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
|