the "anon" parameter is meant for the "s3" part of
the implementation, not to the caching implementation. Note that this caches to a temporary
directory for the duration of the session only, but you can also specify
a permanent store.

.. _fsspec: https://filesystem-spec.readthedocs.io/en/latest/
.. _fsimpl1: https://filesystem-spec.readthedocs.io/en/latest/api.html#built-in-implementations
.. _fsimpl2: https://filesystem-spec.readthedocs.io/en/latest/api.html#other-known-implementations

Writing out data
''''''''''''''''

.. _io.store_in_csv:

Writing to CSV format
+++++++++++++++++++++

The ``Series`` and ``DataFrame`` objects have an instance method ``to_csv`` which
allows storing the contents of the object as a comma-separated-values file. The
function takes a number of arguments. Only the first is required.

* ``path_or_buf``: A string path to the file to write or a file object.  If a file object it must be opened with ``newline=''``
* ``sep`` : Field delimiter for the output file (default ",")
* ``na_rep``: A string representation of a missing value (default '')
* ``float_format``: Format string for floating point numbers
* ``columns``: Columns to write (default None)
* ``header``: Whether to write out the column names (default True)
* ``index``: whether to write row (index) names (default True)
* ``index_label``: Column label(s) for index column(s) if desired. If None
  (default), and ``header`` and ``index`` are True, then the index names are
  used. (A sequence should be given if the ``DataFrame`` uses MultiIndex).
* ``mode`` : Python write mode, default 'w'
* ``encoding``: a string representing the encoding to use if the contents are
  non-ASCII, for Python versions prior to 3
* ``lineterminator``: Character sequence denoting line end (default ``os.linesep``)
* ``quoting``: Set quoting rules as in csv module (default csv.QUOTE_MINIMAL). Note that if you have set a ``float_format`` then floats are converted to strings and csv.QUOTE_NONNUMERIC will treat them as non-numeric
* ``quotechar``: Character used to quote fields (default '"')
* ``doublequote``: Control quoting of ``quotechar`` in fields (default True)
* ``escapechar``: Character used to escape ``sep`` and ``quotechar`` when
  appropriate (default None)
* ``chunksize``: Number of rows to write at a time
* ``date_format``: Format string for datetime objects

Writing a formatted string
++++++++++++++++++++++++++

.. _io.formatting:

The ``DataFrame`` object has an instance method ``to_string`` which allows control
over the string representation of the object. All arguments are optional:

* ``buf`` default None, for example a StringIO object
* ``columns`` default None, which columns to write
* ``col_space`` default None, minimum width of each column.
* ``na_rep`` default ``NaN``, representation of NA value
* ``formatters`` default None, a dictionary (by column) of functions each of
  which takes a single argument and returns a formatted string
* ``float_format`` default None, a function which takes a single (float)
  argument and returns a formatted string; to be applied to floats in the
  ``DataFrame``.
* ``sparsify`` default True, set to False for a ``DataFrame`` with a hierarchical
  index to print every MultiIndex key at each row.
* ``index_names`` default True, will print the names of the indices
* ``index`` default True, will print the index (ie, row labels)
* ``header`` default True, will print the column labels
* ``justify`` default ``left``, will print column headers left- or
  right-justified

The ``Series`` object also has a ``to_string`` method, but with only the ``buf``,
``na_rep``, ``float_format`` arguments. There is also a ``length`` argument
which, if set to ``True``, will additionally output the length of the Series.

.. _io.json:

JSON
----

Read and write ``JSON`` format files and strings.

.. _io.json_writer:

Writing JSON
''''''''''''

A ``Series`` or ``DataFrame`` can be converted to a valid JSON string. Use ``to_json``
with optional parameters:

* ``path_or_buf`` : the pathname or buffer to write the output
  This can be ``None`` in which case a JSON string is returned
* ``orient`` :

  ``Series``:
      * default is ``index``
      * allowed values are {``split``, ``records``, ``index``}

  ``DataFrame``:
      * default is ``columns``
      * allowed values are {``split``, ``records``, ``index``, ``columns``, ``values``, ``table``}

  The format of the JSON string

  .. csv-table::
     :widths: 20, 150
     :delim: ;

     ``split``; dict like {index -> [index], columns -> [columns], data -> [values]}
     ``records``; list like [{column -> value}, ... , {column -> value}]
     ``index``; dict like {index -> {column -> value}}
     ``columns``; dict like {column -> {index -> value}}
     ``values``; just the values array
     ``table``; adhering to the JSON `Table Schema`_

* ``date_format`` : string, type of date conversion, 'epoch' for timestamp, 'iso' for ISO8601.
* ``double_precision`` : The number of decimal places to use when encoding floating point values, default 10.
* ``force_ascii`` : force encoded string to be ASCII, default True.
* ``date_unit`` : The time unit to encode to, governs timestamp and ISO8601 precision. One of 's', 'ms', 'us' or 'ns' for seconds, milliseconds, microseconds and nanoseconds respectively. Default 'ms'.
* ``default_handler`` : The handler to call if an object cannot otherwise be converted to a suitable format for JSON. Takes a single argument, which is the object to convert, and returns a serializable object.
* ``lines`` : If ``records`` orient, then will write each record per line as json.

Note ``NaN``'s, ``NaT``'s and ``None`` will be converted to ``null`` and ``datetime`` objects will be converted based on the ``date_format`` and ``date_unit`` parameters.

.. ipython:: python

   dfj = pd.DataFrame(np.random.randn(5, 2), columns=list("AB"))
   json = dfj.to_json()
   json

Orient options
++++++++++++++

There are a number of different options for the format of the resulting JSON
file / string. Consider the following ``DataFrame`` and ``Series``:

.. ipython:: python

  dfjo = pd.DataFrame(
      dict(A=range(1, 4), B=range(4, 7), C=range(7, 10)),
      columns=list("ABC"),
      index=list("xyz"),
  )
  dfjo
  sjo = pd.Series(dict(x=15, y=16, z=17), name="D")
  sjo

**Column oriented** (the default for ``DataFrame``) serializes the data as
nested JSON objects with column labels acting as the primary index:

.. ipython:: python

  dfjo.to_json(orient="columns")
  # Not available for Series

**Index oriented** (the default for ``Series``) similar to column oriented
but the index labels are now primary:

.. ipython:: python

  dfjo.to_json(orient="index")
  sjo.to_json(orient="index")

**Record oriented** serializes the data to a JSON array of column -> value records,
index labels are not included. This is useful for passing ``DataFrame`` data to plotting
libraries, for example the JavaScript library ``d3.js``:

.. ipython:: python

  dfjo.to_json(orient="records")
  sjo.to_json(orient="records")

**Value oriented** is a bare-bones option which serializes to nested JSON arrays of
values only, column and index labels are not included:

.. ipython:: python

  dfjo.to_json(orient="values")
  # Not available for Series

**Split oriented** serializes to a JSON object containing separate entries for
values, index and columns. Name is also included for ``Series``:

.. ipython:: python

  dfjo.to_json(orient="split")
  sjo.to_json(orient="split")

**Table oriented** serializes to the JSON `Table Schema`_, allowing for the
preservation of metadata including but not limited to dtypes and index names.

.. note::

  Any orient option that encodes to a JSON object will not preserve the ordering of
  index and column labels during round-trip serialization. If you wish to preserve
  label ordering use the ``split`` option as it uses ordered containers.

Date handling
+++++++++++++

Writing in ISO date format:

.. ipython:: python

   dfd = pd.DataFrame(np.random.randn(5, 2), columns=list("AB"))
   dfd["date"] = pd.Timestamp("20130101")
   dfd = dfd.sort_index(axis=1, ascending=False)
   json = dfd.to_json(date_format="iso")
   json

Writing in ISO date format, with microseconds:

.. ipython:: python

   json = dfd.to_json(date_format="iso", date_unit="us")
   json

Epoch timestamps, in seconds:

.. ipython:: python

   json = dfd.to_json(date_format="epoch", date_unit="s")
   json

Writing to a file, with a date index and a date column:

.. ipython:: python

   dfj2 = dfj.copy()
   dfj2["date"] = pd.Timestamp("20130101")
   dfj2["ints"] = list(range(5))
   dfj2["bools"] = True
   dfj2.index = pd.date_range("20130101", periods=5)
   dfj2.to_json("test.json")

   with open("test.json") as fh:
       print(fh.read())

Fallback behavior
+++++++++++++++++

If the JSON serializer cannot handle the container contents directly it will
fall back in the following manner:

* if the dtype is unsupported (e.g. ``np.complex_``) then the ``default_handler``, if provided, will be called
  for each value, otherwise an exception is raised.

* if an object is unsupported it will attempt the following:


    * check if the object has defined a ``toDict`` method and call it.
      A ``toDict`` method should return a ``dict`` which will then be JSON serialized.

    * invoke the ``default_handler`` if one was provided.

    * convert the object to a ``dict`` by traversing its contents. However this will often fail
      with an ``OverflowError`` or give unexpected results.

In general the best approach for unsupported objects or dtypes is to provide a ``default_handler``.
For example:

.. code-block:: python

  >>> DataFrame([1.0, 2.0, complex(1.0, 2.0)]).to_json()  # raises
  RuntimeError: Unhandled numpy dtype 15

can be dealt with by specifying a simple ``default_handler``:

.. ipython:: python

   pd.DataFrame([1.0, 2.0, complex(1.0, 2.0)]).to_json(default_handler=str)

.. _io.json_reader:

Reading JSON
''''''''''''

Reading a JSON string to pandas object can take a number of parameters.
The parser will try to parse a ``DataFrame`` if ``typ`` is not supplied or
is ``None``. To explicitly force ``Series`` parsing, pass ``typ=series``

* ``filepath_or_buffer`` : a **VALID** JSON string or file handle / StringIO. The string could be
  a URL. Valid URL schemes include http, ftp, S3, and file. For file URLs, a host
  is expected. For instance, a local file could be
  file ://localhost/path/to/table.json
* ``typ``    : type of object to recover (series or frame), default 'frame'
* ``orient`` :

  Series :
      * default is ``index``
      * allowed values are {``split``, ``records``, ``index``}

  DataFrame
      * default is ``columns``
      * allowed values are {``split``, ``records``, ``index``, ``columns``, ``values``, ``table``}

  The format of the JSON string

  .. csv-table::
     :widths: 20, 150
     :delim: ;

     ``split``; dict like {index -> [index], columns -> [columns], data -> [values]}
     ``records``; list like [{column -> value}, ... , {column -> value}]
     ``index``; dict like {index -> {column -> value}}
     ``columns``; dict like {column -> {index -> value}}
     ``values``; just the values array
     ``table``; adhering to the JSON `Table Schema`_


* ``dtype`` : if True, infer dtypes, if a dict of column to dtype, then use those, if ``False``, then don't infer dtypes at all, default is True, apply only to the data.
* ``convert_axes`` : boolean, try to convert the axes to the proper dtypes, default is ``True``
* ``convert_dates`` : a list of columns to parse for dates; If ``True``, then try to parse date-like columns, default is ``True``.
* ``keep_default_dates`` : boolean, default ``True``. If parsing dates, then parse the default date-like columns.
* ``numpy`` : direct decoding to NumPy arrays. default is ``False``;
  Supports numeric data only, although labels may be non-numeric. Also note that the JSON ordering **MUST** be the same for each term if ``numpy=True``.
* ``precise_float`` : boolean, default ``False``. Set to enable usage of higher precision (strtod) function when decoding string to double values. Default (``False``) is to use fast but less precise builtin functionality.
* ``date_unit`` : string, the timestamp unit to detect if converting dates. Default
  None. By default the timestamp precision will be detected, if this is not desired
  then pass one of 's', 'ms', 'us' or 'ns' to force timestamp precision to
  seconds, milliseconds, microseconds or nanoseconds respectively.
* ``lines`` : reads file as one json object per line.
* ``encoding`` : The encoding to use to decode py3 bytes.
* ``chunksize`` : when used in combination with ``lines=True``, return a JsonReader which reads in ``chunksize`` lines per iteration.

The parser will raise one of ``ValueError/TypeError/AssertionError`` if the JSON is not parseable.

If a non-default ``orient`` was used when encoding to JSON be sure to pass the same
option here so that decoding produces sensible results, see `Orient Options`_ for an
overview.

Data conversion
+++++++++++++++

The default of ``convert_axes=True``, ``dtype=True``, and ``convert_dates=True``
will try to parse the axes, and all of the data into appropriate types,
including dates. If you need to override specific dtypes, pass a dict to
``dtype``. ``convert_axes`` should only be set to ``False`` if you need to
preserve string-like numbers (e.g. '1', '2') in an axes.

.. note::

  Large integer values may be converted to dates if ``convert_dates=True`` and the data and / or column labels appear 'date-like'. The exact threshold depends on the ``date_unit`` specified. 'date-like' means that the column label meets one of the following criteria:

     * it ends with ``'_at'``
     * it ends with ``'_time'``
     * it begins with ``'timestamp'``
     * it is ``'modified'``
     * it is ``'date'``

.. warning::

   When reading JSON data, automatic coercing into dtypes has some quirks:

     * an index can be reconstructed in a different order from serialization, that is, the returned order is not guaranteed to be the same as before serialization
     * a column that was ``float`` data will be converted to ``integer`` if it can be done safely, e.g. a column of ``1.``
     * bool columns will be converted to ``integer`` on reconstruction

   Thus there are times where you may want to specify specific dtypes via the ``dtype`` keyword argument.

Reading from a JSON string:

.. ipython:: python

   pd.read_json(json)

Reading from a file:

.. ipython:: python

   pd.read_json("test.json")

Don't convert any data (but still convert axes and dates):

.. ipython:: python

   pd.read_json("test.json", dtype=object).dtypes

Specify dtypes for conversion:

.. ipython:: python

   pd.read_json("test.json", dtype={"A": "float32", "bools": "int8"}).dtypes

Preserve string indices:

.. ipython:: python

   si = pd.DataFrame(
       np.zeros((4, 4)), columns=list(range(4)), index=[str(i) for i in range(4)]
   )
   si
   si.index
   si.columns
   json = si.to_json()

   sij = pd.read_json(json, convert_axes=False)
   sij
   sij.index
   sij.columns

Dates written in nanoseconds need to be read back in nanoseconds:

.. ipython:: python

   json = dfj2.to_json(date_unit="ns")

   # Try to parse timestamps as milliseconds -> Won't Work
   dfju = pd.read_json(json, date_unit="ms")
   dfju

   # Let pandas detect the correct precision
   dfju = pd.read_json(json)
   dfju

   # Or specify that all timestamps are in nanoseconds
   dfju = pd.read_json(json, date_unit="ns")
   dfju

The Numpy parameter
+++++++++++++++++++

.. note::
  This param has been deprecated as of version 1.0.0 and will raise a ``FutureWarning``.

  This supports numeric data only. Index and columns labels may be non-numeric, e.g. strings, dates etc.

If ``numpy=True`` is passed to ``read_json`` an attempt will be made to sniff
an appropriate dtype during deserialization and to subsequently decode directly
to NumPy arrays, bypassing the need for intermediate Python objects.

This can provide speedups if you are deserialising a large amount of numeric
data:

.. ipython:: python

   randfloats = np.random.uniform(-100, 1000, 10000)
   randfloats.shape = (1000, 10)
   dffloats = pd.DataFrame(randfloats, columns=list("ABCDEFGHIJ"))

   jsonfloats = dffloats.to_json()

.. ipython:: python

   %timeit pd.read_json(jsonfloats)

.. ipython:: python
   :okwarning:

   %timeit pd.read_json(jsonfloats, numpy=True)

The speedup is less noticeable for smaller datasets:

.. ipython:: python

   jsonfloats = dffloats.head(100).to_json()

.. ipython:: python

   %timeit pd.read_json(jsonfloats)

.. ipython:: python
   :okwarning:

   %timeit pd.read_json(jsonfloats, numpy=True)

.. warning::

   Direct NumPy decoding makes a number of assumptions and may fail or produce
   unexpected output if these assumptions are not satisfied:

    - data is numeric.

    - data is uniform. The dtype is sniffed from the first value decoded.
      A ``ValueError`` may be raised, or incorrect output may be produced
      if this condition is not satisfied.

    - labels are ordered. Labels are only read from the first container, it is assumed
      that each subsequent row / column has been encoded in the same order. This should be satisfied if the
      data was encoded using ``to_json`` but may not be the case if the JSON
      is from another source.

.. ipython:: python
   :suppress:

   os.remove("test.json")

.. _io.json_normalize:

Normalization
'''''''''''''

pandas provides a utility function to take a dict or list of dicts and *normalize* this semi-structured data
into a flat table.

.. ipython:: python

   data = [
       {"id": 1, "name": {"first": "Coleen", "last": "Volk"}},
       {"name": {"given": "Mark", "family": "Regner"}},
       {"id": 2, "name": "Faye Raker"},
   ]
   pd.json_normalize(data)

.. ipython:: python

   data = [
       {
           "state": "Florida",
           "shortname": "FL",
           "info": {"governor": "Rick Scott"},
           "county": [
               {"name": "Dade", "population": 12345},
               {"name": "Broward", "population": 40000},
               {"name": "Palm Beach", "population": 60000},
           ],
       },
       {
           "state": "Ohio",
           "shortname": "OH",
           "info": {"governor": "John Kasich"},
           "county": [
               {"name": "Summit", "population": 1234},
               {"name": "Cuyahoga", "population": 1337},
           ],
       },
   ]

   pd.json_normalize(data, "county", ["state", "shortname", ["info", "governor"]])

The max_level parameter provides more control over which level to end normalization.
With max_level=1 the following snippet normalizes until 1st nesting level of the provided dict.

.. ipython:: python

    data = [
        {
            "CreatedBy": {"Name": "User001"},
            "Lookup": {
                "TextField": "Some text",
                "UserField": {"Id": "ID001", "Name": "Name001"},
            },
            "Image": {"a": "b"},
        }
    ]
    pd.json_normalize(data, max_level=1)

.. _io.jsonl:

Line delimited json
'''''''''''''''''''

pandas is able to read and write line-delimited json files that are common in data processing pipelines
using Hadoop or Spark.

For line-delimited json files, pandas can also return an iterator which reads in ``chunksize`` lines at a time. This can be useful for large files or to read from a stream.

.. ipython:: python

  jsonl = """
      {"a": 1, "b": 2}
      {"a": 3, "b": 4}
  """
  df = pd.read_json(jsonl, lines=True)
  df
  df.to_json(orient="records", lines=True)

  # reader is an iterator that returns ``chunksize`` lines each iteration
  with pd.read_json(StringIO(jsonl), lines=True, chunksize=1) as reader:
      reader
      for chunk in reader:
          print(chunk)

.. _io.table_schema:

Table schema
''''''''''''

`Table Schema`_ is a spec for describing tabular datasets as a JSON
object. The JSON includes information on the field names, types, and
other attributes. You can use the orient ``table`` to build
a JSON string with two fields, ``schema`` and ``data``.

.. ipython:: python

   df = pd.DataFrame(
       {
           "A": [1, 2, 3],
           "B": ["a", "b", "c"],
           "C": pd.date_range("2016-01-01", freq="d", periods=3),
       },
       index=pd.Index(range(3), name="idx"),
   )
   df
   df.to_json(orient="table", date_format="iso")

The ``schema`` field contains the ``fields`` key, which itself contains
a list of column name to type pairs, including the ``Index`` or ``MultiIndex``
(see below for a list of types).
The ``schema`` field also contains a ``primaryKey`` field if the (Multi)index
is unique.

The second field, ``data``, contains the serialized data with the ``records``
orient.
The index is included, and any datetimes are ISO 8601 formatted, as required
by the Table Schema spec.

The full list of types supported are described in the Table Schema
spec. This table shows the mapping from pandas types:

=============== =================
pandas type     Table Schema type
=============== =================
int64           integer
float64         number
bool            boolean
datetime64[ns]  datetime
timedelta64[ns] duration
categorical     any
object          str
=============== =================

A few notes on the generated table schema:

* The ``schema`` object contains a ``pandas_version`` field. This contains
  the version of pandas' dialect of the schema, and will be incremented
  with each revision.
* All dates are converted to UTC when serializing. Even timezone naive values,
  which are treated as UTC with an offset of 0.

  .. ipython:: python

     from pandas.io.json import build_table_schema

     s = pd.Series(pd.date_range("2016", periods=4))
     build_table_schema(s)

* datetimes with a timezone (before serializing), include an additional field
  ``tz`` with the time zone name (e.g. ``'US/Central'``).

  .. ipython:: python

     s_tz = pd.Series(pd.date_range("2016", periods=12, tz="US/Central"))
     build_table_schema(s_tz)

* Periods are converted to timestamps before serialization, and so have the
  same behavior of being converted to UTC. In addition, periods will contain
  and additional field ``freq`` with the period's frequency, e.g. ``'A-DEC'``.

  .. ipython:: python

     s_per = pd.Series(1, index=pd.period_range("2016", freq="A-DEC", periods=4))
     build_table_schema(s_per)

* Categoricals use the ``any`` type and an ``enum`` constraint listing
  the set of possible values. Additionally, an ``ordered`` field is included:

  .. ipython:: python

     s_cat = pd.Series(pd.Categorical(["a", "b", "a"]))
     build_table_schema(s_cat)

* A ``primaryKey`` field, containing an array of labels, is included
  *if the index is unique*:

  .. ipython:: python

     s_dupe = pd.Series([1, 2], index=[1, 1])
     build_table_schema(s_dupe)

* The ``primaryKey`` behavior is the same with MultiIndexes, but in this
  case the ``primaryKey`` is an array:

  .. ipython:: python

     s_multi = pd.Series(1, index=pd.MultiIndex.from_product([("a", "b"), (0, 1)]))
     build_table_schema(s_multi)

* The default naming roughly follows these rules:

    * For series, the ``object.name`` is used. If that's none, then the
      name is ``values``
    * For ``DataFrames``, the stringified version of the column name is used
    * For ``Index`` (not ``MultiIndex``), ``index.name`` is used, with a
      fallback to ``index`` if that is None.
    * For ``MultiIndex``, ``mi.names`` is used. If any level has no name,
      then ``level_<i>`` is used.

``read_json`` also accepts ``orient='table'`` as an argument. This allows for
the preservation of metadata such as dtypes and index names in a
round-trippable manner.

  .. ipython:: python

   df = pd.DataFrame(
       {
           "foo": [1, 2, 3, 4],
           "bar": ["a", "b", "c", "d"],
           "baz": pd.date_range("2018-01-01", freq="d", periods=4),
           "qux": pd.Categorical(["a", "b", "c", "c"]),
       },
       index=pd.Index(range(4), name="idx"),
   )
   df
   df.dtypes

   df.to_json("test.json", orient="table")
   new_df = pd.read_json("test.json", orient="table")
   new_df
   new_df.dtypes

Please note that the literal string 'index' as the name of an :class:`Index`
is not round-trippable, nor are any names beginning with ``'level_'`` within a
:class:`MultiIndex`. These are used by default in :func:`DataFrame.to_json` to
indicate missing values and the subsequent read cannot distinguish the intent.

.. ipython:: python
   :okwarning:

   df.index.name = "index"
   df.to_json("test.json", orient="table")
   new_df = pd.read_json("test.json", orient="table")
   print(new_df.index.name)

.. ipython:: python
   :suppress:

   os.remove("test.json")

When using ``orient='table'`` along with user-defined ``ExtensionArray``,
the generated schema will contain an additional ``extDtype`` key in the respective
``fields`` element. This extra key is not standard but does enable JSON roundtrips
for extension types (e.g. ``read_json(df.to_json(orient="table"), orient="table")``).

The ``extDtype`` key carries the name of the extension, if you have properly registered
the ``ExtensionDtype``, pandas will use said name to perform a lookup into the registry
and re-convert the serialized data into your custom dtype.

.. _Table Schema: https://specs.frictionlessdata.io/table-schema/


HTML
----

.. _io.read_html:

Reading HTML content
''''''''''''''''''''''

.. warning::

   We **highly encourage** you to read the :ref:`HTML Table Parsing gotchas <io.html.gotchas>`
   below regarding the issues surrounding the BeautifulSoup4/html5lib/lxml parsers.

The top-level :func:`~pandas.io.html.read_html` function can accept an HTML
string/file/URL and will parse HTML tables into list of pandas ``DataFrames``.
Let's look at a few examples.

.. note::

   ``read_html`` returns a ``list`` of ``DataFrame`` objects, even if there is
   only a single table contained in the HTML content.

Read a URL with no options:

.. code-block:: ipython

   In [320]: "https://www.fdic.gov/resources/resolutions/bank-failures/failed-bank-list"
   In [321]: pd.read_html(url)
   Out[321]:
   [                         Bank NameBank           CityCity StateSt  ...              Acquiring InstitutionAI Closing DateClosing FundFund
    0                    Almena State Bank             Almena      KS  ...                          Equity Bank    October 23, 2020    10538
    1           First City Bank of Florida  Fort Walton Beach      FL  ...            United Fidelity Bank, fsb    October 16, 2020    10537
    2                 The First State Bank      Barboursville      WV  ...                       MVB Bank, Inc.       April 3, 2020    10536
    3                   Ericson State Bank            Ericson      NE  ...           Farmers and Merchants Bank   February 14, 2020    10535
    4     City National Bank of New Jersey             Newark      NJ  ...                      Industrial Bank    November 1, 2019    10534
    ..                                 ...                ...     ...  ...                                  ...                 ...      ...
    558                 Superior Bank, FSB           Hinsdale      IL  ...                Superior Federal, FSB       July 27, 2001     6004
    559                Malta National Bank              Malta      OH  ...                    North Valley Bank         May 3, 2001     4648
    560    First Alliance Bank & Trust Co.         Manchester      NH  ...  Southern New Hampshire Bank & Trust    February 2, 2001     4647
    561  National State Bank of Metropolis         Metropolis      IL  ...              Banterra Bank of Marion   December 14, 2000     4646
    562                   Bank of Honolulu           Honolulu      HI  ...                   Bank of the Orient    October 13, 2000     4645

    [563 rows x 7 columns]]

.. note::

   The data from the above URL changes every Monday so the resulting data above may be slightly different.

Read in the content of the file from the above URL and pass it to ``read_html``
as a string:

.. ipython:: python

   html_str = """
            <table>
                <tr>
                    <th>A</th>
                    <th colspan="1">B</th>
                    <th rowspan="1">C</th>
                </tr>
                <tr>
                    <td>a</td>
                    <td>b</td>
                    <td>c</td>
                </tr>
            </table>
        """

   with open("tmp.html", "w") as f:
       f.write(html_str)
   df = pd.read_html("tmp.html")
   df[0]

.. ipython:: python
   :suppress:

   os.remove("tmp.html")

You can even pass in an instance of ``StringIO`` if you so desire:

.. ipython:: python

   dfs = pd.read_html(StringIO(html_str))
   dfs[0]

.. note::

   The following examples are not run by the IPython evaluator due to the fact
   that having so many network-accessing functions slows down the documentation
   build. If you spot an error or an example that doesn't run, please do not
   hesitate to report it over on `pandas GitHub issues page
   <https://github.com/pandas-dev/pandas/issues>`__.


Read a URL and match a table that contains specific text:

.. code-block:: python

   match = "Metcalf Bank"
   df_list = pd.read_html(url, match=match)

Specify a header row (by default ``<th>`` or ``<td>`` elements located within a
``<thead>`` are used to form the column index, if multiple rows are contained within
``<thead>`` then a MultiIndex is created); if specified, the header row is taken
from the data minus the parsed header elements (``<th>`` elements).

.. code-block:: python

   dfs = pd.read_html(url, header=0)

Specify an index column:

.. code-block:: python

   dfs = pd.read_html(url, index_col=0)

Specify a number of rows to skip:

.. code-block:: python

   dfs = pd.read_html(url, skiprows=0)

Specify a number of rows to skip using a list (``range`` works
as well):

.. code-block:: python

   dfs = pd.read_html(url, skiprows=range(2))

Specify an HTML attribute:

.. code-block:: python

   dfs1 = pd.read_html(url, attrs={"id": "table"})
   dfs2 = pd.read_html(url, attrs={"class": "sortable"})
   print(np.array_equal(dfs1[0], dfs2[0]))  # Should be True

Specify values that should be converted to NaN:

.. code-block:: python

   dfs = pd.read_html(url, na_values=["No Acquirer"])

Specify whether to keep the default set of NaN values:

.. code-block:: python

   dfs = pd.read_html(url, keep_default_na=False)

Specify converters for columns. This is useful for numerical text data that has
leading zeros.  By default columns that are numerical are cast to numeric
types and the leading zeros are lost. To avoid this, we can convert these
columns to strings.

.. code-block:: python

   url_mcc = "https://en.wikipedia.org/wiki/Mobile_country_code"
   dfs = pd.read_html(
       url_mcc,
       match="Telekom Albania",
       header=0,
       converters={"MNC": str},
   )

Use some combination of the above:

.. code-block:: python

   dfs = pd.read_html(url, match="Metcalf Bank", index_col=0)

Read in pandas ``to_html`` output (with some loss of floating point precision):

.. code-block:: python

   df = pd.DataFrame(np.random.randn(2, 2))
   s = df.to_html(float_format="{0:.40g}".format)
   dfin = pd.read_html(s, index_col=0)

The ``lxml`` backend will raise an error on a failed parse if that is the only
parser you provide. If you only have a single parser you can provide just a
string, but it is considered good practice to pass a list with one string if,
for example, the function expects a sequence of strings. You may use:

.. code-block:: python

   dfs = pd.read_html(url, "Metcalf Bank", index_col=0, flavor=["lxml"])

Or you could pass ``flavor='lxml'`` without a list:

.. code-block:: python

   dfs = pd.read_html(url, "Metcalf Bank", index_col=0, flavor="lxml")

However, if you have bs4 and html5lib installed and pass ``None`` or ``['lxml',
'bs4']`` then the parse will most likely succeed. Note that *as soon as a parse
succeeds, the function will return*.

.. code-block:: python

   dfs = pd.read_html(url, "Metcalf Bank", index_col=0, flavor=["lxml", "bs4"])

Links can be extracted from cells along with the text using ``extract_links="all"``.

.. ipython:: python

    html_table = """
    <table>
      <tr>
        <th>GitHub</th>
      </tr>
      <tr>
        <td><a href="https://github.com/pandas-dev/pandas">pandas</a></td>
      </tr>
    </table>
    """

    df = pd.read_html(
        html_table,
        extract_links="all"
    )[0]
    df
    df[("GitHub", None)]
    df[("GitHub", None)].str[1]

.. versionadded:: 1.5.0

.. _io.html:

Writing to HTML files
''''''''''''''''''''''

``DataFrame`` objects have an instance method ``to_html`` which renders the
contents of the ``DataFrame`` as an HTML table. The function arguments are as
in the method ``to_string`` described above.

.. note::

   Not all of the possible options for ``DataFrame.to_html`` are shown here for
   brevity's sake. See :func:`~pandas.core.frame.DataFrame.to_html` for the
   full set of options.

.. note::

   In an HTML-rendering supported environment like a Jupyter Notebook, ``display(HTML(...))```
   will render the raw HTML into the environment.

.. ipython:: python

   from IPython.display import display, HTML

   df = pd.DataFrame(np.random.randn(2, 2))
   df
   html = df.to_html()
   print(html)  # raw html
   display(HTML(html))

The ``columns`` argument will limit the columns shown:

.. ipython:: python

   html = df.to_html(columns=[0])
   print(html)
   display(HTML(html))

``float_format`` takes a Python callable to control the precision of floating
point values:

.. ipython:: python

   html = df.to_html(float_format="{0:.10f}".format)
   print(html)
   display(HTML(html))


``bold_rows`` will make the row labels bold by default, but you can turn that
off:

.. ipython:: python

   html = df.to_html(bold_rows=False)
   print(html)
   display(HTML(html))


The ``classes`` argument provides the ability to give the resulting HTML
table CSS classes. Note that these classes are *appended* to the existing
``'dataframe'`` class.

.. ipython:: python

   print(df.to_html(classes=["awesome_table_class", "even_more_awesome_class"]))

The ``render_links`` argument provides the ability to add hyperlinks to cells
that contain URLs.

.. ipython:: python

   url_df = pd.DataFrame(
       {
           "name": ["Python", "pandas"],
           "url": ["https://www.python.org/", "https://pandas.pydata.org"],
       }
   )
   html = url_df.to_html(render_links=True)
   print(html)
   display(HTML(html))

Finally, the ``escape`` argument allows you to control whether the
"<", ">" and "&" characters escaped in the resulting HTML (by default it is
``True``). So to get the HTML without escaped characters pass ``escape=False``

.. ipython:: python

   df = pd.DataFrame({"a": list("&<>"), "b": np.random.randn(3)})

Escaped:

.. ipython:: python

   html = df.to_html()
   print(html)
   display(HTML(html))

Not escaped:

.. ipython:: python

   html = df.to_html(escape=False)
   print(html)
   display(HTML(html))

.. note::

   Some browsers may not show a difference in the rendering of the previous two
   HTML tables.


.. _io.html.gotchas:

HTML Table Parsing Gotchas
''''''''''''''''''''''''''

There are some versioning issues surrounding the libraries that are used to
parse HTML tables in the top-level pandas io function ``read_html``.

**Issues with** |lxml|_

* Benefits

    * |lxml|_ is very fast.

    * |lxml|_ requires Cython to install correctly.

* Drawbacks

    * |lxml|_ does *not* make any guarantees about the results of its parse
      *unless* it is given |svm|_.

    * In light of the above, we have chosen to allow you, the user, to use the
      |lxml|_ backend, but **this backend will use** |html5lib|_ if |lxml|_
      fails to parse

    * It is therefore *highly recommended* that you install both
      |BeautifulSoup4|_ and |html5lib|_, so that you will still get a valid
      result (provided everything else is valid) even if |lxml|_ fails.

**Issues with** |BeautifulSoup4|_ **using** |lxml|_ **as a backend**

* The above issues hold here as well since |BeautifulSoup4|_ is essentially
  just a wrapper around a parser backend.

**Issues with** |BeautifulSoup4|_ **using** |html5lib|_ **as a backend**

* Benefits

    * |html5lib|_ is far more lenient than |lxml|_ and consequently deals
      with *real-life markup* in a much saner way rather than just, e.g.,
      dropping an element without notifying you.

    * |html5lib|_ *generates valid HTML5 markup from invalid markup
      automatically*. This is extremely important for parsing HTML tables,
      since it guarantees a valid document. However, that does NOT mean that
      it is "correct", since the process of fixing markup does not have a
      single definition.

    * |html5lib|_ is pure Python and requires no additional build steps beyond
      its own installation.

* Drawbacks

    * The biggest drawback to using |html5lib|_ is that it is slow as
      molasses.  However consider the fact that many tables on the web are not
      big enough for the parsing algorithm runtime to matter. It is more
      likely that the bottleneck will be in the process of reading the raw
      text from the URL over the web, i.e., IO (input-output). For very large
      tables, this might not be true.


.. |svm| replace:: **strictly valid markup**
.. _svm: https://validator.w3.org/docs/help.html#validation_basics

.. |html5lib| replace:: **html5lib**
.. _html5lib: https://github.com/html5lib/html5lib-python

.. |BeautifulSoup4| replace:: **BeautifulSoup4**
.. _BeautifulSoup4: https://www.crummy.com/software/BeautifulSoup

.. |lxml| replace:: **lxml**
.. _lxml: https://lxml.de

.. _io.latex:

LaTeX
-----

.. versionadded:: 1.3.0

Currently there are no methods to read from LaTeX, only output methods.

Writing to LaTeX files
''''''''''''''''''''''

.. note::

   DataFrame *and* Styler objects currently have a ``to_latex`` method. We recommend
   using the `Styler.to_latex() <../reference/api/pandas.io.formats.style.Styler.to_latex.rst>`__ method
   over `DataFrame.to_latex() <../reference/api/pandas.DataFrame.to_latex.rst>`__ due to the former's greater flexibility with
   conditional styling, and the latter's possible future deprecation.

Review the documentation for `Styler.to_latex <../reference/api/pandas.io.formats.style.Styler.to_latex.rst>`__,
which gives examples of conditional styling and explains the operation of its keyword
arguments.

For simple application the following pattern is sufficient.

.. ipython:: python

   df = pd.DataFrame([[1, 2], [3, 4]], index=["a", "b"], columns=["c", "d"])
   print(df.style.to_latex())

To format values before output, chain the `Styler.format <../reference/api/pandas.io.formats.style.Styler.format.rst>`__
method.

.. ipython:: python

   print(df.style.format("€ {}").to_latex())

XML
---

.. _io.read_xml:

Reading XML
'''''''''''

.. versionadded:: 1.3.0

The top-level :func:`~pandas.io.xml.read_xml` function can accept an XML
string/file/URL and will parse nodes and attributes into a pandas ``DataFrame``.

.. note::

   Since there is no standard XML structure where design types can vary in
   many ways, ``read_xml`` works best with flatter, shallow versions. If
   an XML document is deeply nested, use the ``stylesheet`` feature to
   transform XML into a flatter version.

Let's look at a few examples.

Read an XML string:

.. ipython:: python

   xml = """<?xml version="1.0" encoding="UTF-8"?>
   <bookstore>
     <book category="cooking">
       <title lang="en">Everyday Italian</title>
       <author>Giada De Laurentiis</author>
       <year>2005</year>
       <price>30.00</price>
     </book>
     <book category="children">
       <title lang="en">Harry Potter</title>
       <author>J K. Rowling</author>
       <year>2005</year>
       <price>29.99</price>
     </book>
     <book category="web">
       <title lang="en">Learning XML</title>
       <author>Erik T. Ray</author>
       <year>2003</year>
       <price>39.95</price>
     </book>
   </bookstore>"""

   df = pd.read_xml(xml)
   df

Read a URL with no options:

.. ipython:: python

   df = pd.read_xml("https://www.w3schools.com/xml/books.xml")
   df

Read in the content of the "books.xml" file and pass it to ``read_xml``
as a string:

.. ipython:: python

   file_path = "books.xml"
   with open(file_path, "w") as f:
       f.write(xml)

   with open(file_path, "r") as f:
       df = pd.read_xml(f.read())
   df

Read in the content of the "books.xml" as instance of ``StringIO`` or
``BytesIO`` and pass it to ``read_xml``:

.. ipython:: python

   with open(file_path, "r") as f:
       sio = StringIO(f.read())

   df = pd.read_xml(sio)
   df

.. ipython:: python

   with open(file_path, "rb") as f:
       bio = BytesIO(f.read())

   df = pd.read_xml(bio)
   df

Even read XML from AWS S3 buckets such as NIH NCBI PMC Article Datasets providing
Biomedical and Life Science Jorurnals:

.. ipython:: python
   :okwarning:

   df = pd.read_xml(
       "s3://pmc-oa-opendata/oa_comm/xml/all/PMC1236943.xml",
       xpath=".//journal-meta",
   )
   df

With `lxml`_ as default ``parser``, you access the full-featured XML library
that extends Python's ElementTree API. One powerful tool is ability to query
nodes selectively or conditionally with more expressive XPath:

.. _lxml: https://lxml.de

.. ipython:: python

   df = pd.read_xml(file_path, xpath="//book[year=2005]")
   df

Specify only elements or only attributes to parse:

.. ipython:: python

   df = pd.read_xml(file_path, elems_only=True)
   df

.. ipython:: python

   df = pd.read_xml(file_path, attrs_only=True)
   df

.. ipython:: python
   :suppress:

   os.remove("books.xml")

XML documents can have namespaces with prefixes and default namespaces without
prefixes both of which are denoted with a special attribute ``xmlns``. In order
to parse by node under a namespace context, ``xpath`` must reference a prefix.

For example, below XML contains a namespace with prefix, ``doc``, and URI at
``https://example.com``. In order to parse ``doc:row`` nodes,
``namespaces`` must be used.

.. ipython:: python

   xml = """<?xml version='1.0' encoding='utf-8'?>
   <doc:data xmlns:doc="https://example.com">
     <doc:row>
       <doc:shape>square</doc:shape>
       <doc:degrees>360</doc:degrees>
       <doc:sides>4.0</doc:sides>
     </doc:row>
     <doc:row>
       <doc:shape>circle</doc:shape>
       <doc:degrees>360</doc:degrees>
       <doc:sides/>
     </doc:row>
     <doc:row>
       <doc:shape>triangle</doc:shape>
       <doc:degrees>180</doc:degrees>
       <doc:sides>3.0</doc:sides>
     </doc:row>
   </doc:data>"""

   df = pd.read_xml(xml,
                    xpath="//doc:row",
                    namespaces={"doc": "https://example.com"})
   df

Similarly, an XML document can have a default namespace without prefix. Failing
to assign a temporary prefix will return no nodes and raise a ``ValueError``.
But assigning *any* temporary name to correct URI allows parsing by nodes.

.. ipython:: python

   xml = """<?xml version='1.0' encoding='utf-8'?>
   <data xmlns="https://example.com">
    <row>
      <shape>square</shape>
      <degrees>360</degrees>
      <sides>4.0</sides>
    </row>
    <row>
      <shape>circle</shape>
      <degrees>360</degrees>
      <sides/>
    </row>
    <row>
      <shape>triangle</shape>
      <degrees>180</degrees>
      <sides>3.0</sides>
    </row>
   </data>"""

   df = pd.read_xml(xml,
                    xpath="//pandas:row",
                    namespaces={"pandas": "https://example.com"})
   df

However, if XPath does not reference node names such as default, ``/*``, then
``namespaces`` is not required.

With `lxml`_ as parser, you can flatten nested XML documents with an XSLT
script which also can be string/file/URL types. As background, `XSLT`_ is
a special-purpose language written in a special XML file that can transform
original XML documents into other XML, HTML, even text (CSV, JSON, etc.)
using an XSLT processor.

.. _lxml: https://lxml.de
.. _XSLT: https://www.w3.org/TR/xslt/

For example, consider this somewhat nested structure of Chicago "L" Rides
where station and rides elements encapsulate data in their own sections.
With below XSLT, ``lxml`` can transform original nested document into a flatter
output (as shown below for demonstration) for easier parse into ``DataFrame``:

.. ipython:: python

   xml = """<?xml version='1.0' encoding='utf-8'?>
    <response>
     <row>
       <station id="40850" name="Library"/>
       <month>2020-09-01T00:00:00</month>
       <rides>
         <avg_weekday_rides>864.2</avg_weekday_rides>
         <avg_saturday_rides>534</avg_saturday_rides>
         <avg_sunday_holiday_rides>417.2</avg_sunday_holiday_rides>
       </rides>
     </row>
     <row>
       <station id="41700" name="Washington/Wabash"/>
       <month>2020-09-01T00:00:00</month>
       <rides>
         <avg_weekday_rides>2707.4</avg_weekday_rides>
         <avg_saturday_rides>1909.8</avg_saturday_rides>
         <avg_sunday_holiday_rides>1438.6</avg_sunday_holiday_rides>
       </rides>
     </row>
     <row>
       <station id="40380" name="Clark/Lake"/>
       <month>2020-09-01T00:00:00</month>
       <rides>
         <avg_weekday_rides>2949.6</avg_weekday_rides>
         <avg_saturday_rides>1657</avg_saturday_rides>
         <avg_sunday_holiday_rides>1453.8</avg_sunday_holiday_rides>
       </rides>
     </row>
    </response>"""

   xsl = """<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
      <xsl:output method="xml" omit-xml-declaration="no" indent="yes"/>
      <xsl:strip-space elements="*"/>
      <xsl:template match="/response">
         <xsl:copy>
           <xsl:apply-templates select="row"/>
         </xsl:copy>
      </xsl:template>
      <xsl:template match="row">
         <xsl:copy>
           <station_id><xsl:value-of select="station/@id"/></station_id>
           <station_name><xsl:value-of select="station/@name"/></station_name>
           <xsl:copy-of select="month|rides/*"/>
         </xsl:copy>
      </xsl:template>
    </xsl:stylesheet>"""

   output = """<?xml version='1.0' encoding='utf-8'?>
    <response>
      <row>
         <station_id>40850</station_id>
         <station_name>Library</station_name>
         <month>2020-09-01T00:00:00</month>
         <avg_weekday_rides>864.2</avg_weekday_rides>
         <avg_saturday_rides>534</avg_saturday_rides>
         <avg_sunday_holiday_rides>417.2</avg_sunday_holiday_rides>
      </row>
      <row>
         <station_id>41700</station_id>
         <station_name>Washington/Wabash</station_name>
         <month>2020-09-01T00:00:00</month>
         <avg_weekday_rides>2707.4</avg_weekday_rides>
         <avg_saturday_rides>1909.8</avg_saturday_rides>
         <avg_sunday_holiday_rides>1438.6</avg_sunday_holiday_rides>
      </row>
      <row>
         <station_id>40380</station_id>
         <station_name>Clark/Lake</station_name>
         <month>2020-09-01T00:00:00</month>
         <avg_weekday_rides>2949.6</avg_weekday_rides>
         <avg_saturday_rides>1657</avg_saturday_rides>
         <avg_sunday_holiday_rides>1453.8</avg_sunday_holiday_rides>
      </row>
    </response>"""

   df = pd.read_xml(xml, stylesheet=xsl)
   df

For very large XML files that can range in hundreds of megabytes to gigabytes, :func:`pandas.read_xml`
supports parsing such sizeable files using `lxml's iterparse`_ and `etree's iterparse`_
which are memory-efficient methods to iterate through an XML tree and extract specific elements and attributes.
without holding entire tree in memory.

    .. versionadded:: 1.5.0

.. _`lxml's iterparse`: https://lxml.de/3.2/parsing.html#iterparse-and-iterwalk
.. _`etree's iterparse`: https://docs.python.org/3/library/xml.etree.elementtree.html#xml.etree.ElementTree.iterparse

To use this feature, you must pass a physical XML file path into ``read_xml`` and use the ``iterparse`` argument.
Files should not be compressed or point to online sources but stored on local disk. Also, ``iterparse`` should be
a dictionary where the key is the repeating nodes in document (which become the rows) and the value is a list of
any element or attribute that is a descendant (i.e., child, grandchild) of repeating node. Since XPath is not
used in this method, descendants do not need to share same relationship with one another. Below shows example
of reading in Wikipedia's very large (12 GB+) latest article data dump.

.. code-block:: ipython

    In [1]: df = pd.read_xml(
    ...         "/path/to/downloaded/enwikisource-latest-pages-articles.xml",
    ...         iterparse = {"page": ["title", "ns", "id"]}
    ...     )
    ...     df
    Out[2]:
                                                         title   ns        id
    0                                       Gettysburg Address    0     21450
    1                                                Main Page    0     42950
    2                            Declaration by United Nations    0      8435
    3             Constitution of the United States of America    0      8435
    4                     Declaration of Independence (Israel)    0     17858
    ...                                                    ...  ...       ...
    3578760               Page:Black cat 1897 07 v2 n10.pdf/17  104    219649
    3578761               Page:Black cat 1897 07 v2 n10.pdf/43  104    219649
    3578762               Page:Black cat 1897 07 v2 n10.pdf/44  104    219649
    3578763      The History of Tom Jones, a Foundling/Book IX    0  12084291
    3578764  Page:Shakespeare of Stratford (1926) Yale.djvu/91  104     21450

    [3578765 rows x 3 columns]

.. _io.xml:

Writing XML
'''''''''''

.. versionadded:: 1.3.0

``DataFrame`` objects have an instance method ``to_xml`` which renders the
contents of the ``DataFrame`` as an XML document.

.. note::

   This method does not support special properties of XML including DTD,
   CData, XSD schemas, processing instructions, comments, and others.
   Only namespaces at the root level is supported. However, ``stylesheet``
   allows design changes after initial output.

Let's look at a few examples.

Write an XML without options:

.. ipython:: python

   geom_df = pd.DataFrame(
       {
           "shape": ["square", "circle", "triangle"],
           "degrees": [360, 360, 180],
           "sides": [4, np.nan, 3],
       }
   )

   print(geom_df.to_xml())


Write an XML with new root and row name:

.. ipython:: python

   print(geom_df.to_xml(root_name="geometry", row_name="objects"))

Write an attribute-centric XML:

.. ipython:: python

   print(geom_df.to_xml(attr_cols=geom_df.columns.tolist()))

Write a mix of elements and attributes:

.. ipython:: python

   print(
       geom_df.to_xml(
           index=False,
           attr_cols=['shape'],
           elem_cols=['degrees', 'sides'])
   )

Any ``DataFrames`` with hierarchical columns will be flattened for XML element names
with levels delimited by underscores:

.. ipython:: python

   ext_geom_df = pd.DataFrame(
       {
           "type": ["polygon", "other", "polygon"],
           "shape": ["square", "circle", "triangle"],
           "degrees": [360, 360, 180],
           "sides": [4, np.nan, 3],
       }
   )

   pvt_df = ext_geom_df.pivot_table(index='shape',
                                    columns='type',
                                    values=['degrees', 'sides'],
                                    aggfunc='sum')
   pvt_df

   print(pvt_df.to_xml())

Write an XML with default namespace:

.. ipython:: python

   print(geom_df.to_xml(namespaces={"": "https://example.com"}))

Write an XML with namespace prefix:

.. ipython:: python

   print(
       geom_df.to_xml(namespaces={"doc": "https://example.com"},
                      prefix="doc")
   )

Write an XML without declaration or pretty print:

.. ipython:: python

   print(
       geom_df.to_xml(xml_declaration=False,
                      pretty_print=False)
   )

Write an XML and transform with stylesheet:

.. ipython:: python

   xsl = """<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
      <xsl:output method="xml" omit-xml-declaration="no" indent="yes"/>
      <xsl:strip-space elements="*"/>
      <xsl:template match="/data">
        <geometry>
          <xsl:apply-templates select="row"/>
        </geometry>
      </xsl:template>
      <xsl:template match="row">
        <object index="{index}">
          <xsl:if test="shape!='circle'">
              <xsl:attribute name="type">polygon</xsl:attribute>
          </xsl:if>
          <xsl:copy-of select="shape"/>
          <property>
              <xsl:copy-of select="degrees|sides"/>
          </property>
        </object>
      </xsl:template>
    </xsl:stylesheet>"""

   print(geom_df.to_xml(stylesheet=xsl))


XML Final Notes
'''''''''''''''

* All XML documents adhere to `W3C specifications`_. Both ``etree`` and ``lxml``
  parsers will fail to parse any markup document that is not well-formed or
  follows XML syntax rules. Do be aware HTML is not an XML document unless it
  follows XHTML specs. However, other popular markup types including KML, XAML,
  RSS, MusicML, MathML are compliant `XML schemas`_.

* For above reason, if your application builds XML prior to pandas operations,
  use appropriate DOM libraries like ``etree`` and ``lxml`` to build the necessary
  document and not by string concatenation or regex adjustments. Always remember
  XML is a *special* text file with markup rules.

* With very large XML files (several hundred MBs to GBs), XPath and XSLT
  can become memory-intensive operations. Be sure to have enough available
  RAM for reading and writing to large XML files (roughly about 5 times the
  size of text).

* Because XSLT is a programming language, use it with caution since such scripts
  can pose a security risk in your environment and can run large or infinite
  recursive operations. Always test scripts on small fragments before full run.

* The `etree`_ parser supports all functionality of both ``read_xml`` and
  ``to_xml`` except for complex XPath and any XSLT. Though limited in features,
  ``etree`` is still a reliable and capable parser and tree builder. Its
  performance may trail ``lxml`` to a certain degree for larger files but
  relatively unnoticeable on small to medium size files.

.. _`W3C specifications`: https://www.w3.org/TR/xml/
.. _`XML schemas`: https://en.wikipedia.org/wiki/List_of_types_of_XML_schemas
.. _`etree`: https://docs.python.org/3/library/xml.etree.elementtree.html



.. _io.excel:

Excel files
-----------

The :func:`~pandas.read_excel` method can read Excel 2007+ (``.xlsx``) files
using the ``openpyxl`` Python module. Excel 2003 (``.xls``) files
can be read using ``xlrd``. Binary Excel (``.xlsb``)
files can be read using ``pyxlsb``.
The :meth:`~DataFrame.to_excel` instance method is used for
saving a ``DataFrame`` to Excel.  Generally the semantics are
similar to working with :ref:`csv<io.read_csv_table>` data.
See the :ref:`cookbook<cookbook.excel>` for some advanced strategies.

.. warning::

   The `xlwt <https://xlwt.readthedocs.io/en/latest/>`__ package for writing old-style ``.xls``
   excel files is no longer maintained.
   The `xlrd <https://xlrd.readthedocs.io/en/latest/>`__ package is now only for reading
   old-style ``.xls`` files.

   Before pandas 1.3.0, the default argument ``engine=None`` to :func:`~pandas.read_excel`
   would result in using the ``xlrd`` engine in many cases, including new
   Excel 2007+ (``.xlsx``) files. pandas will now default to using the
   `openpyxl <https://openpyxl.readthedocs.io/en/stable/>`__ engine.

   It is strongly encouraged to install ``openpyxl`` to read Excel 2007+
   (``.xlsx``) files.
   **Please do not report issues when using ``xlrd`` to read ``.xlsx`` files.**
   This is no longer supported, switch to using ``openpyxl`` instead.

   Attempting to use the ``xlwt`` engine will raise a ``FutureWarning``
   unless the option :attr:`io.excel.xls.writer` is set to ``"xlwt"``.
   While this option is now deprecated and will also raise a ``FutureWarning``,
   it can be globally set and the warning suppressed. Users are recommended to
   write ``.xlsx`` files using the ``openpyxl`` engine instead.

.. _io.excel_reader:

Reading Excel files
'''''''''''''''''''

In the most basic use-case, ``read_excel`` takes a path to an Excel
file, and the ``sheet_name`` indicating which sheet to parse.

.. code-block:: python

   # Returns a DataFrame
   pd.read_excel("path_to_file.xls", sheet_name="Sheet1")


.. _io.excel.excelfile_class:

``ExcelFile`` class
+++++++++++++++++++

To facilitate working with multiple sheets from the same file, the ``ExcelFile``
class can be used to wrap the file and can be passed into ``read_excel``
There will be a performance benefit for reading multiple sheets as the file is
read into memory only once.

.. code-block:: python

   xlsx = pd.ExcelFile("path_to_file.xls")
   df = pd.read_excel(xlsx, "Sheet1")

The ``ExcelFile`` class can also be used as a context manager.

.. code-block:: python

   with pd.ExcelFile("path_to_file.xls") as xls:
       df1 = pd.read_excel(xls, "Sheet1")
       df2 = pd.read_excel(xls, "Sheet2")

The ``sheet_names`` property will generate
a list of the sheet names in the file.

The primary use-case for an ``ExcelFile`` is parsing multiple sheets with
different parameters:

.. code-block:: python

    data = {}
    # For when Sheet1's format differs from Sheet2
    with pd.ExcelFile("path_to_file.xls") as xls:
        data["Sheet1"] = pd.read_excel(xls, "Sheet1", index_col=None, na_values=["NA"])
        data["Sheet2"] = pd.read_excel(xls, "Sheet2", index_col=1)

Note that if the same parsing parameters are used for all sheets, a list
of sheet names can simply be passed to ``read_excel`` with no loss in performance.

.. code-block:: python

    # using the ExcelFile class
    data = {}
    with pd.ExcelFile("path_to_file.xls") as xls:
        data["Sheet1"] = pd.read_excel(xls, "Sheet1", index_col=None, na_values=["NA"])
        data["Sheet2"] = pd.read_excel(xls, "Sheet2", index_col=None, na_values=["NA"])

    # equivalent using the read_excel function
    data = pd.read_excel(
        "path_to_file.xls", ["Sheet1", "Sheet2"], index_col=None, na_values=["NA"]
    )

``ExcelFile`` can also be called with a ``xlrd.book.Book`` object
as a parameter. This allows the user to control how the excel file is read.
For example, sheets can be loaded on demand by calling ``xlrd.open_workbook()``
with ``on_demand=True``.

.. code-block:: python

    import xlrd

    xlrd_book = xlrd.open_workbook("path_to_file.xls", on_demand=True)
    with pd.ExcelFile(xlrd_book) as xls:
        df1 = pd.read_excel(xls, "Sheet1")
        df2 = pd.read_excel(xls, "Sheet2")

.. _io.excel.specifying_sheets:

Specifying sheets
+++++++++++++++++

.. note:: The second argument is ``sheet_name``, not to be confused with ``ExcelFile.sheet_names``.

.. note:: An ExcelFile's attribute ``sheet_names`` provides access to a list of sheets.

* The arguments ``sheet_name`` allows specifying the sheet or sheets to read.
* The default value for ``sheet_name`` is 0, indicating to read the first sheet
* Pass a string to refer to the name of a particular sheet in the workbook.
* Pass an integer to refer to the index of a sheet. Indices follow Python
  convention, beginning at 0.
* Pass a list of either strings or integers, to return a dictionary of specified sheets.
* Pass a ``None`` to return a dictionary of all available sheets.

.. code-block:: python

   # Returns a DataFrame
   pd.read_excel("path_to_file.xls", "Sheet1", index_col=None, na_values=["NA"])

Using the sheet index:

.. code-block:: python

   # Returns a DataFrame
   pd.read_excel("path_to_file.xls", 0, index_col=None, na_values=["NA"])

Using all default values:

.. code-block:: python

   # Returns a DataFrame
   pd.read_excel("path_to_file.xls")

Using None to get all sheets:

.. code-block:: python

   # Returns a dictionary of DataFrames
   pd.read_excel("path_to_file.xls", sheet_name=None)

Using a list to get multiple sheets:

.. code-block:: python

   # Returns the 1st and 4th sheet, as a dictionary of DataFrames.
   pd.read_excel("path_to_file.xls", sheet_name=["Sheet1", 3])

``read_excel`` can read more than one sheet, by setting ``sheet_name`` to either
a list of sheet names, a list of sheet positions, or ``None`` to read all sheets.
Sheets can be specified by sheet index or sheet name, using an integer or string,
respectively.

.. _io.excel.reading_multiindex:

Reading a ``MultiIndex``
++++++++++++++++++++++++

``read_excel`` can read a ``MultiIndex`` index, by passing a list of columns to ``index_col``
and a ``MultiIndex`` column by passing a list of rows to ``header``.  If either the ``index``
or ``columns`` have serialized level names those will be read in as well by specifying
the rows/columns that make up the levels.

For example, to read in a ``MultiIndex`` index without names:

.. ipython:: python

   df = pd.DataFrame(
       {"a": [1, 2, 3, 4], "b": [5, 6, 7, 8]},
       index=pd.MultiIndex.from_product([["a", "b"], ["c", "d"]]),
   )
   df.to_excel("path_to_file.xlsx")
   df = pd.read_excel("path_to_file.xlsx", index_col=[0, 1])
   df

If the index has level names, they will parsed as well, using the same
parameters.

.. ipython:: python

   df.index = df.index.set_names(["lvl1", "lvl2"])
   df.to_excel("path_to_file.xlsx")
   df = pd.read_excel("path_to_file.xlsx", index_col=[0, 1])
   df


If the source file has both ``MultiIndex`` index and columns, lists specifying each
should be passed to ``index_col`` and ``header``:

.. ipython:: python

   df.columns = pd.MultiIndex.from_product([["a"], ["b", "d"]], names=["c1", "c2"])
   df.to_excel("path_to_file.xlsx")
   df = pd.read_excel("path_to_file.xlsx", index_col=[0, 1], header=[0, 1])
   df

.. ipython:: python
   :suppress:

   os.remove("path_to_file.xlsx")

Missing values in columns specified in ``index_col`` will be forward filled to
allow roundtripping with ``to_excel`` for ``merged_cells=True``. To avoid forward
filling the missing values use ``set_index`` after reading the data instead of
``index_col``.

Parsing specific columns
++++++++++++++++++++++++

It is often the case that users will insert columns to do temporary computations
in Excel and you may not want to read in those columns. ``read_excel`` takes
a ``usecols`` keyword to allow you to specify a subset of columns to parse.

.. versionchanged:: 1.0.0

Passing in an integer for ``usecols`` will no longer work. Please pass in a list
of ints from 0 to ``usecols`` inclusive instead.

You can specify a comma-delimited set of Excel columns and ranges as a string:

.. code-block:: python

   pd.read_excel("path_to_file.xls", "Sheet1", usecols="A,C:E")

If ``usecols`` is a list of integers, then it is assumed to be the file column
indices to be parsed.

.. code-block:: python

   pd.read_excel("path_to_file.xls", "Sheet1", usecols=[0, 2, 3])

Element order is ignored, so ``usecols=[0, 1]`` is the same as ``[1, 0]``.

If ``usecols`` is a list of strings, it is assumed that each string corresponds
to a column name provided either by the user in ``names`` or inferred from the
document header row(s). Those strings define which columns will be parsed:

.. code-block:: python

    pd.read_excel("path_to_file.xls", "Sheet1", usecols=["foo", "bar"])

Element order is ignored, so ``usecols=['baz', 'joe']`` is the same as ``['joe', 'baz']``.

If ``usecols`` is callable, the callable function will be evaluated against
the column names, returning names where the callable function evaluates to ``True``.

.. code-block:: python

    pd.read_excel("path_to_file.xls", "Sheet1", usecols=lambda x: x.isalpha())

Parsing dates
+++++++++++++

Datetime-like values are normally automatically converted to the appropriate
dtype when reading the excel file. But if you have a column of strings that
*look* like dates (but are not actually formatted as dates in excel), you can
use the ``parse_dates`` keyword to parse those strings to datetimes:

.. code-block:: python

   pd.read_excel("path_to_file.xls", "Sheet1", parse_dates=["date_strings"])


Cell converters
+++++++++++++++

It is possible to transform the contents of Excel cells via the ``converters``
option. For instance, to convert a column to boolean:

.. code-block:: python

   pd.read_excel("path_to_file.xls", "Sheet1", converters={"MyBools": bool})

This options handles missing values and treats exceptions in the converters
as missing data. Transformations are applied cell by cell rather than to the
column as a whole, so the array dtype is not guaranteed. For instance, a
column of integers with missing values cannot be transformed to an array
with integer dtype, because NaN is strictly a float. You can manually mask
missing data to recover integer dtype:

.. code-block:: python

   def cfun(x):
       return int(x) if x else -1


   pd.read_excel("path_to_file.xls", "Sheet1", converters={"MyInts": cfun})

Dtype specifications
++++++++++++++++++++

As an alternative to converters, the type for an entire column can
be specified using the ``dtype`` keyword, which takes a dictionary
mapping column names to types.  To interpret data with
no type inference, use the type ``str`` or ``object``.

.. code-block:: python

   pd.read_excel("path_to_file.xls", dtype={"MyInts": "int64", "MyText": str})

.. _io.excel_writer:

Writing Excel files
'''''''''''''''''''

Writing Excel files to disk
+++++++++++++++++++++++++++

To write a ``DataFrame`` object to a sheet of an Excel file, you can use the
``to_excel`` instance method.  The arguments are largely the same as ``to_csv``
described above, the first argument being the name of the excel file, and the
optional second argument the name of the sheet to which the ``DataFrame`` should be
written. For example:

.. code-block:: python

   df.to_excel("path_to_file.xlsx", sheet_name="Sheet1")

Files with a ``.xls`` extension will be written using ``xlwt`` and those with a
``.xlsx`` extension will be written using ``xlsxwriter`` (if available) or
``openpyxl``.

The ``DataFrame`` will be written in a way that tries to mimic the REPL output.
The ``index_label`` will be placed in the second
row instead of the first. You can place it in the first row by setting the
``merge_cells`` option in ``to_excel()`` to ``False``:

.. code-block:: python

   df.to_excel("path_to_file.xlsx", index_label="label", merge_cells=False)

In order to write separate ``DataFrames`` to separate sheets in a single Excel file,
one can pass an :class:`~pandas.io.excel.ExcelWriter`.

.. code-block:: python

   with pd.ExcelWriter("path_to_file.xlsx") as writer:
       df1.to_excel(writer, sheet_name="Sheet1")
       df2.to_excel(writer, sheet_name="Sheet2")

.. _io.excel_writing_buffer:

Writing Excel files to memory
+++++++++++++++++++++++++++++

pandas supports writing Excel files to buffer-like objects such as ``StringIO`` or
``BytesIO`` using :class:`~pandas.io.excel.ExcelWriter`.

.. code-block:: python

   from io import BytesIO

   bio = BytesIO()

   # By setting the 'engine' in the ExcelWriter constructor.
   writer = pd.ExcelWriter(bio, engine="xlsxwriter")
   df.to_excel(writer, sheet_name="Sheet1")

   # Save the workbook
   writer.save()

   # Seek to the beginning and read to copy the workbook to a variable in memory
   bio.seek(0)
   workbook = bio.read()

.. note::

    ``engine`` is optional but recommended.  Setting the engine determines
    the version of workbook produced. Setting ``engine='xlrd'`` will produce an
    Excel 2003-format workbook (xls).  Using either ``'openpyxl'`` or
    ``'xlsxwriter'`` will produce an Excel 2007-format workbook (xlsx). If
    omitted, an Excel 2007-formatted workbook is produced.


.. _io.excel.writers:

Excel writer engines
''''''''''''''''''''

.. deprecated:: 1.2.0

   As the `xlwt <https://pypi.org/project/xlwt/>`__ package is no longer
   maintained, the ``xlwt`` engine will be removed from a future version
   of pandas. This is the only engine in pandas that supports writing to
   ``.xls`` files.

pandas chooses an Excel writer via two methods:

1. the ``engine`` keyword argument
2. the filename extension (via the default specified in config options)

By default, pandas uses the `XlsxWriter`_  for ``.xlsx``, `openpyxl`_
for ``.xlsm``, and `xlwt`_ for ``.xls`` files. If you have multiple
engines installed, you can set the default engine through :ref:`setting the
config options <options>` ``io.excel.xlsx.writer`` and
``io.excel.xls.writer``. pandas will fall back on `openpyxl`_ for ``.xlsx``
files if `Xlsxwriter`_ is not available.

.. _XlsxWriter: https://xlsxwriter.readthedocs.io
.. _openpyxl: https://openpyxl.readthedocs.io/
.. _xlwt: http://www.python-excel.org

To specify which writer you want to use, you can pass an engine keyword
argument to ``to_excel`` and to ``ExcelWriter``. The built-in engines are:

* ``openpyxl``: version 2.4 or higher is required
* ``xlsxwriter``
* ``xlwt``

.. code-block:: python

   # By setting the 'engine' in the DataFrame 'to_excel()' methods.
   df.to_excel("path_to_file.xlsx", sheet_name="Sheet1", engine="xlsxwriter")

   # By setting the 'engine' in the ExcelWriter constructor.
   writer = pd.ExcelWriter("path_to_file.xlsx", engine="xlsxwriter")

   # Or via pandas configuration.
   from pandas import options  # noqa: E402

   options.io.excel.xlsx.writer = "xlsxwriter"

   df.to_excel("path_to_file.xlsx", sheet_name="Sheet1")

.. _io.excel.style:

Style and formatting
''''''''''''''''''''

The look and feel of Excel worksheets created from pandas can be modified using the following parameters on the ``DataFrame``'s ``to_excel`` method.

* ``float_format`` : Format string for floating point numbers (default ``None``).
* ``freeze_panes`` : A tuple of two integers representing the bottommost row and rightmost column to freeze. Each of these parameters is one-based, so (1, 1) will freeze the first row and first column (default ``None``).

Using the `Xlsxwriter`_ engine provides many options for controlling the
format of an Excel worksheet created with the ``to_excel`` method.  Excellent examples can be found in the
`Xlsxwriter`_ documentation here: https://xlsxwriter.readthedocs.io/working_with_pandas.html

.. _io.ods:

OpenDocument Spreadsheets
-------------------------

.. versionadded:: 0.25

The :func:`~pandas.read_excel` method can also read OpenDocument spreadsheets
using the ``odfpy`` module. The semantics and features for reading
OpenDocument spreadsheets match what can be done for `Excel files`_ using
``engine='odf'``.

.. code-block:: python

   # Returns a DataFrame
   pd.read_excel("path_to_file.ods", engine="odf")

.. note::

   Currently pandas only supports *reading* OpenDocument spreadsheets. Writing
   is not implemented.

.. _io.xlsb:

Binary Excel (.xlsb) files
--------------------------

.. versionadded:: 1.0.0

The :func:`~pandas.read_excel` method can also read binary Excel files
using the ``pyxlsb`` module. The semantics and features for reading
binary Excel files mostly match what can be done for `Excel files`_ using
``engine='pyxlsb'``. ``pyxlsb`` does not recognize datetime types
in files and will return floats instead.

.. code-block:: python

   # Returns a DataFrame
   pd.read_excel("path_to_file.xlsb", engine="pyxlsb")

.. note::

   Currently pandas only supports *reading* binary Excel files. Writing
   is not implemented.


.. _io.clipboard:

Clipboard
---------

A handy way to grab data is to use the :meth:`~DataFrame.read_clipboard` method,
which takes the contents of the clipboard buffer and passes them to the
``read_csv`` method. For instance, you can copy the following text to the
clipboard (CTRL-C on many operating systems):

.. code-block:: console

     A B C
   x 1 4 p
   y 2 5 q
   z 3 6 r

And then import the data directly to a ``DataFrame`` by calling:

.. code-block:: python

    >>> clipdf = pd.read_clipboard()
    >>> clipdf
      A B C
    x 1 4 p
    y 2 5 q
    z 3 6 r

The ``to_clipboard`` method can be used to write the contents of a ``DataFrame`` to
the clipboard. Following which you can paste the clipboard contents into other
applications (CTRL-V on many operating systems). Here we illustrate writing a
``DataFrame`` into clipboard and reading it back.

.. code-block:: python

    >>> df = pd.DataFrame(
    ...     {"A": [1, 2, 3], "B": [4, 5, 6], "C": ["p", "q", "r"]}, index=["x", "y", "z"]
    ... )

    >>> df
      A B C
    x 1 4 p
    y 2 5 q
    z 3 6 r
    >>> df.to_clipboard()
    >>> pd.read_clipboard()
      A B C
    x 1 4 p
    y 2 5 q
    z 3 6 r

We can see that we got the same content back, which we had earlier written to the clipboard.

.. note::

   You may need to install xclip or xsel (with PyQt5, PyQt4 or qtpy) on Linux to use these methods.

.. _io.pickle:

Pickling
--------

All pandas objects are equipped with ``to_pickle`` methods which use Python's
``cPickle`` module to save data structures to disk using the pickle format.

.. ipython:: python

   df
   df.to_pickle("foo.pkl")

The ``read_pickle`` function in the ``pandas`` namespace can be used to load
any pickled pandas object (or any other pickled object) from file:


.. ipython:: python

   pd.read_pickle("foo.pkl")

.. ipython:: python
   :suppress:

   os.remove("foo.pkl")

.. warning::

   Loading pickled data received from untrusted sources can be unsafe.

   See: https://docs.python.org/3/library/pickle.html

.. warning::

   :func:`read_pickle` is only guaranteed backwards compatible back to pandas version 0.20.3

.. _io.pickle.compression:

Compressed pickle files
'''''''''''''''''''''''

:func:`read_pickle`, :meth:`DataFrame.to_pickle` and :meth:`Series.to_pickle` can read
and write compressed pickle files. The compression types of ``gzip``, ``bz2``, ``xz``, ``zstd`` are supported for reading and writing.
The ``zip`` file format only supports reading and must contain only one data file
to be read.

The compression type can be an explicit parameter or be inferred from the file extension.
If 'infer', then use ``gzip``, ``bz2``, ``zip``, ``xz``, ``zstd`` if filename ends in ``'.gz'``, ``'.bz2'``, ``'.zip'``,
``'.xz'``, or ``'.zst'``, respectively.

The compression parameter can also be a ``dict`` in order to pass options to the
compression protocol. It must have a ``'method'`` key set to the name
of the compression protocol, which must be one of
{``'zip'``, ``'gzip'``, ``'bz2'``, ``'xz'``, ``'zstd'``}. All other key-value pairs are passed to
the underlying compression library.

.. ipython:: python

   df = pd.DataFrame(
       {
           "A": np.random.randn(1000),
           "B": "foo",
           "C": pd.date_range("20130101", periods=1000, freq="s"),
       }
   )
   df

Using an explicit compression type:

.. ipython:: python

   df.to_pickle("data.pkl.compress", compression="gzip")
   rt = pd.read_pickle("data.pkl.compress", compression="gzip")
   rt

Inferring compression type from the extension:

.. ipython:: python

   df.to_pickle("data.pkl.xz", compression="infer")
   rt = pd.read_pickle("data.pkl.xz", compression="infer")
   rt

The default is to 'infer':

.. ipython:: python

   df.to_pickle("data.pkl.gz")
   rt = pd.read_pickle("data.pkl.gz")
   rt

   df["A"].to_pickle("s1.pkl.bz2")
   rt = pd.read_pickle("s1.pkl.bz2")
   rt

Passing options to the compression protocol in order to speed up compression:

.. ipython:: python

   df.to_pickle("data.pkl.gz", compression={"method": "gzip", "compresslevel": 1})

.. ipython:: python
   :suppress:

   os.remove("data.pkl.compress")
   os.remove("data.pkl.xz")
   os.remove("data.pkl.gz")
   os.remove("s1.pkl.bz2")

.. _io.msgpack:

msgpack
-------

pandas support for ``msgpack`` has been removed in version 1.0.0. It is
recommended to use :ref:`pickle <io.pickle>` instead.

Alternatively, you can also the Arrow IPC serialization format for on-the-wire
transmission of pandas objects. For documentation on pyarrow, see
`here <https://arrow.apache.org/docs/python/ipc.html>`__.


.. _io.hdf5:

HDF5 (PyTables)
---------------

``HDFStore`` is a dict-like object which reads and writes pandas using
the high performance HDF5 format using the excellent `PyTables
<https://www.pytables.org/>`__ library. See the :ref:`cookbook <cookbook.hdf>`
for some advanced strategies

.. warning::

   pandas uses PyTables for reading and writing HDF5 files, which allows
   serializing object-dtype data with pickle. Loading pickled data received from
   untrusted sources can be unsafe.

   See: https://docs.python.org/3/library/pickle.html for more.

.. ipython:: python
   :suppress:
   :okexcept:

   os.remove("store.h5")

.. ipython:: python

   store = pd.HDFStore("store.h5")
   print(store)

Objects can be written to the file just like adding key-value pairs to a
dict:

.. ipython:: python

   index = pd.date_range("1/1/2000", periods=8)
   s = pd.Series(np.random.randn(5), index=["a", "b", "c", "d", "e"])
   df = pd.DataFrame(np.random.randn(8, 3), index=index, columns=["A", "B", "C"])

   # store.put('s', s) is an equivalent method
   store["s"] = s

   store["df"] = df

   store

In a current or later Python session, you can retrieve stored objects:

.. ipython:: python

   # store.get('df') is an equivalent method
   store["df"]

   # dotted (attribute) access provides get as well
   store.df

Deletion of the object specified by the key:

.. ipython:: python

   # store.remove('df') is an equivalent method
   del store["df"]

   store

Closing a Store and using a context manager:

.. ipython:: python

   store.close()
   store
   store.is_open

   # Working with, and automatically closing the store using a context manager
   with pd.HDFStore("store.h5") as store:
       store.keys()

.. ipython:: python
   :suppress:

   store.close()
   os.remove("store.h5")



Read/write API
''''''''''''''

``HDFStore`` supports a top-level API using  ``read_hdf`` for reading and ``to_hdf`` for writing,
similar to how ``read_csv`` and ``to_csv`` work.

.. ipython:: python

   df_tl = pd.DataFrame({"A": list(range(5)), "B": list(range(5))})
   df_tl.to_hdf("store_tl.h5", "table", append=True)
   pd.read_hdf("store_tl.h5", "table", where=["index>2"])

.. ipython:: python
   :suppress:
   :okexcept:

   os.remove("store_tl.h5")


HDFStore will by default not drop rows that are all missing. This behavior can be changed by setting ``dropna=True``.


.. ipython:: python

   df_with_missing = pd.DataFrame(
       {
           "col1": [0, np.nan, 2],
           "col2": [1, np.nan, np.nan],
       }
   )
   df_with_missing

   df_with_missing.to_hdf("file.h5", "df_with_missing", format="table", mode="w")

   pd.read_hdf("file.h5", "df_with_missing")

   df_with_missing.to_hdf(
       "file.h5", "df_with_missing", format="table", mode="w", dropna=True
   )
   pd.read_hdf("file.h5", "df_with_missing")


.. ipython:: python
   :suppress:

   os.remove("file.h5")


.. _io.hdf5-fixed:

Fixed format
''''''''''''

The examples above show storing using ``put``, which write the HDF5 to ``PyTables`` in a fixed array format, called
the ``fixed`` format. These types of stores are **not** appendable once written (though you can simply
remove them and rewrite). Nor are they **queryable**; they must be
retrieved in their entirety. They also do not support dataframes with non-unique column names.
The ``fixed`` format stores offer very fast writing and slightly faster reading than ``table`` stores.
This format is specified by default when using ``put`` or ``to_hdf`` or by ``format='fixed'`` or ``format='f'``.

.. warning::

   A ``fixed`` format will raise a ``TypeError`` if you try to retrieve using a ``where``:

   .. code-block:: python

       >>> pd.DataFrame(np.random.randn(10, 2)).to_hdf("test_fixed.h5", "df")
       >>> pd.read_hdf("test_fixed.h5", "df", where="index>5")
       TypeError: cannot pass a where specification when reading a fixed format.
                  this store must be selected in its entirety


.. _io.hdf5-table:

Table format
''''''''''''

``HDFStore`` supports another ``PyTables`` format on disk, the ``table``
format. Conceptually a ``table`` is shaped very much like a DataFrame,
with rows and columns. A ``table`` may be appended to in the same or
other sessions.  In addition, delete and query type operations are
supported. This format is specified by ``format='table'`` or ``format='t'``
to ``append`` or ``put`` or ``to_hdf``.

This format can be set as an option as well ``pd.set_option('io.hdf.default_format','table')`` to
enable ``put/append/to_hdf`` to by default store in the ``table`` format.

.. ipython:: python
   :suppress:
   :okexcept:

   os.remove("store.h5")

.. ipython:: python

   store = pd.HDFStore("store.h5")
   df1 = df[0:4]
   df2 = df[4:]

   # append data (creates a table automatically)
   store.append("df", df1)
   store.append("df", df2)
   store

   # select the entire object
   store.select("df")

   # the type of stored data
   store.root.df._v_attrs.pandas_type

.. note::

   You can also create a ``table`` by passing ``format='table'`` or ``format='t'`` to a ``put`` operation.

.. _io.hdf5-keys:

Hierarchical keys
'''''''''''''''''

Keys to a store can be specified as a string. These can be in a
hierarchical path-name like format (e.g. ``foo/bar/bah``), which will
generate a hierarchy of sub-stores (or ``Groups`` in PyTables
parlance). Keys can be specified without the leading '/' and are **always**
absolute (e.g. 'foo' refers to '/foo'). Removal operations can remove
everything in the sub-store and **below**, so be *careful*.

.. ipython:: python

   store.put("foo/bar/bah", df)
   store.append("food/orange", df)
   store.append("food/apple", df)
   store

   # a list of keys are returned
   store.keys()

   # remove all nodes under this level
   store.remove("food")
   store


You can walk through the group hierarchy using the ``walk`` method which
will yield a tuple for each group key along with the relative keys of its contents.

.. ipython:: python

   for (path, subgroups, subkeys) in store.walk():
       for subgroup in subgroups:
           print("GROUP: {}/{}".format(path, subgroup))
       for subkey in subkeys:
           key = "/".join([path, subkey])
           print("KEY: {}".format(key))
           print(store.get(key))



.. warning::

    Hierarchical keys cannot be retrieved as dotted (attribute) access as described above for items stored under the root node.

    .. code-block:: ipython

       In [8]: store.foo.bar.bah
       AttributeError: 'HDFStore' object has no attribute 'foo'

       # you can directly access the actual PyTables node but using the root node
       In [9]: store.root.foo.bar.bah
       Out[9]:
       /foo/bar/bah (Group) ''
         children := ['block0_items' (Array), 'block0_values' (Array), 'axis0' (Array), 'axis1' (Array)]

    Instead, use explicit string based keys:

    .. ipython:: python

       store["foo/bar/bah"]


.. _io.hdf5-types:

Storing types
'''''''''''''

Storing mixed types in a table
++++++++++++++++++++++++++++++

Storing mixed-dtype data is supported. Strings are stored as a
fixed-width using the maximum size of the appended column. Subsequent attempts
at appending longer strings will raise a ``ValueError``.

Passing ``min_itemsize={`values`: size}`` as a parameter to append
will set a larger minimum for the string columns. Storing ``floats,
strings, ints, bools, datetime64`` are currently supported. For string
columns, passing ``nan_rep = 'nan'`` to append will change the default
nan representation on disk (which converts to/from ``np.nan``), this
defaults to ``nan``.

.. ipython:: python

    df_mixed = pd.DataFrame(
        {
            "A": np.random.randn(8),
            "B": np.random.randn(8),
            "C": np.array(np.random.randn(8), dtype="float32"),
            "string": "string",
            "int": 1,
            "bool": True,
            "datetime64": pd.Timestamp("20010102"),
        },
        index=list(range(8)),
    )
    df_mixed.loc[df_mixed.index[3:5], ["A", "B", "string", "datetime64"]] = np.nan

    store.append("df_mixed", df_mixed, min_itemsize={"values": 50})
    df_mixed1 = store.select("df_mixed")
    df_mixed1
    df_mixed1.dtypes.value_counts()

    # we have provided a minimum string column size
    store.root.df_mixed.table

Storing MultiIndex DataFrames
+++++++++++++++++++++++++++++

Storing MultiIndex ``DataFrames`` as tables is very similar to
storing/selecting from homogeneous index ``DataFrames``.

.. ipython:: python

        index = pd.MultiIndex(
            levels=[["foo", "bar", "baz", "qux"], ["one", "two", "three"]],
            codes=[[0, 0, 0, 1, 1, 2, 2, 3, 3, 3], [0, 1, 2, 0, 1, 1, 2, 0, 1, 2]],
            names=["foo", "bar"],
        )
        df_mi = pd.DataFrame(np.random.randn(10, 3), index=index, columns=["A", "B", "C"])
        df_mi

        store.append("df_mi", df_mi)
        store.select("df_mi")

        # the levels are automatically included as data columns
        store.select("df_mi", "foo=bar")

.. note::
   The ``index`` keyword is reserved and cannot be use as a level name.

.. _io.hdf5-query:

Querying
''''''''

Querying a table
++++++++++++++++

``select`` and ``delete`` operations have an optional criterion that can
be specified to select/delete only a subset of the data. This allows one
to have a very large on-disk table and retrieve only a portion of the
data.

A query is specified using the ``Term`` class under the hood, as a boolean expression.

* ``index`` and ``columns`` are supported indexers of ``DataFrames``.
* if ``data_columns`` are specified, these can be used as additional indexers.
* level name in a MultiIndex, with default name  ``level_0``, ``level_1``, … if not provided.

Valid comparison operators are:

``=, ==, !=, >, >=, <, <=``

Valid boolean expressions are combined with:

* ``|`` : or
* ``&`` : and
* ``(`` and ``)`` : for grouping

These rules are similar to how boolean expressions are used in pandas for indexing.

.. note::

   - ``=`` will be automatically expanded to the comparison operator ``==``
   - ``~`` is the not operator, but can only be used in very limited
     circumstances
   - If a list/tuple of expressions is passed they will be combined via ``&``

The following are valid expressions:

* ``'index >= date'``
* ``"columns = ['A', 'D']"``
* ``"columns in ['A', 'D']"``
* ``'columns = A'``
* ``'columns == A'``
* ``"~(columns = ['A', 'B'])"``
* ``'index > df.index[3] & string = "bar"'``
* ``'(index > df.index[3] & index <= df.index[6]) | string = "bar"'``
* ``"ts >= Timestamp('2012-02-01')"``
* ``"major_axis>=20130101"``

The ``indexers`` are on the left-hand side of the sub-expression:

``columns``, ``major_axis``, ``ts``

The right-hand side of the sub-expression (after a comparison operator) can be:

* functions that will be evaluated, e.g. ``Timestamp('2012-02-01')``
* strings, e.g. ``"bar"``
* date-like, e.g. ``20130101``, or ``"20130101"``
* lists, e.g. ``"['A', 'B']"``
* variables that are defined in the local names space, e.g. ``date``

.. note::

   Passing a string to a query by interpolating it into the query
   expression is not recommended. Simply assign the string of interest to a
   variable and use that variable in an expression. For example, do this

   .. code-block:: python

      string = "HolyMoly'"
      store.select("df", "index == string")

   instead of this

   .. code-block:: ipython

      string = "HolyMoly'"
      store.select('df', f'index == {string}')

   The latter will **not** work and will raise a ``SyntaxError``.Note that
   there's a single quote followed by a double quote in the ``string``
   variable.

   If you *must* interpolate, use the ``'%r'`` format specifier

   .. code-block:: python

      store.select("df", "index == %r" % string)

   which will quote ``string``.


Here are some examples:

.. ipython:: python

    dfq = pd.DataFrame(
        np.random.randn(10, 4),
        columns=list("ABCD"),
        index=pd.date_range("20130101", periods=10),
    )
    store.append("dfq", dfq, format="table", data_columns=True)

Use boolean expressions, with in-line function evaluation.

.. ipython:: python

    store.select("dfq", "index>pd.Timestamp('20130104') & columns=['A', 'B']")

Use inline column reference.

.. ipython:: python

   store.select("dfq", where="A>0 or C>0")

The ``columns`` keyword can be supplied to select a list of columns to be
returned, this is equivalent to passing a
``'columns=list_of_columns_to_filter'``:

.. ipython:: python

   store.select("df", "columns=['A', 'B']")

``start`` and ``stop`` parameters can be specified to limit the total search
space. These are in terms of the total number of rows in a table.

.. note::

   ``select`` will raise a ``ValueError`` if the query expression has an unknown
   variable reference. Usually this means that you are trying to select on a column
   that is **not** a data_column.

   ``select`` will raise a ``SyntaxError`` if the query expression is not valid.


.. _io.hdf5-timedelta:

Query timedelta64[ns]
+++++++++++++++++++++

You can store and query using the ``timedelta64[ns]`` type. Terms can be
specified in the format: ``<float>(<unit>)``, where float may be signed (and fractional), and unit can be
``D,s,ms,us,ns`` for the timedelta. Here's an example:

.. ipython:: python

   from datetime import timedelta

   dftd = pd.DataFrame(
       {
           "A": pd.Timestamp("20130101"),
           "B": [
               pd.Timestamp("20130101") + timedelta(days=i, seconds=10)
               for i in range(10)
           ],
       }
   )
   dftd["C"] = dftd["A"] - dftd["B"]
   dftd
   store.append("dftd", dftd, data_columns=True)
   store.select("dftd", "C<'-3.5D'")

.. _io.query_multi:

Query MultiIndex
++++++++++++++++

Selecting from a ``MultiIndex`` can be achieved by using the name of the level.

.. ipython:: python

   df_mi.index.names
   store.select("df_mi", "foo=baz and bar=two")

If the ``MultiIndex`` levels names are ``None``, the levels are automatically made available via
the ``level_n`` keyword with ``n`` the level of the ``MultiIndex`` you want to select from.

.. ipython:: python

   index = pd.MultiIndex(
       levels=[["foo", "bar", "baz", "qux"], ["one", "two", "three"]],
       codes=[[0, 0, 0, 1, 1, 2, 2, 3, 3, 3], [0, 1, 2, 0, 1, 1, 2, 0, 1, 2]],
   )
   df_mi_2 = pd.DataFrame(np.random.randn(10, 3), index=index, columns=["A", "B", "C"])
   df_mi_2

   store.append("df_mi_2", df_mi_2)

   # the levels are automatically included as data columns with keyword level_n
   store.select("df_mi_2", "level_0=foo and level_1=two")


Indexing
++++++++

You can create/modify an index for a table with ``create_table_index``
after data is already in the table (after and ``append/put``
operation). Creating a table index is **highly** encouraged. This will
speed your queries a great deal when you use a ``select`` with the
indexed dimension as the ``where``.

.. note::

   Indexes are automagically created on the indexables
   and any data columns you specify. This behavior can be turned off by passing
   ``index=False`` to ``append``.

.. ipython:: python

   # we have automagically already created an index (in the first section)
   i = store.root.df.table.cols.index.index
   i.optlevel, i.kind

   # change an index by passing new parameters
   store.create_table_index("df", optlevel=9, kind="full")
   i = store.root.df.table.cols.index.index
   i.optlevel, i.kind

Oftentimes when appending large amounts of data to a store, it is useful to turn off index creation for each append, then recreate at the end.

.. ipython:: python

   df_1 = pd.DataFrame(np.random.randn(10, 2), columns=list("AB"))
   df_2 = pd.DataFrame(np.random.randn(10, 2), columns=list("AB"))

   st = pd.HDFStore("appends.h5", mode="w")
   st.append("df", df_1, data_columns=["B"], index=False)
   st.append("df", df_2, data_columns=["B"], index=False)
   st.get_storer("df").table

Then create the index when finished appending.

.. ipython:: python

   st.create_table_index("df", columns=["B"], optlevel=9, kind="full")
   st.get_storer("df").table

   st.close()

.. ipython:: python
   :suppress:
   :okexcept:

   os.remove("appends.h5")

See `here <https://stackoverflow.com/questions/17893370/ptrepack-sortby-needs-full-index>`__ for how to create a completely-sorted-index (CSI) on an existing store.

.. _io.hdf5-query-data-columns:

Query via data columns
++++++++++++++++++++++

You can designate (and index) certain columns that you want to be able
to perform queries (other than the ``indexable`` columns, which you can
always query). For instance say you want to perform this common
operation, on-disk, and return just the frame that matches this
query. You can specify ``data_columns = True`` to force all columns to
be ``data_columns``.

.. ipython:: python

   df_dc = df.copy()
   df_dc["string"] = "foo"
   df_dc.loc[df_dc.index[4:6], "string"] = np.nan
   df_dc.loc[df_dc.index[7:9], "string"] = "bar"
   df_dc["string2"] = "cool"
   df_dc.loc[df_dc.index[1:3], ["B", "C"]] = 1.0
   df_dc

   # on-disk operations
   store.append("df_dc", df_dc, data_columns=["B", "C", "string", "string2"])
   store.select("df_dc", where="B > 0")

   # getting creative
   store.select("df_dc", "B > 0 & C > 0 & string == foo")

   # this is in-memory version of this type of selection
   df_dc[(df_dc.B > 0) & (df_dc.C > 0) & (df_dc.string == "foo")]

   # we have automagically created this index and the B/C/string/string2
   # columns are stored separately as ``PyTables`` columns
   store.root.df_dc.table

There is some performance degradation by making lots of columns into
``data columns``, so it is up to the user to designate these. In addition,
you cannot change data columns (nor indexables) after the first
append/put operation (Of course you can simply read in the data and
create a new table!).

Iterator
++++++++

You can pass ``iterator=True`` or ``chunksize=number_in_a_chunk``
to ``select`` and ``select_as_multiple`` to return an iterator on the results.
The default is 50,000 rows returned in a chunk.

.. ipython:: python

   for df in store.select("df", chunksize=3):
       print(df)

.. note::

   You can also use the iterator with ``read_hdf`` which will open, then
   automatically close the store when finished iterating.

   .. code-block:: python

      for df in pd.read_hdf("store.h5", "df", chunksize=3):
          print(df)

Note, that the chunksize keyword applies to the **source** rows. So if you
are doing a query, then the chunksize will subdivide the total rows in the table
and the query applied, returning an iterator on potentially unequal sized chunks.

Here is a recipe for generating a query and using it to create equal sized return
chunks.

.. ipython:: python

   dfeq = pd.DataFrame({"number": np.arange(1, 11)})
   dfeq

   store.append("dfeq", dfeq, data_columns=["number"])

   def chunks(l, n):
       return [l[i: i + n] for i in range(0, len(l), n)]

   evens = [2, 4, 6, 8, 10]
   coordinates = store.select_as_coordinates("dfeq", "number=evens")
   for c in chunks(coordinates, 2):
       print(store.select("dfeq", where=c))

Advanced queries
++++++++++++++++

Select a single column
^^^^^^^^^^^^^^^^^^^^^^

To retrieve a single indexable or data column, use the
method ``select_column``. This will, for example, enable you to get the index
very quickly. These return a ``Series`` of the result, indexed by the row number.
These do not currently accept the ``where`` selector.

.. ipython:: python

   store.select_column("df_dc", "index")
   store.select_column("df_dc", "string")

.. _io.hdf5-selecting_coordinates:

Selecting coordinates
^^^^^^^^^^^^^^^^^^^^^

Sometimes you want to get the coordinates (a.k.a the index locations) of your query. This returns an
``Int64Index`` of the resulting locations. These coordinates can also be passed to subsequent
``where`` operations.

.. ipython:: python

   df_coord = pd.DataFrame(
       np.random.randn(1000, 2), index=pd.date_range("20000101", periods=1000)
   )
   store.append("df_coord", df_coord)
   c = store.select_as_coordinates("df_coord", "index > 20020101")
   c
   store.select("df_coord", where=c)

.. _io.hdf5-where_mask:

Selecting using a where mask
^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Sometime your query can involve creating a list of rows to select. Usually this ``mask`` would
be a resulting ``index`` from an indexing operation. This example selects the months of
a datetimeindex which are 5.

.. ipython:: python

   df_mask = pd.DataFrame(
       np.random.randn(1000, 2), index=pd.date_range("20000101", periods=1000)
   )
   store.append("df_mask", df_mask)
   c = store.select_column("df_mask", "index")
   where = c[pd.DatetimeIndex(c).month == 5].index
   store.select("df_mask", where=where)

Storer object
^^^^^^^^^^^^^

If you want to inspect the stored object, retrieve via
``get_storer``. You could use this programmatically to say get the number
of rows in an object.

.. ipython:: python

   store.get_storer("df_dc").nrows


Multiple table queries
++++++++++++++++++++++

The methods ``append_to_multiple`` and
``select_as_multiple`` can perform appending/selecting from
multiple tables at once. The idea is to have one table (call it the
selector table) that you index most/all of the columns, and perform your
queries. The other table(s) are data tables with an index matching the
selector table's index. You can then perform a very fast query
on the selector table, yet get lots of data back. This method is similar to
having a very wide table, but enables more efficient queries.

The ``append_to_multiple`` method splits a given single DataFrame
into multiple tables according to ``d``, a dictionary that maps the
table names to a list of 'columns' you want in that table. If ``None``
is used in place of a list, that table will have the remaining
unspecified columns of the given DataFrame. The argument ``selector``
defines which table is the selector table (which you can make queries from).
The argument ``dropna`` will drop rows from the input ``DataFrame`` to ensure
tables are synchronized.  This means that if a row for one of the tables
being written to is entirely ``np.NaN``, that row will be dropped from all tables.

If ``dropna`` is False, **THE USER IS RESPONSIBLE FOR SYNCHRONIZING THE TABLES**.
Remember that entirely ``np.Nan`` rows are not written to the HDFStore, so if
you choose to call ``dropna=False``, some tables may have more rows than others,
and therefore ``select_as_multiple`` may not work or it may return unexpected
results.

.. ipython:: python

   df_mt = pd.DataFrame(
       np.random.randn(8, 6),
       index=pd.date_range("1/1/2000", periods=8),
       columns=["A", "B", "C", "D", "E", "F"],
   )
   df_mt["foo"] = "bar"
   df_mt.loc[df_mt.index[1], ("A", "B")] = np.nan

   # you can also create the tables individually
   store.append_to_multiple(
       {"df1_mt": ["A", "B"], "df2_mt": None}, df_mt, selector="df1_mt"
   )
   store

   # individual tables were created
   store.select("df1_mt")
   store.select("df2_mt")

   # as a multiple
   store.select_as_multiple(
       ["df1_mt", "df2_mt"],
       where=["A>0", "B>0"],
       selector="df1_mt",
   )


Delete from a table
'''''''''''''''''''

You can delete from a table selectively by specifying a ``where``. In
deleting rows, it is important to understand the ``PyTables`` deletes
rows by erasing the rows, then **moving** the following data. Thus
deleting can potentially be a very expensive operation depending on the
orientation of your data. To get optimal performance, it's
worthwhile to have the dimension you are deleting be the first of the
``indexables``.

Data is ordered (on the disk) in terms of the ``indexables``. Here's a
simple use case. You store panel-type data, with dates in the
``major_axis`` and ids in the ``minor_axis``. The data is then
interleaved like this:

* date_1
    * id_1
    * id_2
    *  .
    * id_n
* date_2
    * id_1
    *  .
    * id_n

It should be clear that a delete operation on the ``major_axis`` will be
fairly quick, as one chunk is removed, then the following data moved. On
the other hand a delete operation on the ``minor_axis`` will be very
expensive. In this case it would almost certainly be faster to rewrite
the table using a ``where`` that selects all but the missing data.

.. warning::

   Please note that HDF5 **DOES NOT RECLAIM SPACE** in the h5 files
   automatically. Thus, repeatedly deleting (or removing nodes) and adding
   again, **WILL TEND TO INCREASE THE FILE SIZE**.

   To *repack and clean* the file, use :ref:`ptrepack <io.hdf5-ptrepack>`.

.. _io.hdf5-notes:

Notes & caveats
'''''''''''''''


Compression
+++++++++++

``PyTables`` allows the stored data to be compressed. This applies to
all kinds of stores, not just tables. Two parameters are used to
control compression: ``complevel`` and ``complib``.

* ``complevel`` specifies if and how hard data is to be compressed.
  ``complevel=0`` and ``complevel=None`` disables compression and
  ``0<complevel<10`` enables compression.

* ``complib`` specifies which compression library to use.
  If nothing is  specified the default library ``zlib`` is used. A
  compression library usually optimizes for either good compression rates
  or speed and the results will depend on the type of data. Which type of
  compression to choose depends on your specific needs and data. The list
  of supported compression libraries:

  - `zlib <https://zlib.net/>`_: The default compression library.
    A classic in terms of compression, achieves good compression
    rates but is somewhat slow.
  - `lzo <https://www.oberhumer.com/opensource/lzo/>`_: Fast
    compression and decompression.
  - `bzip2 <https://sourceware.org/bzip2/>`_: Good compression rates.
  - `blosc <https://www.blosc.org/>`_: Fast compression and
    decompression.

    Support for alternative blosc compressors:

    - `blosc:blosclz <https://www.blosc.org/>`_ This is the
      default compressor for ``blosc``
    - `blosc:lz4
      <https://fastcompression.blogspot.com/p/lz4.html>`_:
      A compact, very popular and fast compressor.
    - `blosc:lz4hc
      <https://fastcompression.blogspot.com/p/lz4.html>`_:
      A tweaked version of LZ4, produces better
      compression ratios at the expense of speed.
    - `blosc:snappy <https://google.github.io/snappy/>`_:
      A popular compressor used in many places.
    - `blosc:zlib <https://zlib.net/>`_: A classic;
      somewhat slower than the previous ones, but
      achieving better compression ratios.
    - `blosc:zstd <https://facebook.github.io/zstd/>`_: An
      extremely well balanced codec; it provides the best
      compression ratios among the others above, and at
      reasonably fast speed.

  If ``complib`` is defined as something other than the listed libraries a
  ``ValueError`` exception is issued.

.. note::

   If the library specified with the ``complib`` option is missing on your platform,
   compression defaults to ``zlib`` without further ado.

Enable compression for all objects within the file:

.. code-block:: python

   store_compressed = pd.HDFStore(
       "store_compressed.h5", complevel=9, complib="blosc:blosclz"
   )

Or on-the-fly compression (this only applies to tables) in stores where compression is not enabled:

.. code-block:: python

   store.append("df", df, complib="zlib", complevel=5)

.. _io.hdf5-ptrepack:

ptrepack
++++++++

``PyTables`` offers better write performance when tables are compressed after
they are written, as opposed to turning on compression at the very
beginning. You can use the supplied ``PyTables`` utility
``ptrepack``. In addition, ``ptrepack`` can change compression levels
after the fact.

.. code-block:: console

   ptrepack --chunkshape=auto --propindexes --complevel=9 --complib=blosc in.h5 out.h5

Furthermore ``ptrepack in.h5 out.h5`` will *repack* the file to allow
you to reuse previously deleted space. Alternatively, one can simply
remove the file and write again, or use the ``copy`` method.

.. _io.hdf5-caveats:

Caveats
+++++++

.. warning::

   ``HDFStore`` is **not-threadsafe for writing**. The underlying
   ``PyTables`` only supports concurrent reads (via threading or
   processes). If you need reading and writing *at the same time*, you
   need to serialize these operations in a single thread in a single
   process. You will corrupt your data otherwise. See the (:issue:`2397`) for more information.

* If you use locks to manage write access between multiple processes, you
  may want to use :py:func:`~os.fsync` before releasing write locks. For
  convenience you can use ``store.flush(fsync=True)`` to do this for you.
* Once a ``table`` is created columns (DataFrame)
  are fixed; only exactly the same columns can be appended
* Be aware that timezones (e.g., ``pytz.timezone('US/Eastern')``)
  are not necessarily equal across timezone versions.  So if data is
  localized to a specific timezone in the HDFStore using one version
  of a timezone library and that data is updated with another version, the data
  will be converted to UTC since these timezones are not considered
  equal.  Either use the same version of timezone library or use ``tz_convert`` with
  the updated timezone definition.

.. warning::

   ``PyTables`` will show a ``NaturalNameWarning`` if a column name
   cannot be used as an attribute selector.
   *Natural* identifiers contain only letters, numbers, and underscores,
   and may not begin with a number.
   Other identifiers cannot be used in a ``where`` clause
   and are generally a bad idea.

.. _io.hdf5-data_types:

DataTypes
'''''''''

``HDFStore`` will map an object dtype to the ``PyTables`` underlying
dtype. This means the following types are known to work:

======================================================  =========================
Type                                                    Represents missing values
======================================================  =========================
floating : ``float64, float32, float16``                ``np.nan``
integer : ``int64, int32, int8, uint64,uint32, uint8``
boolean
``datetime64[ns]``                                      ``NaT``
``timedelta64[ns]``                                     ``NaT``
categorical : see the section below
object : ``strings``                                    ``np.nan``
======================================================  =========================

``unicode`` columns are not supported, and **WILL FAIL**.

.. _io.hdf5-categorical:

Categorical data
++++++++++++++++

You can write data that contains ``category`` dtypes to a ``HDFStore``.
Queries work the same as if it was an object array. However, the ``category`` dtyped data is
stored in a more efficient manner.

.. ipython:: python

   dfcat = pd.DataFrame(
       {"A": pd.Series(list("aabbcdba")).astype("category"), "B": np.random.randn(8)}
   )
   dfcat
   dfcat.dtypes
   cstore = pd.HDFStore("cats.h5", mode="w")
   cstore.append("dfcat", dfcat, format="table", data_columns=["A"])
   result = cstore.select("dfcat", where="A in ['b', 'c']")
   result
   result.dtypes

.. ipython:: python
   :suppress:
   :okexcept:

   cstore.close()
   os.remove("cats.h5")


String columns
++++++++++++++

**min_itemsize**

The underlying implementation of ``HDFStore`` uses a fixed column width (itemsize) for string columns.
A string column itemsize is calculated as the maximum of the
length of data (for that column) that is passed to the ``HDFStore``, **in the first append**. Subsequent appends,
may introduce a string for a column **larger** than the column can hold, an Exception will be raised (otherwise you
could have a silent truncation of these columns, leading to loss of information). In the future we may relax this and
allow a user-specified truncation to occur.

Pass ``min_itemsize`` on the first table creation to a-priori specify the minimum length of a particular string column.
``min_itemsize`` can be an integer, or a dict mapping a column name to an integer. You can pass ``values`` as a key to
allow all *indexables* or *data_columns* to have this min_itemsize.

Passing a ``min_itemsize`` dict will cause all passed columns to be created as *data_columns* automatically.

.. note::

   If you are not passing any ``data_columns``, then the ``min_itemsize`` will be the maximum of the length of any string passed

.. ipython:: python

   dfs = pd.DataFrame({"A": "foo", "B": "bar"}, index=list(range(5)))
   dfs

   # A and B have a size of 30
   store.append("dfs", dfs, min_itemsize=30)
   store.get_storer("dfs").table

   # A is created as a data_column with a size of 30
   # B is size is calculated
   store.append("dfs2", dfs, min_itemsize={"A": 30})
   store.get_storer("dfs2").table

**nan_rep**

String columns will serialize a ``np.nan`` (a missing value) with the ``nan_rep`` string representation. This defaults to the string value ``nan``.
You could inadvertently turn an actual ``nan`` value into a missing value.

.. ipython:: python

   dfss = pd.DataFrame({"A": ["foo", "bar", "nan"]})
   dfss

   store.append("dfss", dfss)
   store.select("dfss")

   # here you need to specify a different nan rep
   store.append("dfss2", dfss, nan_rep="_nan_")
   store.select("dfss2")

.. _io.external_compatibility:

External compatibility
''''''''''''''''''''''

``HDFStore`` writes ``table`` format objects in specific formats suitable for
producing loss-less round trips to pandas objects. For external
compatibility, ``HDFStore`` can read native ``PyTables`` format
tables.

It is possible to write an ``HDFStore`` object that can easily be imported into ``R`` using the
``rhdf5`` library (`Package website`_). Create a table format store like this:

.. _package website: https://www.bioconductor.org/packages/release/bioc/html/rhdf5.html

.. ipython:: python

   df_for_r = pd.DataFrame(
       {
           "first": np.random.rand(100),
           "second": np.random.rand(100),
           "class": np.random.randint(0, 2, (100,)),
       },
       index=range(100),
   )
   df_for_r.head()

   store_export = pd.HDFStore("export.h5")
   store_export.append("df_for_r", df_for_r, data_columns=df_dc.columns)
   store_export

.. ipython:: python
   :suppress:

   store_export.close()
   os.remove("export.h5")

In R this file can be read into a ``data.frame`` object using the ``rhdf5``
library. The following example function reads the corresponding column names
and data values from the values and assembles them into a ``data.frame``:

.. code-block:: R

   # Load values and column names for all datasets from corresponding nodes and
   # insert them into one data.frame object.

   library(rhdf5)

   loadhdf5data <- function(h5File) {

   listing <- h5ls(h5File)
   # Find all data nodes, values are stored in *_values and corresponding column
   # titles in *_items
   data_nodes <- grep("_values", listing$name)
   name_nodes <- grep("_items", listing$name)
   data_paths = paste(listing$group[data_nodes], listing$name[data_nodes], sep = "/")
   name_paths = paste(listing$group[name_nodes], listing$name[name_nodes], sep = "/")
   columns = list()
   for (idx in seq(data_paths)) {
     # NOTE: matrices returned by h5read have to be transposed to obtain
     # required Fortran order!
     data <- data.frame(t(h5read(h5File, data_paths[idx])))
     names <- t(h5read(h5File, name_paths[idx]))
     entry <- data.frame(data)
     colnames(entry) <- names
     columns <- append(columns, entry)
   }

   data <- data.frame(columns)

   return(data)
   }

Now you can import the ``DataFrame`` into R:

.. code-block:: R

   > data = loadhdf5data("transfer.hdf5")
   > head(data)
            first    second class
   1 0.4170220047 0.3266449     0
   2 0.7203244934 0.5270581     0
   3 0.0001143748 0.8859421     1
   4 0.3023325726 0.3572698     1
   5 0.1467558908 0.9085352     1
   6 0.0923385948 0.6233601     1

.. note::
   The R function lists the entire HDF5 file's contents and assembles the
   ``data.frame`` object from all matching nodes, so use this only as a
   starting point if you have stored multiple ``DataFrame`` objects to a
   single HDF5 file.


Performance
'''''''''''

* ``tables`` format come with a writing performance penalty as compared to
  ``fixed`` stores. The benefit is the ability to append/delete and
  query (potentially very large amounts of data).  Write times are
  generally longer as compared with regular stores. Query times can
  be quite fast, especially on an indexed axis.
* You can pass ``chunksize=<int>`` to ``append``, specifying the
  write chunksize (default is 50000). This will significantly lower
  your memory usage on writing.
* You can pass ``expectedrows=<int>`` to the first ``append``,
  to set the TOTAL number of rows that ``PyTables`` will expect.
  This will optimize read/write performance.
* Duplicate rows can be written to tables, but are filtered out in
  selection (with the last items being selected; thus a table is
  unique on major, minor pairs)
* A ``PerformanceWarning`` will be raised if you are attempting to
  store types that will be pickled by PyTables (rather than stored as
  endemic types). See
  `Here <https://stackoverflow.com/questions/14355151/how-to-make-pandas-hdfstore-put-operation-faster/14370190#14370190>`__
  for more information and some solutions.


.. ipython:: python
   :suppress:

   store.close()
   os.remove("store.h5")


.. _io.feather:

Feather
-------

Feather provides binary columnar serialization for data frames. It is designed to make reading and writing data
frames efficient, and to make sharing data across data analysis languages easy.

Feather is designed to faithfully serialize and de-serialize DataFrames, supporting all of the pandas
dtypes, including extension dtypes such as categorical and datetime with tz.

Several caveats:

* The format will NOT write an ``Index``, or ``MultiIndex`` for the
  ``DataFrame`` and will raise an error if a non-default one is provided. You
  can ``.reset_index()`` to store the index or ``.reset_index(drop=True)`` to
  ignore it.
* Duplicate column names and non-string columns names are not supported
* Actual Python objects in object dtype columns are not supported. These will
  raise a helpful error message on an attempt at serialization.

See the `Full Documentation <https://github.com/wesm/feather>`__.

.. ipython:: python

   df = pd.DataFrame(
       {
           "a": list("abc"),
           "b": list(range(1, 4)),
           "c": np.arange(3, 6).astype("u1"),
           "d": np.arange(4.0, 7.0, dtype="float64"),
           "e": [True, False, True],
           "f": pd.Categorical(list("abc")),
           "g": pd.date_range("20130101", periods=3),
           "h": pd.date_range("20130101", periods=3, tz="US/Eastern"),
           "i": pd.date_range("20130101", periods=3, freq="ns"),
       }
   )

   df
   df.dtypes

Write to a feather file.

.. ipython:: python

   df.to_feather("example.feather")

Read from a feather file.

.. ipython:: python
   :okwarning:

   result = pd.read_feather("example.feather")
   result

   # we preserve dtypes
   result.dtypes

.. ipython:: python
   :suppress:

   os.remove("example.feather")


.. _io.parquet:

Parquet
-------

`Apache Parquet <https://parquet.apache.org/>`__ provides a partitioned binary columnar serialization for data frames. It is designed to
make reading and writing data frames efficient, and to make sharing data across data analysis
languages easy. Parquet can use a variety of compression techniques to shrink the file size as much as possible
while still maintaining good read performance.

Parquet is designed to faithfully serialize and de-serialize ``DataFrame`` s, supporting all of the pandas
dtypes, including extension dtypes such as datetime with tz.

Several caveats.

* Duplicate column names and non-string columns names are not supported.
* The ``pyarrow`` engine always writes the index to the output, but ``fastparquet`` only writes non-default
  indexes. This extra column can cause problems for non-pandas consumers that are not expecting it. You can
  force including or omitting indexes with the ``index`` argument, regardless of the underlying engine.
* Index level names, if specified, must be strings.
* In the ``pyarrow`` engine, categorical dtypes for non-string types can be serialized to parquet, but will de-serialize as their primitive dtype.
* The ``pyarrow`` engine preserves the ``ordered`` flag of categorical dtypes with string types. ``fastparquet`` does not preserve the ``ordered`` flag.
* Non supported types include ``Interval`` and actual Python object types. These will raise a helpful error message
  on an attempt at serialization. ``Period`` type is supported with pyarrow >= 0.16.0.
* The ``pyarrow`` engine preserves extension data types such as the nullable integer and string data
  type (requiring pyarrow >= 0.16.0, and requiring the extension type to implement the needed protocols,
  see the :ref:`extension types documentation <extending.extension.arrow>`).

You can specify an ``engine`` to direct the serialization. This can be one of ``pyarrow``, or ``fastparquet``, or ``auto``.
If the engine is NOT specified, then the ``pd.options.io.parquet.engine`` option is checked; if this is also ``auto``,
then ``pyarrow`` is tried, and falling back to ``fastparquet``.

See the documentation for `pyarrow <https://arrow.apache.org/docs/python/>`__ and `fastparquet <https://fastparquet.readthedocs.io/en/latest/>`__.

.. note::

   These engines are very similar and should read/write nearly identical parquet format files.
   ``pyarrow>=8.0.0`` supports timedelta data, ``fastparquet>=0.1.4`` supports timezone aware datetimes.
   These libraries differ by having different underlying dependencies (``fastparquet`` by using ``numba``, while ``pyarrow`` uses a c-library).

.. ipython:: python

   df = pd.DataFrame(
       {
           "a": list("abc"),
           "b": list(range(1, 4)),
           "c": np.arange(3, 6).astype("u1"),
           "d": np.arange(4.0, 7.0, dtype="float64"),
           "e": [True, False, True],
           "f": pd.date_range("20130101", periods=3),
           "g": pd.date_range("20130101", periods=3, tz="US/Eastern"),
           "h": pd.Categorical(list("abc")),
           "i": pd.Categorical(list("abc"), ordered=True),
       }
   )

   df
   df.dtypes

Write to a parquet file.

.. ipython:: python
   :okwarning:

   df.to_parquet("example_pa.parquet", engine="pyarrow")
   df.to_parquet("example_fp.parquet", engine="fastparquet")

Read from a parquet file.

.. ipython:: python
   :okwarning:

   result = pd.read_parquet("example_fp.parquet", engine="fastparquet")
   result = pd.read_parquet("example_pa.parquet", engine="pyarrow")

   result.dtypes

Read only certain columns of a parquet file.

.. ipython:: python

   result = pd.read_parquet(
       "example_fp.parquet",
       engine="fastparquet",
       columns=["a", "b"],
   )
   result = pd.read_parquet(
       "example_pa.parquet",
       engine="pyarrow",
       columns=["a", "b"],
   )
   result.dtypes


.. ipython:: python
   :suppress:

   os.remove("example_pa.parquet")
   os.remove("example_fp.parquet")


Handling indexes
''''''''''''''''

Serializing a ``DataFrame`` to parquet may include the implicit index as one or
more columns in the output file. Thus, this code:

.. ipython:: python

    df = pd.DataFrame({"a": [1, 2], "b": [3, 4]})
    df.to_parquet("test.parquet", engine="pyarrow")

creates a parquet file with *three* columns if you use ``pyarrow`` for serialization:
``a``, ``b``, and ``__index_level_0__``. If you're using ``fastparquet``, the
index `may or may not <https://fastparquet.readthedocs.io/en/latest/api.html#fastparquet.write>`_
be written to the file.

This unexpected extra column causes some databases like Amazon Redshift to reject
the file, because that column doesn't exist in the target table.

If you want to omit a dataframe's indexes when writing, pass ``index=False`` to
:func:`~pandas.DataFrame.to_parquet`:

.. ipython:: python

    df.to_parquet("test.parquet", index=False)

This creates a parquet file with just the two expected columns, ``a`` and ``b``.
If your ``DataFrame`` has a custom index, you won't get it back when you load
this file into a ``DataFrame``.

Passing ``index=True`` will *always* write the index, even if that's not the
underlying engine's default behavior.

.. ipython:: python
   :suppress:

   os.remove("test.parquet")


Partitioning Parquet files
''''''''''''''''''''''''''

Parquet supports partitioning of data based on the values of one or more columns.

.. ipython:: python

    df = pd.DataFrame({"a": [0, 0, 1, 1], "b": [0, 1, 0, 1]})
    df.to_parquet(path="test", engine="pyarrow", partition_cols=["a"], compression=None)

The ``path`` specifies the parent directory to which data will be saved.
The ``partition_cols`` are the column names by which the dataset will be partitioned.
Columns are partitioned in the order they are given. The partition splits are
determined by the unique values in the partition columns.
The above example creates a partitioned dataset that may look like:

.. code-block:: text

    test
    ├── a=0
    │   ├── 0bac803e32dc42ae83fddfd029cbdebc.parquet
    │   └──  ...
    └── a=1
        ├── e6ab24a4f45147b49b54a662f0c412a3.parquet
        └── ...

.. ipython:: python
   :suppress:

   from shutil import rmtree

   try:
       rmtree("test")
   except OSError:
       pass

.. _io.orc:

ORC
---

.. versionadded:: 1.0.0

Similar to the :ref:`parquet <io.parquet>` format, the `ORC Format <https://orc.apache.org/>`__ is a binary columnar serialization
for data frames. It is designed to make reading data frames efficient. pandas provides both the reader and the writer for the
ORC format, :func:`~pandas.read_orc` and :func:`~pandas.DataFrame.to_orc`. This requires the `pyarrow <https://arrow.apache.org/docs/python/>`__ library.

.. warning::

   * It is *highly recommended* to install pyarrow using conda due to some issues occurred by pyarrow.
   * :func:`~pandas.DataFrame.to_orc` requires pyarrow>=7.0.0.
   * :func:`~pandas.read_orc` and :func:`~pandas.DataFrame.to_orc` are not supported on Windows yet, you can find valid environments on :ref:`install optional dependencies <install.warn_orc>`.
   * For supported dtypes please refer to `supported ORC features in Arrow <https://arrow.apache.org/docs/cpp/orc.html#data-types>`__.
   * Currently timezones in datetime columns are not preserved when a dataframe is converted into ORC files.

.. ipython:: python

   df = pd.DataFrame(
       {
           "a": list("abc"),
           "b": list(range(1, 4)),
           "c": np.arange(4.0, 7.0, dtype="float64"),
           "d": [True, False, True],
           "e": pd.date_range("20130101", periods=3),
       }
   )

   df
   df.dtypes

Write to an orc file.

.. ipython:: python
   :okwarning:

   df.to_orc("example_pa.orc", engine="pyarrow")

Read from an orc file.

.. ipython:: python
   :okwarning:

   result = pd.read_orc("example_pa.orc")

   result.dtypes

Read only certain columns of an orc file.

.. ipython:: python

   result = pd.read_orc(
       "example_pa.orc",
       columns=["a", "b"],
   )
   result.dtypes


.. ipython:: python
   :suppress:

   os.remove("example_pa.orc")


.. _io.sql:

SQL queries
-----------

The :mod:`pandas.io.sql` module provides a collection of query wrappers to both
facilitate data retrieval and to reduce dependency on DB-specific API. Database abstraction
is provided by SQLAlchemy if installed. In addition you will need a driver library for
your database. Examples of such drivers are `psycopg2 <https://www.psycopg.org/>`__
for PostgreSQL or `pymysql <https://github.com/PyMySQL/PyMySQL>`__ for MySQL.
For `SQLite <https://docs.python.org/3/library/sqlite3.html>`__ this is
included in Python's standard library by default.
You can find an overview of supported drivers for each SQL dialect in the
`SQLAlchemy docs <https://docs.sqlalchemy.org/en/latest/dialects/index.html>`__.

If SQLAlchemy is not installed, a fallback is only provided for sqlite (and
for mysql for backwards compatibility, but this is deprecated and will be
removed in a future version).
This mode requires a Python database adapter which respect the `Python
DB-API <https://www.python.org/dev/peps/pep-0249/>`__.

See also some :ref:`cookbook examples <cookbook.sql>` for some advanced strategies.

The key functions are:

.. autosummary::

    read_sql_table
    read_sql_query
    read_sql
    DataFrame.to_sql

.. note::

    The function :func:`~pandas.read_sql` is a convenience wrapper around
    :func:`~pandas.read_sql_table` and :func:`~pandas.read_sql_query` (and for
    backward compatibility) and will delegate to specific function depending on
    the provided input (database table name or sql query).
    Table names do not need to be quoted if they have special characters.

In the following example, we use the `SQlite <https://www.sqlite.org/index.html>`__ SQL database
engine. You can use a temporary SQLite database where data are stored in
"memory".

To connect with SQLAlchemy you use the :func:`create_engine` function to create an engine
object from database URI. You only need to create the engine once per database you are
connecting to.
For more information on :func:`create_engine` and the URI formatting, see the examples
below and the SQLAlchemy `documentation <https://docs.sqlalchemy.org/en/latest/core/engines.html>`__

.. ipython:: python

   from sqlalchemy import create_engine

   # Create your engine.
   engine = create_engine("sqlite:///:memory:")

If you want to manage your own connections you can pass one of those instead. The example below opens a
connection to the database using a Python context manager that automatically closes the connection after
the block has completed.
See the `SQLAlchemy docs <https://docs.sqlalchemy.org/en/latest/core/connections.html#basic-usage>`__
for an explanation of how the database connection is handled.

.. code-block:: python

   with engine.connect() as conn, conn.begin():
       data = pd.read_sql_table("data", conn)

.. warning::

        When you open a connection to a database you are also responsible for closing it.
        Side effects of leaving a connection open may include locking the database or
        other breaking behaviour.

Writing DataFrames
''''''''''''''''''

Assuming the following data is in a ``DataFrame`` ``data``, we can insert it into
the database using :func:`~pandas.DataFrame.to_sql`.

+-----+------------+-------+-------+-------+
| id  |    Date    | Col_1 | Col_2 | Col_3 |
+=====+============+=======+=======+=======+
| 26  | 2012-10-18 |   X   |  25.7 | True  |
+-----+------------+-------+-------+-------+
| 42  | 2012-10-19 |   Y   | -12.4 | False |
+-----+------------+-------+-------+-------+
| 63  | 2012-10-20 |   Z   |  5.73 | True  |
+-----+------------+-------+-------+-------+


.. ipython:: python

   import datetime

   c = ["id", "Date", "Col_1", "Col_2", "Col_3"]
   d = [
       (26, datetime.datetime(2010, 10, 18), "X", 27.5, True),
       (42, datetime.datetime(2010, 10, 19), "Y", -12.5, False),
       (63, datetime.datetime(2010, 10, 20), "Z", 5.73, True),
   ]

   data = pd.DataFrame(d, columns=c)

   data
   data.to_sql("data", engine)

With some databases, writing large DataFrames can result in errors due to
packet size limitations being exceeded. This can be avoided by setting the
``chunksize`` parameter when calling ``to_sql``.  For example, the following
writes ``data`` to the database in batches of 1000 rows at a time:

.. ipython:: python

    data.to_sql("data_chunked", engine, chunksize=1000)

SQL data types
++++++++++++++

:func:`~pandas.DataFrame.to_sql` will try to map your data to an appropriate
SQL data type based on the dtype of the data. When you have columns of dtype
``object``, pandas will try to infer the data type.

You can always override the default type by specifying the desired SQL type of
any of the columns by using the ``dtype`` argument. This argument needs a
dictionary mapping column names to SQLAlchemy types (or strings for the sqlite3
fallback mode).
For example, specifying to use the sqlalchemy ``String`` type instead of the
default ``Text`` type for string columns:

.. ipython:: python

    from sqlalchemy.types import String

    data.to_sql("data_dtype", engine, dtype={"Col_1": String})

.. note::

    Due to the limited support for timedelta's in the different database
    flavors, columns with type ``timedelta64`` will be written as integer
    values as nanoseconds to the database and a warning will be raised.

.. note::

    Columns of ``category`` dtype will be converted to the dense representation
    as you would get with ``np.asarray(categorical)`` (e.g. for string categories
    this gives an array of strings).
    Because of this, reading the database table back in does **not** generate
    a categorical.

.. _io.sql_datetime_data:

Datetime data types
'''''''''''''''''''

Using SQLAlchemy, :func:`~pandas.DataFrame.to_sql` is capable of writing
datetime data that is timezone naive or timezone aware. However, the resulting
data stored in the database ultimately depends on the supported data type
for datetime data of the database system being used.

The following table lists supported data types for datetime data for some
common databases. Other database dialects may have different data types for
datetime data.

===========   =============================================  ===================
Database      SQL Datetime Types                             Timezone Support
===========   =============================================  ===================
SQLite        ``TEXT``                                       No
MySQL         ``TIMESTAMP`` or ``DATETIME``                  No
PostgreSQL    ``TIMESTAMP`` or ``TIMESTAMP WITH TIME ZONE``  Yes
===========   =============================================  ===================

When writing timezone aware data to databases that do not support timezones,
the data will be written as timezone naive timestamps that are in local time
with respect to the timezone.

:func:`~pandas.read_sql_table` is also capable of reading datetime data that is
timezone aware or naive. When reading ``TIMESTAMP WITH TIME ZONE`` types, pandas
will convert the data to UTC.

.. _io.sql.method:

Insertion method
++++++++++++++++

The parameter ``method`` controls the SQL insertion clause used.
Possible values are:

- ``None``: Uses standard SQL ``INSERT`` clause (one per row).
- ``'multi'``: Pass multiple values in a single ``INSERT`` clause.
  It uses a *special* SQL syntax not supported by all backends.
  This usually provides better performance for analytic databases
  like *Presto* and *Redshift*, but has worse performance for
  traditional SQL backend if the table contains many columns.
  For more information check the SQLAlchemy `documentation
  <https://docs.sqlalchemy.org/en/latest/core/dml.html#sqlalchemy.sql.expression.Insert.values.params.*args>`__.
- callable with signature ``(pd_table, conn, keys, data_iter)``:
  This can be used to implement a more performant insertion method based on
  specific backend dialect features.

Example of a callable using PostgreSQL `COPY clause
<https://www.postgresql.org/docs/current/sql-copy.html>`__::

  # Alternative to_sql() *method* for DBs that support COPY FROM
  import csv
  from io import StringIO

  def psql_insert_copy(table, conn, keys, data_iter):
      """
      Execute SQL statement inserting data

      Parameters
      ----------
      table : pandas.io.sql.SQLTable
      conn : sqlalchemy.engine.Engine or sqlalchemy.engine.Connection
      keys : list of str
          Column names
      data_iter : Iterable that iterates the values to be inserted
      """
      # gets a DBAPI connection that can provide a cursor
      dbapi_conn = conn.connection
      with dbapi_conn.cursor() as cur:
          s_buf = StringIO()
          writer = csv.writer(s_buf)
          writer.writerows(data_iter)
          s_buf.seek(0)

          columns = ', '.join(['"{}"'.format(k) for k in keys])
          if table.schema:
              table_name = '{}.{}'.format(table.schema, table.name)
          else:
              table_name = table.name

          sql = 'COPY {} ({}) FROM STDIN WITH CSV'.format(
              table_name, columns)
          cur.copy_expert(sql=sql, file=s_buf)

Reading tables
''''''''''''''

:func:`~pandas.read_sql_table` will read a database table given the
table name and optionally a subset of columns to read.

.. note::

    In order to use :func:`~pandas.read_sql_table`, you **must** have the
    SQLAlchemy optional dependency installed.

.. ipython:: python

   pd.read_sql_table("data", engine)

.. note::

  Note that pandas infers column dtypes from query outputs, and not by looking
  up data types in the physical database schema. For example, assume ``userid``
  is an integer column in a table. Then, intuitively, ``select userid ...`` will
  return integer-valued series, while ``select cast(userid as text) ...`` will
  return object-valued (str) series. Accordingly, if the query output is empty,
  then all resulting columns will be returned as object-valued (since they are
  most general). If you foresee that your query will sometimes generate an empty
  result, you may want to explicitly typecast afterwards to ensure dtype
  integrity.

You can also specify the name of the column as the ``DataFrame`` index,
and specify a subset of columns to be read.

.. ipython:: python

   pd.read_sql_table("data", engine, index_col="id")
   pd.read_sql_table("data", engine, columns=["Col_1", "Col_2"])

And you can explicitly force columns to be parsed as dates:

.. ipython:: python

   pd.read_sql_table("data", engine, parse_dates=["Date"])

If needed you can explicitly specify a format string, or a dict of arguments
to pass to :func:`pandas.to_datetime`:

.. code-block:: python

   pd.read_sql_table("data", engine, parse_dates={"Date": "%Y-%m-%d"})
   pd.read_sql_table(
       "data",
       engine,
       parse_dates={"Date": {"format": "%Y-%m-%d %H:%M:%S"}},
   )


You can check if a table exists using :func:`~pandas.io.sql.has_table`

Schema support
''''''''''''''

Reading from and writing to different schema's is supported through the ``schema``
keyword in the :func:`~pandas.read_sql_table` and :func:`~pandas.DataFrame.to_sql`
functions. Note however that this depends on the database flavor (sqlite does not
have schema's). For example:

.. code-block:: python

   df.to_sql("table", engine, schema="other_schema")
   pd.read_sql_table("table", engine, schema="other_schema")

Querying
''''''''

You can query using raw SQL in the :func:`~pandas.read_sql_query` function.
In this case you must use the SQL variant appropriate for your database.
When using SQLAlchemy, you can also pass SQLAlchemy Expression language constructs,
which are database-agnostic.

.. ipython:: python

   pd.read_sql_query("SELECT * FROM data", engine)

Of course, you can specify a more "complex" query.

.. ipython:: python

   pd.read_sql_query("SELECT id, Col_1, Col_2 FROM data WHERE id = 42;", engine)

The :func:`~pandas.read_sql_query` function supports a ``chunksize`` argument.
Specifying this will return an iterator through chunks of the query result:

.. ipython:: python

    df = pd.DataFrame(np.random.randn(20, 3), columns=list("abc"))
    df.to_sql("data_chunks", engine, index=False)

.. ipython:: python

    for chunk in pd.read_sql_query("SELECT * FROM data_chunks", engine, chunksize=5):
        print(chunk)

You can also run a plain query without creating a ``DataFrame`` with
:func:`~pandas.io.sql.execute`. This is useful for queries that don't return values,
such as INSERT. This is functionally equivalent to calling ``execute`` on the
SQLAlchemy engine or db connection object. Again, you must use the SQL syntax
variant appropriate for your database.

.. code-block:: python

   from pandas.io import sql

   sql.execute("SELECT * FROM table_name", engine)
   sql.execute(
       "INSERT INTO table_name VALUES(?, ?, ?)", engine, params=[("id", 1, 12.2, True)]
   )


Engine connection examples
''''''''''''''''''''''''''

To connect with SQLAlchemy you use the :func:`create_engine` function to create an engine
object from database URI. You only need to create the engine once per database you are
connecting to.

.. code-block:: python

   from sqlalchemy import create_engine

   engine = create_engine("postgresql://scott:tiger@localhost:5432/mydatabase")

   engine = create_engine("mysql+mysqldb://scott:tiger@localhost/foo")

   engine = create_engine("oracle://scott:tiger@127.0.0.1:1521/sidname")

   engine = create_engine("mssql+pyodbc://mydsn")

   # sqlite://<nohostname>/<path>
   # where <path> is relative:
   engine = create_engine("sqlite:///foo.db")

   # or absolute, starting with a slash:
   engine = create_engine("sqlite:////absolute/path/to/foo.db")

For more information see the examples the SQLAlchemy `documentation <https://docs.sqlalchemy.org/en/latest/core/engines.html>`__


Advanced SQLAlchemy queries
'''''''''''''''''''''''''''

You can use SQLAlchemy constructs to describe your query.

Use :func:`sqlalchemy.text` to specify query parameters in a backend-neutral way

.. ipython:: python

   import sqlalchemy as sa

   pd.read_sql(
       sa.text("SELECT * FROM data where Col_1=:col1"), engine, params={"col1": "X"}
   )

If you have an SQLAlchemy description of your database you can express where conditions using SQLAlchemy expressions

.. ipython:: python

   metadata = sa.MetaData()
   data_table = sa.Table(
       "data",
       metadata,
       sa.Column("index", sa.Integer),
       sa.Column("Date", sa.DateTime),
       sa.Column("Col_1", sa.String),
       sa.Column("Col_2", sa.Float),
       sa.Column("Col_3", sa.Boolean),
   )

   pd.read_sql(sa.select([data_table]).where(data_table.c.Col_3 is True), engine)

You can combine SQLAlchemy expressions with parameters passed to :func:`read_sql` using :func:`sqlalchemy.bindparam`

.. ipython:: python

    import datetime as dt

    expr = sa.select([data_table]).where(data_table.c.Date > sa.bindparam("date"))
    pd.read_sql(expr, engine, params={"date": dt.datetime(2010, 10, 18)})


Sqlite fallback
'''''''''''''''

The use of sqlite is supported without using SQLAlchemy.
This mode requires a Python database adapter which respect the `Python
DB-API <https://www.python.org/dev/peps/pep-0249/>`__.

You can create connections like so:

.. code-block:: python

   import sqlite3

   con = sqlite3.connect(":memory:")

And then issue the following queries:

.. code-block:: python

   data.to_sql("data", con)
   pd.read_sql_query("SELECT * FROM data", con)


.. _io.bigquery:

Google BigQuery
---------------

.. warning::

   Starting in 0.20.0, pandas has split off Google BigQuery support into the
   separate package ``pandas-gbq``. You can ``pip install pandas-gbq`` to get it.

The ``pandas-gbq`` package provides functionality to read/write from Google BigQuery.

pandas integrates with this external package. if ``pandas-gbq`` is installed, you can
use the pandas methods ``pd.read_gbq`` and ``DataFrame.to_gbq``, which will call the
respective functions from ``pandas-gbq``.

Full documentation can be found `here <https://pandas-gbq.readthedocs.io/en/latest/>`__.

.. _io.stata:

Stata format
------------

.. _io.stata_writer:

Writing to stata format
'''''''''''''''''''''''

The method :func:`~pandas.core.frame.DataFrame.to_stata` will write a DataFrame
into a .dta file. The format version of this file is always 115 (Stata 12).

.. ipython:: python

   df = pd.DataFrame(np.random.randn(10, 2), columns=list("AB"))
   df.to_stata("stata.dta")

*Stata* data files have limited data type support; only strings with
244 or fewer characters, ``int8``, ``int16``, ``int32``, ``float32``
and ``float64`` can be stored in ``.dta`` files.  Additionally,
*Stata* reserves certain values to represent missing data. Exporting a
non-missing value that is outside of the permitted range in Stata for
a particular data type will retype the variable to the next larger
size.  For example, ``int8`` values are restricted to lie between -127
and 100 in Stata, and so variables with values above 100 will trigger
a conversion to ``int16``. ``nan`` values in floating points data
types are stored as the basic missing data type (``.`` in *Stata*).

.. note::

    It is not possible to export missing data values for integer data types.


The *Stata* writer gracefully handles other data types including ``int64``,
``bool``, ``uint8``, ``uint16``, ``uint32`` by casting to
the smallest supported type that can represent the data.  For example, data
with a type of ``uint8`` will be cast to ``int8`` if all values are less than
100 (the upper bound for non-missing ``int8`` data in *Stata*), or, if values are
outside of this range, the variable is cast to ``int16``.


.. warning::

   Conversion from ``int64`` to ``float64`` may result in a loss of precision
   if ``int64`` values are larger than 2**53.

.. warning::

  :class:`~pandas.io.stata.StataWriter` and
  :func:`~pandas.core.frame.DataFrame.to_stata` only support fixed width
  strings containing up to 244 characters, a limitation imposed by the version
  115 dta file format. Attempting to write *Stata* dta files with strings
  longer than 244 characters raises a ``ValueError``.

.. _io.stata_reader:

Reading from Stata format
'''''''''''''''''''''''''

The top-level function ``read_stata`` will read a dta file and return
either a ``DataFrame`` or a :class:`~pandas.io.stata.StataReader` that can
be used to read the file incrementally.

.. ipython:: python

   pd.read_stata("stata.dta")

Specifying a ``chunksize`` yields a
:class:`~pandas.io.stata.StataReader` instance that can be used to
read ``chunksize`` lines from the file at a time.  The ``StataReader``
object can be used as an iterator.

.. ipython:: python

  with pd.read_stata("stata.dta", chunksize=3) as reader:
      for df in reader:
          print(df.shape)

For more fine-grained control, use ``iterator=True`` and specify
``chunksize`` with each call to
:func:`~pandas.io.stata.StataReader.read`.

.. ipython:: python

  with pd.read_stata("stata.dta", iterator=True) as reader:
      chunk1 = reader.read(5)
      chunk2 = reader.read(5)

Currently the ``index`` is retrieved as a column.

The parameter ``convert_categoricals`` indicates whether value labels should be
read and used to create a ``Categorical`` variable from them. Value labels can
also be retrieved by the function ``value_labels``, which requires :func:`~pandas.io.stata.StataReader.read`
to be called before use.

The parameter ``convert_missing`` indicates whether missing value
representations in Stata should be preserved.  If ``False`` (the default),
missing values are represented as ``np.nan``.  If ``True``, missing values are
represented using ``StataMissingValue`` objects, and columns containing missing
values will have ``object`` data type.

.. note::

   :func:`~pandas.read_stata` and
   :class:`~pandas.io.stata.StataReader` support .dta formats 113-115
   (Stata 10-12), 117 (Stata 13), and 118 (Stata 14).

.. note::

   Setting ``preserve_dtypes=False`` will upcast to the standard pandas data types:
   ``int64`` for all integer types and ``float64`` for floating point data.  By default,
   the Stata data types are preserved when importing.

.. ipython:: python
   :suppress:

   os.remove("stata.dta")

.. _io.stata-categorical:

Categorical data
++++++++++++++++

``Categorical`` data can be exported to *Stata* data files as value labeled data.
The exported data consists of the underlying category codes as integer data values
and the categories as value labels.  *Stata* does not have an explicit equivalent
to a ``Categorical`` and information about *whether* the variable is ordered
is lost when exporting.

.. warning::

    *Stata* only supports string value labels, and so ``str`` is called on the
    categories when exporting data.  Exporting ``Categorical`` variables with
    non-string categories produces a warning, and can result a loss of
    information if the ``str`` representations of the categories are not unique.

Labeled data can similarly be imported from *Stata* data files as ``Categorical``
variables using the keyword argument ``convert_categoricals`` (``True`` by default).
The keyword argument ``order_categoricals`` (``True`` by default) determines
whether imported ``Categorical`` variables are ordered.

.. note::

    When importing categorical data, the values of the variables in the *Stata*
    data file are not preserved since ``Categorical`` variables always
    use integer data types between ``-1`` and ``n-1`` where ``n`` is the number
    of categories. If the original values in the *Stata* data file are required,
    these can be imported by setting ``convert_categoricals=False``, which will
    import original data (but not the variable labels). The original values can
    be matched to the imported categorical data since there is a simple mapping
    between the original *Stata* data values and the category codes of imported
    Categorical variables: missing values are assigned code ``-1``, and the
    smallest original value is assigned ``0``, the second smallest is assigned
    ``1`` and so on until the largest original value is assigned the code ``n-1``.

.. note::

    *Stata* supports partially labeled series. These series have value labels for
    some but not all data values. Importing a partially labeled series will produce
    a ``Categorical`` with string categories for the values that are labeled and
    numeric categories for values with no label.

.. _io.sas:

.. _io.sas_reader:

SAS formats
-----------

The top-level function :func:`read_sas` can read (but not write) SAS
XPORT (.xpt) and (since *v0.18.0*) SAS7BDAT (.sas7bdat) format files.

SAS files only contain two value types: ASCII text and floating point
values (usually 8 bytes but sometimes truncated).  For xport files,
there is no automatic type conversion to integers, dates, or
categoricals.  For SAS7BDAT files, the format codes may allow date
variables to be automatically converted to dates.  By default the
whole file is read and returned as a ``DataFrame``.

Specify a ``chunksize`` or use ``iterator=True`` to obtain reader
objects (``XportReader`` or ``SAS7BDATReader``) for incrementally
reading the file.  The reader objects also have attributes that
contain additional information about the file and its variables.

Read a SAS7BDAT file:

.. code-block:: python

    df = pd.read_sas("sas_data.sas7bdat")

Obtain an iterator and read an XPORT file 100,000 lines at a time:

.. code-block:: python

    def do_something(chunk):
        pass


    with pd.read_sas("sas_xport.xpt", chunk=100000) as rdr:
        for chunk in rdr:
            do_something(chunk)

The specification_ for the xport file format is available from the SAS
web site.

.. _specification: https://support.sas.com/content/dam/SAS/support/en/technical-papers/record-layout-of-a-sas-version-5-or-6-data-set-in-sas-transport-xport-format.pdf

No official documentation is available for the SAS7BDAT format.

.. _io.spss:

.. _io.spss_reader:

SPSS formats
------------

.. versionadded:: 0.25.0

The top-level function :func:`read_spss` can read (but not write) SPSS
SAV (.sav) and  ZSAV (.zsav) format files.

SPSS files contain column names. By default the
whole file is read, categorical columns are converted into ``pd.Categorical``,
and a ``DataFrame`` with all columns is returned.

Specify the ``usecols`` parameter to obtain a subset of columns. Specify ``convert_categoricals=False``
to avoid converting categorical columns into ``pd.Categorical``.

Read an SPSS file:

.. code-block:: python

    df = pd.read_spss("spss_data.sav")

Extract a subset of columns contained in ``usecols`` from an SPSS file and
avoid converting categorical columns into ``pd.Categorical``:

.. code-block:: python

    df = pd.read_spss(
        "spss_data.sav",
        usecols=["foo", "bar"],
        convert_categoricals=False,
    )

More information about the SAV and ZSAV file formats is available here_.

.. _here: https://www.ibm.com/docs/en/spss-statistics/22.0.0

.. _io.other:

Other file formats
------------------

pandas itself only supports IO with a limited set of file formats that map
cleanly to its tabular data model. For reading and writing other file formats
into and from pandas, we recommend these packages from the broader community.

netCDF
''''''

xarray_ provides data structures inspired by the pandas ``DataFrame`` for working
with multi-dimensional datasets, with a focus on the netCDF file format and
easy conversion to and from pandas.

.. _xarray: https://xarray.pydata.org/en/stable/

.. _io.perf:

Performance considerations
--------------------------

This is an informal comparison of various IO methods, using pandas
0.24.2. Timings are machine dependent and small differences should be
ignored.

.. code-block:: ipython

   In [1]: sz = 1000000
   In [2]: df = pd.DataFrame({'A': np.random.randn(sz), 'B': [1] * sz})

   In [3]: df.info()
   <class 'pandas.core.frame.DataFrame'>
   RangeIndex: 1000000 entries, 0 to 999999
   Data columns (total 2 columns):
   A    1000000 non-null float64
   B    1000000 non-null int64
   dtypes: float64(1), int64(1)
   memory usage: 15.3 MB

The following test functions will be used below to compare the performance of several IO methods:

.. code-block:: python



   import numpy as np

   import os

   sz = 1000000
   df = pd.DataFrame({"A": np.random.randn(sz), "B": [1] * sz})

   sz = 1000000
   np.random.seed(42)
   df = pd.DataFrame({"A": np.random.randn(sz), "B": [1] * sz})


   def test_sql_write(df):
       if os.path.exists("test.sql"):
           os.remove("test.sql")
       sql_db = sqlite3.connect("test.sql")
       df.to_sql(name="test_table", con=sql_db)
       sql_db.close()


   def test_sql_read():
       sql_db = sqlite3.connect("test.sql")
       pd.read_sql_query("select * from test_table", sql_db)
       sql_db.close()


   def test_hdf_fixed_write(df):
       df.to_hdf("test_fixed.hdf", "test", mode="w")


   def test_hdf_fixed_read():
       pd.read_hdf("test_fixed.hdf", "test")


   def test_hdf_fixed_write_compress(df):
       df.to_hdf("test_fixed_compress.hdf", "test", mode="w", complib="blosc")


   def test_hdf_fixed_read_compress():
       pd.read_hdf("test_fixed_compress.hdf", "test")


   def test_hdf_table_write(df):
       df.to_hdf("test_table.hdf", "test", mode="w", format="table")


   def test_hdf_table_read():
       pd.read_hdf("test_table.hdf", "test")


   def test_hdf_table_write_compress(df):
       df.to_hdf(
           "test_table_compress.hdf", "test", mode="w", complib="blosc", format="table"
       )


   def test_hdf_table_read_compress():
       pd.read_hdf("test_table_compress.hdf", "test")


   def test_csv_write(df):
       df.to_csv("test.csv", mode="w")


   def test_csv_read():
       pd.read_csv("test.csv", index_col=0)


   def test_feather_write(df):
       df.to_feather("test.feather")


   def test_feather_read():
       pd.read_feather("test.feather")


   def test_pickle_write(df):
       df.to_pickle("test.pkl")


   def test_pickle_read():
       pd.read_pickle("test.pkl")


   def test_pickle_write_compress(df):
       df.to_pickle("test.pkl.compress", compression="xz")


   def test_pickle_read_compress():
       pd.read_pickle("test.pkl.compress", compression="xz")


   def test_parquet_write(df):
       df.to_parquet("test.parquet")


   def test_parquet_read():
       pd.read_parquet("test.parquet")

When writing, the top three functions in terms of speed are ``test_feather_write``, ``test_hdf_fixed_write`` and ``test_hdf_fixed_write_compress``.

.. code-block:: ipython

   In [4]: %timeit test_sql_write(df)
   3.29 s ± 43.2 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

   In [5]: %timeit test_hdf_fixed_write(df)
   19.4 ms ± 560 µs per loop (mean ± std. dev. of 7 runs, 1 loop each)

   In [6]: %timeit test_hdf_fixed_write_compress(df)
   19.6 ms ± 308 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)

   In [7]: %timeit test_hdf_table_write(df)
   449 ms ± 5.61 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

   In [8]: %timeit test_hdf_table_write_compress(df)
   448 ms ± 11.9 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

   In [9]: %timeit test_csv_write(df)
   3.66 s ± 26.2 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

   In [10]: %timeit test_feather_write(df)
   9.75 ms ± 117 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

   In [11]: %timeit test_pickle_write(df)
   30.1 ms ± 229 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)

   In [12]: %timeit test_pickle_write_compress(df)
   4.29 s ± 15.9 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

   In [13]: %timeit test_parquet_write(df)
   67.6 ms ± 706 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)

When reading, the top three functions in terms of speed are ``test_feather_read``, ``test_pickle_read`` and
``test_hdf_fixed_read``.


.. code-block:: ipython

   In [14]: %timeit test_sql_read()
   1.77 s ± 17.7 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

   In [15]: %timeit test_hdf_fixed_read()
   19.4 ms ± 436 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)

   In [16]: %timeit test_hdf_fixed_read_compress()
   19.5 ms ± 222 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)

   In [17]: %timeit test_hdf_table_read()
   38.6 ms ± 857 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)

   In [18]: %timeit test_hdf_table_read_compress()
   38.8 ms ± 1.49 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)

   In [19]: %timeit test_csv_read()
   452 ms ± 9.04 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

   In [20]: %timeit test_feather_read()
   12.4 ms ± 99.7 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

   In [21]: %timeit test_pickle_read()
   18.4 ms ± 191 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

   In [22]: %timeit test_pickle_read_compress()
   915 ms ± 7.48 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

   In [23]: %timeit test_parquet_read()
   24.4 ms ± 146 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)


The files ``test.pkl.compress``, ``test.parquet`` and ``test.feather`` took the least space on disk (in bytes).

.. code-block:: none

    29519500 Oct 10 06:45 test.csv
    16000248 Oct 10 06:45 test.feather
    8281983  Oct 10 06:49 test.parquet
    16000857 Oct 10 06:47 test.pkl
    7552144  Oct 10 06:48 test.pkl.compress
    34816000 Oct 10 06:42 test.sql
    24009288 Oct 10 06:43 test_fixed.hdf
    24009288 Oct 10 06:43 test_fixed_compress.hdf
    24458940 Oct 10 06:44 test_table.hdf
    24458940 Oct 10 06:44 test_table_compress.hdf