This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Make isSTRICT_UTF8_CHAR() an inline function
[perl5.git] / inline.h
CommitLineData
25468daa
FC
1/* inline.h
2 *
3 * Copyright (C) 2012 by Larry Wall and others
4 *
5 * You may distribute under the terms of either the GNU General Public
6 * License or the Artistic License, as specified in the README file.
7 *
8ed185f9
KW
8 * This file contains tables and code adapted from
9 * http://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which requires this
10 * copyright notice:
11
12Copyright (c) 2008-2009 Bjoern Hoehrmann <bjoern@hoehrmann.de>
13
14Permission is hereby granted, free of charge, to any person obtaining a copy of
15this software and associated documentation files (the "Software"), to deal in
16the Software without restriction, including without limitation the rights to
17use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
18of the Software, and to permit persons to whom the Software is furnished to do
19so, subject to the following conditions:
20
21The above copyright notice and this permission notice shall be included in all
22copies or substantial portions of the Software.
23
24THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
25IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
26FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
27AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
28LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
29OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
30SOFTWARE.
31
32 *
25468daa 33 * This file is a home for static inline functions that cannot go in other
e15e54ff 34 * header files, because they depend on proto.h (included after most other
25468daa
FC
35 * headers) or struct definitions.
36 *
37 * Each section names the header file that the functions "belong" to.
38 */
27669aa4 39
be3a7a5d
KW
40/* ------------------------------- av.h ------------------------------- */
41
c70927a6 42PERL_STATIC_INLINE SSize_t
be3a7a5d
KW
43S_av_top_index(pTHX_ AV *av)
44{
45 PERL_ARGS_ASSERT_AV_TOP_INDEX;
46 assert(SvTYPE(av) == SVt_PVAV);
47
48 return AvFILL(av);
49}
50
1afe1db1
FC
51/* ------------------------------- cv.h ------------------------------- */
52
ae77754a
FC
53PERL_STATIC_INLINE GV *
54S_CvGV(pTHX_ CV *sv)
55{
56 return CvNAMED(sv)
57 ? Perl_cvgv_from_hek(aTHX_ sv)
58 : ((XPVCV*)MUTABLE_PTR(SvANY(sv)))->xcv_gv_u.xcv_gv;
59}
60
1afe1db1
FC
61PERL_STATIC_INLINE I32 *
62S_CvDEPTHp(const CV * const sv)
63{
64 assert(SvTYPE(sv) == SVt_PVCV || SvTYPE(sv) == SVt_PVFM);
8de47657 65 return &((XPVCV*)SvANY(sv))->xcv_depth;
1afe1db1
FC
66}
67
d16269d8
PM
68/*
69 CvPROTO returns the prototype as stored, which is not necessarily what
70 the interpreter should be using. Specifically, the interpreter assumes
71 that spaces have been stripped, which has been the case if the prototype
72 was added by toke.c, but is generally not the case if it was added elsewhere.
73 Since we can't enforce the spacelessness at assignment time, this routine
74 provides a temporary copy at parse time with spaces removed.
75 I<orig> is the start of the original buffer, I<len> is the length of the
76 prototype and will be updated when this returns.
77 */
78
5b67adb8 79#ifdef PERL_CORE
d16269d8
PM
80PERL_STATIC_INLINE char *
81S_strip_spaces(pTHX_ const char * orig, STRLEN * const len)
82{
83 SV * tmpsv;
84 char * tmps;
85 tmpsv = newSVpvn_flags(orig, *len, SVs_TEMP);
86 tmps = SvPVX(tmpsv);
87 while ((*len)--) {
88 if (!isSPACE(*orig))
89 *tmps++ = *orig;
90 orig++;
91 }
92 *tmps = '\0';
93 *len = tmps - SvPVX(tmpsv);
94 return SvPVX(tmpsv);
95}
5b67adb8 96#endif
d16269d8 97
25fdce4a
FC
98/* ------------------------------- mg.h ------------------------------- */
99
100#if defined(PERL_CORE) || defined(PERL_EXT)
101/* assumes get-magic and stringification have already occurred */
102PERL_STATIC_INLINE STRLEN
103S_MgBYTEPOS(pTHX_ MAGIC *mg, SV *sv, const char *s, STRLEN len)
104{
105 assert(mg->mg_type == PERL_MAGIC_regex_global);
106 assert(mg->mg_len != -1);
107 if (mg->mg_flags & MGf_BYTES || !DO_UTF8(sv))
108 return (STRLEN)mg->mg_len;
109 else {
110 const STRLEN pos = (STRLEN)mg->mg_len;
111 /* Without this check, we may read past the end of the buffer: */
112 if (pos > sv_or_pv_len_utf8(sv, s, len)) return len+1;
113 return sv_or_pv_pos_u2b(sv, s, pos, NULL);
114 }
115}
116#endif
117
03414f05
FC
118/* ------------------------------- pad.h ------------------------------ */
119
120#if defined(PERL_IN_PAD_C) || defined(PERL_IN_OP_C)
121PERL_STATIC_INLINE bool
122PadnameIN_SCOPE(const PADNAME * const pn, const U32 seq)
123{
124 /* is seq within the range _LOW to _HIGH ?
125 * This is complicated by the fact that PL_cop_seqmax
126 * may have wrapped around at some point */
127 if (COP_SEQ_RANGE_LOW(pn) == PERL_PADSEQ_INTRO)
128 return FALSE; /* not yet introduced */
129
130 if (COP_SEQ_RANGE_HIGH(pn) == PERL_PADSEQ_INTRO) {
131 /* in compiling scope */
132 if (
133 (seq > COP_SEQ_RANGE_LOW(pn))
134 ? (seq - COP_SEQ_RANGE_LOW(pn) < (U32_MAX >> 1))
135 : (COP_SEQ_RANGE_LOW(pn) - seq > (U32_MAX >> 1))
136 )
137 return TRUE;
138 }
139 else if (
140 (COP_SEQ_RANGE_LOW(pn) > COP_SEQ_RANGE_HIGH(pn))
141 ?
142 ( seq > COP_SEQ_RANGE_LOW(pn)
143 || seq <= COP_SEQ_RANGE_HIGH(pn))
144
145 : ( seq > COP_SEQ_RANGE_LOW(pn)
146 && seq <= COP_SEQ_RANGE_HIGH(pn))
147 )
148 return TRUE;
149 return FALSE;
150}
151#endif
152
33a4312b
FC
153/* ------------------------------- pp.h ------------------------------- */
154
155PERL_STATIC_INLINE I32
156S_TOPMARK(pTHX)
157{
158 DEBUG_s(DEBUG_v(PerlIO_printf(Perl_debug_log,
147e3846 159 "MARK top %p %" IVdf "\n",
33a4312b
FC
160 PL_markstack_ptr,
161 (IV)*PL_markstack_ptr)));
162 return *PL_markstack_ptr;
163}
164
165PERL_STATIC_INLINE I32
166S_POPMARK(pTHX)
167{
168 DEBUG_s(DEBUG_v(PerlIO_printf(Perl_debug_log,
147e3846 169 "MARK pop %p %" IVdf "\n",
33a4312b
FC
170 (PL_markstack_ptr-1),
171 (IV)*(PL_markstack_ptr-1))));
172 assert((PL_markstack_ptr > PL_markstack) || !"MARK underflow");
173 return *PL_markstack_ptr--;
174}
175
8d919b0a
FC
176/* ----------------------------- regexp.h ----------------------------- */
177
178PERL_STATIC_INLINE struct regexp *
179S_ReANY(const REGEXP * const re)
180{
df6b4bd5 181 XPV* const p = (XPV*)SvANY(re);
8d919b0a 182 assert(isREGEXP(re));
df6b4bd5
DM
183 return SvTYPE(re) == SVt_PVLV ? p->xpv_len_u.xpvlenu_rx
184 : (struct regexp *)p;
8d919b0a
FC
185}
186
27669aa4
FC
187/* ------------------------------- sv.h ------------------------------- */
188
189PERL_STATIC_INLINE SV *
190S_SvREFCNT_inc(SV *sv)
191{
2439e033 192 if (LIKELY(sv != NULL))
27669aa4
FC
193 SvREFCNT(sv)++;
194 return sv;
195}
196PERL_STATIC_INLINE SV *
197S_SvREFCNT_inc_NN(SV *sv)
198{
199 SvREFCNT(sv)++;
200 return sv;
201}
202PERL_STATIC_INLINE void
203S_SvREFCNT_inc_void(SV *sv)
204{
2439e033 205 if (LIKELY(sv != NULL))
27669aa4
FC
206 SvREFCNT(sv)++;
207}
75e16a44
FC
208PERL_STATIC_INLINE void
209S_SvREFCNT_dec(pTHX_ SV *sv)
210{
2439e033 211 if (LIKELY(sv != NULL)) {
75a9bf96 212 U32 rc = SvREFCNT(sv);
79e2a32a 213 if (LIKELY(rc > 1))
75a9bf96
DM
214 SvREFCNT(sv) = rc - 1;
215 else
216 Perl_sv_free2(aTHX_ sv, rc);
75e16a44
FC
217 }
218}
541377b1
FC
219
220PERL_STATIC_INLINE void
4a9a56a7
DM
221S_SvREFCNT_dec_NN(pTHX_ SV *sv)
222{
223 U32 rc = SvREFCNT(sv);
79e2a32a 224 if (LIKELY(rc > 1))
4a9a56a7
DM
225 SvREFCNT(sv) = rc - 1;
226 else
227 Perl_sv_free2(aTHX_ sv, rc);
228}
229
230PERL_STATIC_INLINE void
541377b1
FC
231SvAMAGIC_on(SV *sv)
232{
233 assert(SvROK(sv));
234 if (SvOBJECT(SvRV(sv))) HvAMAGIC_on(SvSTASH(SvRV(sv)));
235}
236PERL_STATIC_INLINE void
237SvAMAGIC_off(SV *sv)
238{
239 if (SvROK(sv) && SvOBJECT(SvRV(sv)))
240 HvAMAGIC_off(SvSTASH(SvRV(sv)));
241}
242
243PERL_STATIC_INLINE U32
541377b1
FC
244S_SvPADSTALE_on(SV *sv)
245{
c0683843 246 assert(!(SvFLAGS(sv) & SVs_PADTMP));
541377b1
FC
247 return SvFLAGS(sv) |= SVs_PADSTALE;
248}
249PERL_STATIC_INLINE U32
250S_SvPADSTALE_off(SV *sv)
251{
c0683843 252 assert(!(SvFLAGS(sv) & SVs_PADTMP));
541377b1
FC
253 return SvFLAGS(sv) &= ~SVs_PADSTALE;
254}
25fdce4a 255#if defined(PERL_CORE) || defined (PERL_EXT)
4ddea69a 256PERL_STATIC_INLINE STRLEN
6964422a 257S_sv_or_pv_pos_u2b(pTHX_ SV *sv, const char *pv, STRLEN pos, STRLEN *lenp)
4ddea69a 258{
25fdce4a 259 PERL_ARGS_ASSERT_SV_OR_PV_POS_U2B;
4ddea69a
FC
260 if (SvGAMAGIC(sv)) {
261 U8 *hopped = utf8_hop((U8 *)pv, pos);
262 if (lenp) *lenp = (STRLEN)(utf8_hop(hopped, *lenp) - hopped);
263 return (STRLEN)(hopped - (U8 *)pv);
264 }
265 return sv_pos_u2b_flags(sv,pos,lenp,SV_CONST_RETURN);
266}
267#endif
f019c49e 268
d1decf2b
TC
269/* ------------------------------- handy.h ------------------------------- */
270
271/* saves machine code for a common noreturn idiom typically used in Newx*() */
7347ee54 272GCC_DIAG_IGNORE_DECL(-Wunused-function);
d1decf2b
TC
273static void
274S_croak_memory_wrap(void)
275{
276 Perl_croak_nocontext("%s",PL_memory_wrap);
277}
7347ee54 278GCC_DIAG_RESTORE_DECL;
d1decf2b 279
a8a2ceaa
KW
280/* ------------------------------- utf8.h ------------------------------- */
281
2fe720e2
KW
282/*
283=head1 Unicode Support
284*/
285
55d09dc8
KW
286PERL_STATIC_INLINE void
287S_append_utf8_from_native_byte(const U8 byte, U8** dest)
288{
289 /* Takes an input 'byte' (Latin1 or EBCDIC) and appends it to the UTF-8
290 * encoded string at '*dest', updating '*dest' to include it */
291
55d09dc8
KW
292 PERL_ARGS_ASSERT_APPEND_UTF8_FROM_NATIVE_BYTE;
293
6f2d5cbc 294 if (NATIVE_BYTE_IS_INVARIANT(byte))
a09ec51a 295 *((*dest)++) = byte;
55d09dc8 296 else {
a09ec51a
KW
297 *((*dest)++) = UTF8_EIGHT_BIT_HI(byte);
298 *((*dest)++) = UTF8_EIGHT_BIT_LO(byte);
55d09dc8
KW
299 }
300}
301
e123187a 302/*
2fe720e2 303=for apidoc valid_utf8_to_uvchr
2717076a 304Like C<L</utf8_to_uvchr_buf>>, but should only be called when it is known that
2fe720e2
KW
305the next character in the input UTF-8 string C<s> is well-formed (I<e.g.>,
306it passes C<L</isUTF8_CHAR>>. Surrogates, non-character code points, and
307non-Unicode code points are allowed.
308
309=cut
310
311 */
312
313PERL_STATIC_INLINE UV
314Perl_valid_utf8_to_uvchr(const U8 *s, STRLEN *retlen)
315{
c41b2540 316 const UV expectlen = UTF8SKIP(s);
2fe720e2
KW
317 const U8* send = s + expectlen;
318 UV uv = *s;
319
320 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR;
321
322 if (retlen) {
323 *retlen = expectlen;
324 }
325
326 /* An invariant is trivially returned */
327 if (expectlen == 1) {
328 return uv;
329 }
330
331 /* Remove the leading bits that indicate the number of bytes, leaving just
332 * the bits that are part of the value */
333 uv = NATIVE_UTF8_TO_I8(uv) & UTF_START_MASK(expectlen);
334
335 /* Now, loop through the remaining bytes, accumulating each into the
336 * working total as we go. (I khw tried unrolling the loop for up to 4
337 * bytes, but there was no performance improvement) */
338 for (++s; s < send; s++) {
339 uv = UTF8_ACCUMULATE(uv, *s);
340 }
341
342 return UNI_TO_NATIVE(uv);
343
344}
345
1e599354
KW
346/*
347=for apidoc is_utf8_invariant_string
348
82c5d941 349Returns TRUE if the first C<len> bytes of the string C<s> are the same
1e599354 350regardless of the UTF-8 encoding of the string (or UTF-EBCDIC encoding on
82c5d941
KW
351EBCDIC machines); otherwise it returns FALSE. That is, it returns TRUE if they
352are UTF-8 invariant. On ASCII-ish machines, all the ASCII characters and only
353the ASCII characters fit this definition. On EBCDIC machines, the ASCII-range
354characters are invariant, but so also are the C1 controls.
1e599354
KW
355
356If C<len> is 0, it will be calculated using C<strlen(s)>, (which means if you
357use this option, that C<s> can't have embedded C<NUL> characters and has to
358have a terminating C<NUL> byte).
359
9f2abfde
KW
360See also
361C<L</is_utf8_string>>,
362C<L</is_utf8_string_flags>>,
363C<L</is_utf8_string_loc>>,
364C<L</is_utf8_string_loc_flags>>,
365C<L</is_utf8_string_loclen>>,
366C<L</is_utf8_string_loclen_flags>>,
8bc127bf
KW
367C<L</is_utf8_fixed_width_buf_flags>>,
368C<L</is_utf8_fixed_width_buf_loc_flags>>,
369C<L</is_utf8_fixed_width_buf_loclen_flags>>,
9f2abfde
KW
370C<L</is_strict_utf8_string>>,
371C<L</is_strict_utf8_string_loc>>,
372C<L</is_strict_utf8_string_loclen>>,
373C<L</is_c9strict_utf8_string>>,
374C<L</is_c9strict_utf8_string_loc>>,
375and
376C<L</is_c9strict_utf8_string_loclen>>.
1e599354
KW
377
378=cut
0cbf5865
KW
379
380*/
381
382#define is_utf8_invariant_string(s, len) \
383 is_utf8_invariant_string_loc(s, len, NULL)
384
385/*
386=for apidoc is_utf8_invariant_string_loc
387
388Like C<L</is_utf8_invariant_string>> but upon failure, stores the location of
389the first UTF-8 variant character in the C<ep> pointer; if all characters are
390UTF-8 invariant, this function does not change the contents of C<*ep>.
391
392=cut
393
1e599354
KW
394*/
395
396PERL_STATIC_INLINE bool
e17544a6 397S_is_utf8_invariant_string_loc(const U8* const s, STRLEN len, const U8 ** ep)
1e599354 398{
e17544a6 399 const U8* send;
1e599354
KW
400 const U8* x = s;
401
0cbf5865
KW
402 PERL_ARGS_ASSERT_IS_UTF8_INVARIANT_STRING_LOC;
403
e17544a6
KW
404 if (len == 0) {
405 len = strlen((const char *)s);
406 }
407
408 send = s + len;
409
4ab2fd9b 410/* This looks like 0x010101... */
2c5c8af5 411# define PERL_COUNT_MULTIPLIER (~ (UINTMAX_C(0)) / 0xFF)
4ab2fd9b
KW
412
413/* This looks like 0x808080... */
2c5c8af5 414# define PERL_VARIANTS_WORD_MASK (PERL_COUNT_MULTIPLIER * 0x80)
e099ea69 415# define PERL_WORDSIZE sizeof(PERL_UINTMAX_T)
2c5c8af5 416# define PERL_WORD_BOUNDARY_MASK (PERL_WORDSIZE - 1)
e17544a6 417
099e59a4
KW
418/* Evaluates to 0 if 'x' is at a word boundary; otherwise evaluates to 1, by
419 * or'ing together the lowest bits of 'x'. Hopefully the final term gets
420 * optimized out completely on a 32-bit system, and its mask gets optimized out
421 * on a 64-bit system */
2c5c8af5 422# define PERL_IS_SUBWORD_ADDR(x) (1 & ( PTR2nat(x) \
5eabe374
KW
423 | ( PTR2nat(x) >> 1) \
424 | ( ( (PTR2nat(x) \
425 & PERL_WORD_BOUNDARY_MASK) >> 2))))
099e59a4 426
3f515a2e
KW
427#ifndef EBCDIC
428
099e59a4
KW
429 /* Do the word-at-a-time iff there is at least one usable full word. That
430 * means that after advancing to a word boundary, there still is at least a
431 * full word left. The number of bytes needed to advance is 'wordsize -
432 * offset' unless offset is 0. */
433 if ((STRLEN) (send - x) >= PERL_WORDSIZE
434
435 /* This term is wordsize if subword; 0 if not */
436 + PERL_WORDSIZE * PERL_IS_SUBWORD_ADDR(x)
437
438 /* 'offset' */
439 - (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK))
440 {
b40579ff 441
46bb68f6
KW
442 /* Process per-byte until reach word boundary. XXX This loop could be
443 * eliminated if we knew that this platform had fast unaligned reads */
b40579ff 444 while (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK) {
46bb68f6
KW
445 if (! UTF8_IS_INVARIANT(*x)) {
446 if (ep) {
447 *ep = x;
448 }
e17544a6 449
46bb68f6
KW
450 return FALSE;
451 }
452 x++;
e17544a6 453 }
e17544a6 454
099e59a4
KW
455 /* Here, we know we have at least one full word to process. Process
456 * per-word as long as we have at least a full word left */
457 do {
4ab2fd9b 458 if ((* (PERL_UINTMAX_T *) x) & PERL_VARIANTS_WORD_MASK) {
e17544a6 459
46bb68f6
KW
460 /* Found a variant. Just return if caller doesn't want its
461 * exact position */
462 if (! ep) {
463 return FALSE;
464 }
e17544a6 465
2c5c8af5
KW
466# if BYTEORDER == 0x1234 || BYTEORDER == 0x12345678 \
467 || BYTEORDER == 0x4321 || BYTEORDER == 0x87654321
1d2af574
KW
468
469 *ep = x + _variant_byte_number(* (PERL_UINTMAX_T *) x);
470 assert(*ep >= s && *ep < send);
471
472 return FALSE;
473
2c5c8af5 474# else /* If weird byte order, drop into next loop to do byte-at-a-time
1d2af574
KW
475 checks. */
476
46bb68f6 477 break;
2c5c8af5 478# endif
46bb68f6 479 }
1d2af574 480
46bb68f6 481 x += PERL_WORDSIZE;
1d2af574 482
099e59a4 483 } while (x + PERL_WORDSIZE <= send);
b40579ff 484 }
e17544a6 485
0b08cab0 486#endif /* End of ! EBCDIC */
e17544a6
KW
487
488 /* Process per-byte */
489 while (x < send) {
490 if (! UTF8_IS_INVARIANT(*x)) {
491 if (ep) {
492 *ep = x;
493 }
0cbf5865 494
e17544a6 495 return FALSE;
0cbf5865 496 }
1e599354 497
e17544a6 498 x++;
1e599354
KW
499 }
500
501 return TRUE;
502}
503
23a7ee81
KW
504#ifndef EBCDIC
505
1d2af574
KW
506PERL_STATIC_INLINE unsigned int
507S__variant_byte_number(PERL_UINTMAX_T word)
508{
509
510 /* This returns the position in a word (0..7) of the first variant byte in
511 * it. This is a helper function. Note that there are no branches */
512
513 assert(word);
514
515 /* Get just the msb bits of each byte */
516 word &= PERL_VARIANTS_WORD_MASK;
517
597ee3f4
KW
518# ifdef USING_MSVC6 /* VC6 has some issues with the normal code, and the
519 easiest thing is to hide that from the callers */
520 {
521 unsigned int i;
522 const U8 * s = (U8 *) &word;
523 dTHX;
524
525 for (i = 0; i < sizeof(word); i++ ) {
526 if (s[i]) {
527 return i;
528 }
529 }
530
531 Perl_croak(aTHX_ "panic: %s: %d: unexpected zero word\n",
532 __FILE__, __LINE__);
533 }
534
535# elif BYTEORDER == 0x1234 || BYTEORDER == 0x12345678
1d2af574
KW
536
537 /* Bytes are stored like
538 * Byte8 ... Byte2 Byte1
539 * 63..56...15...8 7...0
540 *
541 * Isolate the lsb;
542 * https://stackoverflow.com/questions/757059/position-of-least-significant-bit-that-is-set
543 *
544 * The word will look this this, with a rightmost set bit in position 's':
545 * ('x's are don't cares)
546 * s
547 * x..x100..0
548 * x..xx10..0 Right shift (rightmost 0 is shifted off)
549 * x..xx01..1 Subtract 1, turns all the trailing zeros into 1's and
550 * the 1 just to their left into a 0; the remainder is
551 * untouched
552 * 0..0011..1 The xor with x..xx10..0 clears that remainder, sets
553 * bottom to all 1
554 * 0..0100..0 Add 1 to clear the word except for the bit in 's'
555 *
556 * Another method is to do 'word &= -word'; but it generates a compiler
557 * message on some platforms about taking the negative of an unsigned */
558
559 word >>= 1;
560 word = 1 + (word ^ (word - 1));
561
562# elif BYTEORDER == 0x4321 || BYTEORDER == 0x87654321
563
564 /* Bytes are stored like
565 * Byte1 Byte2 ... Byte8
566 * 63..56 55..47 ... 7...0
567 *
568 * Isolate the msb; http://codeforces.com/blog/entry/10330
569 *
570 * Only the most significant set bit matters. Or'ing word with its right
571 * shift of 1 makes that bit and the next one to its right both 1. Then
572 * right shifting by 2 makes for 4 1-bits in a row. ... We end with the
573 * msb and all to the right being 1. */
574 word |= word >> 1;
575 word |= word >> 2;
576 word |= word >> 4;
577 word |= word >> 8;
578 word |= word >> 16;
579 word |= word >> 32; /* This should get optimized out on 32-bit systems. */
580
581 /* Then subtracting the right shift by 1 clears all but the left-most of
582 * the 1 bits, which is our desired result */
583 word -= (word >> 1);
584
585# else
586# error Unexpected byte order
587# endif
588
7cf2d6c7
KW
589 /* Here 'word' has a single bit set: the msb of the first byte in which it
590 * is set. Calculate that position in the word. We can use this
1d2af574 591 * specialized solution: https://stackoverflow.com/a/32339674/1626653,
67e12c5c
KW
592 * assumes an 8-bit byte. (On a 32-bit machine, the larger numbers should
593 * just get shifted off at compile time) */
594 word = (word >> 7) * ((UINTMAX_C( 7) << 56) | (UINTMAX_C(15) << 48)
595 | (UINTMAX_C(23) << 40) | (UINTMAX_C(31) << 32)
596 | (39 << 24) | (47 << 16)
597 | (55 << 8) | (63 << 0));
1d2af574
KW
598 word >>= PERL_WORDSIZE * 7; /* >> by either 56 or 24 */
599
600 /* Here, word contains the position 7..63 of that bit. Convert to 0..7 */
601 word = ((word + 1) >> 3) - 1;
602
603# if BYTEORDER == 0x4321 || BYTEORDER == 0x87654321
604
605 /* And invert the result */
606 word = CHARBITS - word - 1;
607
608# endif
609
610 return (unsigned int) word;
611}
612
23a7ee81 613#endif
03c1e4ab
KW
614#if defined(PERL_CORE) || defined(PERL_EXT)
615
616/*
617=for apidoc variant_under_utf8_count
618
619This function looks at the sequence of bytes between C<s> and C<e>, which are
620assumed to be encoded in ASCII/Latin1, and returns how many of them would
621change should the string be translated into UTF-8. Due to the nature of UTF-8,
622each of these would occupy two bytes instead of the single one in the input
623string. Thus, this function returns the precise number of bytes the string
624would expand by when translated to UTF-8.
625
626Unlike most of the other functions that have C<utf8> in their name, the input
627to this function is NOT a UTF-8-encoded string. The function name is slightly
628I<odd> to emphasize this.
629
630This function is internal to Perl because khw thinks that any XS code that
631would want this is probably operating too close to the internals. Presenting a
632valid use case could change that.
633
634See also
635C<L<perlapi/is_utf8_invariant_string>>
636and
637C<L<perlapi/is_utf8_invariant_string_loc>>,
638
639=cut
640
641*/
642
643PERL_STATIC_INLINE Size_t
644S_variant_under_utf8_count(const U8* const s, const U8* const e)
645{
646 const U8* x = s;
647 Size_t count = 0;
648
649 PERL_ARGS_ASSERT_VARIANT_UNDER_UTF8_COUNT;
650
651# ifndef EBCDIC
652
5d0379de
KW
653 /* Test if the string is long enough to use word-at-a-time. (Logic is the
654 * same as for is_utf8_invariant_string()) */
03c1e4ab
KW
655 if ((STRLEN) (e - x) >= PERL_WORDSIZE
656 + PERL_WORDSIZE * PERL_IS_SUBWORD_ADDR(x)
657 - (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK))
658 {
659
660 /* Process per-byte until reach word boundary. XXX This loop could be
661 * eliminated if we knew that this platform had fast unaligned reads */
662 while (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK) {
663 count += ! UTF8_IS_INVARIANT(*x++);
664 }
665
666 /* Process per-word as long as we have at least a full word left */
74472cc2
KW
667 do { /* Commit 03c1e4ab1d6ee9062fb3f94b0ba31db6698724b1 contains an
668 explanation of how this works */
e5863284
KW
669 PERL_UINTMAX_T increment
670 = ((((* (PERL_UINTMAX_T *) x) & PERL_VARIANTS_WORD_MASK) >> 7)
03c1e4ab
KW
671 * PERL_COUNT_MULTIPLIER)
672 >> ((PERL_WORDSIZE - 1) * CHARBITS);
e5863284 673 count += (Size_t) increment;
03c1e4ab
KW
674 x += PERL_WORDSIZE;
675 } while (x + PERL_WORDSIZE <= e);
676 }
677
678# endif
679
680 /* Process per-byte */
681 while (x < e) {
682 if (! UTF8_IS_INVARIANT(*x)) {
683 count++;
684 }
685
686 x++;
687 }
688
689 return count;
690}
691
692#endif
693
aff4cafe
KW
694#ifndef PERL_IN_REGEXEC_C /* Keep these around for that file */
695# undef PERL_WORDSIZE
696# undef PERL_COUNT_MULTIPLIER
697# undef PERL_WORD_BOUNDARY_MASK
698# undef PERL_VARIANTS_WORD_MASK
699#endif
03c1e4ab 700
7c93d8f0 701/*
5ff889fb
KW
702=for apidoc is_utf8_string
703
82c5d941
KW
704Returns TRUE if the first C<len> bytes of string C<s> form a valid
705Perl-extended-UTF-8 string; returns FALSE otherwise. If C<len> is 0, it will
706be calculated using C<strlen(s)> (which means if you use this option, that C<s>
707can't have embedded C<NUL> characters and has to have a terminating C<NUL>
708byte). Note that all characters being ASCII constitute 'a valid UTF-8 string'.
709
2717076a
KW
710This function considers Perl's extended UTF-8 to be valid. That means that
711code points above Unicode, surrogates, and non-character code points are
9f2abfde
KW
712considered valid by this function. Use C<L</is_strict_utf8_string>>,
713C<L</is_c9strict_utf8_string>>, or C<L</is_utf8_string_flags>> to restrict what
714code points are considered valid.
5ff889fb 715
9f2abfde
KW
716See also
717C<L</is_utf8_invariant_string>>,
0cbf5865 718C<L</is_utf8_invariant_string_loc>>,
9f2abfde
KW
719C<L</is_utf8_string_loc>>,
720C<L</is_utf8_string_loclen>>,
8bc127bf
KW
721C<L</is_utf8_fixed_width_buf_flags>>,
722C<L</is_utf8_fixed_width_buf_loc_flags>>,
723C<L</is_utf8_fixed_width_buf_loclen_flags>>,
5ff889fb
KW
724
725=cut
726*/
727
dd237e82 728#define is_utf8_string(s, len) is_utf8_string_loclen(s, len, NULL, NULL)
5ff889fb 729
c9cd936b
KW
730#if defined(PERL_CORE) || defined (PERL_EXT)
731
732/*
733=for apidoc is_utf8_non_invariant_string
734
735Returns TRUE if L<perlapi/is_utf8_invariant_string> returns FALSE for the first
736C<len> bytes of the string C<s>, but they are, nonetheless, legal Perl-extended
737UTF-8; otherwise returns FALSE.
738
739A TRUE return means that at least one code point represented by the sequence
740either is a wide character not representable as a single byte, or the
741representation differs depending on whether the sequence is encoded in UTF-8 or
742not.
743
744See also
745C<L<perlapi/is_utf8_invariant_string>>,
746C<L<perlapi/is_utf8_string>>
747
748=cut
749
750This is commonly used to determine if a SV's UTF-8 flag should be turned on.
b3b93dfe
KW
751It generally needn't be if its string is entirely UTF-8 invariant, and it
752shouldn't be if it otherwise contains invalid UTF-8.
c9cd936b
KW
753
754It is an internal function because khw thinks that XS code shouldn't be working
755at this low a level. A valid use case could change that.
756
757*/
758
759PERL_STATIC_INLINE bool
760S_is_utf8_non_invariant_string(const U8* const s, STRLEN len)
761{
762 const U8 * first_variant;
763
764 PERL_ARGS_ASSERT_IS_UTF8_NON_INVARIANT_STRING;
765
766 if (is_utf8_invariant_string_loc(s, len, &first_variant)) {
767 return FALSE;
768 }
769
770 return is_utf8_string(first_variant, len - (first_variant - s));
771}
772
773#endif
774
5ff889fb 775/*
9f2abfde
KW
776=for apidoc is_strict_utf8_string
777
778Returns TRUE if the first C<len> bytes of string C<s> form a valid
779UTF-8-encoded string that is fully interchangeable by any application using
780Unicode rules; otherwise it returns FALSE. If C<len> is 0, it will be
781calculated using C<strlen(s)> (which means if you use this option, that C<s>
782can't have embedded C<NUL> characters and has to have a terminating C<NUL>
783byte). Note that all characters being ASCII constitute 'a valid UTF-8 string'.
784
785This function returns FALSE for strings containing any
786code points above the Unicode max of 0x10FFFF, surrogate code points, or
787non-character code points.
788
789See also
790C<L</is_utf8_invariant_string>>,
0cbf5865 791C<L</is_utf8_invariant_string_loc>>,
9f2abfde
KW
792C<L</is_utf8_string>>,
793C<L</is_utf8_string_flags>>,
794C<L</is_utf8_string_loc>>,
795C<L</is_utf8_string_loc_flags>>,
796C<L</is_utf8_string_loclen>>,
797C<L</is_utf8_string_loclen_flags>>,
8bc127bf
KW
798C<L</is_utf8_fixed_width_buf_flags>>,
799C<L</is_utf8_fixed_width_buf_loc_flags>>,
800C<L</is_utf8_fixed_width_buf_loclen_flags>>,
9f2abfde
KW
801C<L</is_strict_utf8_string_loc>>,
802C<L</is_strict_utf8_string_loclen>>,
803C<L</is_c9strict_utf8_string>>,
804C<L</is_c9strict_utf8_string_loc>>,
805and
806C<L</is_c9strict_utf8_string_loclen>>.
807
808=cut
809*/
810
dd237e82 811#define is_strict_utf8_string(s, len) is_strict_utf8_string_loclen(s, len, NULL, NULL)
9f2abfde
KW
812
813/*
814=for apidoc is_c9strict_utf8_string
815
816Returns TRUE if the first C<len> bytes of string C<s> form a valid
817UTF-8-encoded string that conforms to
818L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>;
819otherwise it returns FALSE. If C<len> is 0, it will be calculated using
820C<strlen(s)> (which means if you use this option, that C<s> can't have embedded
821C<NUL> characters and has to have a terminating C<NUL> byte). Note that all
822characters being ASCII constitute 'a valid UTF-8 string'.
823
824This function returns FALSE for strings containing any code points above the
825Unicode max of 0x10FFFF or surrogate code points, but accepts non-character
826code points per
827L<Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>.
828
829See also
830C<L</is_utf8_invariant_string>>,
0cbf5865 831C<L</is_utf8_invariant_string_loc>>,
9f2abfde
KW
832C<L</is_utf8_string>>,
833C<L</is_utf8_string_flags>>,
834C<L</is_utf8_string_loc>>,
835C<L</is_utf8_string_loc_flags>>,
836C<L</is_utf8_string_loclen>>,
837C<L</is_utf8_string_loclen_flags>>,
8bc127bf
KW
838C<L</is_utf8_fixed_width_buf_flags>>,
839C<L</is_utf8_fixed_width_buf_loc_flags>>,
840C<L</is_utf8_fixed_width_buf_loclen_flags>>,
9f2abfde
KW
841C<L</is_strict_utf8_string>>,
842C<L</is_strict_utf8_string_loc>>,
843C<L</is_strict_utf8_string_loclen>>,
844C<L</is_c9strict_utf8_string_loc>>,
845and
846C<L</is_c9strict_utf8_string_loclen>>.
847
848=cut
849*/
850
dd237e82 851#define is_c9strict_utf8_string(s, len) is_c9strict_utf8_string_loclen(s, len, NULL, 0)
9f2abfde
KW
852
853/*
854=for apidoc is_utf8_string_flags
855
856Returns TRUE if the first C<len> bytes of string C<s> form a valid
857UTF-8 string, subject to the restrictions imposed by C<flags>;
858returns FALSE otherwise. If C<len> is 0, it will be calculated
859using C<strlen(s)> (which means if you use this option, that C<s> can't have
860embedded C<NUL> characters and has to have a terminating C<NUL> byte). Note
861that all characters being ASCII constitute 'a valid UTF-8 string'.
862
863If C<flags> is 0, this gives the same results as C<L</is_utf8_string>>; if
864C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
865as C<L</is_strict_utf8_string>>; and if C<flags> is
866C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives the same results as
867C<L</is_c9strict_utf8_string>>. Otherwise C<flags> may be any
868combination of the C<UTF8_DISALLOW_I<foo>> flags understood by
869C<L</utf8n_to_uvchr>>, with the same meanings.
870
871See also
872C<L</is_utf8_invariant_string>>,
0cbf5865 873C<L</is_utf8_invariant_string_loc>>,
9f2abfde
KW
874C<L</is_utf8_string>>,
875C<L</is_utf8_string_loc>>,
876C<L</is_utf8_string_loc_flags>>,
877C<L</is_utf8_string_loclen>>,
878C<L</is_utf8_string_loclen_flags>>,
8bc127bf
KW
879C<L</is_utf8_fixed_width_buf_flags>>,
880C<L</is_utf8_fixed_width_buf_loc_flags>>,
881C<L</is_utf8_fixed_width_buf_loclen_flags>>,
9f2abfde
KW
882C<L</is_strict_utf8_string>>,
883C<L</is_strict_utf8_string_loc>>,
884C<L</is_strict_utf8_string_loclen>>,
885C<L</is_c9strict_utf8_string>>,
886C<L</is_c9strict_utf8_string_loc>>,
887and
888C<L</is_c9strict_utf8_string_loclen>>.
889
890=cut
891*/
892
893PERL_STATIC_INLINE bool
f60f61fd 894S_is_utf8_string_flags(const U8 *s, STRLEN len, const U32 flags)
9f2abfde 895{
33756530 896 const U8 * first_variant;
9f2abfde
KW
897
898 PERL_ARGS_ASSERT_IS_UTF8_STRING_FLAGS;
899 assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE
d044b7a7 900 |UTF8_DISALLOW_PERL_EXTENDED)));
9f2abfde 901
f60f61fd
KW
902 if (len == 0) {
903 len = strlen((const char *)s);
904 }
905
9f2abfde
KW
906 if (flags == 0) {
907 return is_utf8_string(s, len);
908 }
909
d044b7a7 910 if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED)
9f2abfde
KW
911 == UTF8_DISALLOW_ILLEGAL_INTERCHANGE)
912 {
913 return is_strict_utf8_string(s, len);
914 }
915
d044b7a7 916 if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED)
9f2abfde
KW
917 == UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE)
918 {
919 return is_c9strict_utf8_string(s, len);
920 }
921
33756530
KW
922 if (! is_utf8_invariant_string_loc(s, len, &first_variant)) {
923 const U8* const send = s + len;
924 const U8* x = first_variant;
925
a0d7f935
KW
926 while (x < send) {
927 STRLEN cur_len = isUTF8_CHAR_flags(x, send, flags);
928 if (UNLIKELY(! cur_len)) {
929 return FALSE;
930 }
931 x += cur_len;
9f2abfde 932 }
33756530 933 }
9f2abfde
KW
934
935 return TRUE;
936}
937
938/*
5ff889fb
KW
939
940=for apidoc is_utf8_string_loc
941
2717076a 942Like C<L</is_utf8_string>> but stores the location of the failure (in the
5ff889fb 943case of "utf8ness failure") or the location C<s>+C<len> (in the case of
82c5d941 944"utf8ness success") in the C<ep> pointer.
5ff889fb 945
2717076a 946See also C<L</is_utf8_string_loclen>>.
5ff889fb 947
3964c812
KW
948=cut
949*/
950
951#define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0)
952
953/*
954
5ff889fb
KW
955=for apidoc is_utf8_string_loclen
956
2717076a 957Like C<L</is_utf8_string>> but stores the location of the failure (in the
5ff889fb 958case of "utf8ness failure") or the location C<s>+C<len> (in the case of
9f2abfde 959"utf8ness success") in the C<ep> pointer, and the number of UTF-8
82c5d941 960encoded characters in the C<el> pointer.
5ff889fb 961
2717076a 962See also C<L</is_utf8_string_loc>>.
5ff889fb
KW
963
964=cut
965*/
966
56e4cf64 967PERL_STATIC_INLINE bool
33756530 968Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
5ff889fb 969{
33756530 970 const U8 * first_variant;
5ff889fb
KW
971
972 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
973
33756530
KW
974 if (len == 0) {
975 len = strlen((const char *) s);
976 }
977
978 if (is_utf8_invariant_string_loc(s, len, &first_variant)) {
979 if (el)
980 *el = len;
981
982 if (ep) {
983 *ep = s + len;
984 }
985
986 return TRUE;
987 }
988
989 {
990 const U8* const send = s + len;
991 const U8* x = first_variant;
992 STRLEN outlen = first_variant - s;
993
a0d7f935
KW
994 while (x < send) {
995 const STRLEN cur_len = isUTF8_CHAR(x, send);
996 if (UNLIKELY(! cur_len)) {
997 break;
998 }
999 x += cur_len;
1000 outlen++;
5ff889fb 1001 }
5ff889fb 1002
a0d7f935
KW
1003 if (el)
1004 *el = outlen;
5ff889fb 1005
a0d7f935
KW
1006 if (ep) {
1007 *ep = x;
1008 }
5ff889fb 1009
a0d7f935 1010 return (x == send);
33756530 1011 }
5ff889fb
KW
1012}
1013
1014/*
9f2abfde 1015
8ed185f9
KW
1016=for apidoc Am|STRLEN|isUTF8_CHAR|const U8 *s|const U8 *e
1017
1018Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1019looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1020that represents some code point; otherwise it evaluates to 0. If non-zero, the
1021value gives how many bytes starting at C<s> comprise the code point's
1022representation. Any bytes remaining before C<e>, but beyond the ones needed to
1023form the first code point in C<s>, are not examined.
1024
13aab5dd 1025The code point can be any that will fit in an IV on this machine, using Perl's
8ed185f9
KW
1026extension to official UTF-8 to represent those higher than the Unicode maximum
1027of 0x10FFFF. That means that this macro is used to efficiently decide if the
1028next few bytes in C<s> is legal UTF-8 for a single character.
1029
1030Use C<L</isSTRICT_UTF8_CHAR>> to restrict the acceptable code points to those
1031defined by Unicode to be fully interchangeable across applications;
1032C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1033#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1034code points; and C<L</isUTF8_CHAR_flags>> for a more customized definition.
1035
1036Use C<L</is_utf8_string>>, C<L</is_utf8_string_loc>>, and
1037C<L</is_utf8_string_loclen>> to check entire strings.
1038
13aab5dd
KW
1039Note also that a UTF-8 "invariant" character (i.e. ASCII on non-EBCDIC
1040machines) is a valid UTF-8 character.
8ed185f9
KW
1041
1042=cut
1043
1044This uses an adaptation of the table and algorithm given in
1045http://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides comprehensive
1046documentation of the original version. A copyright notice for the original
1047version is given at the beginning of this file. The Perl adapation is
1048documented at the definition of perl_extended_utf8_dfa_tab[].
1049
1050*/
1051
1052PERL_STATIC_INLINE Size_t
1053S_isUTF8_CHAR(const U8 * const s0, const U8 * const e)
1054{
1055 const U8 * s = s0;
1056 UV state = 0;
1057
1058 PERL_ARGS_ASSERT_ISUTF8_CHAR;
1059
1060 /* This dfa is fast. If it accepts the input, it was for a well-formed,
1061 * code point, which can be returned immediately. Otherwise, it is either
1062 * malformed, or for the start byte FF which the dfa doesn't handle (except
1063 * on 32-bit ASCII platforms where it trivially is an error). Call a
1064 * helper function for the other platforms. */
1065
1066 while (s < e && LIKELY(state != 1)) {
1067 state = perl_extended_utf8_dfa_tab[256
1068 + state
1069 + perl_extended_utf8_dfa_tab[*s]];
1070 if (state != 0) {
1071 s++;
1072 continue;
1073 }
1074
1075 return s - s0 + 1;
1076 }
1077
1078#if defined(UV_IS_QUAD) || defined(EBCDIC)
1079
1080 if (NATIVE_UTF8_TO_I8(*s0) == 0xFF && e - s0 >= UTF8_MAXBYTES) {
1081 return _is_utf8_char_helper(s0, e, 0);
1082 }
1083
1084#endif
1085
1086 return 0;
1087}
1088
1089/*
1090
67049a5f
KW
1091=for apidoc isSTRICT_UTF8_CHAR
1092
1093Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1094looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1095Unicode code point completely acceptable for open interchange between all
1096applications; otherwise it evaluates to 0. If non-zero, the value gives how
1097many bytes starting at C<s> comprise the code point's representation. Any
1098bytes remaining before C<e>, but beyond the ones needed to form the first code
1099point in C<s>, are not examined.
1100
1101The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not
1102be a surrogate nor a non-character code point. Thus this excludes any code
1103point from Perl's extended UTF-8.
1104
1105This is used to efficiently decide if the next few bytes in C<s> is
1106legal Unicode-acceptable UTF-8 for a single character.
1107
1108Use C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1109#9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1110code points; C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8;
1111and C<L</isUTF8_CHAR_flags>> for a more customized definition.
1112
1113Use C<L</is_strict_utf8_string>>, C<L</is_strict_utf8_string_loc>>, and
1114C<L</is_strict_utf8_string_loclen>> to check entire strings.
1115
1116=cut
1117
1118This uses an adaptation of the tables and algorithm given in
1119http://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides comprehensive
1120documentation of the original version. A copyright notice for the original
1121version is given at the beginning of this file. The Perl adapation is
1122documented at the definition of strict_extended_utf8_dfa_tab[].
1123
1124*/
1125
1126PERL_STATIC_INLINE Size_t
1127S_isSTRICT_UTF8_CHAR(const U8 * const s0, const U8 * const e)
1128{
1129 const U8 * s = s0;
1130 UV state = 0;
1131
1132 PERL_ARGS_ASSERT_ISSTRICT_UTF8_CHAR;
1133
1134 while (s < e && LIKELY(state != 1)) {
1135 state = strict_utf8_dfa_tab[256 + state + strict_utf8_dfa_tab[*s]];
1136
1137 if (state != 0) {
1138 s++;
1139 continue;
1140 }
1141
1142 return s - s0 + 1;
1143 }
1144
1145#ifndef EBCDIC
1146
1147 /* The dfa above drops out for certain Hanguls; handle them specially */
1148 if (is_HANGUL_ED_utf8_safe(s0, e)) {
1149 return 3;
1150 }
1151
1152#endif
1153
1154 return 0;
1155}
1156
1157/*
1158
9f2abfde
KW
1159=for apidoc is_strict_utf8_string_loc
1160
1161Like C<L</is_strict_utf8_string>> but stores the location of the failure (in the
1162case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1163"utf8ness success") in the C<ep> pointer.
1164
1165See also C<L</is_strict_utf8_string_loclen>>.
1166
1167=cut
1168*/
1169
1170#define is_strict_utf8_string_loc(s, len, ep) \
1171 is_strict_utf8_string_loclen(s, len, ep, 0)
1172
1173/*
1174
1175=for apidoc is_strict_utf8_string_loclen
1176
1177Like C<L</is_strict_utf8_string>> but stores the location of the failure (in the
1178case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1179"utf8ness success") in the C<ep> pointer, and the number of UTF-8
1180encoded characters in the C<el> pointer.
1181
1182See also C<L</is_strict_utf8_string_loc>>.
1183
1184=cut
1185*/
1186
1187PERL_STATIC_INLINE bool
33756530 1188S_is_strict_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
9f2abfde 1189{
33756530 1190 const U8 * first_variant;
9f2abfde
KW
1191
1192 PERL_ARGS_ASSERT_IS_STRICT_UTF8_STRING_LOCLEN;
1193
33756530
KW
1194 if (len == 0) {
1195 len = strlen((const char *) s);
1196 }
1197
1198 if (is_utf8_invariant_string_loc(s, len, &first_variant)) {
1199 if (el)
1200 *el = len;
1201
1202 if (ep) {
1203 *ep = s + len;
1204 }
1205
1206 return TRUE;
1207 }
1208
1209 {
1210 const U8* const send = s + len;
1211 const U8* x = first_variant;
1212 STRLEN outlen = first_variant - s;
1213
a0d7f935
KW
1214 while (x < send) {
1215 const STRLEN cur_len = isSTRICT_UTF8_CHAR(x, send);
1216 if (UNLIKELY(! cur_len)) {
1217 break;
1218 }
1219 x += cur_len;
1220 outlen++;
9f2abfde 1221 }
9f2abfde 1222
a0d7f935
KW
1223 if (el)
1224 *el = outlen;
9f2abfde 1225
a0d7f935
KW
1226 if (ep) {
1227 *ep = x;
1228 }
9f2abfde 1229
a0d7f935 1230 return (x == send);
33756530 1231 }
9f2abfde
KW
1232}
1233
1234/*
1235
1236=for apidoc is_c9strict_utf8_string_loc
1237
1238Like C<L</is_c9strict_utf8_string>> but stores the location of the failure (in
1239the case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1240"utf8ness success") in the C<ep> pointer.
1241
1242See also C<L</is_c9strict_utf8_string_loclen>>.
1243
1244=cut
1245*/
1246
1247#define is_c9strict_utf8_string_loc(s, len, ep) \
1248 is_c9strict_utf8_string_loclen(s, len, ep, 0)
1249
1250/*
1251
1252=for apidoc is_c9strict_utf8_string_loclen
1253
1254Like C<L</is_c9strict_utf8_string>> but stores the location of the failure (in
1255the case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1256"utf8ness success") in the C<ep> pointer, and the number of UTF-8 encoded
1257characters in the C<el> pointer.
1258
1259See also C<L</is_c9strict_utf8_string_loc>>.
1260
1261=cut
1262*/
1263
1264PERL_STATIC_INLINE bool
33756530 1265S_is_c9strict_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
9f2abfde 1266{
33756530 1267 const U8 * first_variant;
9f2abfde
KW
1268
1269 PERL_ARGS_ASSERT_IS_C9STRICT_UTF8_STRING_LOCLEN;
1270
33756530
KW
1271 if (len == 0) {
1272 len = strlen((const char *) s);
1273 }
1274
1275 if (is_utf8_invariant_string_loc(s, len, &first_variant)) {
1276 if (el)
1277 *el = len;
1278
1279 if (ep) {
1280 *ep = s + len;
1281 }
1282
1283 return TRUE;
1284 }
1285
1286 {
1287 const U8* const send = s + len;
1288 const U8* x = first_variant;
1289 STRLEN outlen = first_variant - s;
1290
a0d7f935
KW
1291 while (x < send) {
1292 const STRLEN cur_len = isC9_STRICT_UTF8_CHAR(x, send);
1293 if (UNLIKELY(! cur_len)) {
1294 break;
1295 }
1296 x += cur_len;
1297 outlen++;
9f2abfde 1298 }
9f2abfde 1299
a0d7f935
KW
1300 if (el)
1301 *el = outlen;
9f2abfde 1302
a0d7f935
KW
1303 if (ep) {
1304 *ep = x;
1305 }
9f2abfde 1306
a0d7f935 1307 return (x == send);
33756530 1308 }
9f2abfde
KW
1309}
1310
1311/*
1312
1313=for apidoc is_utf8_string_loc_flags
1314
1315Like C<L</is_utf8_string_flags>> but stores the location of the failure (in the
1316case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1317"utf8ness success") in the C<ep> pointer.
1318
1319See also C<L</is_utf8_string_loclen_flags>>.
1320
1321=cut
1322*/
1323
1324#define is_utf8_string_loc_flags(s, len, ep, flags) \
1325 is_utf8_string_loclen_flags(s, len, ep, 0, flags)
1326
1327
1328/* The above 3 actual functions could have been moved into the more general one
1329 * just below, and made #defines that call it with the right 'flags'. They are
1330 * currently kept separate to increase their chances of getting inlined */
1331
1332/*
1333
1334=for apidoc is_utf8_string_loclen_flags
1335
1336Like C<L</is_utf8_string_flags>> but stores the location of the failure (in the
1337case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1338"utf8ness success") in the C<ep> pointer, and the number of UTF-8
1339encoded characters in the C<el> pointer.
1340
1341See also C<L</is_utf8_string_loc_flags>>.
1342
1343=cut
1344*/
1345
1346PERL_STATIC_INLINE bool
f60f61fd 1347S_is_utf8_string_loclen_flags(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el, const U32 flags)
9f2abfde 1348{
33756530 1349 const U8 * first_variant;
9f2abfde
KW
1350
1351 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN_FLAGS;
1352 assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE
d044b7a7 1353 |UTF8_DISALLOW_PERL_EXTENDED)));
9f2abfde 1354
f60f61fd 1355 if (len == 0) {
a0d7f935 1356 len = strlen((const char *) s);
f60f61fd
KW
1357 }
1358
9f2abfde
KW
1359 if (flags == 0) {
1360 return is_utf8_string_loclen(s, len, ep, el);
1361 }
1362
d044b7a7 1363 if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED)
9f2abfde
KW
1364 == UTF8_DISALLOW_ILLEGAL_INTERCHANGE)
1365 {
1366 return is_strict_utf8_string_loclen(s, len, ep, el);
1367 }
1368
d044b7a7 1369 if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED)
9f2abfde
KW
1370 == UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE)
1371 {
1372 return is_c9strict_utf8_string_loclen(s, len, ep, el);
1373 }
1374
33756530
KW
1375 if (is_utf8_invariant_string_loc(s, len, &first_variant)) {
1376 if (el)
1377 *el = len;
1378
1379 if (ep) {
1380 *ep = s + len;
1381 }
1382
1383 return TRUE;
1384 }
1385
1386 {
1387 const U8* send = s + len;
1388 const U8* x = first_variant;
1389 STRLEN outlen = first_variant - s;
1390
a0d7f935
KW
1391 while (x < send) {
1392 const STRLEN cur_len = isUTF8_CHAR_flags(x, send, flags);
1393 if (UNLIKELY(! cur_len)) {
1394 break;
1395 }
1396 x += cur_len;
1397 outlen++;
9f2abfde 1398 }
9f2abfde 1399
a0d7f935
KW
1400 if (el)
1401 *el = outlen;
9f2abfde 1402
a0d7f935
KW
1403 if (ep) {
1404 *ep = x;
1405 }
9f2abfde 1406
a0d7f935 1407 return (x == send);
33756530 1408 }
9f2abfde
KW
1409}
1410
1411/*
7c93d8f0
KW
1412=for apidoc utf8_distance
1413
1414Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
1415and C<b>.
1416
1417WARNING: use only if you *know* that the pointers point inside the
1418same UTF-8 buffer.
1419
1420=cut
1421*/
1422
1423PERL_STATIC_INLINE IV
1424Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
1425{
1426 PERL_ARGS_ASSERT_UTF8_DISTANCE;
1427
1428 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
1429}
1430
1431/*
1432=for apidoc utf8_hop
1433
1434Return the UTF-8 pointer C<s> displaced by C<off> characters, either
1435forward or backward.
1436
1437WARNING: do not use the following unless you *know* C<off> is within
1438the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
1439on the first byte of character or just after the last byte of a character.
1440
1441=cut
1442*/
1443
1444PERL_STATIC_INLINE U8 *
1445Perl_utf8_hop(const U8 *s, SSize_t off)
1446{
1447 PERL_ARGS_ASSERT_UTF8_HOP;
1448
1449 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1450 * the bitops (especially ~) can create illegal UTF-8.
1451 * In other words: in Perl UTF-8 is not just for Unicode. */
1452
1453 if (off >= 0) {
1454 while (off--)
1455 s += UTF8SKIP(s);
1456 }
1457 else {
1458 while (off++) {
1459 s--;
1460 while (UTF8_IS_CONTINUATION(*s))
1461 s--;
1462 }
1463 }
e099ea69 1464 GCC_DIAG_IGNORE(-Wcast-qual)
7c93d8f0 1465 return (U8 *)s;
e099ea69 1466 GCC_DIAG_RESTORE
7c93d8f0
KW
1467}
1468
4dab108f 1469/*
65df57a8
TC
1470=for apidoc utf8_hop_forward
1471
1472Return the UTF-8 pointer C<s> displaced by up to C<off> characters,
1473forward.
1474
1475C<off> must be non-negative.
1476
1477C<s> must be before or equal to C<end>.
1478
1479When moving forward it will not move beyond C<end>.
1480
1481Will not exceed this limit even if the string is not valid "UTF-8".
1482
1483=cut
1484*/
1485
1486PERL_STATIC_INLINE U8 *
1487Perl_utf8_hop_forward(const U8 *s, SSize_t off, const U8 *end)
1488{
1489 PERL_ARGS_ASSERT_UTF8_HOP_FORWARD;
1490
1491 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1492 * the bitops (especially ~) can create illegal UTF-8.
1493 * In other words: in Perl UTF-8 is not just for Unicode. */
1494
1495 assert(s <= end);
1496 assert(off >= 0);
1497
1498 while (off--) {
1499 STRLEN skip = UTF8SKIP(s);
de979548 1500 if ((STRLEN)(end - s) <= skip) {
e099ea69 1501 GCC_DIAG_IGNORE(-Wcast-qual)
65df57a8 1502 return (U8 *)end;
e099ea69 1503 GCC_DIAG_RESTORE
de979548 1504 }
65df57a8
TC
1505 s += skip;
1506 }
1507
e099ea69 1508 GCC_DIAG_IGNORE(-Wcast-qual)
65df57a8 1509 return (U8 *)s;
e099ea69 1510 GCC_DIAG_RESTORE
65df57a8
TC
1511}
1512
1513/*
1514=for apidoc utf8_hop_back
1515
1516Return the UTF-8 pointer C<s> displaced by up to C<off> characters,
1517backward.
1518
1519C<off> must be non-positive.
1520
1521C<s> must be after or equal to C<start>.
1522
1523When moving backward it will not move before C<start>.
1524
1525Will not exceed this limit even if the string is not valid "UTF-8".
1526
1527=cut
1528*/
1529
1530PERL_STATIC_INLINE U8 *
1531Perl_utf8_hop_back(const U8 *s, SSize_t off, const U8 *start)
1532{
1533 PERL_ARGS_ASSERT_UTF8_HOP_BACK;
1534
1535 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1536 * the bitops (especially ~) can create illegal UTF-8.
1537 * In other words: in Perl UTF-8 is not just for Unicode. */
1538
1539 assert(start <= s);
1540 assert(off <= 0);
1541
1542 while (off++ && s > start) {
1543 s--;
1544 while (UTF8_IS_CONTINUATION(*s) && s > start)
1545 s--;
1546 }
1547
e099ea69 1548 GCC_DIAG_IGNORE(-Wcast-qual)
65df57a8 1549 return (U8 *)s;
e099ea69 1550 GCC_DIAG_RESTORE
65df57a8
TC
1551}
1552
1553/*
1554=for apidoc utf8_hop_safe
1555
1556Return the UTF-8 pointer C<s> displaced by up to C<off> characters,
1557either forward or backward.
1558
1559When moving backward it will not move before C<start>.
1560
1561When moving forward it will not move beyond C<end>.
1562
1563Will not exceed those limits even if the string is not valid "UTF-8".
1564
1565=cut
1566*/
1567
1568PERL_STATIC_INLINE U8 *
1569Perl_utf8_hop_safe(const U8 *s, SSize_t off, const U8 *start, const U8 *end)
1570{
1571 PERL_ARGS_ASSERT_UTF8_HOP_SAFE;
1572
1573 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1574 * the bitops (especially ~) can create illegal UTF-8.
1575 * In other words: in Perl UTF-8 is not just for Unicode. */
1576
1577 assert(start <= s && s <= end);
1578
1579 if (off >= 0) {
1580 return utf8_hop_forward(s, off, end);
1581 }
1582 else {
1583 return utf8_hop_back(s, off, start);
1584 }
1585}
1586
1587/*
4dab108f
KW
1588
1589=for apidoc is_utf8_valid_partial_char
1590
6cbb9248
KW
1591Returns 0 if the sequence of bytes starting at C<s> and looking no further than
1592S<C<e - 1>> is the UTF-8 encoding, as extended by Perl, for one or more code
1593points. Otherwise, it returns 1 if there exists at least one non-empty
1594sequence of bytes that when appended to sequence C<s>, starting at position
1595C<e> causes the entire sequence to be the well-formed UTF-8 of some code point;
1596otherwise returns 0.
1597
1598In other words this returns TRUE if C<s> points to a partial UTF-8-encoded code
1599point.
1600
1601This is useful when a fixed-length buffer is being tested for being well-formed
1602UTF-8, but the final few bytes in it don't comprise a full character; that is,
1603it is split somewhere in the middle of the final code point's UTF-8
1604representation. (Presumably when the buffer is refreshed with the next chunk
1605of data, the new first bytes will complete the partial code point.) This
1606function is used to verify that the final bytes in the current buffer are in
1607fact the legal beginning of some code point, so that if they aren't, the
1608failure can be signalled without having to wait for the next read.
4dab108f
KW
1609
1610=cut
1611*/
2717076a
KW
1612#define is_utf8_valid_partial_char(s, e) \
1613 is_utf8_valid_partial_char_flags(s, e, 0)
f1c999a7
KW
1614
1615/*
1616
1617=for apidoc is_utf8_valid_partial_char_flags
1618
1619Like C<L</is_utf8_valid_partial_char>>, it returns a boolean giving whether
1620or not the input is a valid UTF-8 encoded partial character, but it takes an
1621extra parameter, C<flags>, which can further restrict which code points are
1622considered valid.
1623
1624If C<flags> is 0, this behaves identically to
1625C<L</is_utf8_valid_partial_char>>. Otherwise C<flags> can be any combination
1626of the C<UTF8_DISALLOW_I<foo>> flags accepted by C<L</utf8n_to_uvchr>>. If
1627there is any sequence of bytes that can complete the input partial character in
1628such a way that a non-prohibited character is formed, the function returns
2717076a
KW
1629TRUE; otherwise FALSE. Non character code points cannot be determined based on
1630partial character input. But many of the other possible excluded types can be
f1c999a7
KW
1631determined from just the first one or two bytes.
1632
1633=cut
1634 */
1635
56e4cf64 1636PERL_STATIC_INLINE bool
f1c999a7 1637S_is_utf8_valid_partial_char_flags(const U8 * const s, const U8 * const e, const U32 flags)
4dab108f 1638{
f1c999a7 1639 PERL_ARGS_ASSERT_IS_UTF8_VALID_PARTIAL_CHAR_FLAGS;
4dab108f 1640
f1c999a7 1641 assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE
d044b7a7 1642 |UTF8_DISALLOW_PERL_EXTENDED)));
4dab108f 1643
8875bd48 1644 if (s >= e || s + UTF8SKIP(s) <= e) {
4dab108f
KW
1645 return FALSE;
1646 }
1647
f1c999a7 1648 return cBOOL(_is_utf8_char_helper(s, e, flags));
4dab108f
KW
1649}
1650
8bc127bf
KW
1651/*
1652
1653=for apidoc is_utf8_fixed_width_buf_flags
1654
1655Returns TRUE if the fixed-width buffer starting at C<s> with length C<len>
1656is entirely valid UTF-8, subject to the restrictions given by C<flags>;
1657otherwise it returns FALSE.
1658
1659If C<flags> is 0, any well-formed UTF-8, as extended by Perl, is accepted
1660without restriction. If the final few bytes of the buffer do not form a
1661complete code point, this will return TRUE anyway, provided that
1662C<L</is_utf8_valid_partial_char_flags>> returns TRUE for them.
1663
1664If C<flags> in non-zero, it can be any combination of the
1665C<UTF8_DISALLOW_I<foo>> flags accepted by C<L</utf8n_to_uvchr>>, and with the
1666same meanings.
1667
1668This function differs from C<L</is_utf8_string_flags>> only in that the latter
1669returns FALSE if the final few bytes of the string don't form a complete code
1670point.
1671
1672=cut
1673 */
1674#define is_utf8_fixed_width_buf_flags(s, len, flags) \
1675 is_utf8_fixed_width_buf_loclen_flags(s, len, 0, 0, flags)
1676
1677/*
1678
1679=for apidoc is_utf8_fixed_width_buf_loc_flags
1680
1681Like C<L</is_utf8_fixed_width_buf_flags>> but stores the location of the
1682failure in the C<ep> pointer. If the function returns TRUE, C<*ep> will point
1683to the beginning of any partial character at the end of the buffer; if there is
1684no partial character C<*ep> will contain C<s>+C<len>.
1685
1686See also C<L</is_utf8_fixed_width_buf_loclen_flags>>.
1687
1688=cut
1689*/
1690
1691#define is_utf8_fixed_width_buf_loc_flags(s, len, loc, flags) \
1692 is_utf8_fixed_width_buf_loclen_flags(s, len, loc, 0, flags)
1693
1694/*
1695
1696=for apidoc is_utf8_fixed_width_buf_loclen_flags
1697
1698Like C<L</is_utf8_fixed_width_buf_loc_flags>> but stores the number of
1699complete, valid characters found in the C<el> pointer.
1700
1701=cut
1702*/
1703
1704PERL_STATIC_INLINE bool
1705S_is_utf8_fixed_width_buf_loclen_flags(const U8 * const s,
33756530 1706 STRLEN len,
8bc127bf
KW
1707 const U8 **ep,
1708 STRLEN *el,
1709 const U32 flags)
1710{
1711 const U8 * maybe_partial;
1712
1713 PERL_ARGS_ASSERT_IS_UTF8_FIXED_WIDTH_BUF_LOCLEN_FLAGS;
1714
1715 if (! ep) {
1716 ep = &maybe_partial;
1717 }
1718
1719 /* If it's entirely valid, return that; otherwise see if the only error is
1720 * that the final few bytes are for a partial character */
1721 return is_utf8_string_loclen_flags(s, len, ep, el, flags)
1722 || is_utf8_valid_partial_char_flags(*ep, s + len, flags);
1723}
1724
e6a4ffc3
KW
1725PERL_STATIC_INLINE UV
1726S_utf8n_to_uvchr_msgs(const U8 *s,
1727 STRLEN curlen,
1728 STRLEN *retlen,
1729 const U32 flags,
1730 U32 * errors,
1731 AV ** msgs)
1732{
1733 /* This is the inlined portion of utf8n_to_uvchr_msgs. It handles the
1734 * simple cases, and, if necessary calls a helper function to deal with the
1735 * more complex ones. Almost all well-formed non-problematic code points
1736 * are considered simple, so that it's unlikely that the helper function
1737 * will need to be called.
1738 *
1739 * This is an adaptation of the tables and algorithm given in
1740 * http://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides
1741 * comprehensive documentation of the original version. A copyright notice
1742 * for the original version is given at the beginning of this file. The
1743 * Perl adapation is documented at the definition of strict_utf8_dfa_tab[].
1744 */
1745
1746 const U8 * const s0 = s;
1747 const U8 * send = s0 + curlen;
1748 UV uv;
1749 UV state = 0;
1750
1751 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_MSGS;
1752
1753 /* This dfa is fast. If it accepts the input, it was for a well-formed,
1754 * non-problematic code point, which can be returned immediately.
1755 * Otherwise we call a helper function to figure out the more complicated
1756 * cases. */
1757
1758 while (s < send && LIKELY(state != 1)) {
1759 UV type = strict_utf8_dfa_tab[*s];
1760
1761 uv = (state == 0)
1762 ? ((0xff >> type) & NATIVE_UTF8_TO_I8(*s))
1763 : UTF8_ACCUMULATE(uv, *s);
1764 state = strict_utf8_dfa_tab[256 + state + type];
1765
1766 if (state != 0) {
1767 s++;
1768 continue;
1769 }
1770
1771 if (retlen) {
1772 *retlen = s - s0 + 1;
1773 }
1774 if (errors) {
1775 *errors = 0;
1776 }
1777 if (msgs) {
1778 *msgs = NULL;
1779 }
1780
1781 return uv;
1782 }
1783
1784 /* Here is potentially problematic. Use the full mechanism */
1785 return _utf8n_to_uvchr_msgs_helper(s0, curlen, retlen, flags, errors, msgs);
1786}
1787
c8028aa6
TC
1788/* ------------------------------- perl.h ----------------------------- */
1789
1790/*
dcccc8ff
KW
1791=head1 Miscellaneous Functions
1792
41188aa0 1793=for apidoc AiR|bool|is_safe_syscall|const char *pv|STRLEN len|const char *what|const char *op_name
c8028aa6 1794
6602b933 1795Test that the given C<pv> doesn't contain any internal C<NUL> characters.
796b6530 1796If it does, set C<errno> to C<ENOENT>, optionally warn, and return FALSE.
c8028aa6
TC
1797
1798Return TRUE if the name is safe.
1799
796b6530 1800Used by the C<IS_SAFE_SYSCALL()> macro.
c8028aa6
TC
1801
1802=cut
1803*/
1804
1805PERL_STATIC_INLINE bool
41188aa0 1806S_is_safe_syscall(pTHX_ const char *pv, STRLEN len, const char *what, const char *op_name) {
c8028aa6
TC
1807 /* While the Windows CE API provides only UCS-16 (or UTF-16) APIs
1808 * perl itself uses xce*() functions which accept 8-bit strings.
1809 */
1810
1811 PERL_ARGS_ASSERT_IS_SAFE_SYSCALL;
1812
6c4650b3 1813 if (len > 1) {
c8028aa6 1814 char *null_at;
41188aa0 1815 if (UNLIKELY((null_at = (char *)memchr(pv, 0, len-1)) != NULL)) {
c8028aa6 1816 SETERRNO(ENOENT, LIB_INVARG);
1d505182 1817 Perl_ck_warner(aTHX_ packWARN(WARN_SYSCALLS),
c8028aa6 1818 "Invalid \\0 character in %s for %s: %s\\0%s",
41188aa0 1819 what, op_name, pv, null_at+1);
c8028aa6
TC
1820 return FALSE;
1821 }
1822 }
1823
1824 return TRUE;
1825}
1826
1827/*
7cb3f959
TC
1828
1829Return true if the supplied filename has a newline character
fa6c7d00 1830immediately before the first (hopefully only) NUL.
7cb3f959
TC
1831
1832My original look at this incorrectly used the len from SvPV(), but
1833that's incorrect, since we allow for a NUL in pv[len-1].
1834
1835So instead, strlen() and work from there.
1836
1837This allow for the user reading a filename, forgetting to chomp it,
1838then calling:
1839
1840 open my $foo, "$file\0";
1841
1842*/
1843
1844#ifdef PERL_CORE
1845
1846PERL_STATIC_INLINE bool
1847S_should_warn_nl(const char *pv) {
1848 STRLEN len;
1849
1850 PERL_ARGS_ASSERT_SHOULD_WARN_NL;
1851
1852 len = strlen(pv);
1853
1854 return len > 0 && pv[len-1] == '\n';
1855}
1856
1857#endif
1858
81d52ecd
JH
1859/* ------------------ pp.c, regcomp.c, toke.c, universal.c ------------ */
1860
1861#define MAX_CHARSET_NAME_LENGTH 2
1862
1863PERL_STATIC_INLINE const char *
1864get_regex_charset_name(const U32 flags, STRLEN* const lenp)
1865{
1866 /* Returns a string that corresponds to the name of the regex character set
1867 * given by 'flags', and *lenp is set the length of that string, which
1868 * cannot exceed MAX_CHARSET_NAME_LENGTH characters */
1869
1870 *lenp = 1;
1871 switch (get_regex_charset(flags)) {
1872 case REGEX_DEPENDS_CHARSET: return DEPENDS_PAT_MODS;
1873 case REGEX_LOCALE_CHARSET: return LOCALE_PAT_MODS;
1874 case REGEX_UNICODE_CHARSET: return UNICODE_PAT_MODS;
1875 case REGEX_ASCII_RESTRICTED_CHARSET: return ASCII_RESTRICT_PAT_MODS;
1876 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
1877 *lenp = 2;
1878 return ASCII_MORE_RESTRICT_PAT_MODS;
1879 }
1880 /* The NOT_REACHED; hides an assert() which has a rather complex
1881 * definition in perl.h. */
1882 NOT_REACHED; /* NOTREACHED */
1883 return "?"; /* Unknown */
1884}
1885
7cb3f959 1886/*
ed382232
TC
1887
1888Return false if any get magic is on the SV other than taint magic.
1889
1890*/
1891
1892PERL_STATIC_INLINE bool
1893S_sv_only_taint_gmagic(SV *sv) {
1894 MAGIC *mg = SvMAGIC(sv);
1895
1896 PERL_ARGS_ASSERT_SV_ONLY_TAINT_GMAGIC;
1897
1898 while (mg) {
1899 if (mg->mg_type != PERL_MAGIC_taint
1900 && !(mg->mg_flags & MGf_GSKIP)
1901 && mg->mg_virtual->svt_get) {
1902 return FALSE;
1903 }
1904 mg = mg->mg_moremagic;
1905 }
1906
1907 return TRUE;
1908}
1909
ed8ff0f3
DM
1910/* ------------------ cop.h ------------------------------------------- */
1911
1912
1913/* Enter a block. Push a new base context and return its address. */
1914
1915PERL_STATIC_INLINE PERL_CONTEXT *
1916S_cx_pushblock(pTHX_ U8 type, U8 gimme, SV** sp, I32 saveix)
1917{
1918 PERL_CONTEXT * cx;
1919
1920 PERL_ARGS_ASSERT_CX_PUSHBLOCK;
1921
1922 CXINC;
1923 cx = CX_CUR();
1924 cx->cx_type = type;
1925 cx->blk_gimme = gimme;
1926 cx->blk_oldsaveix = saveix;
4caf7d8c 1927 cx->blk_oldsp = (I32)(sp - PL_stack_base);
ed8ff0f3 1928 cx->blk_oldcop = PL_curcop;
4caf7d8c 1929 cx->blk_oldmarksp = (I32)(PL_markstack_ptr - PL_markstack);
ed8ff0f3
DM
1930 cx->blk_oldscopesp = PL_scopestack_ix;
1931 cx->blk_oldpm = PL_curpm;
ce8bb8d8 1932 cx->blk_old_tmpsfloor = PL_tmps_floor;
ed8ff0f3
DM
1933
1934 PL_tmps_floor = PL_tmps_ix;
1935 CX_DEBUG(cx, "PUSH");
1936 return cx;
1937}
1938
1939
1940/* Exit a block (RETURN and LAST). */
1941
1942PERL_STATIC_INLINE void
1943S_cx_popblock(pTHX_ PERL_CONTEXT *cx)
1944{
1945 PERL_ARGS_ASSERT_CX_POPBLOCK;
1946
1947 CX_DEBUG(cx, "POP");
1948 /* these 3 are common to cx_popblock and cx_topblock */
1949 PL_markstack_ptr = PL_markstack + cx->blk_oldmarksp;
1950 PL_scopestack_ix = cx->blk_oldscopesp;
1951 PL_curpm = cx->blk_oldpm;
1952
1953 /* LEAVE_SCOPE() should have made this true. /(?{})/ cheats
1954 * and leaves a CX entry lying around for repeated use, so
1955 * skip for multicall */ \
1956 assert( (CxTYPE(cx) == CXt_SUB && CxMULTICALL(cx))
1957 || PL_savestack_ix == cx->blk_oldsaveix);
1958 PL_curcop = cx->blk_oldcop;
ce8bb8d8 1959 PL_tmps_floor = cx->blk_old_tmpsfloor;
ed8ff0f3
DM
1960}
1961
1962/* Continue a block elsewhere (e.g. NEXT, REDO, GOTO).
1963 * Whereas cx_popblock() restores the state to the point just before
1964 * cx_pushblock() was called, cx_topblock() restores it to the point just
1965 * *after* cx_pushblock() was called. */
1966
1967PERL_STATIC_INLINE void
1968S_cx_topblock(pTHX_ PERL_CONTEXT *cx)
1969{
1970 PERL_ARGS_ASSERT_CX_TOPBLOCK;
1971
1972 CX_DEBUG(cx, "TOP");
1973 /* these 3 are common to cx_popblock and cx_topblock */
1974 PL_markstack_ptr = PL_markstack + cx->blk_oldmarksp;
1975 PL_scopestack_ix = cx->blk_oldscopesp;
1976 PL_curpm = cx->blk_oldpm;
1977
1978 PL_stack_sp = PL_stack_base + cx->blk_oldsp;
1979}
1980
1981
a73d8813
DM
1982PERL_STATIC_INLINE void
1983S_cx_pushsub(pTHX_ PERL_CONTEXT *cx, CV *cv, OP *retop, bool hasargs)
1984{
1985 U8 phlags = CX_PUSHSUB_GET_LVALUE_MASK(Perl_was_lvalue_sub);
1986
1987 PERL_ARGS_ASSERT_CX_PUSHSUB;
1988
3f6bd23a 1989 PERL_DTRACE_PROBE_ENTRY(cv);
a73d8813
DM
1990 cx->blk_sub.cv = cv;
1991 cx->blk_sub.olddepth = CvDEPTH(cv);
1992 cx->blk_sub.prevcomppad = PL_comppad;
1993 cx->cx_type |= (hasargs) ? CXp_HASARGS : 0;
1994 cx->blk_sub.retop = retop;
1995 SvREFCNT_inc_simple_void_NN(cv);
1996 cx->blk_u16 = PL_op->op_private & (phlags|OPpDEREF);
1997}
1998
1999
2000/* subsets of cx_popsub() */
2001
2002PERL_STATIC_INLINE void
2003S_cx_popsub_common(pTHX_ PERL_CONTEXT *cx)
2004{
2005 CV *cv;
2006
2007 PERL_ARGS_ASSERT_CX_POPSUB_COMMON;
2008 assert(CxTYPE(cx) == CXt_SUB);
2009
2010 PL_comppad = cx->blk_sub.prevcomppad;
2011 PL_curpad = LIKELY(PL_comppad) ? AvARRAY(PL_comppad) : NULL;
2012 cv = cx->blk_sub.cv;
2013 CvDEPTH(cv) = cx->blk_sub.olddepth;
2014 cx->blk_sub.cv = NULL;
2015 SvREFCNT_dec(cv);
2016}
2017
2018
2019/* handle the @_ part of leaving a sub */
2020
2021PERL_STATIC_INLINE void
2022S_cx_popsub_args(pTHX_ PERL_CONTEXT *cx)
2023{
2024 AV *av;
2025
2026 PERL_ARGS_ASSERT_CX_POPSUB_ARGS;
2027 assert(CxTYPE(cx) == CXt_SUB);
2028 assert(AvARRAY(MUTABLE_AV(
2029 PadlistARRAY(CvPADLIST(cx->blk_sub.cv))[
2030 CvDEPTH(cx->blk_sub.cv)])) == PL_curpad);
2031
2032 CX_POP_SAVEARRAY(cx);
2033 av = MUTABLE_AV(PAD_SVl(0));
2034 if (UNLIKELY(AvREAL(av)))
2035 /* abandon @_ if it got reified */
2036 clear_defarray(av, 0);
2037 else {
2038 CLEAR_ARGARRAY(av);
2039 }
2040}
2041
2042
2043PERL_STATIC_INLINE void
2044S_cx_popsub(pTHX_ PERL_CONTEXT *cx)
2045{
2046 PERL_ARGS_ASSERT_CX_POPSUB;
2047 assert(CxTYPE(cx) == CXt_SUB);
2048
3f6bd23a 2049 PERL_DTRACE_PROBE_RETURN(cx->blk_sub.cv);
a73d8813
DM
2050
2051 if (CxHASARGS(cx))
2052 cx_popsub_args(cx);
2053 cx_popsub_common(cx);
2054}
2055
2056
6a7d52cc
DM
2057PERL_STATIC_INLINE void
2058S_cx_pushformat(pTHX_ PERL_CONTEXT *cx, CV *cv, OP *retop, GV *gv)
2059{
2060 PERL_ARGS_ASSERT_CX_PUSHFORMAT;
2061
2062 cx->blk_format.cv = cv;
2063 cx->blk_format.retop = retop;
2064 cx->blk_format.gv = gv;
2065 cx->blk_format.dfoutgv = PL_defoutgv;
2066 cx->blk_format.prevcomppad = PL_comppad;
2067 cx->blk_u16 = 0;
2068
2069 SvREFCNT_inc_simple_void_NN(cv);
2070 CvDEPTH(cv)++;
2071 SvREFCNT_inc_void(cx->blk_format.dfoutgv);
2072}
2073
2074
2075PERL_STATIC_INLINE void
2076S_cx_popformat(pTHX_ PERL_CONTEXT *cx)
2077{
2078 CV *cv;
2079 GV *dfout;
2080
2081 PERL_ARGS_ASSERT_CX_POPFORMAT;
2082 assert(CxTYPE(cx) == CXt_FORMAT);
2083
2084 dfout = cx->blk_format.dfoutgv;
2085 setdefout(dfout);
2086 cx->blk_format.dfoutgv = NULL;
2087 SvREFCNT_dec_NN(dfout);
2088
2089 PL_comppad = cx->blk_format.prevcomppad;
2090 PL_curpad = LIKELY(PL_comppad) ? AvARRAY(PL_comppad) : NULL;
2091 cv = cx->blk_format.cv;
2092 cx->blk_format.cv = NULL;
2093 --CvDEPTH(cv);
2094 SvREFCNT_dec_NN(cv);
2095}
2096
2097
13febba5
DM
2098PERL_STATIC_INLINE void
2099S_cx_pusheval(pTHX_ PERL_CONTEXT *cx, OP *retop, SV *namesv)
2100{
2101 PERL_ARGS_ASSERT_CX_PUSHEVAL;
2102
2103 cx->blk_eval.retop = retop;
2104 cx->blk_eval.old_namesv = namesv;
2105 cx->blk_eval.old_eval_root = PL_eval_root;
2106 cx->blk_eval.cur_text = PL_parser ? PL_parser->linestr : NULL;
2107 cx->blk_eval.cv = NULL; /* later set by doeval_compile() */
2108 cx->blk_eval.cur_top_env = PL_top_env;
2109
4c57ced5 2110 assert(!(PL_in_eval & ~ 0x3F));
13febba5 2111 assert(!(PL_op->op_type & ~0x1FF));
4c57ced5 2112 cx->blk_u16 = (PL_in_eval & 0x3F) | ((U16)PL_op->op_type << 7);
13febba5
DM
2113}
2114
2115
2116PERL_STATIC_INLINE void
2117S_cx_popeval(pTHX_ PERL_CONTEXT *cx)
2118{
2119 SV *sv;
2120
2121 PERL_ARGS_ASSERT_CX_POPEVAL;
2122 assert(CxTYPE(cx) == CXt_EVAL);
2123
2124 PL_in_eval = CxOLD_IN_EVAL(cx);
4c57ced5 2125 assert(!(PL_in_eval & 0xc0));
13febba5
DM
2126 PL_eval_root = cx->blk_eval.old_eval_root;
2127 sv = cx->blk_eval.cur_text;
4c57ced5 2128 if (sv && CxEVAL_TXT_REFCNTED(cx)) {
13febba5
DM
2129 cx->blk_eval.cur_text = NULL;
2130 SvREFCNT_dec_NN(sv);
2131 }
2132
2133 sv = cx->blk_eval.old_namesv;
2a1e0dfe
DM
2134 if (sv) {
2135 cx->blk_eval.old_namesv = NULL;
2136 SvREFCNT_dec_NN(sv);
2137 }
13febba5 2138}
6a7d52cc 2139
a73d8813 2140
d1b6bf72
DM
2141/* push a plain loop, i.e.
2142 * { block }
2143 * while (cond) { block }
2144 * for (init;cond;continue) { block }
2145 * This loop can be last/redo'ed etc.
2146 */
2147
2148PERL_STATIC_INLINE void
2149S_cx_pushloop_plain(pTHX_ PERL_CONTEXT *cx)
2150{
2151 PERL_ARGS_ASSERT_CX_PUSHLOOP_PLAIN;
2152 cx->blk_loop.my_op = cLOOP;
2153}
2154
2155
2156/* push a true for loop, i.e.
2157 * for var (list) { block }
2158 */
2159
2160PERL_STATIC_INLINE void
2161S_cx_pushloop_for(pTHX_ PERL_CONTEXT *cx, void *itervarp, SV* itersave)
2162{
2163 PERL_ARGS_ASSERT_CX_PUSHLOOP_FOR;
2164
2165 /* this one line is common with cx_pushloop_plain */
2166 cx->blk_loop.my_op = cLOOP;
2167
2168 cx->blk_loop.itervar_u.svp = (SV**)itervarp;
2169 cx->blk_loop.itersave = itersave;
2170#ifdef USE_ITHREADS
2171 cx->blk_loop.oldcomppad = PL_comppad;
2172#endif
2173}
2174
2175
2176/* pop all loop types, including plain */
2177
2178PERL_STATIC_INLINE void
2179S_cx_poploop(pTHX_ PERL_CONTEXT *cx)
2180{
2181 PERL_ARGS_ASSERT_CX_POPLOOP;
2182
2183 assert(CxTYPE_is_LOOP(cx));
2184 if ( CxTYPE(cx) == CXt_LOOP_ARY
2185 || CxTYPE(cx) == CXt_LOOP_LAZYSV)
2186 {
2187 /* Free ary or cur. This assumes that state_u.ary.ary
2188 * aligns with state_u.lazysv.cur. See cx_dup() */
2189 SV *sv = cx->blk_loop.state_u.lazysv.cur;
2190 cx->blk_loop.state_u.lazysv.cur = NULL;
2191 SvREFCNT_dec_NN(sv);
2192 if (CxTYPE(cx) == CXt_LOOP_LAZYSV) {
2193 sv = cx->blk_loop.state_u.lazysv.end;
2194 cx->blk_loop.state_u.lazysv.end = NULL;
2195 SvREFCNT_dec_NN(sv);
2196 }
2197 }
2198 if (cx->cx_type & (CXp_FOR_PAD|CXp_FOR_GV)) {
2199 SV *cursv;
2200 SV **svp = (cx)->blk_loop.itervar_u.svp;
2201 if ((cx->cx_type & CXp_FOR_GV))
2202 svp = &GvSV((GV*)svp);
2203 cursv = *svp;
2204 *svp = cx->blk_loop.itersave;
2205 cx->blk_loop.itersave = NULL;
2206 SvREFCNT_dec(cursv);
2207 }
2208}
2209
2a7b7c61
DM
2210
2211PERL_STATIC_INLINE void
7896dde7 2212S_cx_pushwhen(pTHX_ PERL_CONTEXT *cx)
2a7b7c61 2213{
7896dde7 2214 PERL_ARGS_ASSERT_CX_PUSHWHEN;
2a7b7c61 2215
7896dde7 2216 cx->blk_givwhen.leave_op = cLOGOP->op_other;
2a7b7c61
DM
2217}
2218
2219
2220PERL_STATIC_INLINE void
7896dde7 2221S_cx_popwhen(pTHX_ PERL_CONTEXT *cx)
2a7b7c61 2222{
7896dde7
Z
2223 PERL_ARGS_ASSERT_CX_POPWHEN;
2224 assert(CxTYPE(cx) == CXt_WHEN);
2a7b7c61
DM
2225
2226 PERL_UNUSED_ARG(cx);
59a14f30 2227 PERL_UNUSED_CONTEXT;
2a7b7c61
DM
2228 /* currently NOOP */
2229}
2230
2231
7896dde7
Z
2232PERL_STATIC_INLINE void
2233S_cx_pushgiven(pTHX_ PERL_CONTEXT *cx, SV *orig_defsv)
2234{
2235 PERL_ARGS_ASSERT_CX_PUSHGIVEN;
2236
2237 cx->blk_givwhen.leave_op = cLOGOP->op_other;
2238 cx->blk_givwhen.defsv_save = orig_defsv;
2239}
2240
2241
2242PERL_STATIC_INLINE void
2243S_cx_popgiven(pTHX_ PERL_CONTEXT *cx)
2244{
2245 SV *sv;
2246
2247 PERL_ARGS_ASSERT_CX_POPGIVEN;
2248 assert(CxTYPE(cx) == CXt_GIVEN);
2249
2250 sv = GvSV(PL_defgv);
2251 GvSV(PL_defgv) = cx->blk_givwhen.defsv_save;
2252 cx->blk_givwhen.defsv_save = NULL;
2253 SvREFCNT_dec(sv);
2254}
2255
ec2c235b
KW
2256/* ------------------ util.h ------------------------------------------- */
2257
2258/*
2259=head1 Miscellaneous Functions
2260
2261=for apidoc foldEQ
2262
2263Returns true if the leading C<len> bytes of the strings C<s1> and C<s2> are the
2264same
2265case-insensitively; false otherwise. Uppercase and lowercase ASCII range bytes
2266match themselves and their opposite case counterparts. Non-cased and non-ASCII
2267range bytes match only themselves.
2268
2269=cut
2270*/
2271
2272PERL_STATIC_INLINE I32
2273Perl_foldEQ(const char *s1, const char *s2, I32 len)
2274{
2275 const U8 *a = (const U8 *)s1;
2276 const U8 *b = (const U8 *)s2;
2277
2278 PERL_ARGS_ASSERT_FOLDEQ;
2279
2280 assert(len >= 0);
2281
2282 while (len--) {
2283 if (*a != *b && *a != PL_fold[*b])
2284 return 0;
2285 a++,b++;
2286 }
2287 return 1;
2288}
2289
0f9cb40c 2290PERL_STATIC_INLINE I32
ec2c235b
KW
2291Perl_foldEQ_latin1(const char *s1, const char *s2, I32 len)
2292{
2293 /* Compare non-utf8 using Unicode (Latin1) semantics. Does not work on
2294 * MICRO_SIGN, LATIN_SMALL_LETTER_SHARP_S, nor
2295 * LATIN_SMALL_LETTER_Y_WITH_DIAERESIS, and does not check for these. Nor
2296 * does it check that the strings each have at least 'len' characters */
2297
2298 const U8 *a = (const U8 *)s1;
2299 const U8 *b = (const U8 *)s2;
2300
2301 PERL_ARGS_ASSERT_FOLDEQ_LATIN1;
2302
2303 assert(len >= 0);
2304
2305 while (len--) {
2306 if (*a != *b && *a != PL_fold_latin1[*b]) {
2307 return 0;
2308 }
2309 a++, b++;
2310 }
2311 return 1;
2312}
2313
2314/*
2315=for apidoc foldEQ_locale
2316
2317Returns true if the leading C<len> bytes of the strings C<s1> and C<s2> are the
2318same case-insensitively in the current locale; false otherwise.
2319
2320=cut
2321*/
2322
0f9cb40c 2323PERL_STATIC_INLINE I32
ec2c235b
KW
2324Perl_foldEQ_locale(const char *s1, const char *s2, I32 len)
2325{
2326 dVAR;
2327 const U8 *a = (const U8 *)s1;
2328 const U8 *b = (const U8 *)s2;
2329
2330 PERL_ARGS_ASSERT_FOLDEQ_LOCALE;
2331
2332 assert(len >= 0);
2333
2334 while (len--) {
2335 if (*a != *b && *a != PL_fold_locale[*b])
2336 return 0;
2337 a++,b++;
2338 }
2339 return 1;
2340}
2341
6dba01e2
KW
2342#if ! defined (HAS_MEMRCHR) && (defined(PERL_CORE) || defined(PERL_EXT))
2343
2344PERL_STATIC_INLINE void *
2345S_my_memrchr(const char * s, const char c, const STRLEN len)
2346{
2347 /* memrchr(), since many platforms lack it */
2348
2349 const char * t = s + len - 1;
2350
2351 PERL_ARGS_ASSERT_MY_MEMRCHR;
2352
2353 while (t >= s) {
2354 if (*t == c) {
2355 return (void *) t;
2356 }
2357 t--;
2358 }
2359
2360 return NULL;
2361}
2362
2363#endif
2364
ed382232 2365/*
c8028aa6
TC
2366 * ex: set ts=8 sts=4 sw=4 et:
2367 */