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
|
From: Antonio Valentino <antonio.valentino@tiscali.it>
Date: Thu, 10 Apr 2025 05:42:56 +0000
Subject: Fix vlen enc
Forwarded: https://github.com/zarr-developers/numcodecs/pull/736
---
numcodecs/compat_ext.pxd | 2 +
numcodecs/fletcher32.pyx | 4 +-
numcodecs/tests/test_vlen_bytes.py | 4 --
numcodecs/vlen.pyx | 140 ++++++++++++++++---------------------
4 files changed, 66 insertions(+), 84 deletions(-)
diff --git a/numcodecs/compat_ext.pxd b/numcodecs/compat_ext.pxd
index 436c23f..129c1d8 100644
--- a/numcodecs/compat_ext.pxd
+++ b/numcodecs/compat_ext.pxd
@@ -3,7 +3,9 @@
cdef extern from *:
"""
+ #ifndef PyBytes_RESIZE
#define PyBytes_RESIZE(b, n) _PyBytes_Resize(&b, n)
+ #endif
"""
int PyBytes_RESIZE(object b, Py_ssize_t n) except -1
diff --git a/numcodecs/fletcher32.pyx b/numcodecs/fletcher32.pyx
index 387295b..56a5f79 100644
--- a/numcodecs/fletcher32.pyx
+++ b/numcodecs/fletcher32.pyx
@@ -76,7 +76,7 @@ class Fletcher32(Codec):
"""Return buffer plus a footer with the fletcher checksum (4-bytes)"""
buf = ensure_contiguous_ndarray(buf).ravel().view('uint8')
cdef const uint8_t[::1] b_mv = buf
- cdef uint8_t* b_ptr = &b_mv[0]
+ cdef const uint8_t* b_ptr = &b_mv[0]
cdef Py_ssize_t b_len = len(b_mv)
cdef Py_ssize_t out_len = b_len + FOOTER_LENGTH
@@ -92,7 +92,7 @@ class Fletcher32(Codec):
"""Check fletcher checksum, and return buffer without it"""
b = ensure_contiguous_ndarray(buf).view('uint8')
cdef const uint8_t[::1] b_mv = b
- cdef uint8_t* b_ptr = &b_mv[0]
+ cdef const uint8_t* b_ptr = &b_mv[0]
cdef Py_ssize_t b_len = len(b_mv)
val = _fletcher32(b_mv[:-FOOTER_LENGTH])
diff --git a/numcodecs/tests/test_vlen_bytes.py b/numcodecs/tests/test_vlen_bytes.py
index 467c9a8..3546dba 100644
--- a/numcodecs/tests/test_vlen_bytes.py
+++ b/numcodecs/tests/test_vlen_bytes.py
@@ -1,4 +1,3 @@
-import sys
import unittest
import numpy as np
@@ -85,9 +84,6 @@ def test_decode_errors():
codec.decode(enc, out=np.zeros(10, dtype='i4'))
-# TODO: fix this test on GitHub actions somehow...
-# See https://github.com/zarr-developers/numcodecs/issues/683
-@pytest.mark.skipif(sys.platform == "darwin", reason="Test is failing on macOS on GitHub actions.")
def test_encode_none():
a = np.array([b'foo', None, b'bar'], dtype=object)
codec = VLenBytes()
diff --git a/numcodecs/vlen.pyx b/numcodecs/vlen.pyx
index f564a66..6c04260 100644
--- a/numcodecs/vlen.pyx
+++ b/numcodecs/vlen.pyx
@@ -7,26 +7,13 @@
cimport cython
-from libc.stdint cimport uint8_t
+from libc.stdint cimport uint8_t, uint32_t
from libc.string cimport memcpy
-from cpython.buffer cimport PyBuffer_IsContiguous
-from cpython.bytearray cimport (
- PyByteArray_AS_STRING,
- PyByteArray_FromStringAndSize,
-)
-from cpython.bytes cimport (
- PyBytes_AS_STRING,
- PyBytes_GET_SIZE,
- PyBytes_Check,
- PyBytes_FromStringAndSize,
-)
+from cpython.bytearray cimport PyByteArray_FromStringAndSize
+from cpython.bytes cimport PyBytes_FromStringAndSize
from cpython.memoryview cimport PyMemoryView_GET_BUFFER
-from cpython.unicode cimport (
- PyUnicode_AsUTF8String,
- PyUnicode_Check,
- PyUnicode_FromStringAndSize,
-)
+from cpython.unicode cimport PyUnicode_FromStringAndSize
from numpy cimport ndarray
@@ -39,8 +26,12 @@ from .abc import Codec
from .compat import ensure_contiguous_ndarray
-# 4 bytes to store number of items
-cdef Py_ssize_t HEADER_LENGTH = 4
+# Define header size used to store number of items that follow.
+cdef extern from *:
+ """
+ const Py_ssize_t HEADER_LENGTH = sizeof(uint32_t);
+ """
+ const Py_ssize_t HEADER_LENGTH
def check_out_param(out, n_items):
@@ -89,11 +80,11 @@ class VLenUTF8(Codec):
ndarray[object, ndim=1] input_values
object[:] encoded_values
int[:] encoded_lengths
- char* encv
bytes b
bytearray out
char* data
- object u
+ object o
+ unicode u
# normalise input
input_values = np.asarray(buf, dtype=object).reshape(-1, order='A')
@@ -108,15 +99,13 @@ class VLenUTF8(Codec):
# first iteration to convert to bytes
data_length = 0
for i in range(n_items):
- u = input_values[i]
- if u is None or u == 0: # treat these as missing value, normalize
- u = ''
- elif not PyUnicode_Check(u):
- raise TypeError('expected unicode string, found %r' % u)
- b = PyUnicode_AsUTF8String(u)
- l = PyBytes_GET_SIZE(b)
+ o = input_values[i]
+ # replace missing value and coerce to typed data
+ u = "" if o is None or o == 0 else o
+ b = u.encode("utf-8")
+ l = len(b)
encoded_values[i] = b
- data_length += l + 4 # 4 bytes to store item length
+ data_length += l + HEADER_LENGTH
encoded_lengths[i] = l
# setup output
@@ -124,7 +113,7 @@ class VLenUTF8(Codec):
out = PyByteArray_FromStringAndSize(NULL, total_length)
# write header
- data = PyByteArray_AS_STRING(out)
+ data = out
store_le32(<uint8_t*>data, n_items)
# second iteration, store data
@@ -132,9 +121,9 @@ class VLenUTF8(Codec):
for i in range(n_items):
l = encoded_lengths[i]
store_le32(<uint8_t*>data, l)
- data += 4
- encv = PyBytes_AS_STRING(encoded_values[i])
- memcpy(data, encv, l)
+ data += HEADER_LENGTH
+ b = encoded_values[i]
+ memcpy(data, <const char*>b, l)
data += l
return out
@@ -151,12 +140,10 @@ class VLenUTF8(Codec):
# obtain memoryview
buf = ensure_contiguous_ndarray(buf)
- buf_mv = memoryview(buf)
+ buf_mv = ensure_continguous_memoryview(buf)
buf_pb = PyMemoryView_GET_BUFFER(buf_mv)
# sanity checks
- if not PyBuffer_IsContiguous(buf_pb, b'A'):
- raise BufferError("`buf` must contain contiguous memory")
if buf_pb.len < HEADER_LENGTH:
raise ValueError('corrupt buffer, missing or truncated header')
@@ -178,10 +165,10 @@ class VLenUTF8(Codec):
# https://github.com/cython/cython/issues/1608
data += HEADER_LENGTH
for i in range(n_items):
- if data + 4 > data_end:
+ if data + HEADER_LENGTH > data_end:
raise ValueError('corrupt buffer, data seem truncated')
l = load_le32(<uint8_t*>data)
- data += 4
+ data += HEADER_LENGTH
if data + l > data_end:
raise ValueError('corrupt buffer, data seem truncated')
out[i] = PyUnicode_FromStringAndSize(data, l)
@@ -223,9 +210,10 @@ class VLenBytes(Codec):
cdef:
Py_ssize_t i, l, n_items, data_length, total_length
object[:] values
+ object[:] normed_values
int[:] lengths
- char* encv
- object b
+ object o
+ bytes b
bytearray out
char* data
@@ -236,18 +224,18 @@ class VLenBytes(Codec):
n_items = values.shape[0]
# setup intermediates
+ normed_values = np.empty(n_items, dtype=object)
lengths = np.empty(n_items, dtype=np.intc)
# first iteration to find lengths
data_length = 0
for i in range(n_items):
- b = values[i]
- if b is None or b == 0: # treat these as missing value, normalize
- b = b''
- elif not PyBytes_Check(b):
- raise TypeError('expected byte string, found %r' % b)
- l = PyBytes_GET_SIZE(b)
- data_length += l + 4 # 4 bytes to store item length
+ o = values[i]
+ # replace missing value and coerce to typed data
+ b = b"" if o is None or o == 0 else o
+ normed_values[i] = b
+ l = len(b)
+ data_length += l + HEADER_LENGTH
lengths[i] = l
# setup output
@@ -255,7 +243,7 @@ class VLenBytes(Codec):
out = PyByteArray_FromStringAndSize(NULL, total_length)
# write header
- data = PyByteArray_AS_STRING(out)
+ data = out
store_le32(<uint8_t*>data, n_items)
# second iteration, store data
@@ -263,9 +251,9 @@ class VLenBytes(Codec):
for i in range(n_items):
l = lengths[i]
store_le32(<uint8_t*>data, l)
- data += 4
- encv = PyBytes_AS_STRING(values[i])
- memcpy(data, encv, l)
+ data += HEADER_LENGTH
+ b = normed_values[i]
+ memcpy(data, <const char*>b, l)
data += l
return out
@@ -282,12 +270,10 @@ class VLenBytes(Codec):
# obtain memoryview
buf = ensure_contiguous_ndarray(buf)
- buf_mv = memoryview(buf)
+ buf_mv = ensure_continguous_memoryview(buf)
buf_pb = PyMemoryView_GET_BUFFER(buf_mv)
# sanity checks
- if not PyBuffer_IsContiguous(buf_pb, b'A'):
- raise BufferError("`buf` must contain contiguous memory")
if buf_pb.len < HEADER_LENGTH:
raise ValueError('corrupt buffer, missing or truncated header')
@@ -309,10 +295,10 @@ class VLenBytes(Codec):
# https://github.com/cython/cython/issues/1608
data += HEADER_LENGTH
for i in range(n_items):
- if data + 4 > data_end:
+ if data + HEADER_LENGTH > data_end:
raise ValueError('corrupt buffer, data seem truncated')
l = load_le32(<uint8_t*>data)
- data += 4
+ data += HEADER_LENGTH
if data + l > data_end:
raise ValueError('corrupt buffer, data seem truncated')
out[i] = PyBytes_FromStringAndSize(data, l)
@@ -369,13 +355,12 @@ class VLenArray(Codec):
object[:] values
object[:] normed_values
int[:] lengths
- const char* encv
bytes b
bytearray out
char* data
memoryview value_mv
const Py_buffer* value_pb
- object v
+ object o
# normalise input
values = np.asarray(buf, dtype=object).reshape(-1, order='A')
@@ -390,16 +375,18 @@ class VLenArray(Codec):
# first iteration to convert to bytes
data_length = 0
for i in range(n_items):
- v = values[i]
- if v is None:
- v = np.array([], dtype=self.dtype)
- else:
- v = np.ascontiguousarray(v, self.dtype)
- if v.ndim != 1:
- raise ValueError('only 1-dimensional arrays are supported')
- l = v.nbytes
- normed_values[i] = v
- data_length += l + 4 # 4 bytes to store item length
+ o = values[i]
+ # replace missing value and coerce to typed data
+ value_mv = ensure_continguous_memoryview(
+ np.array([], dtype=self.dtype) if o is None
+ else np.ascontiguousarray(o, self.dtype)
+ )
+ value_pb = PyMemoryView_GET_BUFFER(value_mv)
+ if value_pb.ndim != 1:
+ raise ValueError("only 1-dimensional arrays are supported")
+ l = value_pb.len
+ normed_values[i] = value_mv
+ data_length += l + HEADER_LENGTH
lengths[i] = l
# setup output
@@ -407,7 +394,7 @@ class VLenArray(Codec):
out = PyByteArray_FromStringAndSize(NULL, total_length)
# write header
- data = PyByteArray_AS_STRING(out)
+ data = out
store_le32(<uint8_t*>data, n_items)
# second iteration, store data
@@ -415,13 +402,12 @@ class VLenArray(Codec):
for i in range(n_items):
l = lengths[i]
store_le32(<uint8_t*>data, l)
- data += 4
+ data += HEADER_LENGTH
- value_mv = ensure_continguous_memoryview(normed_values[i])
+ value_mv = normed_values[i]
value_pb = PyMemoryView_GET_BUFFER(value_mv)
- encv = <const char*>value_pb.buf
- memcpy(data, encv, l)
+ memcpy(data, value_pb.buf, l)
data += l
return out
@@ -441,12 +427,10 @@ class VLenArray(Codec):
# obtain memoryview
buf = ensure_contiguous_ndarray(buf)
- buf_mv = memoryview(buf)
+ buf_mv = ensure_continguous_memoryview(buf)
buf_pb = PyMemoryView_GET_BUFFER(buf_mv)
# sanity checks
- if not PyBuffer_IsContiguous(buf_pb, b'A'):
- raise BufferError("`buf` must contain contiguous memory")
if buf_pb.len < HEADER_LENGTH:
raise ValueError('corrupt buffer, missing or truncated header')
@@ -468,10 +452,10 @@ class VLenArray(Codec):
# https://github.com/cython/cython/issues/1608
data += HEADER_LENGTH
for i in range(n_items):
- if data + 4 > data_end:
+ if data + HEADER_LENGTH > data_end:
raise ValueError('corrupt buffer, data seem truncated')
l = load_le32(<uint8_t*>data)
- data += 4
+ data += HEADER_LENGTH
if data + l > data_end:
raise ValueError('corrupt buffer, data seem truncated')
|
