3 * Copyright (C) 2012 by Larry Wall and others
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.
8 * This file contains tables and code adapted from
9 * https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which requires this
12 Copyright (c) 2008-2009 Bjoern Hoehrmann <bjoern@hoehrmann.de>
14 Permission is hereby granted, free of charge, to any person obtaining a copy of
15 this software and associated documentation files (the "Software"), to deal in
16 the Software without restriction, including without limitation the rights to
17 use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
18 of the Software, and to permit persons to whom the Software is furnished to do
19 so, subject to the following conditions:
21 The above copyright notice and this permission notice shall be included in all
22 copies or substantial portions of the Software.
24 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
25 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
26 FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
27 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
28 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
29 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
33 * This file is a home for static inline functions that cannot go in other
34 * header files, because they depend on proto.h (included after most other
35 * headers) or struct definitions.
37 * Each section names the header file that the functions "belong" to.
40 /* ------------------------------- av.h ------------------------------- */
42 PERL_STATIC_INLINE SSize_t
43 Perl_av_top_index(pTHX_ AV *av)
45 PERL_ARGS_ASSERT_AV_TOP_INDEX;
46 assert(SvTYPE(av) == SVt_PVAV);
51 /* ------------------------------- cv.h ------------------------------- */
53 PERL_STATIC_INLINE GV *
54 Perl_CvGV(pTHX_ CV *sv)
56 PERL_ARGS_ASSERT_CVGV;
59 ? Perl_cvgv_from_hek(aTHX_ sv)
60 : ((XPVCV*)MUTABLE_PTR(SvANY(sv)))->xcv_gv_u.xcv_gv;
63 PERL_STATIC_INLINE I32 *
64 Perl_CvDEPTH(const CV * const sv)
66 PERL_ARGS_ASSERT_CVDEPTH;
67 assert(SvTYPE(sv) == SVt_PVCV || SvTYPE(sv) == SVt_PVFM);
69 return &((XPVCV*)SvANY(sv))->xcv_depth;
73 CvPROTO returns the prototype as stored, which is not necessarily what
74 the interpreter should be using. Specifically, the interpreter assumes
75 that spaces have been stripped, which has been the case if the prototype
76 was added by toke.c, but is generally not the case if it was added elsewhere.
77 Since we can't enforce the spacelessness at assignment time, this routine
78 provides a temporary copy at parse time with spaces removed.
79 I<orig> is the start of the original buffer, I<len> is the length of the
80 prototype and will be updated when this returns.
84 PERL_STATIC_INLINE char *
85 S_strip_spaces(pTHX_ const char * orig, STRLEN * const len)
89 tmpsv = newSVpvn_flags(orig, *len, SVs_TEMP);
97 *len = tmps - SvPVX(tmpsv);
102 /* ------------------------------- mg.h ------------------------------- */
104 #if defined(PERL_CORE) || defined(PERL_EXT)
105 /* assumes get-magic and stringification have already occurred */
106 PERL_STATIC_INLINE STRLEN
107 S_MgBYTEPOS(pTHX_ MAGIC *mg, SV *sv, const char *s, STRLEN len)
109 assert(mg->mg_type == PERL_MAGIC_regex_global);
110 assert(mg->mg_len != -1);
111 if (mg->mg_flags & MGf_BYTES || !DO_UTF8(sv))
112 return (STRLEN)mg->mg_len;
114 const STRLEN pos = (STRLEN)mg->mg_len;
115 /* Without this check, we may read past the end of the buffer: */
116 if (pos > sv_or_pv_len_utf8(sv, s, len)) return len+1;
117 return sv_or_pv_pos_u2b(sv, s, pos, NULL);
122 /* ------------------------------- pad.h ------------------------------ */
124 #if defined(PERL_IN_PAD_C) || defined(PERL_IN_OP_C)
125 PERL_STATIC_INLINE bool
126 S_PadnameIN_SCOPE(const PADNAME * const pn, const U32 seq)
128 PERL_ARGS_ASSERT_PADNAMEIN_SCOPE;
130 /* is seq within the range _LOW to _HIGH ?
131 * This is complicated by the fact that PL_cop_seqmax
132 * may have wrapped around at some point */
133 if (COP_SEQ_RANGE_LOW(pn) == PERL_PADSEQ_INTRO)
134 return FALSE; /* not yet introduced */
136 if (COP_SEQ_RANGE_HIGH(pn) == PERL_PADSEQ_INTRO) {
137 /* in compiling scope */
139 (seq > COP_SEQ_RANGE_LOW(pn))
140 ? (seq - COP_SEQ_RANGE_LOW(pn) < (U32_MAX >> 1))
141 : (COP_SEQ_RANGE_LOW(pn) - seq > (U32_MAX >> 1))
146 (COP_SEQ_RANGE_LOW(pn) > COP_SEQ_RANGE_HIGH(pn))
148 ( seq > COP_SEQ_RANGE_LOW(pn)
149 || seq <= COP_SEQ_RANGE_HIGH(pn))
151 : ( seq > COP_SEQ_RANGE_LOW(pn)
152 && seq <= COP_SEQ_RANGE_HIGH(pn))
159 /* ------------------------------- pp.h ------------------------------- */
161 PERL_STATIC_INLINE I32
164 DEBUG_s(DEBUG_v(PerlIO_printf(Perl_debug_log,
165 "MARK top %p %" IVdf "\n",
167 (IV)*PL_markstack_ptr)));
168 return *PL_markstack_ptr;
171 PERL_STATIC_INLINE I32
174 DEBUG_s(DEBUG_v(PerlIO_printf(Perl_debug_log,
175 "MARK pop %p %" IVdf "\n",
176 (PL_markstack_ptr-1),
177 (IV)*(PL_markstack_ptr-1))));
178 assert((PL_markstack_ptr > PL_markstack) || !"MARK underflow");
179 return *PL_markstack_ptr--;
182 /* ----------------------------- regexp.h ----------------------------- */
184 PERL_STATIC_INLINE struct regexp *
185 Perl_ReANY(const REGEXP * const re)
187 XPV* const p = (XPV*)SvANY(re);
189 PERL_ARGS_ASSERT_REANY;
190 assert(isREGEXP(re));
192 return SvTYPE(re) == SVt_PVLV ? p->xpv_len_u.xpvlenu_rx
193 : (struct regexp *)p;
196 /* ------------------------------- sv.h ------------------------------- */
198 PERL_STATIC_INLINE bool
199 Perl_SvTRUE(pTHX_ SV *sv) {
203 return SvTRUE_nomg_NN(sv);
206 PERL_STATIC_INLINE SV *
207 Perl_SvREFCNT_inc(SV *sv)
209 if (LIKELY(sv != NULL))
213 PERL_STATIC_INLINE SV *
214 Perl_SvREFCNT_inc_NN(SV *sv)
216 PERL_ARGS_ASSERT_SVREFCNT_INC_NN;
221 PERL_STATIC_INLINE void
222 Perl_SvREFCNT_inc_void(SV *sv)
224 if (LIKELY(sv != NULL))
227 PERL_STATIC_INLINE void
228 Perl_SvREFCNT_dec(pTHX_ SV *sv)
230 if (LIKELY(sv != NULL)) {
231 U32 rc = SvREFCNT(sv);
233 SvREFCNT(sv) = rc - 1;
235 Perl_sv_free2(aTHX_ sv, rc);
239 PERL_STATIC_INLINE void
240 Perl_SvREFCNT_dec_NN(pTHX_ SV *sv)
242 U32 rc = SvREFCNT(sv);
244 PERL_ARGS_ASSERT_SVREFCNT_DEC_NN;
247 SvREFCNT(sv) = rc - 1;
249 Perl_sv_free2(aTHX_ sv, rc);
252 PERL_STATIC_INLINE void
253 Perl_SvAMAGIC_on(SV *sv)
255 PERL_ARGS_ASSERT_SVAMAGIC_ON;
258 if (SvOBJECT(SvRV(sv))) HvAMAGIC_on(SvSTASH(SvRV(sv)));
260 PERL_STATIC_INLINE void
261 Perl_SvAMAGIC_off(SV *sv)
263 PERL_ARGS_ASSERT_SVAMAGIC_OFF;
265 if (SvROK(sv) && SvOBJECT(SvRV(sv)))
266 HvAMAGIC_off(SvSTASH(SvRV(sv)));
269 PERL_STATIC_INLINE U32
270 Perl_SvPADSTALE_on(SV *sv)
272 assert(!(SvFLAGS(sv) & SVs_PADTMP));
273 return SvFLAGS(sv) |= SVs_PADSTALE;
275 PERL_STATIC_INLINE U32
276 Perl_SvPADSTALE_off(SV *sv)
278 assert(!(SvFLAGS(sv) & SVs_PADTMP));
279 return SvFLAGS(sv) &= ~SVs_PADSTALE;
281 #if defined(PERL_CORE) || defined (PERL_EXT)
282 PERL_STATIC_INLINE STRLEN
283 S_sv_or_pv_pos_u2b(pTHX_ SV *sv, const char *pv, STRLEN pos, STRLEN *lenp)
285 PERL_ARGS_ASSERT_SV_OR_PV_POS_U2B;
287 U8 *hopped = utf8_hop((U8 *)pv, pos);
288 if (lenp) *lenp = (STRLEN)(utf8_hop(hopped, *lenp) - hopped);
289 return (STRLEN)(hopped - (U8 *)pv);
291 return sv_pos_u2b_flags(sv,pos,lenp,SV_CONST_RETURN);
295 /* ------------------------------- handy.h ------------------------------- */
297 /* saves machine code for a common noreturn idiom typically used in Newx*() */
298 GCC_DIAG_IGNORE_DECL(-Wunused-function);
300 Perl_croak_memory_wrap(void)
302 Perl_croak_nocontext("%s",PL_memory_wrap);
304 GCC_DIAG_RESTORE_DECL;
306 /* ------------------------------- utf8.h ------------------------------- */
309 =head1 Unicode Support
312 PERL_STATIC_INLINE void
313 Perl_append_utf8_from_native_byte(const U8 byte, U8** dest)
315 /* Takes an input 'byte' (Latin1 or EBCDIC) and appends it to the UTF-8
316 * encoded string at '*dest', updating '*dest' to include it */
318 PERL_ARGS_ASSERT_APPEND_UTF8_FROM_NATIVE_BYTE;
320 if (NATIVE_BYTE_IS_INVARIANT(byte))
323 *((*dest)++) = UTF8_EIGHT_BIT_HI(byte);
324 *((*dest)++) = UTF8_EIGHT_BIT_LO(byte);
329 =for apidoc valid_utf8_to_uvchr
330 Like C<L<perlapi/utf8_to_uvchr_buf>>, but should only be called when it is
331 known that the next character in the input UTF-8 string C<s> is well-formed
332 (I<e.g.>, it passes C<L<perlapi/isUTF8_CHAR>>. Surrogates, non-character code
333 points, and non-Unicode code points are allowed.
339 PERL_STATIC_INLINE UV
340 Perl_valid_utf8_to_uvchr(const U8 *s, STRLEN *retlen)
342 const UV expectlen = UTF8SKIP(s);
343 const U8* send = s + expectlen;
346 PERL_ARGS_ASSERT_VALID_UTF8_TO_UVCHR;
352 /* An invariant is trivially returned */
353 if (expectlen == 1) {
357 /* Remove the leading bits that indicate the number of bytes, leaving just
358 * the bits that are part of the value */
359 uv = NATIVE_UTF8_TO_I8(uv) & UTF_START_MASK(expectlen);
361 /* Now, loop through the remaining bytes, accumulating each into the
362 * working total as we go. (I khw tried unrolling the loop for up to 4
363 * bytes, but there was no performance improvement) */
364 for (++s; s < send; s++) {
365 uv = UTF8_ACCUMULATE(uv, *s);
368 return UNI_TO_NATIVE(uv);
373 =for apidoc is_utf8_invariant_string
375 Returns TRUE if the first C<len> bytes of the string C<s> are the same
376 regardless of the UTF-8 encoding of the string (or UTF-EBCDIC encoding on
377 EBCDIC machines); otherwise it returns FALSE. That is, it returns TRUE if they
378 are UTF-8 invariant. On ASCII-ish machines, all the ASCII characters and only
379 the ASCII characters fit this definition. On EBCDIC machines, the ASCII-range
380 characters are invariant, but so also are the C1 controls.
382 If C<len> is 0, it will be calculated using C<strlen(s)>, (which means if you
383 use this option, that C<s> can't have embedded C<NUL> characters and has to
384 have a terminating C<NUL> byte).
387 C<L</is_utf8_string>>,
388 C<L</is_utf8_string_flags>>,
389 C<L</is_utf8_string_loc>>,
390 C<L</is_utf8_string_loc_flags>>,
391 C<L</is_utf8_string_loclen>>,
392 C<L</is_utf8_string_loclen_flags>>,
393 C<L</is_utf8_fixed_width_buf_flags>>,
394 C<L</is_utf8_fixed_width_buf_loc_flags>>,
395 C<L</is_utf8_fixed_width_buf_loclen_flags>>,
396 C<L</is_strict_utf8_string>>,
397 C<L</is_strict_utf8_string_loc>>,
398 C<L</is_strict_utf8_string_loclen>>,
399 C<L</is_c9strict_utf8_string>>,
400 C<L</is_c9strict_utf8_string_loc>>,
402 C<L</is_c9strict_utf8_string_loclen>>.
408 #define is_utf8_invariant_string(s, len) \
409 is_utf8_invariant_string_loc(s, len, NULL)
412 =for apidoc is_utf8_invariant_string_loc
414 Like C<L</is_utf8_invariant_string>> but upon failure, stores the location of
415 the first UTF-8 variant character in the C<ep> pointer; if all characters are
416 UTF-8 invariant, this function does not change the contents of C<*ep>.
422 PERL_STATIC_INLINE bool
423 Perl_is_utf8_invariant_string_loc(const U8* const s, STRLEN len, const U8 ** ep)
428 PERL_ARGS_ASSERT_IS_UTF8_INVARIANT_STRING_LOC;
431 len = strlen((const char *)s);
436 /* This looks like 0x010101... */
437 # define PERL_COUNT_MULTIPLIER (~ (UINTMAX_C(0)) / 0xFF)
439 /* This looks like 0x808080... */
440 # define PERL_VARIANTS_WORD_MASK (PERL_COUNT_MULTIPLIER * 0x80)
441 # define PERL_WORDSIZE sizeof(PERL_UINTMAX_T)
442 # define PERL_WORD_BOUNDARY_MASK (PERL_WORDSIZE - 1)
444 /* Evaluates to 0 if 'x' is at a word boundary; otherwise evaluates to 1, by
445 * or'ing together the lowest bits of 'x'. Hopefully the final term gets
446 * optimized out completely on a 32-bit system, and its mask gets optimized out
447 * on a 64-bit system */
448 # define PERL_IS_SUBWORD_ADDR(x) (1 & ( PTR2nat(x) \
449 | ( PTR2nat(x) >> 1) \
451 & PERL_WORD_BOUNDARY_MASK) >> 2))))
455 /* Do the word-at-a-time iff there is at least one usable full word. That
456 * means that after advancing to a word boundary, there still is at least a
457 * full word left. The number of bytes needed to advance is 'wordsize -
458 * offset' unless offset is 0. */
459 if ((STRLEN) (send - x) >= PERL_WORDSIZE
461 /* This term is wordsize if subword; 0 if not */
462 + PERL_WORDSIZE * PERL_IS_SUBWORD_ADDR(x)
465 - (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK))
468 /* Process per-byte until reach word boundary. XXX This loop could be
469 * eliminated if we knew that this platform had fast unaligned reads */
470 while (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK) {
471 if (! UTF8_IS_INVARIANT(*x)) {
481 /* Here, we know we have at least one full word to process. Process
482 * per-word as long as we have at least a full word left */
484 if ((* (PERL_UINTMAX_T *) x) & PERL_VARIANTS_WORD_MASK) {
486 /* Found a variant. Just return if caller doesn't want its
492 # if BYTEORDER == 0x1234 || BYTEORDER == 0x12345678 \
493 || BYTEORDER == 0x4321 || BYTEORDER == 0x87654321
495 *ep = x + variant_byte_number(* (PERL_UINTMAX_T *) x);
496 assert(*ep >= s && *ep < send);
500 # else /* If weird byte order, drop into next loop to do byte-at-a-time
509 } while (x + PERL_WORDSIZE <= send);
512 #endif /* End of ! EBCDIC */
514 /* Process per-byte */
516 if (! UTF8_IS_INVARIANT(*x)) {
532 PERL_STATIC_INLINE unsigned int
533 Perl_variant_byte_number(PERL_UINTMAX_T word)
536 /* This returns the position in a word (0..7) of the first variant byte in
537 * it. This is a helper function. Note that there are no branches */
541 /* Get just the msb bits of each byte */
542 word &= PERL_VARIANTS_WORD_MASK;
544 # if BYTEORDER == 0x1234 || BYTEORDER == 0x12345678
546 /* Bytes are stored like
547 * Byte8 ... Byte2 Byte1
548 * 63..56...15...8 7...0
551 * https://stackoverflow.com/questions/757059/position-of-least-significant-bit-that-is-set
553 * The word will look this this, with a rightmost set bit in position 's':
554 * ('x's are don't cares)
557 * x..xx10..0 Right shift (rightmost 0 is shifted off)
558 * x..xx01..1 Subtract 1, turns all the trailing zeros into 1's and
559 * the 1 just to their left into a 0; the remainder is
561 * 0..0011..1 The xor with the original, x..xx10..0, clears that
562 * remainder, sets the bottom to all 1
563 * 0..0100..0 Add 1 to clear the word except for the bit in 's'
565 * Another method is to do 'word &= -word'; but it generates a compiler
566 * message on some platforms about taking the negative of an unsigned */
569 word = 1 + (word ^ (word - 1));
571 # elif BYTEORDER == 0x4321 || BYTEORDER == 0x87654321
573 /* Bytes are stored like
574 * Byte1 Byte2 ... Byte8
575 * 63..56 55..47 ... 7...0
577 * Isolate the msb; http://codeforces.com/blog/entry/10330
579 * Only the most significant set bit matters. Or'ing word with its right
580 * shift of 1 makes that bit and the next one to its right both 1. Then
581 * right shifting by 2 makes for 4 1-bits in a row. ... We end with the
582 * msb and all to the right being 1. */
588 word |= word >> 32; /* This should get optimized out on 32-bit systems. */
590 /* Then subtracting the right shift by 1 clears all but the left-most of
591 * the 1 bits, which is our desired result */
595 # error Unexpected byte order
598 /* Here 'word' has a single bit set: the msb of the first byte in which it
599 * is set. Calculate that position in the word. We can use this
600 * specialized solution: https://stackoverflow.com/a/32339674/1626653,
601 * assumes an 8-bit byte. (On a 32-bit machine, the larger numbers should
602 * just get shifted off at compile time) */
603 word = (word >> 7) * ((UINTMAX_C( 7) << 56) | (UINTMAX_C(15) << 48)
604 | (UINTMAX_C(23) << 40) | (UINTMAX_C(31) << 32)
605 | (39 << 24) | (47 << 16)
606 | (55 << 8) | (63 << 0));
607 word >>= PERL_WORDSIZE * 7; /* >> by either 56 or 24 */
609 /* Here, word contains the position 7..63 of that bit. Convert to 0..7 */
610 word = ((word + 1) >> 3) - 1;
612 # if BYTEORDER == 0x4321 || BYTEORDER == 0x87654321
614 /* And invert the result */
615 word = CHARBITS - word - 1;
619 return (unsigned int) word;
623 #if defined(PERL_CORE) || defined(PERL_EXT)
626 =for apidoc variant_under_utf8_count
628 This function looks at the sequence of bytes between C<s> and C<e>, which are
629 assumed to be encoded in ASCII/Latin1, and returns how many of them would
630 change should the string be translated into UTF-8. Due to the nature of UTF-8,
631 each of these would occupy two bytes instead of the single one in the input
632 string. Thus, this function returns the precise number of bytes the string
633 would expand by when translated to UTF-8.
635 Unlike most of the other functions that have C<utf8> in their name, the input
636 to this function is NOT a UTF-8-encoded string. The function name is slightly
637 I<odd> to emphasize this.
639 This function is internal to Perl because khw thinks that any XS code that
640 would want this is probably operating too close to the internals. Presenting a
641 valid use case could change that.
644 C<L<perlapi/is_utf8_invariant_string>>
646 C<L<perlapi/is_utf8_invariant_string_loc>>,
652 PERL_STATIC_INLINE Size_t
653 S_variant_under_utf8_count(const U8* const s, const U8* const e)
658 PERL_ARGS_ASSERT_VARIANT_UNDER_UTF8_COUNT;
662 /* Test if the string is long enough to use word-at-a-time. (Logic is the
663 * same as for is_utf8_invariant_string()) */
664 if ((STRLEN) (e - x) >= PERL_WORDSIZE
665 + PERL_WORDSIZE * PERL_IS_SUBWORD_ADDR(x)
666 - (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK))
669 /* Process per-byte until reach word boundary. XXX This loop could be
670 * eliminated if we knew that this platform had fast unaligned reads */
671 while (PTR2nat(x) & PERL_WORD_BOUNDARY_MASK) {
672 count += ! UTF8_IS_INVARIANT(*x++);
675 /* Process per-word as long as we have at least a full word left */
676 do { /* Commit 03c1e4ab1d6ee9062fb3f94b0ba31db6698724b1 contains an
677 explanation of how this works */
678 PERL_UINTMAX_T increment
679 = ((((* (PERL_UINTMAX_T *) x) & PERL_VARIANTS_WORD_MASK) >> 7)
680 * PERL_COUNT_MULTIPLIER)
681 >> ((PERL_WORDSIZE - 1) * CHARBITS);
682 count += (Size_t) increment;
684 } while (x + PERL_WORDSIZE <= e);
689 /* Process per-byte */
691 if (! UTF8_IS_INVARIANT(*x)) {
703 #ifndef PERL_IN_REGEXEC_C /* Keep these around for that file */
704 # undef PERL_WORDSIZE
705 # undef PERL_COUNT_MULTIPLIER
706 # undef PERL_WORD_BOUNDARY_MASK
707 # undef PERL_VARIANTS_WORD_MASK
711 =for apidoc is_utf8_string
713 Returns TRUE if the first C<len> bytes of string C<s> form a valid
714 Perl-extended-UTF-8 string; returns FALSE otherwise. If C<len> is 0, it will
715 be calculated using C<strlen(s)> (which means if you use this option, that C<s>
716 can't have embedded C<NUL> characters and has to have a terminating C<NUL>
717 byte). Note that all characters being ASCII constitute 'a valid UTF-8 string'.
719 This function considers Perl's extended UTF-8 to be valid. That means that
720 code points above Unicode, surrogates, and non-character code points are
721 considered valid by this function. Use C<L</is_strict_utf8_string>>,
722 C<L</is_c9strict_utf8_string>>, or C<L</is_utf8_string_flags>> to restrict what
723 code points are considered valid.
726 C<L</is_utf8_invariant_string>>,
727 C<L</is_utf8_invariant_string_loc>>,
728 C<L</is_utf8_string_loc>>,
729 C<L</is_utf8_string_loclen>>,
730 C<L</is_utf8_fixed_width_buf_flags>>,
731 C<L</is_utf8_fixed_width_buf_loc_flags>>,
732 C<L</is_utf8_fixed_width_buf_loclen_flags>>,
737 #define is_utf8_string(s, len) is_utf8_string_loclen(s, len, NULL, NULL)
739 #if defined(PERL_CORE) || defined (PERL_EXT)
742 =for apidoc is_utf8_non_invariant_string
744 Returns TRUE if L<perlapi/is_utf8_invariant_string> returns FALSE for the first
745 C<len> bytes of the string C<s>, but they are, nonetheless, legal Perl-extended
746 UTF-8; otherwise returns FALSE.
748 A TRUE return means that at least one code point represented by the sequence
749 either is a wide character not representable as a single byte, or the
750 representation differs depending on whether the sequence is encoded in UTF-8 or
754 C<L<perlapi/is_utf8_invariant_string>>,
755 C<L<perlapi/is_utf8_string>>
759 This is commonly used to determine if a SV's UTF-8 flag should be turned on.
760 It generally needn't be if its string is entirely UTF-8 invariant, and it
761 shouldn't be if it otherwise contains invalid UTF-8.
763 It is an internal function because khw thinks that XS code shouldn't be working
764 at this low a level. A valid use case could change that.
768 PERL_STATIC_INLINE bool
769 Perl_is_utf8_non_invariant_string(const U8* const s, STRLEN len)
771 const U8 * first_variant;
773 PERL_ARGS_ASSERT_IS_UTF8_NON_INVARIANT_STRING;
775 if (is_utf8_invariant_string_loc(s, len, &first_variant)) {
779 return is_utf8_string(first_variant, len - (first_variant - s));
785 =for apidoc is_strict_utf8_string
787 Returns TRUE if the first C<len> bytes of string C<s> form a valid
788 UTF-8-encoded string that is fully interchangeable by any application using
789 Unicode rules; otherwise it returns FALSE. If C<len> is 0, it will be
790 calculated using C<strlen(s)> (which means if you use this option, that C<s>
791 can't have embedded C<NUL> characters and has to have a terminating C<NUL>
792 byte). Note that all characters being ASCII constitute 'a valid UTF-8 string'.
794 This function returns FALSE for strings containing any
795 code points above the Unicode max of 0x10FFFF, surrogate code points, or
796 non-character code points.
799 C<L</is_utf8_invariant_string>>,
800 C<L</is_utf8_invariant_string_loc>>,
801 C<L</is_utf8_string>>,
802 C<L</is_utf8_string_flags>>,
803 C<L</is_utf8_string_loc>>,
804 C<L</is_utf8_string_loc_flags>>,
805 C<L</is_utf8_string_loclen>>,
806 C<L</is_utf8_string_loclen_flags>>,
807 C<L</is_utf8_fixed_width_buf_flags>>,
808 C<L</is_utf8_fixed_width_buf_loc_flags>>,
809 C<L</is_utf8_fixed_width_buf_loclen_flags>>,
810 C<L</is_strict_utf8_string_loc>>,
811 C<L</is_strict_utf8_string_loclen>>,
812 C<L</is_c9strict_utf8_string>>,
813 C<L</is_c9strict_utf8_string_loc>>,
815 C<L</is_c9strict_utf8_string_loclen>>.
820 #define is_strict_utf8_string(s, len) is_strict_utf8_string_loclen(s, len, NULL, NULL)
823 =for apidoc is_c9strict_utf8_string
825 Returns TRUE if the first C<len> bytes of string C<s> form a valid
826 UTF-8-encoded string that conforms to
827 L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>;
828 otherwise it returns FALSE. If C<len> is 0, it will be calculated using
829 C<strlen(s)> (which means if you use this option, that C<s> can't have embedded
830 C<NUL> characters and has to have a terminating C<NUL> byte). Note that all
831 characters being ASCII constitute 'a valid UTF-8 string'.
833 This function returns FALSE for strings containing any code points above the
834 Unicode max of 0x10FFFF or surrogate code points, but accepts non-character
836 L<Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>.
839 C<L</is_utf8_invariant_string>>,
840 C<L</is_utf8_invariant_string_loc>>,
841 C<L</is_utf8_string>>,
842 C<L</is_utf8_string_flags>>,
843 C<L</is_utf8_string_loc>>,
844 C<L</is_utf8_string_loc_flags>>,
845 C<L</is_utf8_string_loclen>>,
846 C<L</is_utf8_string_loclen_flags>>,
847 C<L</is_utf8_fixed_width_buf_flags>>,
848 C<L</is_utf8_fixed_width_buf_loc_flags>>,
849 C<L</is_utf8_fixed_width_buf_loclen_flags>>,
850 C<L</is_strict_utf8_string>>,
851 C<L</is_strict_utf8_string_loc>>,
852 C<L</is_strict_utf8_string_loclen>>,
853 C<L</is_c9strict_utf8_string_loc>>,
855 C<L</is_c9strict_utf8_string_loclen>>.
860 #define is_c9strict_utf8_string(s, len) is_c9strict_utf8_string_loclen(s, len, NULL, 0)
863 =for apidoc is_utf8_string_flags
865 Returns TRUE if the first C<len> bytes of string C<s> form a valid
866 UTF-8 string, subject to the restrictions imposed by C<flags>;
867 returns FALSE otherwise. If C<len> is 0, it will be calculated
868 using C<strlen(s)> (which means if you use this option, that C<s> can't have
869 embedded C<NUL> characters and has to have a terminating C<NUL> byte). Note
870 that all characters being ASCII constitute 'a valid UTF-8 string'.
872 If C<flags> is 0, this gives the same results as C<L</is_utf8_string>>; if
873 C<flags> is C<UTF8_DISALLOW_ILLEGAL_INTERCHANGE>, this gives the same results
874 as C<L</is_strict_utf8_string>>; and if C<flags> is
875 C<UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE>, this gives the same results as
876 C<L</is_c9strict_utf8_string>>. Otherwise C<flags> may be any
877 combination of the C<UTF8_DISALLOW_I<foo>> flags understood by
878 C<L</utf8n_to_uvchr>>, with the same meanings.
881 C<L</is_utf8_invariant_string>>,
882 C<L</is_utf8_invariant_string_loc>>,
883 C<L</is_utf8_string>>,
884 C<L</is_utf8_string_loc>>,
885 C<L</is_utf8_string_loc_flags>>,
886 C<L</is_utf8_string_loclen>>,
887 C<L</is_utf8_string_loclen_flags>>,
888 C<L</is_utf8_fixed_width_buf_flags>>,
889 C<L</is_utf8_fixed_width_buf_loc_flags>>,
890 C<L</is_utf8_fixed_width_buf_loclen_flags>>,
891 C<L</is_strict_utf8_string>>,
892 C<L</is_strict_utf8_string_loc>>,
893 C<L</is_strict_utf8_string_loclen>>,
894 C<L</is_c9strict_utf8_string>>,
895 C<L</is_c9strict_utf8_string_loc>>,
897 C<L</is_c9strict_utf8_string_loclen>>.
902 PERL_STATIC_INLINE bool
903 Perl_is_utf8_string_flags(const U8 *s, STRLEN len, const U32 flags)
905 const U8 * first_variant;
907 PERL_ARGS_ASSERT_IS_UTF8_STRING_FLAGS;
908 assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE
909 |UTF8_DISALLOW_PERL_EXTENDED)));
912 len = strlen((const char *)s);
916 return is_utf8_string(s, len);
919 if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED)
920 == UTF8_DISALLOW_ILLEGAL_INTERCHANGE)
922 return is_strict_utf8_string(s, len);
925 if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED)
926 == UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE)
928 return is_c9strict_utf8_string(s, len);
931 if (! is_utf8_invariant_string_loc(s, len, &first_variant)) {
932 const U8* const send = s + len;
933 const U8* x = first_variant;
936 STRLEN cur_len = isUTF8_CHAR_flags(x, send, flags);
937 if (UNLIKELY(! cur_len)) {
949 =for apidoc is_utf8_string_loc
951 Like C<L</is_utf8_string>> but stores the location of the failure (in the
952 case of "utf8ness failure") or the location C<s>+C<len> (in the case of
953 "utf8ness success") in the C<ep> pointer.
955 See also C<L</is_utf8_string_loclen>>.
960 #define is_utf8_string_loc(s, len, ep) is_utf8_string_loclen(s, len, ep, 0)
964 =for apidoc is_utf8_string_loclen
966 Like C<L</is_utf8_string>> but stores the location of the failure (in the
967 case of "utf8ness failure") or the location C<s>+C<len> (in the case of
968 "utf8ness success") in the C<ep> pointer, and the number of UTF-8
969 encoded characters in the C<el> pointer.
971 See also C<L</is_utf8_string_loc>>.
976 PERL_STATIC_INLINE bool
977 Perl_is_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
979 const U8 * first_variant;
981 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN;
984 len = strlen((const char *) s);
987 if (is_utf8_invariant_string_loc(s, len, &first_variant)) {
999 const U8* const send = s + len;
1000 const U8* x = first_variant;
1001 STRLEN outlen = first_variant - s;
1004 const STRLEN cur_len = isUTF8_CHAR(x, send);
1005 if (UNLIKELY(! cur_len)) {
1025 =for apidoc isUTF8_CHAR
1027 Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1028 looking no further than S<C<e - 1>> are well-formed UTF-8, as extended by Perl,
1029 that represents some code point; otherwise it evaluates to 0. If non-zero, the
1030 value gives how many bytes starting at C<s> comprise the code point's
1031 representation. Any bytes remaining before C<e>, but beyond the ones needed to
1032 form the first code point in C<s>, are not examined.
1034 The code point can be any that will fit in an IV on this machine, using Perl's
1035 extension to official UTF-8 to represent those higher than the Unicode maximum
1036 of 0x10FFFF. That means that this macro is used to efficiently decide if the
1037 next few bytes in C<s> is legal UTF-8 for a single character.
1039 Use C<L</isSTRICT_UTF8_CHAR>> to restrict the acceptable code points to those
1040 defined by Unicode to be fully interchangeable across applications;
1041 C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1042 #9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1043 code points; and C<L</isUTF8_CHAR_flags>> for a more customized definition.
1045 Use C<L</is_utf8_string>>, C<L</is_utf8_string_loc>>, and
1046 C<L</is_utf8_string_loclen>> to check entire strings.
1048 Note also that a UTF-8 "invariant" character (i.e. ASCII on non-EBCDIC
1049 machines) is a valid UTF-8 character.
1053 This uses an adaptation of the table and algorithm given in
1054 https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides comprehensive
1055 documentation of the original version. A copyright notice for the original
1056 version is given at the beginning of this file. The Perl adapation is
1057 documented at the definition of PL_extended_utf8_dfa_tab[].
1061 PERL_STATIC_INLINE Size_t
1062 Perl_isUTF8_CHAR(const U8 * const s0, const U8 * const e)
1067 PERL_ARGS_ASSERT_ISUTF8_CHAR;
1069 /* This dfa is fast. If it accepts the input, it was for a well-formed,
1070 * code point, which can be returned immediately. Otherwise, it is either
1071 * malformed, or for the start byte FF which the dfa doesn't handle (except
1072 * on 32-bit ASCII platforms where it trivially is an error). Call a
1073 * helper function for the other platforms. */
1075 while (s < e && LIKELY(state != 1)) {
1076 state = PL_extended_utf8_dfa_tab[256
1078 + PL_extended_utf8_dfa_tab[*s]];
1087 #if defined(UV_IS_QUAD) || defined(EBCDIC)
1089 if (NATIVE_UTF8_TO_I8(*s0) == 0xFF && e - s0 >= UTF8_MAXBYTES) {
1090 return is_utf8_char_helper(s0, e, 0);
1100 =for apidoc isSTRICT_UTF8_CHAR
1102 Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1103 looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1104 Unicode code point completely acceptable for open interchange between all
1105 applications; otherwise it evaluates to 0. If non-zero, the value gives how
1106 many bytes starting at C<s> comprise the code point's representation. Any
1107 bytes remaining before C<e>, but beyond the ones needed to form the first code
1108 point in C<s>, are not examined.
1110 The largest acceptable code point is the Unicode maximum 0x10FFFF, and must not
1111 be a surrogate nor a non-character code point. Thus this excludes any code
1112 point from Perl's extended UTF-8.
1114 This is used to efficiently decide if the next few bytes in C<s> is
1115 legal Unicode-acceptable UTF-8 for a single character.
1117 Use C<L</isC9_STRICT_UTF8_CHAR>> to use the L<Unicode Corrigendum
1118 #9|http://www.unicode.org/versions/corrigendum9.html> definition of allowable
1119 code points; C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8;
1120 and C<L</isUTF8_CHAR_flags>> for a more customized definition.
1122 Use C<L</is_strict_utf8_string>>, C<L</is_strict_utf8_string_loc>>, and
1123 C<L</is_strict_utf8_string_loclen>> to check entire strings.
1127 This uses an adaptation of the tables and algorithm given in
1128 https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides comprehensive
1129 documentation of the original version. A copyright notice for the original
1130 version is given at the beginning of this file. The Perl adapation is
1131 documented at the definition of strict_extended_utf8_dfa_tab[].
1135 PERL_STATIC_INLINE Size_t
1136 Perl_isSTRICT_UTF8_CHAR(const U8 * const s0, const U8 * const e)
1141 PERL_ARGS_ASSERT_ISSTRICT_UTF8_CHAR;
1143 while (s < e && LIKELY(state != 1)) {
1144 state = PL_strict_utf8_dfa_tab[256 + state + PL_strict_utf8_dfa_tab[*s]];
1156 /* The dfa above drops out for certain Hanguls; handle them specially */
1157 if (is_HANGUL_ED_utf8_safe(s0, e)) {
1168 =for apidoc isC9_STRICT_UTF8_CHAR
1170 Evaluates to non-zero if the first few bytes of the string starting at C<s> and
1171 looking no further than S<C<e - 1>> are well-formed UTF-8 that represents some
1172 Unicode non-surrogate code point; otherwise it evaluates to 0. If non-zero,
1173 the value gives how many bytes starting at C<s> comprise the code point's
1174 representation. Any bytes remaining before C<e>, but beyond the ones needed to
1175 form the first code point in C<s>, are not examined.
1177 The largest acceptable code point is the Unicode maximum 0x10FFFF. This
1178 differs from C<L</isSTRICT_UTF8_CHAR>> only in that it accepts non-character
1179 code points. This corresponds to
1180 L<Unicode Corrigendum #9|http://www.unicode.org/versions/corrigendum9.html>.
1181 which said that non-character code points are merely discouraged rather than
1182 completely forbidden in open interchange. See
1183 L<perlunicode/Noncharacter code points>.
1185 Use C<L</isUTF8_CHAR>> to check for Perl's extended UTF-8; and
1186 C<L</isUTF8_CHAR_flags>> for a more customized definition.
1188 Use C<L</is_c9strict_utf8_string>>, C<L</is_c9strict_utf8_string_loc>>, and
1189 C<L</is_c9strict_utf8_string_loclen>> to check entire strings.
1193 This uses an adaptation of the tables and algorithm given in
1194 https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides comprehensive
1195 documentation of the original version. A copyright notice for the original
1196 version is given at the beginning of this file. The Perl adapation is
1197 documented at the definition of PL_c9_utf8_dfa_tab[].
1201 PERL_STATIC_INLINE Size_t
1202 Perl_isC9_STRICT_UTF8_CHAR(const U8 * const s0, const U8 * const e)
1207 PERL_ARGS_ASSERT_ISC9_STRICT_UTF8_CHAR;
1209 while (s < e && LIKELY(state != 1)) {
1210 state = PL_c9_utf8_dfa_tab[256 + state + PL_c9_utf8_dfa_tab[*s]];
1225 =for apidoc is_strict_utf8_string_loc
1227 Like C<L</is_strict_utf8_string>> but stores the location of the failure (in the
1228 case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1229 "utf8ness success") in the C<ep> pointer.
1231 See also C<L</is_strict_utf8_string_loclen>>.
1236 #define is_strict_utf8_string_loc(s, len, ep) \
1237 is_strict_utf8_string_loclen(s, len, ep, 0)
1241 =for apidoc is_strict_utf8_string_loclen
1243 Like C<L</is_strict_utf8_string>> but stores the location of the failure (in the
1244 case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1245 "utf8ness success") in the C<ep> pointer, and the number of UTF-8
1246 encoded characters in the C<el> pointer.
1248 See also C<L</is_strict_utf8_string_loc>>.
1253 PERL_STATIC_INLINE bool
1254 Perl_is_strict_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
1256 const U8 * first_variant;
1258 PERL_ARGS_ASSERT_IS_STRICT_UTF8_STRING_LOCLEN;
1261 len = strlen((const char *) s);
1264 if (is_utf8_invariant_string_loc(s, len, &first_variant)) {
1276 const U8* const send = s + len;
1277 const U8* x = first_variant;
1278 STRLEN outlen = first_variant - s;
1281 const STRLEN cur_len = isSTRICT_UTF8_CHAR(x, send);
1282 if (UNLIKELY(! cur_len)) {
1302 =for apidoc is_c9strict_utf8_string_loc
1304 Like C<L</is_c9strict_utf8_string>> but stores the location of the failure (in
1305 the case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1306 "utf8ness success") in the C<ep> pointer.
1308 See also C<L</is_c9strict_utf8_string_loclen>>.
1313 #define is_c9strict_utf8_string_loc(s, len, ep) \
1314 is_c9strict_utf8_string_loclen(s, len, ep, 0)
1318 =for apidoc is_c9strict_utf8_string_loclen
1320 Like C<L</is_c9strict_utf8_string>> but stores the location of the failure (in
1321 the case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1322 "utf8ness success") in the C<ep> pointer, and the number of UTF-8 encoded
1323 characters in the C<el> pointer.
1325 See also C<L</is_c9strict_utf8_string_loc>>.
1330 PERL_STATIC_INLINE bool
1331 Perl_is_c9strict_utf8_string_loclen(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el)
1333 const U8 * first_variant;
1335 PERL_ARGS_ASSERT_IS_C9STRICT_UTF8_STRING_LOCLEN;
1338 len = strlen((const char *) s);
1341 if (is_utf8_invariant_string_loc(s, len, &first_variant)) {
1353 const U8* const send = s + len;
1354 const U8* x = first_variant;
1355 STRLEN outlen = first_variant - s;
1358 const STRLEN cur_len = isC9_STRICT_UTF8_CHAR(x, send);
1359 if (UNLIKELY(! cur_len)) {
1379 =for apidoc is_utf8_string_loc_flags
1381 Like C<L</is_utf8_string_flags>> but stores the location of the failure (in the
1382 case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1383 "utf8ness success") in the C<ep> pointer.
1385 See also C<L</is_utf8_string_loclen_flags>>.
1390 #define is_utf8_string_loc_flags(s, len, ep, flags) \
1391 is_utf8_string_loclen_flags(s, len, ep, 0, flags)
1394 /* The above 3 actual functions could have been moved into the more general one
1395 * just below, and made #defines that call it with the right 'flags'. They are
1396 * currently kept separate to increase their chances of getting inlined */
1400 =for apidoc is_utf8_string_loclen_flags
1402 Like C<L</is_utf8_string_flags>> but stores the location of the failure (in the
1403 case of "utf8ness failure") or the location C<s>+C<len> (in the case of
1404 "utf8ness success") in the C<ep> pointer, and the number of UTF-8
1405 encoded characters in the C<el> pointer.
1407 See also C<L</is_utf8_string_loc_flags>>.
1412 PERL_STATIC_INLINE bool
1413 Perl_is_utf8_string_loclen_flags(const U8 *s, STRLEN len, const U8 **ep, STRLEN *el, const U32 flags)
1415 const U8 * first_variant;
1417 PERL_ARGS_ASSERT_IS_UTF8_STRING_LOCLEN_FLAGS;
1418 assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE
1419 |UTF8_DISALLOW_PERL_EXTENDED)));
1422 len = strlen((const char *) s);
1426 return is_utf8_string_loclen(s, len, ep, el);
1429 if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED)
1430 == UTF8_DISALLOW_ILLEGAL_INTERCHANGE)
1432 return is_strict_utf8_string_loclen(s, len, ep, el);
1435 if ((flags & ~UTF8_DISALLOW_PERL_EXTENDED)
1436 == UTF8_DISALLOW_ILLEGAL_C9_INTERCHANGE)
1438 return is_c9strict_utf8_string_loclen(s, len, ep, el);
1441 if (is_utf8_invariant_string_loc(s, len, &first_variant)) {
1453 const U8* send = s + len;
1454 const U8* x = first_variant;
1455 STRLEN outlen = first_variant - s;
1458 const STRLEN cur_len = isUTF8_CHAR_flags(x, send, flags);
1459 if (UNLIKELY(! cur_len)) {
1478 =for apidoc utf8_distance
1480 Returns the number of UTF-8 characters between the UTF-8 pointers C<a>
1483 WARNING: use only if you *know* that the pointers point inside the
1489 PERL_STATIC_INLINE IV
1490 Perl_utf8_distance(pTHX_ const U8 *a, const U8 *b)
1492 PERL_ARGS_ASSERT_UTF8_DISTANCE;
1494 return (a < b) ? -1 * (IV) utf8_length(a, b) : (IV) utf8_length(b, a);
1498 =for apidoc utf8_hop
1500 Return the UTF-8 pointer C<s> displaced by C<off> characters, either
1501 forward or backward.
1503 WARNING: do not use the following unless you *know* C<off> is within
1504 the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned
1505 on the first byte of character or just after the last byte of a character.
1510 PERL_STATIC_INLINE U8 *
1511 Perl_utf8_hop(const U8 *s, SSize_t off)
1513 PERL_ARGS_ASSERT_UTF8_HOP;
1515 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1516 * the bitops (especially ~) can create illegal UTF-8.
1517 * In other words: in Perl UTF-8 is not just for Unicode. */
1526 while (UTF8_IS_CONTINUATION(*s))
1530 GCC_DIAG_IGNORE(-Wcast-qual)
1536 =for apidoc utf8_hop_forward
1538 Return the UTF-8 pointer C<s> displaced by up to C<off> characters,
1541 C<off> must be non-negative.
1543 C<s> must be before or equal to C<end>.
1545 When moving forward it will not move beyond C<end>.
1547 Will not exceed this limit even if the string is not valid "UTF-8".
1552 PERL_STATIC_INLINE U8 *
1553 Perl_utf8_hop_forward(const U8 *s, SSize_t off, const U8 *end)
1555 PERL_ARGS_ASSERT_UTF8_HOP_FORWARD;
1557 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1558 * the bitops (especially ~) can create illegal UTF-8.
1559 * In other words: in Perl UTF-8 is not just for Unicode. */
1565 STRLEN skip = UTF8SKIP(s);
1566 if ((STRLEN)(end - s) <= skip) {
1567 GCC_DIAG_IGNORE(-Wcast-qual)
1574 GCC_DIAG_IGNORE(-Wcast-qual)
1580 =for apidoc utf8_hop_back
1582 Return the UTF-8 pointer C<s> displaced by up to C<off> characters,
1585 C<off> must be non-positive.
1587 C<s> must be after or equal to C<start>.
1589 When moving backward it will not move before C<start>.
1591 Will not exceed this limit even if the string is not valid "UTF-8".
1596 PERL_STATIC_INLINE U8 *
1597 Perl_utf8_hop_back(const U8 *s, SSize_t off, const U8 *start)
1599 PERL_ARGS_ASSERT_UTF8_HOP_BACK;
1601 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1602 * the bitops (especially ~) can create illegal UTF-8.
1603 * In other words: in Perl UTF-8 is not just for Unicode. */
1608 while (off++ && s > start) {
1611 } while (UTF8_IS_CONTINUATION(*s) && s > start);
1614 GCC_DIAG_IGNORE(-Wcast-qual)
1620 =for apidoc utf8_hop_safe
1622 Return the UTF-8 pointer C<s> displaced by up to C<off> characters,
1623 either forward or backward.
1625 When moving backward it will not move before C<start>.
1627 When moving forward it will not move beyond C<end>.
1629 Will not exceed those limits even if the string is not valid "UTF-8".
1634 PERL_STATIC_INLINE U8 *
1635 Perl_utf8_hop_safe(const U8 *s, SSize_t off, const U8 *start, const U8 *end)
1637 PERL_ARGS_ASSERT_UTF8_HOP_SAFE;
1639 /* Note: cannot use UTF8_IS_...() too eagerly here since e.g
1640 * the bitops (especially ~) can create illegal UTF-8.
1641 * In other words: in Perl UTF-8 is not just for Unicode. */
1643 assert(start <= s && s <= end);
1646 return utf8_hop_forward(s, off, end);
1649 return utf8_hop_back(s, off, start);
1655 =for apidoc is_utf8_valid_partial_char
1657 Returns 0 if the sequence of bytes starting at C<s> and looking no further than
1658 S<C<e - 1>> is the UTF-8 encoding, as extended by Perl, for one or more code
1659 points. Otherwise, it returns 1 if there exists at least one non-empty
1660 sequence of bytes that when appended to sequence C<s>, starting at position
1661 C<e> causes the entire sequence to be the well-formed UTF-8 of some code point;
1662 otherwise returns 0.
1664 In other words this returns TRUE if C<s> points to a partial UTF-8-encoded code
1667 This is useful when a fixed-length buffer is being tested for being well-formed
1668 UTF-8, but the final few bytes in it don't comprise a full character; that is,
1669 it is split somewhere in the middle of the final code point's UTF-8
1670 representation. (Presumably when the buffer is refreshed with the next chunk
1671 of data, the new first bytes will complete the partial code point.) This
1672 function is used to verify that the final bytes in the current buffer are in
1673 fact the legal beginning of some code point, so that if they aren't, the
1674 failure can be signalled without having to wait for the next read.
1678 #define is_utf8_valid_partial_char(s, e) \
1679 is_utf8_valid_partial_char_flags(s, e, 0)
1683 =for apidoc is_utf8_valid_partial_char_flags
1685 Like C<L</is_utf8_valid_partial_char>>, it returns a boolean giving whether
1686 or not the input is a valid UTF-8 encoded partial character, but it takes an
1687 extra parameter, C<flags>, which can further restrict which code points are
1690 If C<flags> is 0, this behaves identically to
1691 C<L</is_utf8_valid_partial_char>>. Otherwise C<flags> can be any combination
1692 of the C<UTF8_DISALLOW_I<foo>> flags accepted by C<L</utf8n_to_uvchr>>. If
1693 there is any sequence of bytes that can complete the input partial character in
1694 such a way that a non-prohibited character is formed, the function returns
1695 TRUE; otherwise FALSE. Non character code points cannot be determined based on
1696 partial character input. But many of the other possible excluded types can be
1697 determined from just the first one or two bytes.
1702 PERL_STATIC_INLINE bool
1703 Perl_is_utf8_valid_partial_char_flags(const U8 * const s, const U8 * const e, const U32 flags)
1705 PERL_ARGS_ASSERT_IS_UTF8_VALID_PARTIAL_CHAR_FLAGS;
1707 assert(0 == (flags & ~(UTF8_DISALLOW_ILLEGAL_INTERCHANGE
1708 |UTF8_DISALLOW_PERL_EXTENDED)));
1710 if (s >= e || s + UTF8SKIP(s) <= e) {
1714 return cBOOL(is_utf8_char_helper(s, e, flags));
1719 =for apidoc is_utf8_fixed_width_buf_flags
1721 Returns TRUE if the fixed-width buffer starting at C<s> with length C<len>
1722 is entirely valid UTF-8, subject to the restrictions given by C<flags>;
1723 otherwise it returns FALSE.
1725 If C<flags> is 0, any well-formed UTF-8, as extended by Perl, is accepted
1726 without restriction. If the final few bytes of the buffer do not form a
1727 complete code point, this will return TRUE anyway, provided that
1728 C<L</is_utf8_valid_partial_char_flags>> returns TRUE for them.
1730 If C<flags> in non-zero, it can be any combination of the
1731 C<UTF8_DISALLOW_I<foo>> flags accepted by C<L</utf8n_to_uvchr>>, and with the
1734 This function differs from C<L</is_utf8_string_flags>> only in that the latter
1735 returns FALSE if the final few bytes of the string don't form a complete code
1740 #define is_utf8_fixed_width_buf_flags(s, len, flags) \
1741 is_utf8_fixed_width_buf_loclen_flags(s, len, 0, 0, flags)
1745 =for apidoc is_utf8_fixed_width_buf_loc_flags
1747 Like C<L</is_utf8_fixed_width_buf_flags>> but stores the location of the
1748 failure in the C<ep> pointer. If the function returns TRUE, C<*ep> will point
1749 to the beginning of any partial character at the end of the buffer; if there is
1750 no partial character C<*ep> will contain C<s>+C<len>.
1752 See also C<L</is_utf8_fixed_width_buf_loclen_flags>>.
1757 #define is_utf8_fixed_width_buf_loc_flags(s, len, loc, flags) \
1758 is_utf8_fixed_width_buf_loclen_flags(s, len, loc, 0, flags)
1762 =for apidoc is_utf8_fixed_width_buf_loclen_flags
1764 Like C<L</is_utf8_fixed_width_buf_loc_flags>> but stores the number of
1765 complete, valid characters found in the C<el> pointer.
1770 PERL_STATIC_INLINE bool
1771 Perl_is_utf8_fixed_width_buf_loclen_flags(const U8 * const s,
1777 const U8 * maybe_partial;
1779 PERL_ARGS_ASSERT_IS_UTF8_FIXED_WIDTH_BUF_LOCLEN_FLAGS;
1782 ep = &maybe_partial;
1785 /* If it's entirely valid, return that; otherwise see if the only error is
1786 * that the final few bytes are for a partial character */
1787 return is_utf8_string_loclen_flags(s, len, ep, el, flags)
1788 || is_utf8_valid_partial_char_flags(*ep, s + len, flags);
1791 PERL_STATIC_INLINE UV
1792 Perl_utf8n_to_uvchr_msgs(const U8 *s,
1799 /* This is the inlined portion of utf8n_to_uvchr_msgs. It handles the
1800 * simple cases, and, if necessary calls a helper function to deal with the
1801 * more complex ones. Almost all well-formed non-problematic code points
1802 * are considered simple, so that it's unlikely that the helper function
1803 * will need to be called.
1805 * This is an adaptation of the tables and algorithm given in
1806 * https://bjoern.hoehrmann.de/utf-8/decoder/dfa/, which provides
1807 * comprehensive documentation of the original version. A copyright notice
1808 * for the original version is given at the beginning of this file. The
1809 * Perl adapation is documented at the definition of PL_strict_utf8_dfa_tab[].
1812 const U8 * const s0 = s;
1813 const U8 * send = s0 + curlen;
1814 UV uv = 0; /* The 0 silences some stupid compilers */
1817 PERL_ARGS_ASSERT_UTF8N_TO_UVCHR_MSGS;
1819 /* This dfa is fast. If it accepts the input, it was for a well-formed,
1820 * non-problematic code point, which can be returned immediately.
1821 * Otherwise we call a helper function to figure out the more complicated
1824 while (s < send && LIKELY(state != 1)) {
1825 UV type = PL_strict_utf8_dfa_tab[*s];
1828 ? ((0xff >> type) & NATIVE_UTF8_TO_I8(*s))
1829 : UTF8_ACCUMULATE(uv, *s);
1830 state = PL_strict_utf8_dfa_tab[256 + state + type];
1838 *retlen = s - s0 + 1;
1847 return UNI_TO_NATIVE(uv);
1850 /* Here is potentially problematic. Use the full mechanism */
1851 return _utf8n_to_uvchr_msgs_helper(s0, curlen, retlen, flags, errors, msgs);
1854 PERL_STATIC_INLINE UV
1855 Perl_utf8_to_uvchr_buf_helper(pTHX_ const U8 *s, const U8 *send, STRLEN *retlen)
1857 PERL_ARGS_ASSERT_UTF8_TO_UVCHR_BUF_HELPER;
1861 if (! ckWARN_d(WARN_UTF8)) {
1863 /* EMPTY is not really allowed, and asserts on debugging builds. But
1864 * on non-debugging we have to deal with it, and this causes it to
1865 * return the REPLACEMENT CHARACTER, as the documentation indicates */
1866 return utf8n_to_uvchr(s, send - s, retlen,
1867 (UTF8_ALLOW_ANY | UTF8_ALLOW_EMPTY));
1870 UV ret = utf8n_to_uvchr(s, send - s, retlen, 0);
1871 if (retlen && ret == 0 && *s != '\0') {
1872 *retlen = (STRLEN) -1;
1879 /* ------------------------------- perl.h ----------------------------- */
1882 =head1 Miscellaneous Functions
1884 =for apidoc is_safe_syscall
1886 Test that the given C<pv> (with length C<len>) doesn't contain any internal
1888 If it does, set C<errno> to C<ENOENT>, optionally warn using the C<syscalls>
1889 category, and return FALSE.
1891 Return TRUE if the name is safe.
1893 C<what> and C<op_name> are used in any warning.
1895 Used by the C<IS_SAFE_SYSCALL()> macro.
1900 PERL_STATIC_INLINE bool
1901 Perl_is_safe_syscall(pTHX_ const char *pv, STRLEN len, const char *what, const char *op_name)
1903 /* While the Windows CE API provides only UCS-16 (or UTF-16) APIs
1904 * perl itself uses xce*() functions which accept 8-bit strings.
1907 PERL_ARGS_ASSERT_IS_SAFE_SYSCALL;
1911 if (UNLIKELY((null_at = (char *)memchr(pv, 0, len-1)) != NULL)) {
1912 SETERRNO(ENOENT, LIB_INVARG);
1913 Perl_ck_warner(aTHX_ packWARN(WARN_SYSCALLS),
1914 "Invalid \\0 character in %s for %s: %s\\0%s",
1915 what, op_name, pv, null_at+1);
1925 Return true if the supplied filename has a newline character
1926 immediately before the first (hopefully only) NUL.
1928 My original look at this incorrectly used the len from SvPV(), but
1929 that's incorrect, since we allow for a NUL in pv[len-1].
1931 So instead, strlen() and work from there.
1933 This allow for the user reading a filename, forgetting to chomp it,
1936 open my $foo, "$file\0";
1942 PERL_STATIC_INLINE bool
1943 S_should_warn_nl(const char *pv)
1947 PERL_ARGS_ASSERT_SHOULD_WARN_NL;
1951 return len > 0 && pv[len-1] == '\n';
1956 #if defined(PERL_IN_PP_C) || defined(PERL_IN_PP_HOT_C)
1958 PERL_STATIC_INLINE bool
1959 S_lossless_NV_to_IV(const NV nv, IV *ivp)
1961 /* This function determines if the input NV 'nv' may be converted without
1962 * loss of data to an IV. If not, it returns FALSE taking no other action.
1963 * But if it is possible, it does the conversion, returning TRUE, and
1964 * storing the converted result in '*ivp' */
1966 PERL_ARGS_ASSERT_LOSSLESS_NV_TO_IV;
1968 # if defined(Perl_isnan)
1970 if (UNLIKELY(Perl_isnan(nv))) {
1976 if (UNLIKELY(nv < IV_MIN) || UNLIKELY(nv > IV_MAX)) {
1980 if ((IV) nv != nv) {
1990 /* ------------------ regcomp.c, toke.c ------------ */
1992 #if defined(PERL_IN_REGCOMP_C) || defined(PERL_IN_TOKE_C)
1995 - regcurly - a little FSA that accepts {\d+,?\d*}
1998 PERL_STATIC_INLINE bool
1999 S_regcurly(const char *s)
2001 PERL_ARGS_ASSERT_REGCURLY;
2020 /* ------------------ pp.c, regcomp.c, toke.c, universal.c ------------ */
2022 #if defined(PERL_IN_PP_C) || defined(PERL_IN_REGCOMP_C) || defined(PERL_IN_TOKE_C) || defined(PERL_IN_UNIVERSAL_C)
2024 #define MAX_CHARSET_NAME_LENGTH 2
2026 PERL_STATIC_INLINE const char *
2027 S_get_regex_charset_name(const U32 flags, STRLEN* const lenp)
2029 PERL_ARGS_ASSERT_GET_REGEX_CHARSET_NAME;
2031 /* Returns a string that corresponds to the name of the regex character set
2032 * given by 'flags', and *lenp is set the length of that string, which
2033 * cannot exceed MAX_CHARSET_NAME_LENGTH characters */
2036 switch (get_regex_charset(flags)) {
2037 case REGEX_DEPENDS_CHARSET: return DEPENDS_PAT_MODS;
2038 case REGEX_LOCALE_CHARSET: return LOCALE_PAT_MODS;
2039 case REGEX_UNICODE_CHARSET: return UNICODE_PAT_MODS;
2040 case REGEX_ASCII_RESTRICTED_CHARSET: return ASCII_RESTRICT_PAT_MODS;
2041 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
2043 return ASCII_MORE_RESTRICT_PAT_MODS;
2045 /* The NOT_REACHED; hides an assert() which has a rather complex
2046 * definition in perl.h. */
2047 NOT_REACHED; /* NOTREACHED */
2048 return "?"; /* Unknown */
2055 Return false if any get magic is on the SV other than taint magic.
2059 PERL_STATIC_INLINE bool
2060 Perl_sv_only_taint_gmagic(SV *sv)
2062 MAGIC *mg = SvMAGIC(sv);
2064 PERL_ARGS_ASSERT_SV_ONLY_TAINT_GMAGIC;
2067 if (mg->mg_type != PERL_MAGIC_taint
2068 && !(mg->mg_flags & MGf_GSKIP)
2069 && mg->mg_virtual->svt_get) {
2072 mg = mg->mg_moremagic;
2078 /* ------------------ cop.h ------------------------------------------- */
2080 /* implement GIMME_V() macro */
2082 PERL_STATIC_INLINE U8
2086 U8 gimme = (PL_op->op_flags & OPf_WANT);
2090 cxix = PL_curstackinfo->si_cxsubix;
2093 assert(cxstack[cxix].blk_gimme & G_WANT);
2094 return (cxstack[cxix].blk_gimme & G_WANT);
2098 /* Enter a block. Push a new base context and return its address. */
2100 PERL_STATIC_INLINE PERL_CONTEXT *
2101 Perl_cx_pushblock(pTHX_ U8 type, U8 gimme, SV** sp, I32 saveix)
2105 PERL_ARGS_ASSERT_CX_PUSHBLOCK;
2110 cx->blk_gimme = gimme;
2111 cx->blk_oldsaveix = saveix;
2112 cx->blk_oldsp = (I32)(sp - PL_stack_base);
2113 cx->blk_oldcop = PL_curcop;
2114 cx->blk_oldmarksp = (I32)(PL_markstack_ptr - PL_markstack);
2115 cx->blk_oldscopesp = PL_scopestack_ix;
2116 cx->blk_oldpm = PL_curpm;
2117 cx->blk_old_tmpsfloor = PL_tmps_floor;
2119 PL_tmps_floor = PL_tmps_ix;
2120 CX_DEBUG(cx, "PUSH");
2125 /* Exit a block (RETURN and LAST). */
2127 PERL_STATIC_INLINE void
2128 Perl_cx_popblock(pTHX_ PERL_CONTEXT *cx)
2130 PERL_ARGS_ASSERT_CX_POPBLOCK;
2132 CX_DEBUG(cx, "POP");
2133 /* these 3 are common to cx_popblock and cx_topblock */
2134 PL_markstack_ptr = PL_markstack + cx->blk_oldmarksp;
2135 PL_scopestack_ix = cx->blk_oldscopesp;
2136 PL_curpm = cx->blk_oldpm;
2138 /* LEAVE_SCOPE() should have made this true. /(?{})/ cheats
2139 * and leaves a CX entry lying around for repeated use, so
2140 * skip for multicall */ \
2141 assert( (CxTYPE(cx) == CXt_SUB && CxMULTICALL(cx))
2142 || PL_savestack_ix == cx->blk_oldsaveix);
2143 PL_curcop = cx->blk_oldcop;
2144 PL_tmps_floor = cx->blk_old_tmpsfloor;
2147 /* Continue a block elsewhere (e.g. NEXT, REDO, GOTO).
2148 * Whereas cx_popblock() restores the state to the point just before
2149 * cx_pushblock() was called, cx_topblock() restores it to the point just
2150 * *after* cx_pushblock() was called. */
2152 PERL_STATIC_INLINE void
2153 Perl_cx_topblock(pTHX_ PERL_CONTEXT *cx)
2155 PERL_ARGS_ASSERT_CX_TOPBLOCK;
2157 CX_DEBUG(cx, "TOP");
2158 /* these 3 are common to cx_popblock and cx_topblock */
2159 PL_markstack_ptr = PL_markstack + cx->blk_oldmarksp;
2160 PL_scopestack_ix = cx->blk_oldscopesp;
2161 PL_curpm = cx->blk_oldpm;
2163 PL_stack_sp = PL_stack_base + cx->blk_oldsp;
2167 PERL_STATIC_INLINE void
2168 Perl_cx_pushsub(pTHX_ PERL_CONTEXT *cx, CV *cv, OP *retop, bool hasargs)
2170 U8 phlags = CX_PUSHSUB_GET_LVALUE_MASK(Perl_was_lvalue_sub);
2172 PERL_ARGS_ASSERT_CX_PUSHSUB;
2174 PERL_DTRACE_PROBE_ENTRY(cv);
2175 cx->blk_sub.old_cxsubix = PL_curstackinfo->si_cxsubix;
2176 PL_curstackinfo->si_cxsubix = cx - PL_curstackinfo->si_cxstack;
2177 cx->blk_sub.cv = cv;
2178 cx->blk_sub.olddepth = CvDEPTH(cv);
2179 cx->blk_sub.prevcomppad = PL_comppad;
2180 cx->cx_type |= (hasargs) ? CXp_HASARGS : 0;
2181 cx->blk_sub.retop = retop;
2182 SvREFCNT_inc_simple_void_NN(cv);
2183 cx->blk_u16 = PL_op->op_private & (phlags|OPpDEREF);
2187 /* subsets of cx_popsub() */
2189 PERL_STATIC_INLINE void
2190 Perl_cx_popsub_common(pTHX_ PERL_CONTEXT *cx)
2194 PERL_ARGS_ASSERT_CX_POPSUB_COMMON;
2195 assert(CxTYPE(cx) == CXt_SUB);
2197 PL_comppad = cx->blk_sub.prevcomppad;
2198 PL_curpad = LIKELY(PL_comppad) ? AvARRAY(PL_comppad) : NULL;
2199 cv = cx->blk_sub.cv;
2200 CvDEPTH(cv) = cx->blk_sub.olddepth;
2201 cx->blk_sub.cv = NULL;
2203 PL_curstackinfo->si_cxsubix = cx->blk_sub.old_cxsubix;
2207 /* handle the @_ part of leaving a sub */
2209 PERL_STATIC_INLINE void
2210 Perl_cx_popsub_args(pTHX_ PERL_CONTEXT *cx)
2214 PERL_ARGS_ASSERT_CX_POPSUB_ARGS;
2215 assert(CxTYPE(cx) == CXt_SUB);
2216 assert(AvARRAY(MUTABLE_AV(
2217 PadlistARRAY(CvPADLIST(cx->blk_sub.cv))[
2218 CvDEPTH(cx->blk_sub.cv)])) == PL_curpad);
2220 CX_POP_SAVEARRAY(cx);
2221 av = MUTABLE_AV(PAD_SVl(0));
2222 if (UNLIKELY(AvREAL(av)))
2223 /* abandon @_ if it got reified */
2224 clear_defarray(av, 0);
2231 PERL_STATIC_INLINE void
2232 Perl_cx_popsub(pTHX_ PERL_CONTEXT *cx)
2234 PERL_ARGS_ASSERT_CX_POPSUB;
2235 assert(CxTYPE(cx) == CXt_SUB);
2237 PERL_DTRACE_PROBE_RETURN(cx->blk_sub.cv);
2241 cx_popsub_common(cx);
2245 PERL_STATIC_INLINE void
2246 Perl_cx_pushformat(pTHX_ PERL_CONTEXT *cx, CV *cv, OP *retop, GV *gv)
2248 PERL_ARGS_ASSERT_CX_PUSHFORMAT;
2250 cx->blk_format.old_cxsubix = PL_curstackinfo->si_cxsubix;
2251 PL_curstackinfo->si_cxsubix= cx - PL_curstackinfo->si_cxstack;
2252 cx->blk_format.cv = cv;
2253 cx->blk_format.retop = retop;
2254 cx->blk_format.gv = gv;
2255 cx->blk_format.dfoutgv = PL_defoutgv;
2256 cx->blk_format.prevcomppad = PL_comppad;
2259 SvREFCNT_inc_simple_void_NN(cv);
2261 SvREFCNT_inc_void(cx->blk_format.dfoutgv);
2265 PERL_STATIC_INLINE void
2266 Perl_cx_popformat(pTHX_ PERL_CONTEXT *cx)
2271 PERL_ARGS_ASSERT_CX_POPFORMAT;
2272 assert(CxTYPE(cx) == CXt_FORMAT);
2274 dfout = cx->blk_format.dfoutgv;
2276 cx->blk_format.dfoutgv = NULL;
2277 SvREFCNT_dec_NN(dfout);
2279 PL_comppad = cx->blk_format.prevcomppad;
2280 PL_curpad = LIKELY(PL_comppad) ? AvARRAY(PL_comppad) : NULL;
2281 cv = cx->blk_format.cv;
2282 cx->blk_format.cv = NULL;
2284 SvREFCNT_dec_NN(cv);
2285 PL_curstackinfo->si_cxsubix = cx->blk_format.old_cxsubix;
2289 PERL_STATIC_INLINE void
2290 Perl_cx_pusheval(pTHX_ PERL_CONTEXT *cx, OP *retop, SV *namesv)
2292 PERL_ARGS_ASSERT_CX_PUSHEVAL;
2294 cx->blk_eval.old_cxsubix = PL_curstackinfo->si_cxsubix;
2295 PL_curstackinfo->si_cxsubix= cx - PL_curstackinfo->si_cxstack;
2296 cx->blk_eval.retop = retop;
2297 cx->blk_eval.old_namesv = namesv;
2298 cx->blk_eval.old_eval_root = PL_eval_root;
2299 cx->blk_eval.cur_text = PL_parser ? PL_parser->linestr : NULL;
2300 cx->blk_eval.cv = NULL; /* later set by doeval_compile() */
2301 cx->blk_eval.cur_top_env = PL_top_env;
2303 assert(!(PL_in_eval & ~ 0x3F));
2304 assert(!(PL_op->op_type & ~0x1FF));
2305 cx->blk_u16 = (PL_in_eval & 0x3F) | ((U16)PL_op->op_type << 7);
2309 PERL_STATIC_INLINE void
2310 Perl_cx_popeval(pTHX_ PERL_CONTEXT *cx)
2314 PERL_ARGS_ASSERT_CX_POPEVAL;
2315 assert(CxTYPE(cx) == CXt_EVAL);
2317 PL_in_eval = CxOLD_IN_EVAL(cx);
2318 assert(!(PL_in_eval & 0xc0));
2319 PL_eval_root = cx->blk_eval.old_eval_root;
2320 sv = cx->blk_eval.cur_text;
2321 if (sv && CxEVAL_TXT_REFCNTED(cx)) {
2322 cx->blk_eval.cur_text = NULL;
2323 SvREFCNT_dec_NN(sv);
2326 sv = cx->blk_eval.old_namesv;
2328 cx->blk_eval.old_namesv = NULL;
2329 SvREFCNT_dec_NN(sv);
2331 PL_curstackinfo->si_cxsubix = cx->blk_eval.old_cxsubix;
2335 /* push a plain loop, i.e.
2337 * while (cond) { block }
2338 * for (init;cond;continue) { block }
2339 * This loop can be last/redo'ed etc.
2342 PERL_STATIC_INLINE void
2343 Perl_cx_pushloop_plain(pTHX_ PERL_CONTEXT *cx)
2345 PERL_ARGS_ASSERT_CX_PUSHLOOP_PLAIN;
2346 cx->blk_loop.my_op = cLOOP;
2350 /* push a true for loop, i.e.
2351 * for var (list) { block }
2354 PERL_STATIC_INLINE void
2355 Perl_cx_pushloop_for(pTHX_ PERL_CONTEXT *cx, void *itervarp, SV* itersave)
2357 PERL_ARGS_ASSERT_CX_PUSHLOOP_FOR;
2359 /* this one line is common with cx_pushloop_plain */
2360 cx->blk_loop.my_op = cLOOP;
2362 cx->blk_loop.itervar_u.svp = (SV**)itervarp;
2363 cx->blk_loop.itersave = itersave;
2365 cx->blk_loop.oldcomppad = PL_comppad;
2370 /* pop all loop types, including plain */
2372 PERL_STATIC_INLINE void
2373 Perl_cx_poploop(pTHX_ PERL_CONTEXT *cx)
2375 PERL_ARGS_ASSERT_CX_POPLOOP;
2377 assert(CxTYPE_is_LOOP(cx));
2378 if ( CxTYPE(cx) == CXt_LOOP_ARY
2379 || CxTYPE(cx) == CXt_LOOP_LAZYSV)
2381 /* Free ary or cur. This assumes that state_u.ary.ary
2382 * aligns with state_u.lazysv.cur. See cx_dup() */
2383 SV *sv = cx->blk_loop.state_u.lazysv.cur;
2384 cx->blk_loop.state_u.lazysv.cur = NULL;
2385 SvREFCNT_dec_NN(sv);
2386 if (CxTYPE(cx) == CXt_LOOP_LAZYSV) {
2387 sv = cx->blk_loop.state_u.lazysv.end;
2388 cx->blk_loop.state_u.lazysv.end = NULL;
2389 SvREFCNT_dec_NN(sv);
2392 if (cx->cx_type & (CXp_FOR_PAD|CXp_FOR_GV)) {
2394 SV **svp = (cx)->blk_loop.itervar_u.svp;
2395 if ((cx->cx_type & CXp_FOR_GV))
2396 svp = &GvSV((GV*)svp);
2398 *svp = cx->blk_loop.itersave;
2399 cx->blk_loop.itersave = NULL;
2400 SvREFCNT_dec(cursv);
2405 PERL_STATIC_INLINE void
2406 Perl_cx_pushwhen(pTHX_ PERL_CONTEXT *cx)
2408 PERL_ARGS_ASSERT_CX_PUSHWHEN;
2410 cx->blk_givwhen.leave_op = cLOGOP->op_other;
2414 PERL_STATIC_INLINE void
2415 Perl_cx_popwhen(pTHX_ PERL_CONTEXT *cx)
2417 PERL_ARGS_ASSERT_CX_POPWHEN;
2418 assert(CxTYPE(cx) == CXt_WHEN);
2420 PERL_UNUSED_ARG(cx);
2421 PERL_UNUSED_CONTEXT;
2422 /* currently NOOP */
2426 PERL_STATIC_INLINE void
2427 Perl_cx_pushgiven(pTHX_ PERL_CONTEXT *cx, SV *orig_defsv)
2429 PERL_ARGS_ASSERT_CX_PUSHGIVEN;
2431 cx->blk_givwhen.leave_op = cLOGOP->op_other;
2432 cx->blk_givwhen.defsv_save = orig_defsv;
2436 PERL_STATIC_INLINE void
2437 Perl_cx_popgiven(pTHX_ PERL_CONTEXT *cx)
2441 PERL_ARGS_ASSERT_CX_POPGIVEN;
2442 assert(CxTYPE(cx) == CXt_GIVEN);
2444 sv = GvSV(PL_defgv);
2445 GvSV(PL_defgv) = cx->blk_givwhen.defsv_save;
2446 cx->blk_givwhen.defsv_save = NULL;
2450 /* ------------------ util.h ------------------------------------------- */
2453 =head1 Miscellaneous Functions
2457 Returns true if the leading C<len> bytes of the strings C<s1> and C<s2> are the
2459 case-insensitively; false otherwise. Uppercase and lowercase ASCII range bytes
2460 match themselves and their opposite case counterparts. Non-cased and non-ASCII
2461 range bytes match only themselves.
2466 PERL_STATIC_INLINE I32
2467 Perl_foldEQ(const char *s1, const char *s2, I32 len)
2469 const U8 *a = (const U8 *)s1;
2470 const U8 *b = (const U8 *)s2;
2472 PERL_ARGS_ASSERT_FOLDEQ;
2477 if (*a != *b && *a != PL_fold[*b])
2484 PERL_STATIC_INLINE I32
2485 Perl_foldEQ_latin1(const char *s1, const char *s2, I32 len)
2487 /* Compare non-UTF-8 using Unicode (Latin1) semantics. Works on all folds
2488 * representable without UTF-8, except for LATIN_SMALL_LETTER_SHARP_S, and
2489 * does not check for this. Nor does it check that the strings each have
2490 * at least 'len' characters. */
2492 const U8 *a = (const U8 *)s1;
2493 const U8 *b = (const U8 *)s2;
2495 PERL_ARGS_ASSERT_FOLDEQ_LATIN1;
2500 if (*a != *b && *a != PL_fold_latin1[*b]) {
2509 =for apidoc foldEQ_locale
2511 Returns true if the leading C<len> bytes of the strings C<s1> and C<s2> are the
2512 same case-insensitively in the current locale; false otherwise.
2517 PERL_STATIC_INLINE I32
2518 Perl_foldEQ_locale(const char *s1, const char *s2, I32 len)
2521 const U8 *a = (const U8 *)s1;
2522 const U8 *b = (const U8 *)s2;
2524 PERL_ARGS_ASSERT_FOLDEQ_LOCALE;
2529 if (*a != *b && *a != PL_fold_locale[*b])
2537 =for apidoc my_strnlen
2539 The C library C<strnlen> if available, or a Perl implementation of it.
2541 C<my_strnlen()> computes the length of the string, up to C<maxlen>
2542 characters. It will will never attempt to address more than C<maxlen>
2543 characters, making it suitable for use with strings that are not
2544 guaranteed to be NUL-terminated.
2548 Description stolen from http://man.openbsd.org/strnlen.3,
2549 implementation stolen from PostgreSQL.
2553 PERL_STATIC_INLINE Size_t
2554 Perl_my_strnlen(const char *str, Size_t maxlen)
2556 const char *end = (char *) memchr(str, '\0', maxlen);
2558 PERL_ARGS_ASSERT_MY_STRNLEN;
2560 if (end == NULL) return maxlen;
2566 #if ! defined (HAS_MEMRCHR) && (defined(PERL_CORE) || defined(PERL_EXT))
2568 PERL_STATIC_INLINE void *
2569 S_my_memrchr(const char * s, const char c, const STRLEN len)
2571 /* memrchr(), since many platforms lack it */
2573 const char * t = s + len - 1;
2575 PERL_ARGS_ASSERT_MY_MEMRCHR;
2589 PERL_STATIC_INLINE char *
2590 Perl_mortal_getenv(const char * str)
2592 /* This implements a (mostly) thread-safe, sequential-call-safe getenv().
2594 * It's (mostly) thread-safe because it uses a mutex to prevent
2595 * simultaneous access from other threads that use the same mutex, and
2596 * makes a copy of the result before releasing that mutex. All of the Perl
2597 * core uses that mutex, but, like all mutexes, everything has to cooperate
2598 * for it to completely work. It is possible for code from, say XS, to not
2599 * use this mutex, defeating the safety.
2601 * On some platforms, getenv() is not sequential-call-safe, because
2602 * subsequent calls destroy the static storage inside the C library
2603 * returned by an earlier call. The result must be copied or completely
2604 * acted upon before a subsequent getenv call. Those calls could come from
2605 * another thread. Again, making a copy while controlling the mutex
2606 * prevents these problems..
2608 * To prevent leaks, the copy is made by creating a new SV containing it,
2609 * mortalizing the SV, and returning the SV's string (the copy). Thus this
2610 * is a drop-in replacement for getenv().
2612 * A complication is that this can be called during phases where the
2613 * mortalization process isn't available. These are in interpreter
2614 * destruction or early in construction. khw believes that at these times
2615 * there shouldn't be anything else going on, so plain getenv is safe AS
2616 * LONG AS the caller acts on the return before calling it again. */
2621 PERL_ARGS_ASSERT_MORTAL_GETENV;
2623 /* Can't mortalize without stacks. khw believes that no other threads
2624 * should be running, so no need to lock things, and this may be during a
2625 * phase when locking isn't even available */
2626 if (UNLIKELY(PL_scopestack_ix == 0)) {
2635 ret = SvPVX(sv_2mortal(newSVpv(ret, 0)));
2643 * ex: set ts=8 sts=4 sw=4 et: