Mercurial > cpython-lru_cache-weakref

annotate Objects/stringlib/codecs.h @ 82004:b414b2dfd3d3 3.3

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
merge #17091: update docstring for _thread.Lock.acquire.
author R David Murray <rdmurray@bitdance.com>
date Mon, 04 Feb 2013 10:22:39 -0500
parents 96f4cee8ea5e
children 6f44f33460cd
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
1 /* stringlib: codec implementations */
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
2
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
3 #if STRINGLIB_IS_UNICODE
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
4
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
5 /* Mask to quickly check whether a C 'long' contains a
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
6 non-ASCII, UTF8-encoded char. */
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
7 #if (SIZEOF_LONG == 8)
77966
ab9e824116eb Use correct types for ASCII_CHAR_MASK integer constants.
Mark Dickinson <mdickinson@enthought.com>
parents: 77649
diff changeset
8 # define ASCII_CHAR_MASK 0x8080808080808080UL
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
9 #elif (SIZEOF_LONG == 4)
77966
ab9e824116eb Use correct types for ASCII_CHAR_MASK integer constants.
Mark Dickinson <mdickinson@enthought.com>
parents: 77649
diff changeset
10 # define ASCII_CHAR_MASK 0x80808080UL
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
11 #else
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
12 # error C 'long' size should be either 4 or 8!
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
13 #endif
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
14
77649
3214c9ebcf5e Issue #14923: Optimize continuation-byte check in UTF-8 decoding. Patch by Serhiy Storchaka.
Mark Dickinson <mdickinson@enthought.com>
parents: 77448
diff changeset
15 /* 10xxxxxx */
3214c9ebcf5e Issue #14923: Optimize continuation-byte check in UTF-8 decoding. Patch by Serhiy Storchaka.
Mark Dickinson <mdickinson@enthought.com>
parents: 77448
diff changeset
16 #define IS_CONTINUATION_BYTE(ch) ((ch) >= 0x80 && (ch) < 0xC0)
3214c9ebcf5e Issue #14923: Optimize continuation-byte check in UTF-8 decoding. Patch by Serhiy Storchaka.
Mark Dickinson <mdickinson@enthought.com>
parents: 77448
diff changeset
17
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
18 Py_LOCAL_INLINE(Py_UCS4)
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
19 STRINGLIB(utf8_decode)(const char **inptr, const char *end,
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
20 STRINGLIB_CHAR *dest,
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
21 Py_ssize_t *outpos)
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
22 {
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
23 Py_UCS4 ch;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
24 const char *s = *inptr;
79065
99112b851b25 Issue #15144: Fix possible integer overflow when handling pointers as integer values, by using Py_uintptr_t instead of size_t.
Antoine Pitrou <solipsis@pitrou.net>
parents: 77966
diff changeset
25 const char *aligned_end = (const char *) _Py_ALIGN_DOWN(end, SIZEOF_LONG);
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
26 STRINGLIB_CHAR *p = dest + *outpos;
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
27
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
28 while (s < end) {
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
29 ch = (unsigned char)*s;
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
30
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
31 if (ch < 0x80) {
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
32 /* Fast path for runs of ASCII characters. Given that common UTF-8
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
33 input will consist of an overwhelming majority of ASCII
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
34 characters, we try to optimize for this case by checking
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
35 as many characters as a C 'long' can contain.
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
36 First, check if we can do an aligned read, as most CPUs have
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
37 a penalty for unaligned reads.
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
38 */
79065
99112b851b25 Issue #15144: Fix possible integer overflow when handling pointers as integer values, by using Py_uintptr_t instead of size_t.
Antoine Pitrou <solipsis@pitrou.net>
parents: 77966
diff changeset
39 if (_Py_IS_ALIGNED(s, SIZEOF_LONG)) {
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
40 /* Help register allocation */
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
41 register const char *_s = s;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
42 register STRINGLIB_CHAR *_p = p;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
43 while (_s < aligned_end) {
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
44 /* Read a whole long at a time (either 4 or 8 bytes),
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
45 and do a fast unrolled copy if it only contains ASCII
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
46 characters. */
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
47 unsigned long value = *(unsigned long *) _s;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
48 if (value & ASCII_CHAR_MASK)
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
49 break;
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
50 #ifdef BYTEORDER_IS_LITTLE_ENDIAN
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
51 _p[0] = (STRINGLIB_CHAR)(value & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
52 _p[1] = (STRINGLIB_CHAR)((value >> 8) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
53 _p[2] = (STRINGLIB_CHAR)((value >> 16) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
54 _p[3] = (STRINGLIB_CHAR)((value >> 24) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
55 # if SIZEOF_LONG == 8
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
56 _p[4] = (STRINGLIB_CHAR)((value >> 32) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
57 _p[5] = (STRINGLIB_CHAR)((value >> 40) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
58 _p[6] = (STRINGLIB_CHAR)((value >> 48) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
59 _p[7] = (STRINGLIB_CHAR)((value >> 56) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
60 # endif
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
61 #else
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
62 # if SIZEOF_LONG == 8
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
63 _p[0] = (STRINGLIB_CHAR)((value >> 56) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
64 _p[1] = (STRINGLIB_CHAR)((value >> 48) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
65 _p[2] = (STRINGLIB_CHAR)((value >> 40) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
66 _p[3] = (STRINGLIB_CHAR)((value >> 32) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
67 _p[4] = (STRINGLIB_CHAR)((value >> 24) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
68 _p[5] = (STRINGLIB_CHAR)((value >> 16) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
69 _p[6] = (STRINGLIB_CHAR)((value >> 8) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
70 _p[7] = (STRINGLIB_CHAR)(value & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
71 # else
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
72 _p[0] = (STRINGLIB_CHAR)((value >> 24) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
73 _p[1] = (STRINGLIB_CHAR)((value >> 16) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
74 _p[2] = (STRINGLIB_CHAR)((value >> 8) & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
75 _p[3] = (STRINGLIB_CHAR)(value & 0xFFu);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
76 # endif
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
77 #endif
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
78 _s += SIZEOF_LONG;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
79 _p += SIZEOF_LONG;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
80 }
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
81 s = _s;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
82 p = _p;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
83 if (s == end)
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
84 break;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
85 ch = (unsigned char)*s;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
86 }
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
87 if (ch < 0x80) {
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
88 s++;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
89 *p++ = ch;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
90 continue;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
91 }
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
92 }
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
93
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
94 if (ch < 0xE0) {
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
95 /* \xC2\x80-\xDF\xBF -- 0080-07FF */
80253
96f4cee8ea5e Issue #8271: Fix compilation on Windows
Victor Stinner <victor.stinner@gmail.com>
parents: 80246
diff changeset
96 Py_UCS4 ch2;
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
97 if (ch < 0xC2) {
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
98 /* invalid sequence
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
99 \x80-\xBF -- continuation byte
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
100 \xC0-\xC1 -- fake 0000-007F */
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
101 goto InvalidStart;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
102 }
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
103 if (end - s < 2) {
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
104 /* unexpected end of data: the caller will decide whether
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
105 it's an error or not */
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
106 break;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
107 }
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
108 ch2 = (unsigned char)s[1];
77649
3214c9ebcf5e Issue #14923: Optimize continuation-byte check in UTF-8 decoding. Patch by Serhiy Storchaka.
Mark Dickinson <mdickinson@enthought.com>
parents: 77448
diff changeset
109 if (!IS_CONTINUATION_BYTE(ch2))
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
110 /* invalid continuation byte */
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
111 goto InvalidContinuation1;
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
112 ch = (ch << 6) + ch2 -
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
113 ((0xC0 << 6) + 0x80);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
114 assert ((ch > 0x007F) && (ch <= 0x07FF));
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
115 s += 2;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
116 if (STRINGLIB_MAX_CHAR <= 0x007F ||
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
117 (STRINGLIB_MAX_CHAR < 0x07FF && ch > STRINGLIB_MAX_CHAR))
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
118 /* Out-of-range */
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
119 goto Return;
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
120 *p++ = ch;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
121 continue;
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
122 }
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
123
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
124 if (ch < 0xF0) {
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
125 /* \xE0\xA0\x80-\xEF\xBF\xBF -- 0800-FFFF */
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
126 Py_UCS4 ch2, ch3;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
127 if (end - s < 3) {
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
128 /* unexpected end of data: the caller will decide whether
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
129 it's an error or not */
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
130 if (end - s < 2)
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
131 break;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
132 ch2 = (unsigned char)s[1];
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
133 if (!IS_CONTINUATION_BYTE(ch2) ||
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
134 (ch2 < 0xA0 ? ch == 0xE0 : ch == 0xED))
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
135 /* for clarification see comments below */
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
136 goto InvalidContinuation1;
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
137 break;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
138 }
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
139 ch2 = (unsigned char)s[1];
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
140 ch3 = (unsigned char)s[2];
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
141 if (!IS_CONTINUATION_BYTE(ch2)) {
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
142 /* invalid continuation byte */
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
143 goto InvalidContinuation1;
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
144 }
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
145 if (ch == 0xE0) {
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
146 if (ch2 < 0xA0)
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
147 /* invalid sequence
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
148 \xE0\x80\x80-\xE0\x9F\xBF -- fake 0000-0800 */
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
149 goto InvalidContinuation1;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
150 } else if (ch == 0xED && ch2 >= 0xA0) {
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
151 /* Decoding UTF-8 sequences in range \xED\xA0\x80-\xED\xBF\xBF
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
152 will result in surrogates in range D800-DFFF. Surrogates are
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
153 not valid UTF-8 so they are rejected.
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
154 See http://www.unicode.org/versions/Unicode5.2.0/ch03.pdf
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
155 (table 3-7) and http://www.rfc-editor.org/rfc/rfc3629.txt */
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
156 goto InvalidContinuation1;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
157 }
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
158 if (!IS_CONTINUATION_BYTE(ch3)) {
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
159 /* invalid continuation byte */
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
160 goto InvalidContinuation2;
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
161 }
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
162 ch = (ch << 12) + (ch2 << 6) + ch3 -
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
163 ((0xE0 << 12) + (0x80 << 6) + 0x80);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
164 assert ((ch > 0x07FF) && (ch <= 0xFFFF));
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
165 s += 3;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
166 if (STRINGLIB_MAX_CHAR <= 0x07FF ||
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
167 (STRINGLIB_MAX_CHAR < 0xFFFF && ch > STRINGLIB_MAX_CHAR))
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
168 /* Out-of-range */
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
169 goto Return;
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
170 *p++ = ch;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
171 continue;
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
172 }
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
173
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
174 if (ch < 0xF5) {
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
175 /* \xF0\x90\x80\x80-\xF4\x8F\xBF\xBF -- 10000-10FFFF */
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
176 Py_UCS4 ch2, ch3, ch4;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
177 if (end - s < 4) {
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
178 /* unexpected end of data: the caller will decide whether
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
179 it's an error or not */
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
180 if (end - s < 2)
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
181 break;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
182 ch2 = (unsigned char)s[1];
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
183 if (!IS_CONTINUATION_BYTE(ch2) ||
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
184 (ch2 < 0x90 ? ch == 0xF0 : ch == 0xF4))
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
185 /* for clarification see comments below */
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
186 goto InvalidContinuation1;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
187 if (end - s < 3)
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
188 break;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
189 ch3 = (unsigned char)s[2];
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
190 if (!IS_CONTINUATION_BYTE(ch3))
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
191 goto InvalidContinuation2;
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
192 break;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
193 }
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
194 ch2 = (unsigned char)s[1];
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
195 ch3 = (unsigned char)s[2];
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
196 ch4 = (unsigned char)s[3];
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
197 if (!IS_CONTINUATION_BYTE(ch2)) {
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
198 /* invalid continuation byte */
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
199 goto InvalidContinuation1;
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
200 }
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
201 if (ch == 0xF0) {
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
202 if (ch2 < 0x90)
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
203 /* invalid sequence
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
204 \xF0\x80\x80\x80-\xF0\x8F\xBF\xBF -- fake 0000-FFFF */
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
205 goto InvalidContinuation1;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
206 } else if (ch == 0xF4 && ch2 >= 0x90) {
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
207 /* invalid sequence
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
208 \xF4\x90\x80\80- -- 110000- overflow */
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
209 goto InvalidContinuation1;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
210 }
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
211 if (!IS_CONTINUATION_BYTE(ch3)) {
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
212 /* invalid continuation byte */
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
213 goto InvalidContinuation2;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
214 }
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
215 if (!IS_CONTINUATION_BYTE(ch4)) {
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
216 /* invalid continuation byte */
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
217 goto InvalidContinuation3;
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
218 }
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
219 ch = (ch << 18) + (ch2 << 12) + (ch3 << 6) + ch4 -
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
220 ((0xF0 << 18) + (0x80 << 12) + (0x80 << 6) + 0x80);
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
221 assert ((ch > 0xFFFF) && (ch <= 0x10FFFF));
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
222 s += 4;
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
223 if (STRINGLIB_MAX_CHAR <= 0xFFFF ||
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
224 (STRINGLIB_MAX_CHAR < 0x10FFFF && ch > STRINGLIB_MAX_CHAR))
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
225 /* Out-of-range */
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
226 goto Return;
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
227 *p++ = ch;
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
228 continue;
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
229 }
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
230 goto InvalidStart;
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
231 }
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
232 ch = 0;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
233 Return:
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
234 *inptr = s;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
235 *outpos = p - dest;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
236 return ch;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
237 InvalidStart:
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
238 ch = 1;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
239 goto Return;
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
240 InvalidContinuation1:
76862
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
241 ch = 2;
e08c3791f035 Issue #14738: Speed-up UTF-8 decoding on non-ASCII data. Patch by Serhiy Storchaka.
Antoine Pitrou <solipsis@pitrou.net>
parents: 74032
diff changeset
242 goto Return;
80246
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
243 InvalidContinuation2:
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
244 ch = 3;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
245 goto Return;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
246 InvalidContinuation3:
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
247 ch = 4;
5962f192a483 #8271: the utf-8 decoder now outputs the correct number of U+FFFD characters when used with the "replace" error handler on invalid utf-8 sequences. Patch by Serhiy Storchaka, tests by Ezio Melotti.
Ezio Melotti <ezio.melotti@gmail.com>
parents: 79065
diff changeset
248 goto Return;
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
249 }
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
250
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
251 #undef ASCII_CHAR_MASK
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
252
74032
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
253
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
254 /* UTF-8 encoder specialized for a Unicode kind to avoid the slow
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
255 PyUnicode_READ() macro. Delete some parts of the code depending on the kind:
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
256 UCS-1 strings don't need to handle surrogates for example. */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
257 Py_LOCAL_INLINE(PyObject *)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
258 STRINGLIB(utf8_encoder)(PyObject *unicode,
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
259 STRINGLIB_CHAR *data,
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
260 Py_ssize_t size,
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
261 const char *errors)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
262 {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
263 #define MAX_SHORT_UNICHARS 300 /* largest size we'll do on the stack */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
264
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
265 Py_ssize_t i; /* index into s of next input byte */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
266 PyObject *result; /* result string object */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
267 char *p; /* next free byte in output buffer */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
268 Py_ssize_t nallocated; /* number of result bytes allocated */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
269 Py_ssize_t nneeded; /* number of result bytes needed */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
270 #if STRINGLIB_SIZEOF_CHAR > 1
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
271 PyObject *errorHandler = NULL;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
272 PyObject *exc = NULL;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
273 PyObject *rep = NULL;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
274 #endif
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
275 #if STRINGLIB_SIZEOF_CHAR == 1
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
276 const Py_ssize_t max_char_size = 2;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
277 char stackbuf[MAX_SHORT_UNICHARS * 2];
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
278 #elif STRINGLIB_SIZEOF_CHAR == 2
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
279 const Py_ssize_t max_char_size = 3;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
280 char stackbuf[MAX_SHORT_UNICHARS * 3];
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
281 #else /* STRINGLIB_SIZEOF_CHAR == 4 */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
282 const Py_ssize_t max_char_size = 4;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
283 char stackbuf[MAX_SHORT_UNICHARS * 4];
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
284 #endif
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
285
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
286 assert(size >= 0);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
287
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
288 if (size <= MAX_SHORT_UNICHARS) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
289 /* Write into the stack buffer; nallocated can't overflow.
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
290 * At the end, we'll allocate exactly as much heap space as it
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
291 * turns out we need.
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
292 */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
293 nallocated = Py_SAFE_DOWNCAST(sizeof(stackbuf), size_t, int);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
294 result = NULL; /* will allocate after we're done */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
295 p = stackbuf;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
296 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
297 else {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
298 if (size > PY_SSIZE_T_MAX / max_char_size) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
299 /* integer overflow */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
300 return PyErr_NoMemory();
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
301 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
302 /* Overallocate on the heap, and give the excess back at the end. */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
303 nallocated = size * max_char_size;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
304 result = PyBytes_FromStringAndSize(NULL, nallocated);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
305 if (result == NULL)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
306 return NULL;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
307 p = PyBytes_AS_STRING(result);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
308 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
309
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
310 for (i = 0; i < size;) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
311 Py_UCS4 ch = data[i++];
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
312
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
313 if (ch < 0x80) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
314 /* Encode ASCII */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
315 *p++ = (char) ch;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
316
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
317 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
318 else
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
319 #if STRINGLIB_SIZEOF_CHAR > 1
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
320 if (ch < 0x0800)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
321 #endif
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
322 {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
323 /* Encode Latin-1 */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
324 *p++ = (char)(0xc0 | (ch >> 6));
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
325 *p++ = (char)(0x80 | (ch & 0x3f));
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
326 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
327 #if STRINGLIB_SIZEOF_CHAR > 1
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
328 else if (Py_UNICODE_IS_SURROGATE(ch)) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
329 Py_ssize_t newpos;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
330 Py_ssize_t repsize, k, startpos;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
331 startpos = i-1;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
332 rep = unicode_encode_call_errorhandler(
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
333 errors, &errorHandler, "utf-8", "surrogates not allowed",
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
334 unicode, &exc, startpos, startpos+1, &newpos);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
335 if (!rep)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
336 goto error;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
337
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
338 if (PyBytes_Check(rep))
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
339 repsize = PyBytes_GET_SIZE(rep);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
340 else
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
341 repsize = PyUnicode_GET_LENGTH(rep);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
342
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
343 if (repsize > max_char_size) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
344 Py_ssize_t offset;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
345
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
346 if (result == NULL)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
347 offset = p - stackbuf;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
348 else
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
349 offset = p - PyBytes_AS_STRING(result);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
350
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
351 if (nallocated > PY_SSIZE_T_MAX - repsize + max_char_size) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
352 /* integer overflow */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
353 PyErr_NoMemory();
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
354 goto error;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
355 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
356 nallocated += repsize - max_char_size;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
357 if (result != NULL) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
358 if (_PyBytes_Resize(&result, nallocated) < 0)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
359 goto error;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
360 } else {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
361 result = PyBytes_FromStringAndSize(NULL, nallocated);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
362 if (result == NULL)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
363 goto error;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
364 Py_MEMCPY(PyBytes_AS_STRING(result), stackbuf, offset);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
365 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
366 p = PyBytes_AS_STRING(result) + offset;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
367 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
368
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
369 if (PyBytes_Check(rep)) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
370 char *prep = PyBytes_AS_STRING(rep);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
371 for(k = repsize; k > 0; k--)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
372 *p++ = *prep++;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
373 } else /* rep is unicode */ {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
374 enum PyUnicode_Kind repkind;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
375 void *repdata;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
376
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
377 if (PyUnicode_READY(rep) < 0)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
378 goto error;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
379 repkind = PyUnicode_KIND(rep);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
380 repdata = PyUnicode_DATA(rep);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
381
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
382 for(k=0; k<repsize; k++) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
383 Py_UCS4 c = PyUnicode_READ(repkind, repdata, k);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
384 if (0x80 <= c) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
385 raise_encode_exception(&exc, "utf-8",
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
386 unicode,
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
387 i-1, i,
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
388 "surrogates not allowed");
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
389 goto error;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
390 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
391 *p++ = (char)c;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
392 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
393 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
394 Py_CLEAR(rep);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
395 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
396 else
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
397 #if STRINGLIB_SIZEOF_CHAR > 2
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
398 if (ch < 0x10000)
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
399 #endif
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
400 {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
401 *p++ = (char)(0xe0 | (ch >> 12));
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
402 *p++ = (char)(0x80 | ((ch >> 6) & 0x3f));
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
403 *p++ = (char)(0x80 | (ch & 0x3f));
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
404 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
405 #if STRINGLIB_SIZEOF_CHAR > 2
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
406 else /* ch >= 0x10000 */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
407 {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
408 assert(ch <= MAX_UNICODE);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
409 /* Encode UCS4 Unicode ordinals */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
410 *p++ = (char)(0xf0 | (ch >> 18));
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
411 *p++ = (char)(0x80 | ((ch >> 12) & 0x3f));
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
412 *p++ = (char)(0x80 | ((ch >> 6) & 0x3f));
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
413 *p++ = (char)(0x80 | (ch & 0x3f));
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
414 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
415 #endif /* STRINGLIB_SIZEOF_CHAR > 2 */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
416 #endif /* STRINGLIB_SIZEOF_CHAR > 1 */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
417 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
418
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
419 if (result == NULL) {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
420 /* This was stack allocated. */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
421 nneeded = p - stackbuf;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
422 assert(nneeded <= nallocated);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
423 result = PyBytes_FromStringAndSize(stackbuf, nneeded);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
424 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
425 else {
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
426 /* Cut back to size actually needed. */
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
427 nneeded = p - PyBytes_AS_STRING(result);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
428 assert(nneeded <= nallocated);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
429 _PyBytes_Resize(&result, nneeded);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
430 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
431
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
432 #if STRINGLIB_SIZEOF_CHAR > 1
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
433 Py_XDECREF(errorHandler);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
434 Py_XDECREF(exc);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
435 #endif
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
436 return result;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
437
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
438 #if STRINGLIB_SIZEOF_CHAR > 1
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
439 error:
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
440 Py_XDECREF(rep);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
441 Py_XDECREF(errorHandler);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
442 Py_XDECREF(exc);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
443 Py_XDECREF(result);
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
444 return NULL;
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
445 #endif
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
446
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
447 #undef MAX_SHORT_UNICHARS
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
448 }
fbd797fc3809 Issue #13624: Write a specialized UTF-8 encoder to allow more optimization
Victor Stinner <victor.stinner@haypocalc.com>
parents: 73668
diff changeset
449
76971
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
450 /* The pattern for constructing UCS2-repeated masks. */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
451 #if SIZEOF_LONG == 8
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
452 # define UCS2_REPEAT_MASK 0x0001000100010001ul
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
453 #elif SIZEOF_LONG == 4
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
454 # define UCS2_REPEAT_MASK 0x00010001ul
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
455 #else
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
456 # error C 'long' size should be either 4 or 8!
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
457 #endif
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
458
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
459 /* The mask for fast checking. */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
460 #if STRINGLIB_SIZEOF_CHAR == 1
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
461 /* The mask for fast checking of whether a C 'long' contains a
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
462 non-ASCII or non-Latin1 UTF16-encoded characters. */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
463 # define FAST_CHAR_MASK (UCS2_REPEAT_MASK * (0xFFFFu & ~STRINGLIB_MAX_CHAR))
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
464 #else
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
465 /* The mask for fast checking of whether a C 'long' may contain
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
466 UTF16-encoded surrogate characters. This is an efficient heuristic,
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
467 assuming that non-surrogate characters with a code point >= 0x8000 are
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
468 rare in most input.
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
469 */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
470 # define FAST_CHAR_MASK (UCS2_REPEAT_MASK * 0x8000u)
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
471 #endif
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
472 /* The mask for fast byte-swapping. */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
473 #define STRIPPED_MASK (UCS2_REPEAT_MASK * 0x00FFu)
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
474 /* Swap bytes. */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
475 #define SWAB(value) ((((value) >> 8) & STRIPPED_MASK) | \
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
476 (((value) & STRIPPED_MASK) << 8))
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
477
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
478 Py_LOCAL_INLINE(Py_UCS4)
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
479 STRINGLIB(utf16_decode)(const unsigned char **inptr, const unsigned char *e,
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
480 STRINGLIB_CHAR *dest, Py_ssize_t *outpos,
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
481 int native_ordering)
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
482 {
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
483 Py_UCS4 ch;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
484 const unsigned char *aligned_end =
79065
99112b851b25 Issue #15144: Fix possible integer overflow when handling pointers as integer values, by using Py_uintptr_t instead of size_t.
Antoine Pitrou <solipsis@pitrou.net>
parents: 77966
diff changeset
485 (const unsigned char *) _Py_ALIGN_DOWN(e, SIZEOF_LONG);
76971
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
486 const unsigned char *q = *inptr;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
487 STRINGLIB_CHAR *p = dest + *outpos;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
488 /* Offsets from q for retrieving byte pairs in the right order. */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
489 #ifdef BYTEORDER_IS_LITTLE_ENDIAN
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
490 int ihi = !!native_ordering, ilo = !native_ordering;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
491 #else
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
492 int ihi = !native_ordering, ilo = !!native_ordering;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
493 #endif
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
494 --e;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
495
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
496 while (q < e) {
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
497 Py_UCS4 ch2;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
498 /* First check for possible aligned read of a C 'long'. Unaligned
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
499 reads are more expensive, better to defer to another iteration. */
79065
99112b851b25 Issue #15144: Fix possible integer overflow when handling pointers as integer values, by using Py_uintptr_t instead of size_t.
Antoine Pitrou <solipsis@pitrou.net>
parents: 77966
diff changeset
500 if (_Py_IS_ALIGNED(q, SIZEOF_LONG)) {
76971
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
501 /* Fast path for runs of in-range non-surrogate chars. */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
502 register const unsigned char *_q = q;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
503 while (_q < aligned_end) {
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
504 unsigned long block = * (unsigned long *) _q;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
505 if (native_ordering) {
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
506 /* Can use buffer directly */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
507 if (block & FAST_CHAR_MASK)
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
508 break;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
509 }
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
510 else {
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
511 /* Need to byte-swap */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
512 if (block & SWAB(FAST_CHAR_MASK))
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
513 break;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
514 #if STRINGLIB_SIZEOF_CHAR == 1
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
515 block >>= 8;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
516 #else
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
517 block = SWAB(block);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
518 #endif
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
519 }
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
520 #ifdef BYTEORDER_IS_LITTLE_ENDIAN
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
521 # if SIZEOF_LONG == 4
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
522 p[0] = (STRINGLIB_CHAR)(block & 0xFFFFu);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
523 p[1] = (STRINGLIB_CHAR)(block >> 16);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
524 # elif SIZEOF_LONG == 8
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
525 p[0] = (STRINGLIB_CHAR)(block & 0xFFFFu);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
526 p[1] = (STRINGLIB_CHAR)((block >> 16) & 0xFFFFu);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
527 p[2] = (STRINGLIB_CHAR)((block >> 32) & 0xFFFFu);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
528 p[3] = (STRINGLIB_CHAR)(block >> 48);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
529 # endif
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
530 #else
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
531 # if SIZEOF_LONG == 4
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
532 p[0] = (STRINGLIB_CHAR)(block >> 16);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
533 p[1] = (STRINGLIB_CHAR)(block & 0xFFFFu);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
534 # elif SIZEOF_LONG == 8
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
535 p[0] = (STRINGLIB_CHAR)(block >> 48);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
536 p[1] = (STRINGLIB_CHAR)((block >> 32) & 0xFFFFu);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
537 p[2] = (STRINGLIB_CHAR)((block >> 16) & 0xFFFFu);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
538 p[3] = (STRINGLIB_CHAR)(block & 0xFFFFu);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
539 # endif
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
540 #endif
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
541 _q += SIZEOF_LONG;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
542 p += SIZEOF_LONG / 2;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
543 }
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
544 q = _q;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
545 if (q >= e)
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
546 break;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
547 }
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
548
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
549 ch = (q[ihi] << 8) | q[ilo];
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
550 q += 2;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
551 if (!Py_UNICODE_IS_SURROGATE(ch)) {
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
552 #if STRINGLIB_SIZEOF_CHAR < 2
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
553 if (ch > STRINGLIB_MAX_CHAR)
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
554 /* Out-of-range */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
555 goto Return;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
556 #endif
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
557 *p++ = (STRINGLIB_CHAR)ch;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
558 continue;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
559 }
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
560
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
561 /* UTF-16 code pair: */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
562 if (q >= e)
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
563 goto UnexpectedEnd;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
564 if (!Py_UNICODE_IS_HIGH_SURROGATE(ch))
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
565 goto IllegalEncoding;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
566 ch2 = (q[ihi] << 8) | q[ilo];
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
567 q += 2;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
568 if (!Py_UNICODE_IS_LOW_SURROGATE(ch2))
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
569 goto IllegalSurrogate;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
570 ch = Py_UNICODE_JOIN_SURROGATES(ch, ch2);
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
571 #if STRINGLIB_SIZEOF_CHAR < 4
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
572 /* Out-of-range */
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
573 goto Return;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
574 #else
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
575 *p++ = (STRINGLIB_CHAR)ch;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
576 #endif
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
577 }
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
578 ch = 0;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
579 Return:
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
580 *inptr = q;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
581 *outpos = p - dest;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
582 return ch;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
583 UnexpectedEnd:
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
584 ch = 1;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
585 goto Return;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
586 IllegalEncoding:
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
587 ch = 2;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
588 goto Return;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
589 IllegalSurrogate:
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
590 ch = 3;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
591 goto Return;
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
592 }
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
593 #undef UCS2_REPEAT_MASK
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
594 #undef FAST_CHAR_MASK
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
595 #undef STRIPPED_MASK
cdcc816dea85 Issue #14624: UTF-16 decoding is now 3x to 4x faster on various inputs.
Antoine Pitrou <solipsis@pitrou.net>
parents: 76862
diff changeset
596 #undef SWAB
77448
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
597
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
598
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
599 Py_LOCAL_INLINE(void)
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
600 STRINGLIB(utf16_encode)(unsigned short *out,
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
601 const STRINGLIB_CHAR *in,
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
602 Py_ssize_t len,
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
603 int native_ordering)
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
604 {
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
605 const STRINGLIB_CHAR *end = in + len;
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
606 #if STRINGLIB_SIZEOF_CHAR == 1
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
607 # define SWAB2(CH) ((CH) << 8)
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
608 #else
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
609 # define SWAB2(CH) (((CH) << 8) | ((CH) >> 8))
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
610 #endif
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
611 #if STRINGLIB_MAX_CHAR < 0x10000
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
612 if (native_ordering) {
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
613 # if STRINGLIB_SIZEOF_CHAR == 2
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
614 Py_MEMCPY(out, in, 2 * len);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
615 # else
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
616 _PyUnicode_CONVERT_BYTES(STRINGLIB_CHAR, unsigned short, in, end, out);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
617 # endif
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
618 } else {
79065
99112b851b25 Issue #15144: Fix possible integer overflow when handling pointers as integer values, by using Py_uintptr_t instead of size_t.
Antoine Pitrou <solipsis@pitrou.net>
parents: 77966
diff changeset
619 const STRINGLIB_CHAR *unrolled_end = in + _Py_SIZE_ROUND_DOWN(len, 4);
77448
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
620 while (in < unrolled_end) {
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
621 out[0] = SWAB2(in[0]);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
622 out[1] = SWAB2(in[1]);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
623 out[2] = SWAB2(in[2]);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
624 out[3] = SWAB2(in[3]);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
625 in += 4; out += 4;
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
626 }
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
627 while (in < end) {
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
628 *out++ = SWAB2(*in);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
629 ++in;
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
630 }
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
631 }
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
632 #else
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
633 if (native_ordering) {
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
634 while (in < end) {
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
635 Py_UCS4 ch = *in++;
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
636 if (ch < 0x10000)
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
637 *out++ = ch;
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
638 else {
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
639 out[0] = Py_UNICODE_HIGH_SURROGATE(ch);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
640 out[1] = Py_UNICODE_LOW_SURROGATE(ch);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
641 out += 2;
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
642 }
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
643 }
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
644 } else {
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
645 while (in < end) {
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
646 Py_UCS4 ch = *in++;
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
647 if (ch < 0x10000)
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
648 *out++ = SWAB2((Py_UCS2)ch);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
649 else {
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
650 Py_UCS2 ch1 = Py_UNICODE_HIGH_SURROGATE(ch);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
651 Py_UCS2 ch2 = Py_UNICODE_LOW_SURROGATE(ch);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
652 out[0] = SWAB2(ch1);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
653 out[1] = SWAB2(ch2);
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
654 out += 2;
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
655 }
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
656 }
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
657 }
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
658 #endif
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
659 #undef SWAB2
acca141fda80 Issue #15026: utf-16 encoding is now significantly faster (up to 10x).
Antoine Pitrou <solipsis@pitrou.net>
parents: 76971
diff changeset
660 }
73668
8e6c4acaf530 Issue #13417: speed up utf-8 decoding by around 2x for the non-fully-ASCII case.
Antoine Pitrou <solipsis@pitrou.net>
parents:
diff changeset
661 #endif /* STRINGLIB_IS_UNICODE */