3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2000,
4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2012 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
11 /* IMPORTANT NOTE: Everything whose name begins with an underscore is for
12 * internal core Perl use only. */
14 #ifndef PERL_HANDY_H_ /* Guard against nested #inclusion */
18 # define Null(type) ((type)NULL)
23 =for apidoc AmU||Nullch
24 Null character pointer. (No longer available when C<PERL_CORE> is
27 =for apidoc AmU||Nullsv
28 Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
33 # define Nullch Null(char*)
34 # define Nullfp Null(PerlIO*)
35 # define Nullsv Null(SV*)
47 /* The MUTABLE_*() macros cast pointers to the types shown, in such a way
48 * (compiler permitting) that casting away const-ness will give a warning;
52 * AV *av1 = (AV*)sv; <== BAD: the const has been silently cast away
53 * AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
56 #if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
57 # define MUTABLE_PTR(p) ({ void *_p = (p); _p; })
59 # define MUTABLE_PTR(p) ((void *) (p))
62 #define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
63 #define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
64 #define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
65 #define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
66 #define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
67 #define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
69 #if defined(I_STDBOOL) && !defined(PERL_BOOL_AS_CHAR)
76 /* bool is built-in for g++-2.6.3 and later, which might be used
77 for extensions. <_G_config.h> defines _G_HAVE_BOOL, but we can't
78 be sure _G_config.h will be included before this file. _G_config.h
79 also defines _G_HAVE_BOOL for both gcc and g++, but only g++
80 actually has bool. Hence, _G_HAVE_BOOL is pretty useless for us.
81 g++ can be identified by __GNUG__.
82 Andy Dougherty February 2000
84 #ifdef __GNUG__ /* GNU g++ has bool built-in */
85 # ifndef PERL_BOOL_AS_CHAR
100 /* cast-to-bool. A simple (bool) cast may not do the right thing: if bool is
101 * defined as char for example, then the cast from int is
102 * implementation-defined (bool)!!(cbool) in a ternary triggers a bug in xlc on
104 #define cBOOL(cbool) ((cbool) ? (bool)1 : (bool)0)
106 /* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
107 * XXX Should really be a Configure probe, with HAS__FUNCTION__
108 * and FUNCTION__ as results.
109 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
110 #if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
111 # define FUNCTION__ __func__
112 #elif (defined(USING_MSVC6)) || /* MSVC6 has neither __func__ nor __FUNCTION and no good workarounds, either. */ \
113 (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
114 # define FUNCTION__ ""
116 # define FUNCTION__ __FUNCTION__ /* Common extension. */
119 /* XXX A note on the perl source internal type system. The
120 original intent was that I32 be *exactly* 32 bits.
122 Currently, we only guarantee that I32 is *at least* 32 bits.
123 Specifically, if int is 64 bits, then so is I32. (This is the case
124 for the Cray.) This has the advantage of meshing nicely with
125 standard library calls (where we pass an I32 and the library is
126 expecting an int), but the disadvantage that an I32 is not 32 bits.
127 Andy Dougherty August 1996
129 There is no guarantee that there is *any* integral type with
130 exactly 32 bits. It is perfectly legal for a system to have
131 sizeof(short) == sizeof(int) == sizeof(long) == 8.
133 Similarly, there is no guarantee that I16 and U16 have exactly 16
136 For dealing with issues that may arise from various 32/64-bit
137 systems, we will ask Configure to check out
139 SHORTSIZE == sizeof(short)
140 INTSIZE == sizeof(int)
141 LONGSIZE == sizeof(long)
142 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
143 PTRSIZE == sizeof(void *)
144 DOUBLESIZE == sizeof(double)
145 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
149 #ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
150 # include <inttypes.h>
151 # ifdef INT32_MIN_BROKEN
153 # define INT32_MIN (-2147483647-1)
155 # ifdef INT64_MIN_BROKEN
157 # define INT64_MIN (-9223372036854775807LL-1)
173 #if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
175 /* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
176 Please search CHAR_MAX in perl.h for further details. */
177 #define U8_MAX UINT8_MAX
178 #define U8_MIN UINT8_MIN
180 #define I16_MAX INT16_MAX
181 #define I16_MIN INT16_MIN
182 #define U16_MAX UINT16_MAX
183 #define U16_MIN UINT16_MIN
185 #define I32_MAX INT32_MAX
186 #define I32_MIN INT32_MIN
187 #ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
188 # define U32_MAX UINT32_MAX
190 # define U32_MAX 4294967295U
192 #define U32_MIN UINT32_MIN
196 /* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
197 Please search CHAR_MAX in perl.h for further details. */
198 #define U8_MAX PERL_UCHAR_MAX
199 #define U8_MIN PERL_UCHAR_MIN
201 #define I16_MAX PERL_SHORT_MAX
202 #define I16_MIN PERL_SHORT_MIN
203 #define U16_MAX PERL_USHORT_MAX
204 #define U16_MIN PERL_USHORT_MIN
207 # define I32_MAX PERL_INT_MAX
208 # define I32_MIN PERL_INT_MIN
209 # define U32_MAX PERL_UINT_MAX
210 # define U32_MIN PERL_UINT_MIN
212 # define I32_MAX PERL_LONG_MAX
213 # define I32_MIN PERL_LONG_MIN
214 # define U32_MAX PERL_ULONG_MAX
215 # define U32_MIN PERL_ULONG_MIN
220 /* log(2) (i.e., log base 10 of 2) is pretty close to 0.30103, just in case
221 * anyone is grepping for it */
222 #define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log10(2) =~ 146/485 */
223 #define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
224 #define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
226 /* Unused by core; should be deprecated */
227 #define Ctl(ch) ((ch) & 037)
229 #if defined(PERL_CORE) || defined(PERL_EXT)
231 # define MIN(a,b) ((a) < (b) ? (a) : (b))
234 # define MAX(a,b) ((a) > (b) ? (a) : (b))
238 /* This is a helper macro to avoid preprocessor issues, replaced by nothing
239 * unless under DEBUGGING, where it expands to an assert of its argument,
240 * followed by a comma (hence the comma operator). If we just used a straight
241 * assert(), we would get a comma with nothing before it when not DEBUGGING.
243 * We also use empty definition under Coverity since the __ASSERT__
244 * checks often check for things that Really Cannot Happen, and Coverity
245 * detects that and gets all excited. */
247 #if defined(DEBUGGING) && !defined(__COVERITY__)
248 # define __ASSERT_(statement) assert(statement),
250 # define __ASSERT_(statement)
254 =head1 SV-Body Allocation
256 =for apidoc Ama|SV*|newSVpvs|"literal string" s
257 Like C<newSVpvn>, but takes a literal string instead of a
260 =for apidoc Ama|SV*|newSVpvs_flags|"literal string" s|U32 flags
261 Like C<newSVpvn_flags>, but takes a literal string instead of
262 a string/length pair.
264 =for apidoc Ama|SV*|newSVpvs_share|"literal string" s
265 Like C<newSVpvn_share>, but takes a literal string instead of
266 a string/length pair and omits the hash parameter.
268 =for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string" s|I32 flags
269 Like C<sv_catpvn_flags>, but takes a literal string instead
270 of a string/length pair.
272 =for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string" s
273 Like C<sv_catpvn_nomg>, but takes a literal string instead of
274 a string/length pair.
276 =for apidoc Am|void|sv_catpvs|SV* sv|"literal string" s
277 Like C<sv_catpvn>, but takes a literal string instead of a
280 =for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string" s
281 Like C<sv_catpvn_mg>, but takes a literal string instead of a
284 =for apidoc Am|void|sv_setpvs|SV* sv|"literal string" s
285 Like C<sv_setpvn>, but takes a literal string instead of a
288 =for apidoc Am|void|sv_setpvs_mg|SV* sv|"literal string" s
289 Like C<sv_setpvn_mg>, but takes a literal string instead of a
292 =for apidoc Am|SV *|sv_setref_pvs|"literal string" s
293 Like C<sv_setref_pvn>, but takes a literal string instead of
294 a string/length pair.
296 =head1 Memory Management
298 =for apidoc Ama|char*|savepvs|"literal string" s
299 Like C<savepvn>, but takes a literal string instead of a
302 =for apidoc Ama|char*|savesharedpvs|"literal string" s
303 A version of C<savepvs()> which allocates the duplicate string in memory
304 which is shared between threads.
308 =for apidoc Am|HV*|gv_stashpvs|"literal string" name|I32 create
309 Like C<gv_stashpvn>, but takes a literal string instead of a
312 =head1 Hash Manipulation Functions
314 =for apidoc Am|SV**|hv_fetchs|HV* tb|"literal string" key|I32 lval
315 Like C<hv_fetch>, but takes a literal string instead of a
318 =for apidoc Am|SV**|hv_stores|HV* tb|"literal string" key|SV* val
319 Like C<hv_store>, but takes a literal string instead of a
321 and omits the hash parameter.
323 =head1 Lexer interface
325 =for apidoc Amx|void|lex_stuff_pvs|"literal string" pv|U32 flags
327 Like L</lex_stuff_pvn>, but takes a literal string instead of
328 a string/length pair.
333 /* concatenating with "" ensures that only literal strings are accepted as
335 #define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
337 /* note that STR_WITH_LEN() can't be used as argument to macros or functions
338 * that under some configurations might be macros, which means that it requires
339 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
342 /* STR_WITH_LEN() shortcuts */
343 #define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
344 #define newSVpvs_flags(str,flags) \
345 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
346 #define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
347 #define sv_catpvs_flags(sv, str, flags) \
348 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
349 #define sv_catpvs_nomg(sv, str) \
350 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
351 #define sv_catpvs(sv, str) \
352 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
353 #define sv_catpvs_mg(sv, str) \
354 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
355 #define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
356 #define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
357 #define sv_setref_pvs(rv, classname, str) \
358 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
359 #define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
360 #define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
361 #define gv_stashpvs(str, create) \
362 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
363 #define gv_fetchpvs(namebeg, add, sv_type) \
364 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
365 #define gv_fetchpvn(namebeg, len, add, sv_type) \
366 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
367 #define sv_catxmlpvs(dsv, str, utf8) \
368 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
371 #define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
373 #define get_cvs(str, flags) \
374 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
377 =head1 Miscellaneous Functions
379 =for apidoc Am|bool|strNE|char* s1|char* s2
380 Test two C<NUL>-terminated strings to see if they are different. Returns true
383 =for apidoc Am|bool|strEQ|char* s1|char* s2
384 Test two C<NUL>-terminated strings to see if they are equal. Returns true or
387 =for apidoc Am|bool|strLT|char* s1|char* s2
388 Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
389 second, C<s2>. Returns true or false.
391 =for apidoc Am|bool|strLE|char* s1|char* s2
392 Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
393 equal to the second, C<s2>. Returns true or false.
395 =for apidoc Am|bool|strGT|char* s1|char* s2
396 Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
397 the second, C<s2>. Returns true or false.
399 =for apidoc Am|bool|strGE|char* s1|char* s2
400 Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
401 or equal to the second, C<s2>. Returns true or false.
403 =for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
404 Test two C<NUL>-terminated strings to see if they are different. The C<len>
405 parameter indicates the number of bytes to compare. Returns true or false. (A
406 wrapper for C<strncmp>).
408 =for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
409 Test two C<NUL>-terminated strings to see if they are equal. The C<len>
410 parameter indicates the number of bytes to compare. Returns true or false. (A
411 wrapper for C<strncmp>).
413 =for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
414 Test two buffers (which may contain embedded C<NUL> characters, to see if they
415 are equal. The C<len> parameter indicates the number of bytes to compare.
416 Returns zero if equal, or non-zero if non-equal.
418 =for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
419 Test two buffers (which may contain embedded C<NUL> characters, to see if they
420 are not equal. The C<len> parameter indicates the number of bytes to compare.
421 Returns zero if non-equal, or non-zero if equal.
425 New macros should use the following conventions for their names (which are
426 based on the underlying C library functions):
428 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
430 Each has two main parameters, string-like operands that are compared
431 against each other, as specified by the macro name. Some macros may
432 additionally have one or potentially even two length parameters. If a length
433 parameter applies to both string parameters, it will be positioned third;
434 otherwise any length parameter immediately follows the string parameter it
437 If the prefix to the name is 'str', the string parameter is a pointer to a C
438 language string. Such a string does not contain embedded NUL bytes; its
439 length may be unknown, but can be calculated by C<strlen()>, since it is
440 terminated by a NUL, which isn't included in its length.
442 The optional 'n' following 'str' means that that there is a third parameter,
443 giving the maximum number of bytes to look at in each string. Even if both
444 strings are longer than the length parameter, those extra bytes will be
447 The 's' suffix means that the 2nd byte string parameter is a literal C
448 double-quoted string. Its length will automatically be calculated by the
449 macro, so no length parameter will ever be needed for it.
451 If the prefix is 'mem', the string parameters don't have to be C strings;
452 they may contain embedded NUL bytes, do not necessarily have a terminating
453 NUL, and their lengths can be known only through other means, which in
454 practice are additional parameter(s) passed to the function. All 'mem'
455 functions have at least one length parameter. Barring any 'l' or 's' suffix,
456 there is a single length parameter, in position 3, which applies to both
457 string parameters. The 's' suffix means, as described above, that the 2nd
458 string is a literal double-quoted C string (hence its length is calculated by
459 the macro, and the length parameter to the function applies just to the first
460 string parameter, and hence is positioned just after it). An 'l' suffix
461 means that the 2nd string parameter has its own length parameter, and the
462 signature will look like memFOOl(s1, l1, s2, l2).
464 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
465 substring of the 1st string. 'P' if present indicates that the substring
466 must be a "proper" one in tha mathematical sense that the first one must be
467 strictly larger than the 2nd.
472 #define strNE(s1,s2) (strcmp(s1,s2) != 0)
473 #define strEQ(s1,s2) (strcmp(s1,s2) == 0)
474 #define strLT(s1,s2) (strcmp(s1,s2) < 0)
475 #define strLE(s1,s2) (strcmp(s1,s2) <= 0)
476 #define strGT(s1,s2) (strcmp(s1,s2) > 0)
477 #define strGE(s1,s2) (strcmp(s1,s2) >= 0)
479 #define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
480 #define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
482 #define memNE(s1,s2,l) (memcmp(s1,s2,l) != 0)
483 #define memEQ(s1,s2,l) (memcmp(s1,s2,l) == 0)
485 /* memEQ and memNE where second comparand is a string constant */
486 #define memEQs(s1, l, s2) \
487 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
488 #define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
490 /* Keep these private until we decide it was a good idea */
491 #if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
493 #define strBEGINs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1) == 0)
495 #define memBEGINs(s1, l, s2) \
496 ( (l) >= sizeof(s2) - 1 \
497 && memEQ(s1, "" s2 "", sizeof(s2)-1))
498 #define memBEGINPs(s1, l, s2) \
499 ( (l) > sizeof(s2) - 1 \
500 && memEQ(s1, "" s2 "", sizeof(s2)-1))
501 #define memENDs(s1, l, s2) \
502 ( (l) >= sizeof(s2) - 1 \
503 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
504 #define memENDPs(s1, l, s2) \
506 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
507 #endif /* End of making macros private */
509 #define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
510 #define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
511 #define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
512 #define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
517 * Unfortunately, the introduction of locales means that we
518 * can't trust isupper(), etc. to tell the truth. And when
519 * it comes to /\w+/ with tainting enabled, we *must* be able
520 * to trust our character classes.
522 * Therefore, the default tests in the text of Perl will be
523 * independent of locale. Any code that wants to depend on
524 * the current locale will use the tests that begin with "lc".
527 #ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
535 =head1 Character classification
536 This section is about functions (really macros) that classify characters
537 into types, such as punctuation versus alphabetic, etc. Most of these are
538 analogous to regular expression character classes. (See
539 L<perlrecharclass/POSIX Character Classes>.) There are several variants for
540 each class. (Not all macros have all variants; each item below lists the
541 ones valid for it.) None are affected by C<use bytes>, and only the ones
542 with C<LC> in the name are affected by the current locale.
544 The base function, e.g., C<isALPHA()>, takes an octet (either a C<char> or a
545 C<U8>) as input and returns a boolean as to whether or not the character
546 represented by that octet is (or on non-ASCII platforms, corresponds to) an
547 ASCII character in the named class based on platform, Unicode, and Perl rules.
548 If the input is a number that doesn't fit in an octet, FALSE is returned.
550 Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
551 with no suffix C<"_A">. This variant is used to emphasize by its name that
552 only ASCII-range characters can return TRUE.
554 Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
555 onto the platform. That is, the code points that are ASCII are unaffected,
556 since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
557 as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
558 true when called with the code point 0xDF, which is a word character in both
559 ASCII and EBCDIC (though it represents different characters in each).
561 Variant C<isI<FOO>_uvchr> is like the C<isI<FOO>_L1> variant, but accepts any UV code
562 point as input. If the code point is larger than 255, Unicode rules are used
563 to determine if it is in the character class. For example,
564 C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
565 WITH MACRON in Unicode, and is a word character.
567 Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
568 encoded strings. Each call classifies one character, even if the string
569 contains many. This variant takes two parameters. The first, C<p>, is a
570 pointer to the first byte of the character to be classified. (Recall that it
571 may take more than one byte to represent a character in UTF-8 strings.) The
572 second parameter, C<e>, points to anywhere in the string beyond the first
573 character, up to one byte past the end of the entire string. The suffix
574 C<_safe> in the function's name indicates that it will not attempt to read
575 beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
576 is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
577 character is malformed in some way, the program may croak, or the function may
578 return FALSE, at the discretion of the implementation, and subject to change in
581 Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
582 parameter, C<p>, which has the same meaning as the corresponding parameter does
583 in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
584 beyond the end of the string. Starting in Perl v5.30, it will take a second
585 parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
586 program that uses it will have to be changed to successfully compile. In the
587 meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
588 program will raise a deprecation warning, enabled by default. You can convert
589 your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
590 extra measure of protection, or you can wait until v5.30, when you'll be forced
591 to add the C<e> parameter.
593 Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants, but the
594 result is based on the current locale, which is what C<LC> in the name stands
595 for. If Perl can determine that the current locale is a UTF-8 locale, it uses
596 the published Unicode rules; otherwise, it uses the C library function that
597 gives the named classification. For example, C<isDIGIT_LC()> when not in a
598 UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
599 returned if the input won't fit into an octet. On some platforms where the C
600 library function is known to be defective, Perl changes its result to follow
601 the POSIX standard's rules.
603 Variant C<isI<FOO>_LC_uvchr> is like C<isI<FOO>_LC>, but is defined on any UV. It
604 returns the same as C<isI<FOO>_LC> for input code points less than 256, and
605 returns the hard-coded, not-affected-by-locale, Unicode results for larger ones.
607 Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
608 encoded strings. Each call classifies one character, even if the string
609 contains many. This variant takes two parameters. The first, C<p>, is a
610 pointer to the first byte of the character to be classified. (Recall that it
611 may take more than one byte to represent a character in UTF-8 strings.) The
612 second parameter, C<e>, points to anywhere in the string beyond the first
613 character, up to one byte past the end of the entire string. The suffix
614 C<_safe> in the function's name indicates that it will not attempt to read
615 beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
616 is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
617 character is malformed in some way, the program may croak, or the function may
618 return FALSE, at the discretion of the implementation, and subject to change in
621 Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
622 parameter, C<p>, which has the same meaning as the corresponding parameter does
623 in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
624 beyond the end of the string. Starting in Perl v5.30, it will take a second
625 parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
626 program that uses it will have to be changed to successfully compile. In the
627 meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
628 the program will raise a deprecation warning, enabled by default. You can
629 convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
630 and get an extra measure of protection, or you can wait until v5.30, when
631 you'll be forced to add the C<e> parameter.
633 =for apidoc Am|bool|isALPHA|char ch
634 Returns a boolean indicating whether the specified character is an
635 alphabetic character, analogous to C<m/[[:alpha:]]/>.
636 See the L<top of this section|/Character classification> for an explanation of
638 C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
639 C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
641 =for apidoc Am|bool|isALPHANUMERIC|char ch
642 Returns a boolean indicating whether the specified character is a either an
643 alphabetic character or decimal digit, analogous to C<m/[[:alnum:]]/>.
644 See the L<top of this section|/Character classification> for an explanation of
646 C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
647 C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
648 and C<isALPHANUMERIC_LC_utf8_safe>.
650 =for apidoc Am|bool|isASCII|char ch
651 Returns a boolean indicating whether the specified character is one of the 128
652 characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
653 On non-ASCII platforms, it returns TRUE iff this
654 character corresponds to an ASCII character. Variants C<isASCII_A()> and
655 C<isASCII_L1()> are identical to C<isASCII()>.
656 See the L<top of this section|/Character classification> for an explanation of
658 C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
659 C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
660 library routine C<isascii()>. In these cases, the variants whose names contain
661 C<LC> are the same as the corresponding ones without.
663 Also note, that because all ASCII characters are UTF-8 invariant (meaning they
664 have the exact same representation (always a single byte) whether encoded in
665 UTF-8 or not), C<isASCII> will give the correct results when called with any
666 byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
667 will work properly on any string encoded or not in UTF-8.
669 =for apidoc Am|bool|isBLANK|char ch
670 Returns a boolean indicating whether the specified character is a
671 character considered to be a blank, analogous to C<m/[[:blank:]]/>.
672 See the L<top of this section|/Character classification> for an explanation of
674 C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
675 C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
676 however, that some platforms do not have the C library routine
677 C<isblank()>. In these cases, the variants whose names contain C<LC> are
678 the same as the corresponding ones without.
680 =for apidoc Am|bool|isCNTRL|char ch
681 Returns a boolean indicating whether the specified character is a
682 control character, analogous to C<m/[[:cntrl:]]/>.
683 See the L<top of this section|/Character classification> for an explanation of
685 C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
686 C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
687 platforms, you almost always want to use the C<isCNTRL_L1> variant.
689 =for apidoc Am|bool|isDIGIT|char ch
690 Returns a boolean indicating whether the specified character is a
691 digit, analogous to C<m/[[:digit:]]/>.
692 Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
693 See the L<top of this section|/Character classification> for an explanation of
695 C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
696 C<isDIGIT_LC_utf8_safe>.
698 =for apidoc Am|bool|isGRAPH|char ch
699 Returns a boolean indicating whether the specified character is a
700 graphic character, analogous to C<m/[[:graph:]]/>.
701 See the L<top of this section|/Character classification> for an explanation of
702 variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
703 C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
705 =for apidoc Am|bool|isLOWER|char ch
706 Returns a boolean indicating whether the specified character is a
707 lowercase character, analogous to C<m/[[:lower:]]/>.
708 See the L<top of this section|/Character classification> for an explanation of
710 C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
711 C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
713 =for apidoc Am|bool|isOCTAL|char ch
714 Returns a boolean indicating whether the specified character is an
716 The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
719 =for apidoc Am|bool|isPUNCT|char ch
720 Returns a boolean indicating whether the specified character is a
721 punctuation character, analogous to C<m/[[:punct:]]/>.
722 Note that the definition of what is punctuation isn't as
723 straightforward as one might desire. See L<perlrecharclass/POSIX Character
724 Classes> for details.
725 See the L<top of this section|/Character classification> for an explanation of
726 variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
727 C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
729 =for apidoc Am|bool|isSPACE|char ch
730 Returns a boolean indicating whether the specified character is a
731 whitespace character. This is analogous
732 to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
733 this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
734 locale forms of this macro (the ones with C<LC> in their names) matched
735 precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
736 in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
737 (See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
738 See the L<top of this section|/Character classification> for an explanation of
740 C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
741 C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
743 =for apidoc Am|bool|isPSXSPC|char ch
744 (short for Posix Space)
745 Starting in 5.18, this is identical in all its forms to the
746 corresponding C<isSPACE()> macros.
747 The locale forms of this macro are identical to their corresponding
748 C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
749 non-locale forms differ from their C<isSPACE()> forms only in that the
750 C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
751 Otherwise they are identical. Thus this macro is analogous to what
752 C<m/[[:space:]]/> matches in a regular expression.
753 See the L<top of this section|/Character classification> for an explanation of
754 variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
755 C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
757 =for apidoc Am|bool|isUPPER|char ch
758 Returns a boolean indicating whether the specified character is an
759 uppercase character, analogous to C<m/[[:upper:]]/>.
760 See the L<top of this section|/Character classification> for an explanation of
761 variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
762 C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
764 =for apidoc Am|bool|isPRINT|char ch
765 Returns a boolean indicating whether the specified character is a
766 printable character, analogous to C<m/[[:print:]]/>.
767 See the L<top of this section|/Character classification> for an explanation of
769 C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
770 C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
772 =for apidoc Am|bool|isWORDCHAR|char ch
773 Returns a boolean indicating whether the specified character is a character
774 that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
775 in a regular expression. A word character is an alphabetic character, a
776 decimal digit, a connecting punctuation character (such as an underscore), or
777 a "mark" character that attaches to one of those (like some sort of accent).
778 C<isALNUM()> is a synonym provided for backward compatibility, even though a
779 word character includes more than the standard C language meaning of
781 See the L<top of this section|/Character classification> for an explanation of
782 variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
783 C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
784 C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
785 include the platform's native underscore.
787 =for apidoc Am|bool|isXDIGIT|char ch
788 Returns a boolean indicating whether the specified character is a hexadecimal
789 digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
790 and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
791 See the L<top of this section|/Character classification> for an explanation of
793 C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
794 and C<isXDIGIT_LC_utf8_safe>.
796 =for apidoc Am|bool|isIDFIRST|char ch
797 Returns a boolean indicating whether the specified character can be the first
798 character of an identifier. This is very close to, but not quite the same as
799 the official Unicode property C<XID_Start>. The difference is that this
800 returns true only if the input character also matches L</isWORDCHAR>.
801 See the L<top of this section|/Character classification> for an explanation of
803 C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
804 C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
806 =for apidoc Am|bool|isIDCONT|char ch
807 Returns a boolean indicating whether the specified character can be the
808 second or succeeding character of an identifier. This is very close to, but
809 not quite the same as the official Unicode property C<XID_Continue>. The
810 difference is that this returns true only if the input character also matches
811 L</isWORDCHAR>. See the L<top of this section|/Character classification> for
813 explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
814 C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
815 C<isIDCONT_LC_utf8_safe>.
817 =head1 Miscellaneous Functions
819 =for apidoc Am|U8|READ_XDIGIT|char str*
820 Returns the value of an ASCII-range hex digit and advances the string pointer.
821 Behaviour is only well defined when isXDIGIT(*str) is true.
823 =head1 Character case changing
824 Perl uses "full" Unicode case mappings. This means that converting a single
825 character to another case may result in a sequence of more than one character.
826 For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
827 character sequence C<SS>. This presents some complications The lowercase of
828 all characters in the range 0..255 is a single character, and thus
829 C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
830 return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
831 an API that does allow every possible legal result to be returned.) Likewise
832 no other function that is crippled by not being able to give the correct
833 results for the full range of possible inputs has been implemented here.
835 =for apidoc Am|U8|toUPPER|U8 ch
836 Converts the specified character to uppercase. If the input is anything but an
837 ASCII lowercase character, that input character itself is returned. Variant
838 C<toUPPER_A> is equivalent.
840 =for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
841 Converts the code point C<cp> to its uppercase version, and
842 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
843 point is interpreted as native if less than 256; otherwise as Unicode. Note
844 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
845 bytes since the uppercase version may be longer than the original character.
847 The first code point of the uppercased version is returned
848 (but note, as explained at L<the top of this section|/Character case
849 changing>, that there may be more.)
851 =for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
852 Converts the first UTF-8 encoded character in the sequence starting at C<p> and
853 extending no further than S<C<e - 1>> to its uppercase version, and
854 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
855 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
856 bytes since the uppercase version may be longer than the original character.
858 The first code point of the uppercased version is returned
859 (but note, as explained at L<the top of this section|/Character case
860 changing>, that there may be more).
862 The suffix C<_safe> in the function's name indicates that it will not attempt
863 to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
864 true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
865 input character is malformed in some way, the program may croak, or the
866 function may return the REPLACEMENT CHARACTER, at the discretion of the
867 implementation, and subject to change in future releases.
869 =for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
870 This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
871 parameter The function therefore can't check if it is reading
872 beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
873 parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
874 program that uses it will have to be changed to successfully compile. In the
875 meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
876 program will raise a deprecation warning, enabled by default. You can convert
877 your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
878 extra measure of protection, or you can wait until v5.30, when you'll be forced
879 to add the C<e> parameter.
881 =for apidoc Am|U8|toFOLD|U8 ch
882 Converts the specified character to foldcase. If the input is anything but an
883 ASCII uppercase character, that input character itself is returned. Variant
884 C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
885 Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
887 =for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
888 Converts the code point C<cp> to its foldcase version, and
889 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
890 point is interpreted as native if less than 256; otherwise as Unicode. Note
891 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
892 bytes since the foldcase version may be longer than the original character.
894 The first code point of the foldcased version is returned
895 (but note, as explained at L<the top of this section|/Character case
896 changing>, that there may be more).
898 =for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
899 Converts the first UTF-8 encoded character in the sequence starting at C<p> and
900 extending no further than S<C<e - 1>> to its foldcase version, and
901 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
902 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
903 bytes since the foldcase version may be longer than the original character.
905 The first code point of the foldcased version is returned
906 (but note, as explained at L<the top of this section|/Character case
907 changing>, that there may be more).
909 The suffix C<_safe> in the function's name indicates that it will not attempt
910 to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
911 true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
912 input character is malformed in some way, the program may croak, or the
913 function may return the REPLACEMENT CHARACTER, at the discretion of the
914 implementation, and subject to change in future releases.
916 =for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
917 This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
918 parameter The function therefore can't check if it is reading
919 beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
920 parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
921 program that uses it will have to be changed to successfully compile. In the
922 meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
923 program will raise a deprecation warning, enabled by default. You can convert
924 your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
925 extra measure of protection, or you can wait until v5.30, when you'll be forced
926 to add the C<e> parameter.
928 =for apidoc Am|U8|toLOWER|U8 ch
929 Converts the specified character to lowercase. If the input is anything but an
930 ASCII uppercase character, that input character itself is returned. Variant
931 C<toLOWER_A> is equivalent.
933 =for apidoc Am|U8|toLOWER_L1|U8 ch
934 Converts the specified Latin1 character to lowercase. The results are
935 undefined if the input doesn't fit in a byte.
937 =for apidoc Am|U8|toLOWER_LC|U8 ch
938 Converts the specified character to lowercase using the current locale's rules,
939 if possible; otherwise returns the input character itself.
941 =for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
942 Converts the code point C<cp> to its lowercase version, and
943 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
944 point is interpreted as native if less than 256; otherwise as Unicode. Note
945 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
946 bytes since the lowercase version may be longer than the original character.
948 The first code point of the lowercased version is returned
949 (but note, as explained at L<the top of this section|/Character case
950 changing>, that there may be more).
953 =for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
954 Converts the first UTF-8 encoded character in the sequence starting at C<p> and
955 extending no further than S<C<e - 1>> to its lowercase version, and
956 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
957 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
958 bytes since the lowercase version may be longer than the original character.
960 The first code point of the lowercased version is returned
961 (but note, as explained at L<the top of this section|/Character case
962 changing>, that there may be more).
964 The suffix C<_safe> in the function's name indicates that it will not attempt
965 to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
966 true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
967 input character is malformed in some way, the program may croak, or the
968 function may return the REPLACEMENT CHARACTER, at the discretion of the
969 implementation, and subject to change in future releases.
971 =for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
972 This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
973 parameter The function therefore can't check if it is reading
974 beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
975 parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
976 program that uses it will have to be changed to successfully compile. In the
977 meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
978 program will raise a deprecation warning, enabled by default. You can convert
979 your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
980 extra measure of protection, or you can wait until v5.30, when you'll be forced
981 to add the C<e> parameter.
983 =for apidoc Am|U8|toTITLE|U8 ch
984 Converts the specified character to titlecase. If the input is anything but an
985 ASCII lowercase character, that input character itself is returned. Variant
986 C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
987 range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
988 not a concept used in locale handling, so there is no functionality for that.)
990 =for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
991 Converts the code point C<cp> to its titlecase version, and
992 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
993 point is interpreted as native if less than 256; otherwise as Unicode. Note
994 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
995 bytes since the titlecase version may be longer than the original character.
997 The first code point of the titlecased version is returned
998 (but note, as explained at L<the top of this section|/Character case
999 changing>, that there may be more).
1001 =for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1002 Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1003 extending no further than S<C<e - 1>> to its titlecase version, and
1004 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1005 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1006 bytes since the titlecase version may be longer than the original character.
1008 The first code point of the titlecased version is returned
1009 (but note, as explained at L<the top of this section|/Character case
1010 changing>, that there may be more).
1012 The suffix C<_safe> in the function's name indicates that it will not attempt
1013 to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1014 true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1015 input character is malformed in some way, the program may croak, or the
1016 function may return the REPLACEMENT CHARACTER, at the discretion of the
1017 implementation, and subject to change in future releases.
1019 =for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1020 This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1021 parameter The function therefore can't check if it is reading
1022 beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
1023 parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1024 program that uses it will have to be changed to successfully compile. In the
1025 meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1026 program will raise a deprecation warning, enabled by default. You can convert
1027 your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
1028 extra measure of protection, or you can wait until v5.30, when you'll be forced
1029 to add the C<e> parameter.
1033 XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1034 really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1035 and aren't general purpose as they don't work on U+DF, and assert against that.
1037 Note that these macros are repeated in Devel::PPPort, so should also be
1038 patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1042 /* Specify the widest unsigned type on the platform. */
1044 # define WIDEST_UTYPE U64
1046 # define WIDEST_UTYPE U32
1049 /* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1050 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1051 * bits of information are lost even on a 64-bit machine, but to get the
1052 * compiler to optimize it out if possible. This is because Configure makes
1053 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1054 * sizeof() guarantees that this evaluates to a constant true at compile time.
1056 * For Coverity, be always true, because otherwise Coverity thinks
1057 * it finds several expressions that are always true, independent
1058 * of operands. Well, they are, but that is kind of the point.
1060 #ifndef __COVERITY__
1061 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1063 #define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
1064 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
1066 #define FITS_IN_8_BITS(c) (1)
1070 # ifndef _ALL_SOURCE
1071 /* The native libc isascii() et.al. functions return the wrong results
1072 * on at least z/OS unless this is defined. */
1073 # error _ALL_SOURCE should probably be defined
1076 /* There is a simple definition of ASCII for ASCII platforms. But the
1077 * EBCDIC one isn't so simple, so is defined using table look-up like the
1078 * other macros below.
1080 * The cast here is used instead of '(c) >= 0', because some compilers emit
1081 * a warning that that test is always true when the parameter is an
1082 * unsigned type. khw supposes that it could be written as
1083 * && ((c) == '\0' || (c) > 0)
1084 * to avoid the message, but the cast will likely avoid extra branches even
1085 * with stupid compilers.
1087 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1089 # define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
1092 /* Take the eight possible bit patterns of the lower 3 bits and you get the
1093 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1094 * can be ignored. If the rest match '0', we have an octal */
1095 #define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
1097 #ifdef H_PERL /* If have access to perl.h, lookup in its table */
1099 /* Character class numbers. For internal core Perl use only. The ones less
1100 * than 32 are used in PL_charclass[] and the ones up through the one that
1101 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1102 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1103 * their actual definitions are here. If that file has a name not used here,
1106 * The first group of these is ordered in what I (khw) estimate to be the
1107 * frequency of their use. This gives a slight edge to exiting a loop earlier
1108 * (in reginclass() in regexec.c) */
1109 # define _CC_WORDCHAR 0 /* \w and [:word:] */
1110 # define _CC_DIGIT 1 /* \d and [:digit:] */
1111 # define _CC_ALPHA 2 /* [:alpha:] */
1112 # define _CC_LOWER 3 /* [:lower:] */
1113 # define _CC_UPPER 4 /* [:upper:] */
1114 # define _CC_PUNCT 5 /* [:punct:] */
1115 # define _CC_PRINT 6 /* [:print:] */
1116 # define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1117 # define _CC_GRAPH 8 /* [:graph:] */
1118 # define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
1120 #define _FIRST_NON_SWASH_CC 10
1121 /* The character classes above are implemented with swashes. The second group
1122 * (just below) contains the ones implemented without. These are also sorted
1123 * in rough order of the frequency of their use, except that \v should be last,
1124 * as it isn't a real Posix character class, and some (small) inefficiencies in
1125 * regular expression handling would be introduced by putting it in the middle
1126 * of those that are. Also, cntrl and ascii come after the others as it may be
1127 * useful to group these which have no members that match above Latin1, (or
1128 * above ASCII in the latter case) */
1130 # define _CC_SPACE 10 /* \s, [:space:] */
1131 # define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1132 when the deprecated isFOO_utf8()
1133 functions are removed */
1134 # define _CC_BLANK 11 /* [:blank:] */
1135 # define _CC_XDIGIT 12 /* [:xdigit:] */
1136 # define _CC_CNTRL 13 /* [:cntrl:] */
1137 # define _CC_ASCII 14 /* [:ascii:] */
1138 # define _CC_VERTSPACE 15 /* \v */
1140 # define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1142 /* The members of the third group below do not need to be coordinated with data
1143 * structures in regcomp.[ch] and regexec.c. */
1144 # define _CC_IDFIRST 16
1145 # define _CC_CHARNAME_CONT 17
1146 # define _CC_NONLATIN1_FOLD 18
1147 # define _CC_NONLATIN1_SIMPLE_FOLD 19
1148 # define _CC_QUOTEMETA 20
1149 # define _CC_NON_FINAL_FOLD 21
1150 # define _CC_IS_IN_SOME_FOLD 22
1151 # define _CC_MNEMONIC_CNTRL 23
1153 # define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1154 isFOO_utf8() functions are removed */
1156 /* This next group is only used on EBCDIC platforms, so theoretically could be
1157 * shared with something entirely different that's only on ASCII platforms */
1158 # define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1159 # define _CC_UTF8_IS_START 29
1160 # define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1161 # define _CC_UTF8_IS_CONTINUATION 31
1163 * If more bits are needed, one could add a second word for non-64bit
1164 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
1165 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1166 * is used only for optimization (as of this writing), and differs in the
1167 * Latin1 range from the ALPHA bit only in two relatively unimportant
1168 * characters: the masculine and feminine ordinal indicators, so removing it
1169 * would just cause /i regexes which match them to run less efficiently.
1170 * Similarly the EBCDIC-only bits are used just for speed, and could be
1171 * replaced by other means */
1173 #if defined(PERL_CORE) || defined(PERL_EXT)
1174 /* An enum version of the character class numbers, to help compilers
1177 _CC_ENUM_ALPHA = _CC_ALPHA,
1178 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1179 _CC_ENUM_ASCII = _CC_ASCII,
1180 _CC_ENUM_BLANK = _CC_BLANK,
1181 _CC_ENUM_CASED = _CC_CASED,
1182 _CC_ENUM_CNTRL = _CC_CNTRL,
1183 _CC_ENUM_DIGIT = _CC_DIGIT,
1184 _CC_ENUM_GRAPH = _CC_GRAPH,
1185 _CC_ENUM_LOWER = _CC_LOWER,
1186 _CC_ENUM_PRINT = _CC_PRINT,
1187 _CC_ENUM_PUNCT = _CC_PUNCT,
1188 _CC_ENUM_SPACE = _CC_SPACE,
1189 _CC_ENUM_UPPER = _CC_UPPER,
1190 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
1191 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
1192 _CC_ENUM_XDIGIT = _CC_XDIGIT
1193 } _char_class_number;
1196 #define POSIX_SWASH_COUNT _FIRST_NON_SWASH_CC
1197 #define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
1199 #if defined(PERL_IN_UTF8_C) \
1200 || defined(PERL_IN_REGCOMP_C) \
1201 || defined(PERL_IN_REGEXEC_C)
1202 # if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 || _CC_LOWER != 3 \
1203 || _CC_UPPER != 4 || _CC_PUNCT != 5 || _CC_PRINT != 6 \
1204 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 || _CC_CASED != 9
1205 #error Need to adjust order of swash_property_names[]
1208 /* This is declared static in each of the few files that this is #defined for
1209 * to keep them from being publicly accessible. Hence there is a small amount
1210 * of wasted space */
1212 static const char* const swash_property_names[] = {
1228 EXTCONST U32 PL_charclass[] = {
1229 # include "l1_char_class_tab.h"
1232 # else /* ! DOINIT */
1233 EXTCONST U32 PL_charclass[];
1237 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
1238 # define _CC_mask(classnum) (1U << (classnum))
1240 /* For internal core Perl use only: the base macro for defining macros like
1242 # define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
1243 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
1245 /* The mask for the _A versions of the macros; it just adds in the bit for
1247 # define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1249 /* For internal core Perl use only: the base macro for defining macros like
1250 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1251 * the ASCII bit are present */
1252 # define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1253 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1254 == _CC_mask_A(classnum)))
1256 # define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
1257 # define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
1258 # define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1259 # define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
1260 # define isDIGIT_A(c) _generic_isCC(c, _CC_DIGIT) /* No non-ASCII digits */
1261 # define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1262 # define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1263 # define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
1264 # define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1265 # define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
1266 # define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1267 # define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
1268 # define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1270 # define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
1271 # define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1272 # define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1273 # define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1275 /* continuation character for legal NAME in \N{NAME} */
1276 # define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1278 # define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1279 # define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1280 # define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1281 # define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
1282 # define isPSXSPC_L1(c) isSPACE_L1(c)
1283 # define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1284 # define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1285 # define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1286 # define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1287 # define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
1290 # define isASCII(c) _generic_isCC(c, _CC_ASCII)
1293 /* Participates in a single-character fold with a character above 255 */
1294 # define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) ((! cBOOL(FITS_IN_8_BITS(c))) || (PL_charclass[(U8) (c)] & _CC_mask(_CC_NONLATIN1_SIMPLE_FOLD)))
1296 /* Like the above, but also can be part of a multi-char fold */
1297 # define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) ((! cBOOL(FITS_IN_8_BITS(c))) || (PL_charclass[(U8) (c)] & _CC_mask(_CC_NONLATIN1_FOLD)))
1299 # define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
1300 # define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1301 _generic_isCC(c, _CC_NON_FINAL_FOLD)
1302 # define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1303 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
1304 # define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1305 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
1306 #else /* else we don't have perl.h H_PERL */
1308 /* If we don't have perl.h, we are compiling a utility program. Below we
1309 * hard-code various macro definitions that wouldn't otherwise be available
1310 * to it. Most are coded based on first principles. These are written to
1311 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
1312 # define isDIGIT_A(c) ((c) <= '9' && (c) >= '0')
1313 # define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
1314 # define isSPACE_A(c) (isBLANK_A(c) \
1319 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1320 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1321 * performance (if optimization isn't on), and allow the same code to be
1322 * used for both platform types */
1323 # define isLOWER_A(c) ((c) >= 'a' && (c) <= 'z' \
1325 || ((c) >= 'j' && (c) <= 'r') \
1327 # define isUPPER_A(c) ((c) >= 'A' && (c) <= 'Z' \
1329 || ((c) >= 'J' && (c) <= 'R') \
1331 # define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1332 # define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
1333 # define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
1334 # define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
1335 # define isXDIGIT_A(c) (isDIGIT_A(c) \
1336 || ((c) >= 'a' && (c) <= 'f') \
1337 || ((c) <= 'F' && (c) >= 'A'))
1338 # define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1339 || (c) == '#' || (c) == '$' || (c) == '%' \
1340 || (c) == '&' || (c) == '\'' || (c) == '(' \
1341 || (c) == ')' || (c) == '*' || (c) == '+' \
1342 || (c) == ',' || (c) == '.' || (c) == '/' \
1343 || (c) == ':' || (c) == ';' || (c) == '<' \
1344 || (c) == '=' || (c) == '>' || (c) == '?' \
1345 || (c) == '@' || (c) == '[' || (c) == '\\' \
1346 || (c) == ']' || (c) == '^' || (c) == '_' \
1347 || (c) == '`' || (c) == '{' || (c) == '|' \
1348 || (c) == '}' || (c) == '~')
1349 # define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1350 # define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
1353 /* The below is accurate for the 3 EBCDIC code pages traditionally
1354 * supported by perl. The only difference between them in the controls
1355 * is the position of \n, and that is represented symbolically below */
1356 # define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1357 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1358 || (c) == '\t' || (c) == '\v' \
1359 || ((c) <= 3 && (c) >= 1) /* SOH, STX, ETX */ \
1360 || (c) == 7 /* U+7F DEL */ \
1361 || ((c) <= 0x13 && (c) >= 0x0E) /* SO, SI */ \
1362 /* DLE, DC[1-3] */ \
1363 || (c) == 0x18 /* U+18 CAN */ \
1364 || (c) == 0x19 /* U+19 EOM */ \
1365 || ((c) <= 0x1F && (c) >= 0x1C) /* [FGRU]S */ \
1366 || (c) == 0x26 /* U+17 ETB */ \
1367 || (c) == 0x27 /* U+1B ESC */ \
1368 || (c) == 0x2D /* U+05 ENQ */ \
1369 || (c) == 0x2E /* U+06 ACK */ \
1370 || (c) == 0x32 /* U+16 SYN */ \
1371 || (c) == 0x37 /* U+04 EOT */ \
1372 || (c) == 0x3C /* U+14 DC4 */ \
1373 || (c) == 0x3D /* U+15 NAK */ \
1374 || (c) == 0x3F)/* U+1A SUB */
1375 # define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
1376 # else /* isASCII is already defined for ASCII platforms, so can use that to
1378 # define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
1381 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
1382 * during debugging, and it seems best to keep them. We may be called
1383 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1384 * do anything anyway, so make it not a problem */
1385 # if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1386 # define NATIVE_TO_LATIN1(ch) (ch)
1388 # define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1389 # define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1390 # define isBLANK_L1(c) (isBLANK_A(c) \
1391 || (FITS_IN_8_BITS(c) \
1392 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1393 # define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1394 # define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1395 # define isLOWER_L1(c) (isLOWER_A(c) \
1396 || (FITS_IN_8_BITS(c) \
1397 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
1398 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1399 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1400 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1401 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1402 # define isPRINT_L1(c) (isPRINT_A(c) \
1403 || (FITS_IN_8_BITS(c) \
1404 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
1405 # define isPUNCT_L1(c) (isPUNCT_A(c) \
1406 || (FITS_IN_8_BITS(c) \
1407 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
1408 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1409 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1410 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1411 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1412 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1413 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1414 # define isSPACE_L1(c) (isSPACE_A(c) \
1415 || (FITS_IN_8_BITS(c) \
1416 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
1417 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1418 # define isUPPER_L1(c) (isUPPER_A(c) \
1419 || (FITS_IN_8_BITS(c) \
1420 && ( NATIVE_TO_LATIN1((U8) c) >= 0xC0 \
1421 && NATIVE_TO_LATIN1((U8) c) <= 0xDE \
1422 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1423 # define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1424 # define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1425 # define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1430 /* The following are not fully accurate in the above-ASCII range. I (khw)
1431 * don't think it's necessary to be so for the purposes where this gets
1433 # define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1434 # define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1436 /* And these aren't accurate at all. They are useful only for above
1437 * Latin1, which utilities and bootstrapping don't deal with */
1438 # define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
1439 # define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1440 # define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1442 /* Many of the macros later in this file are defined in terms of these. By
1443 * implementing them with a function, which converts the class number into
1444 * a call to the desired macro, all of the later ones work. However, that
1445 * function won't be actually defined when building a utility program (no
1446 * perl.h), and so a compiler error will be generated if one is attempted
1447 * to be used. And the above-Latin1 code points require Unicode tables to
1448 * be present, something unlikely to be the case when bootstrapping */
1449 # define _generic_isCC(c, classnum) \
1450 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1451 # define _generic_isCC_A(c, classnum) \
1452 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
1453 #endif /* End of no perl.h H_PERL */
1455 #define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1456 #define isALPHA(c) isALPHA_A(c)
1457 #define isASCII_A(c) isASCII(c)
1458 #define isASCII_L1(c) isASCII(c)
1459 #define isBLANK(c) isBLANK_A(c)
1460 #define isCNTRL(c) isCNTRL_A(c)
1461 #define isDIGIT(c) isDIGIT_A(c)
1462 #define isGRAPH(c) isGRAPH_A(c)
1463 #define isIDFIRST(c) isIDFIRST_A(c)
1464 #define isLOWER(c) isLOWER_A(c)
1465 #define isPRINT(c) isPRINT_A(c)
1466 #define isPSXSPC_A(c) isSPACE_A(c)
1467 #define isPSXSPC(c) isPSXSPC_A(c)
1468 #define isPSXSPC_L1(c) isSPACE_L1(c)
1469 #define isPUNCT(c) isPUNCT_A(c)
1470 #define isSPACE(c) isSPACE_A(c)
1471 #define isUPPER(c) isUPPER_A(c)
1472 #define isWORDCHAR(c) isWORDCHAR_A(c)
1473 #define isXDIGIT(c) isXDIGIT_A(c)
1475 /* ASCII casing. These could also be written as
1476 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1477 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1478 which uses table lookup and mask instead of subtraction. (This would
1479 work because the _MOD does not apply in the ASCII range) */
1480 #define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1481 #define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
1483 /* In the ASCII range, these are equivalent to what they're here defined to be.
1484 * But by creating these definitions, other code doesn't have to be aware of
1486 #define toFOLD(c) toLOWER(c)
1487 #define toTITLE(c) toUPPER(c)
1489 #define toLOWER_A(c) toLOWER(c)
1490 #define toUPPER_A(c) toUPPER(c)
1491 #define toFOLD_A(c) toFOLD(c)
1492 #define toTITLE_A(c) toTITLE(c)
1494 /* Use table lookup for speed; returns the input itself if is out-of-range */
1495 #define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
1497 : PL_latin1_lc[ (U8) (c) ])
1498 #define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1500 /* Modified uc. Is correct uc except for three non-ascii chars which are
1501 * all mapped to one of them, and these need special handling; returns the
1502 * input itself if is out-of-range */
1503 #define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
1505 : PL_mod_latin1_uc[ (U8) (c) ])
1506 #define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
1508 /* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1510 /* For internal core Perl use only: the base macro for defining macros like
1511 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
1512 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1513 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1514 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1515 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1516 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1517 * code points under locale rules, use the macros based on _generic_LC_uvchr
1518 * instead of this. */
1519 #define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1520 (! FITS_IN_8_BITS(c) \
1522 : IN_UTF8_CTYPE_LOCALE \
1523 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
1526 /* For internal core Perl use only: a helper macro for defining macros like
1527 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1528 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1529 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1530 * _generic_LC_base for more info */
1531 #define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1532 _generic_LC_base(c,utf8_locale_classnum, \
1533 non_utf8_func( (_LC_CAST) (c)))
1535 /* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1536 * 'c' is the platform's native underscore character */
1537 #define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1538 _generic_LC_base(c, utf8_locale_classnum, \
1539 (non_utf8_func( (_LC_CAST) (c)) \
1540 || (char)(c) == '_'))
1542 /* These next three are also for internal core Perl use only: case-change
1544 #define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1546 : (IN_UTF8_CTYPE_LOCALE) \
1547 ? PL_latin1_lc[ (U8) (c) ] \
1548 : (cast)function((cast)(c)))
1550 /* Note that the result can be larger than a byte in a UTF-8 locale. It
1551 * returns a single value, so can't adequately return the upper case of LATIN
1552 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1553 * values "SS"); instead it asserts against that under DEBUGGING, and
1554 * otherwise returns its input */
1555 #define _generic_toUPPER_LC(c, function, cast) \
1556 (! FITS_IN_8_BITS(c) \
1558 : ((! IN_UTF8_CTYPE_LOCALE) \
1559 ? (cast)function((cast)(c)) \
1560 : ((((U8)(c)) == MICRO_SIGN) \
1561 ? GREEK_CAPITAL_LETTER_MU \
1562 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1563 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1564 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1565 ? (__ASSERT_(0) (c)) \
1566 : PL_mod_latin1_uc[ (U8) (c) ])))))
1568 /* Note that the result can be larger than a byte in a UTF-8 locale. It
1569 * returns a single value, so can't adequately return the fold case of LATIN
1570 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1571 * values "ss"); instead it asserts against that under DEBUGGING, and
1572 * otherwise returns its input */
1573 #define _generic_toFOLD_LC(c, function, cast) \
1574 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
1575 ? GREEK_SMALL_LETTER_MU \
1576 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1577 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1578 _generic_toLOWER_LC(c, function, cast)))
1580 /* Use the libc versions for these if available. */
1581 #if defined(HAS_ISASCII)
1582 # define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1584 # define isASCII_LC(c) isASCII(c)
1587 #if defined(HAS_ISBLANK)
1588 # define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
1589 #else /* Unlike isASCII, varies if in a UTF-8 locale */
1590 # define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
1596 /* The Windows functions don't bother to follow the POSIX standard, which
1597 * for example says that something can't both be a printable and a control.
1598 * But Windows treats the \t control as a printable, and does such things
1599 * as making superscripts into both digits and punctuation. This tames
1600 * these flaws by assuming that the definitions of both controls and space
1601 * are correct, and then making sure that other definitions don't have
1602 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1603 * Not all possible weirdnesses are checked for, just the ones that were
1604 * detected on actual Microsoft code pages */
1606 # define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1607 # define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1609 # define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1610 && isALPHANUMERIC_LC(c))
1611 # define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1613 # define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1614 isALPHANUMERIC_LC(c))
1615 # define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1616 # define isIDFIRST_LC(c) (((c) == '_') \
1617 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1618 # define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1619 # define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1620 # define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1621 # define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
1622 # define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
1623 # define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1624 && isALPHANUMERIC_LC(c))
1626 # define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1627 # define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1628 # define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1630 #elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
1631 /* For most other platforms */
1633 # define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1634 # define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1635 # define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1636 # define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1637 # define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1638 # define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1639 # define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1640 # define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1641 # define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1642 # define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1643 # define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1644 # define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1645 # define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1648 # define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1649 # define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1650 # define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1652 #else /* The final fallback position */
1654 # define isALPHA_LC(c) (isascii(c) && isalpha(c))
1655 # define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1656 # define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1657 # define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1658 # define isGRAPH_LC(c) (isascii(c) && isgraph(c))
1659 # define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
1660 # define isLOWER_LC(c) (isascii(c) && islower(c))
1661 # define isPRINT_LC(c) (isascii(c) && isprint(c))
1662 # define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1663 # define isSPACE_LC(c) (isascii(c) && isspace(c))
1664 # define isUPPER_LC(c) (isascii(c) && isupper(c))
1665 # define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
1666 # define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
1668 # define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1669 # define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1670 # define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
1674 #define isIDCONT(c) isWORDCHAR(c)
1675 #define isIDCONT_A(c) isWORDCHAR_A(c)
1676 #define isIDCONT_L1(c) isWORDCHAR_L1(c)
1677 #define isIDCONT_LC(c) isWORDCHAR_LC(c)
1678 #define isPSXSPC_LC(c) isSPACE_LC(c)
1680 /* For internal core Perl use only: the base macros for defining macros like
1681 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1682 * number defined earlier in this file. _generic_uvchr() is used for POSIX
1683 * classes where there is a macro or function 'above_latin1' that takes the
1684 * single argument 'c' and returns the desired value. These exist for those
1685 * classes which have simple definitions, avoiding the overhead of a hash
1686 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
1687 * for classes where that overhead is faster than a direct lookup.
1688 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
1689 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1690 * have duplicate checks here, so could create versions of the macros that
1691 * don't, but experiments show that gcc optimizes them out anyway. */
1693 /* Note that all ignore 'use bytes' */
1694 #define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1695 ? _generic_isCC(c, classnum) \
1697 #define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1698 ? _generic_isCC(c, classnum) \
1699 : _is_uni_FOO(classnum, c))
1700 #define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1701 #define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1702 #define isASCII_uvchr(c) isASCII(c)
1703 #define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1704 #define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1705 #define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1706 #define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1707 #define isIDCONT_uvchr(c) \
1708 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1709 #define isIDFIRST_uvchr(c) \
1710 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
1711 #define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1712 #define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1714 #define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1715 #define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1716 #define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1718 #define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1719 #define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1720 #define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1721 #define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1723 #define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1724 #define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1725 #define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1726 #define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1728 /* For backwards compatibility, even though '_uni' should mean official Unicode
1729 * code points, in Perl it means native for those below 256 */
1730 #define isALPHA_uni(c) isALPHA_uvchr(c)
1731 #define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1732 #define isASCII_uni(c) isASCII_uvchr(c)
1733 #define isBLANK_uni(c) isBLANK_uvchr(c)
1734 #define isCNTRL_uni(c) isCNTRL_uvchr(c)
1735 #define isDIGIT_uni(c) isDIGIT_uvchr(c)
1736 #define isGRAPH_uni(c) isGRAPH_uvchr(c)
1737 #define isIDCONT_uni(c) isIDCONT_uvchr(c)
1738 #define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1739 #define isLOWER_uni(c) isLOWER_uvchr(c)
1740 #define isPRINT_uni(c) isPRINT_uvchr(c)
1741 #define isPUNCT_uni(c) isPUNCT_uvchr(c)
1742 #define isSPACE_uni(c) isSPACE_uvchr(c)
1743 #define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1744 #define isUPPER_uni(c) isUPPER_uvchr(c)
1745 #define isVERTWS_uni(c) isVERTWS_uvchr(c)
1746 #define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1747 #define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1748 #define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1749 #define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1750 #define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1751 #define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
1753 /* For internal core Perl use only: the base macros for defining macros like
1754 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
1755 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
1756 * simple class definitions; the other for more complex. These are like
1757 * _generic_uvchr, so see it for more info. */
1758 #define _generic_LC_uvchr(latin1, above_latin1, c) \
1759 (c < 256 ? latin1(c) : above_latin1(c))
1760 #define _generic_LC_swash_uvchr(latin1, classnum, c) \
1761 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1763 #define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1764 #define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1765 _CC_ALPHANUMERIC, c)
1766 #define isASCII_LC_uvchr(c) isASCII_LC(c)
1767 #define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1768 is_HORIZWS_cp_high, c)
1769 #define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
1770 #define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1771 #define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
1772 #define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
1773 _is_uni_perl_idcont, c)
1774 #define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
1775 _is_uni_perl_idstart, c)
1776 #define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1777 #define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
1778 #define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
1779 #define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
1780 #define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
1781 is_XPERLSPACE_cp_high, c)
1782 #define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
1783 #define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
1785 #define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1786 is_XDIGIT_cp_high, c)
1788 #define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
1790 /* For internal core Perl use only: the base macros for defining macros like
1791 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1792 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1793 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1794 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1795 * same representation whether utf8 or not. Note that it assumes that the utf8
1796 * has been validated, and ignores 'use bytes' */
1797 #define _base_generic_utf8(enum_name, name, p, use_locale ) \
1798 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1800 "is" STRINGIFY(name) "_utf8", \
1801 "is" STRINGIFY(name) "_utf8_safe", \
1802 1, use_locale, __FILE__,__LINE__)
1804 #define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
1806 /* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1807 * they don't otherwise go out of their way to look for malformed UTF-8. If
1808 * they can return accurate results without knowing if the input is otherwise
1809 * malformed, they do so. For example isASCII is accurate in spite of any
1810 * non-length malformations because it looks only at a single byte. Likewise
1811 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1812 * variant ones return FALSE. But, if the input has to be well-formed in order
1813 * for the results to be accurate, the macros will test and if malformed will
1814 * call a routine to die
1816 * Except for toke.c, the macros do assume that e > p, asserting that on
1817 * DEBUGGING builds. Much code that calls these depends on this being true,
1818 * for other reasons. toke.c is treated specially as using the regular
1819 * assertion breaks it in many ways. All strings that these operate on there
1820 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1821 * bunch of code in toke.c assumes that this is true, so the assertion allows
1823 #ifdef PERL_IN_TOKE_C
1824 # define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1826 # define _utf8_safe_assert(p,e) ((e) > (p))
1829 #define _generic_utf8_safe(classnum, p, e, above_latin1) \
1830 (__ASSERT_(_utf8_safe_assert(p, e)) \
1831 (UTF8_IS_INVARIANT(*(p))) \
1832 ? _generic_isCC(*(p), classnum) \
1833 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1834 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1835 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1837 : (_force_out_malformed_utf8_message( \
1838 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1840 /* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1841 * 'above_latin1' can be a macro */
1842 #define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1843 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1844 #define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1845 _generic_utf8_safe(classnum, p, e, \
1846 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1847 ? (_force_out_malformed_utf8_message( \
1848 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1850 /* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
1851 * 'above_latin1' parameter */
1852 #define _generic_swash_utf8_safe(classnum, p, e) \
1853 _generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
1855 /* Like the above, but should be used only when it is known that there are no
1856 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1857 * class is TRUE for. Hence it can skip the tests for this range.
1858 * 'above_latin1' should include its arguments */
1859 #define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1860 (__ASSERT_(_utf8_safe_assert(p, e)) \
1861 (UTF8_IS_INVARIANT(*(p))) \
1862 ? _generic_isCC(*(p), classnum) \
1863 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1864 ? 0 /* Note that doesn't check validity for latin1 */ \
1867 /* NOTE that some of these macros have very similar ones in regcharclass.h.
1868 * For example, there is (at the time of this writing) an 'is_SPACE_utf8()'
1869 * there, differing in name only by an underscore from the one here
1870 * 'isSPACE_utf8(). The difference is that the ones here are probably more
1871 * efficient and smaller, using an O(1) array lookup for Latin1-range code
1872 * points; the regcharclass.h ones are implemented as a series of
1873 * "if-else-if-else ..." */
1875 #define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1876 #define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1877 #define isASCII_utf8(p) _generic_utf8(ASCII, p)
1878 #define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1879 #define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1880 #define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1881 #define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1882 #define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1883 #define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1884 #define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1885 #define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1886 #define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1887 #define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1888 #define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1889 #define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1890 #define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1891 #define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1892 #define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
1894 #define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1895 #define isALPHANUMERIC_utf8_safe(p, e) \
1896 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1897 #define isASCII_utf8_safe(p, e) \
1898 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1900 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1901 #define isBLANK_utf8_safe(p, e) \
1902 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1905 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1906 * more efficient macro instead of the more general one */
1907 # define isCNTRL_utf8_safe(p, e) \
1908 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
1910 # define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
1913 #define isDIGIT_utf8_safe(p, e) \
1914 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1915 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1916 #define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1917 #define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1918 _is_utf8_perl_idcont_with_len, p, e)
1920 /* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1921 * IDFIRST is an alnum. See
1922 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
1923 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1924 * This used to be not the XID version, but we decided to go with the more
1925 * modern Unicode definition */
1926 #define isIDFIRST_utf8_safe(p, e) \
1927 _generic_func_utf8_safe(_CC_IDFIRST, \
1928 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1930 #define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1931 #define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1932 #define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1933 #define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1934 #define isSPACE_utf8_safe(p, e) \
1935 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1936 #define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1937 #define isVERTWS_utf8_safe(p, e) \
1938 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1939 #define isWORDCHAR_utf8_safe(p, e) \
1940 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1941 #define isXDIGIT_utf8_safe(p, e) \
1942 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
1943 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1944 ? (_force_out_malformed_utf8_message( \
1945 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1946 : is_XDIGIT_high(p)))
1948 #define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
1949 #define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
1950 #define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
1951 #define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
1953 /* For internal core use only, subject to change */
1954 #define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
1955 #define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
1956 #define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
1957 #define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
1959 #define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
1960 #define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
1961 #define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
1962 #define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
1964 /* For internal core Perl use only: the base macros for defining macros like
1965 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
1966 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
1967 * 'macro' parameter */
1968 #define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
1970 #define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
1971 #define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
1972 #define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
1973 #define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
1974 #define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
1975 #define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
1976 #define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
1977 #define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
1978 #define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
1979 #define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
1980 #define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
1981 #define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
1982 #define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
1983 #define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
1984 #define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
1985 #define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
1986 #define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
1988 /* For internal core Perl use only: the base macros for defining macros like
1989 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
1990 * point in 'p' is within the 0-255 range, it uses locale rules from the
1991 * passed-in 'macro' parameter */
1992 #define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
1993 (__ASSERT_(_utf8_safe_assert(p, e)) \
1994 (UTF8_IS_INVARIANT(*(p))) \
1996 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1997 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1998 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
1999 : (_force_out_malformed_utf8_message( \
2000 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2003 #define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2004 _generic_LC_utf8_safe(macro, p, e, \
2005 _is_utf8_FOO_with_len(classnum, p, e))
2007 #define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2008 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2010 #define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2011 _generic_LC_utf8_safe(classnum, p, e, \
2012 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2013 ? (_force_out_malformed_utf8_message( \
2014 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2017 #define isALPHANUMERIC_LC_utf8_safe(p, e) \
2018 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2019 _CC_ALPHANUMERIC, p, e)
2020 #define isALPHA_LC_utf8_safe(p, e) \
2021 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2022 #define isASCII_LC_utf8_safe(p, e) \
2023 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2024 #define isBLANK_LC_utf8_safe(p, e) \
2025 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2026 #define isCNTRL_LC_utf8_safe(p, e) \
2027 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2028 #define isDIGIT_LC_utf8_safe(p, e) \
2029 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2030 #define isGRAPH_LC_utf8_safe(p, e) \
2031 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2032 #define isIDCONT_LC_utf8_safe(p, e) \
2033 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2034 _is_utf8_perl_idcont_with_len, p, e)
2035 #define isIDFIRST_LC_utf8_safe(p, e) \
2036 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2037 _is_utf8_perl_idstart_with_len, p, e)
2038 #define isLOWER_LC_utf8_safe(p, e) \
2039 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2040 #define isPRINT_LC_utf8_safe(p, e) \
2041 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2042 #define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2043 #define isPUNCT_LC_utf8_safe(p, e) \
2044 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2045 #define isSPACE_LC_utf8_safe(p, e) \
2046 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2047 #define isUPPER_LC_utf8_safe(p, e) \
2048 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2049 #define isWORDCHAR_LC_utf8_safe(p, e) \
2050 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2051 #define isXDIGIT_LC_utf8_safe(p, e) \
2052 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
2054 /* Macros for backwards compatibility and for completeness when the ASCII and
2055 * Latin1 values are identical */
2056 #define isALPHAU(c) isALPHA_L1(c)
2057 #define isDIGIT_L1(c) isDIGIT_A(c)
2058 #define isOCTAL(c) isOCTAL_A(c)
2059 #define isOCTAL_L1(c) isOCTAL_A(c)
2060 #define isXDIGIT_L1(c) isXDIGIT_A(c)
2061 #define isALNUM(c) isWORDCHAR(c)
2062 #define isALNUMU(c) isWORDCHAR_L1(c)
2063 #define isALNUM_LC(c) isWORDCHAR_LC(c)
2064 #define isALNUM_uni(c) isWORDCHAR_uni(c)
2065 #define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
2066 #define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2067 #define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2068 #define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2069 #define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2070 #define isALNUMC(c) isALPHANUMERIC(c)
2071 #define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2072 #define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
2073 #define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
2074 #define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
2075 #define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
2077 /* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2078 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2079 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2080 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2081 * the outlier from the block that contains the other controls, just like
2082 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2083 * block. If it weren't special cased, it would yield a non-control.
2084 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2087 # define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2089 # define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2091 ? (UNLIKELY((c) == '?') \
2092 ? QUESTION_MARK_CTRL \
2093 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2094 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2096 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2099 /* Line numbers are unsigned, 32 bits. */
2101 #define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
2103 /* Helpful alias for version prescan */
2104 #define is_LAX_VERSION(a,b) \
2105 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2107 #define is_STRICT_VERSION(a,b) \
2108 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2110 #define BADVERSION(a,b,c) \
2116 /* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2117 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2118 * which is advanced. The input is validated only by an assert() in DEBUGGING
2119 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2120 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2121 #define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2124 #define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2128 /* Converts a character known to represent an octal digit (0-7) to its numeric
2129 * value. The input is validated only by an assert() in DEBUGGING builds. In
2130 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2131 #define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2133 /* Efficiently returns a boolean as to if two native characters are equivalent
2134 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2135 * the ALPHA in the name is to remind you of that. This is asserted() in
2136 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2137 * works (on valid input) for both non- and UTF-8-encoded bytes.
2139 * When one of the inputs is a compile-time constant and gets folded by the
2140 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2141 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2142 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
2143 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2144 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2145 * just a single 0, in the bit position where the upper- and lowercase differ.
2147 #define isALPHA_FOLD_EQ(c1, c2) \
2148 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2149 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2150 #define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2153 =head1 Memory Management
2155 =for apidoc Am|void|Newx|void* ptr|int nitems|type
2156 The XSUB-writer's interface to the C C<malloc> function.
2158 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2160 In 5.9.3, Newx() and friends replace the older New() API, and drops
2161 the first parameter, I<x>, a debug aid which allowed callers to identify
2162 themselves. This aid has been superseded by a new build option,
2163 PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
2164 there for use in XS modules supporting older perls.
2166 =for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
2167 The XSUB-writer's interface to the C C<malloc> function, with
2168 cast. See also C<L</Newx>>.
2170 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2172 =for apidoc Am|void|Newxz|void* ptr|int nitems|type
2173 The XSUB-writer's interface to the C C<malloc> function. The allocated
2174 memory is zeroed with C<memzero>. See also C<L</Newx>>.
2176 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2178 =for apidoc Am|void|Renew|void* ptr|int nitems|type
2179 The XSUB-writer's interface to the C C<realloc> function.
2181 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2183 =for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2184 The XSUB-writer's interface to the C C<realloc> function, with
2187 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2189 =for apidoc Am|void|Safefree|void* ptr
2190 The XSUB-writer's interface to the C C<free> function.
2192 This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
2194 =for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2195 The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
2196 source, C<dest> is the destination, C<nitems> is the number of items, and
2197 C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
2199 =for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
2200 Like C<Move> but returns C<dest>. Useful
2201 for encouraging compilers to tail-call
2204 =for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2205 The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
2206 source, C<dest> is the destination, C<nitems> is the number of items, and
2207 C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
2209 =for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2211 Like C<Copy> but returns C<dest>. Useful
2212 for encouraging compilers to tail-call
2215 =for apidoc Am|void|Zero|void* dest|int nitems|type
2217 The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2218 destination, C<nitems> is the number of items, and C<type> is the type.
2220 =for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2222 Like C<Zero> but returns dest. Useful
2223 for encouraging compilers to tail-call
2226 =for apidoc Am|void|StructCopy|type *src|type *dest|type
2227 This is an architecture-independent macro to copy one structure to another.
2229 =for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2231 Fill up memory with a byte pattern (a byte repeated over and over
2232 again) that hopefully catches attempts to access uninitialized memory.
2234 =for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2236 PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2238 =for apidoc Am|void|PoisonFree|void* dest|int nitems|type
2240 PoisonWith(0xEF) for catching access to freed memory.
2242 =for apidoc Am|void|Poison|void* dest|int nitems|type
2244 PoisonWith(0xEF) for catching access to freed memory.
2248 /* Maintained for backwards-compatibility only. Use newSV() instead. */
2250 #define NEWSV(x,len) newSV(len)
2253 #define MEM_SIZE_MAX ((MEM_SIZE)-1)
2255 #define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
2257 #ifdef PERL_MALLOC_WRAP
2259 /* This expression will be constant-folded at compile time. It checks
2260 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2261 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2262 * As well as avoiding the need for a run-time check in some cases, it's
2263 * designed to avoid compiler warnings like:
2264 * comparison is always false due to limited range of data type
2265 * It's mathematically equivalent to
2266 * max(n) * sizeof(t) > MEM_SIZE_MAX
2269 # define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
2270 ( sizeof(MEM_SIZE) < sizeof(n) \
2271 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
2273 /* This is written in a slightly odd way to avoid various spurious
2274 * compiler warnings. We *want* to write the expression as
2275 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2276 * (for some compile-time constant C), but even when the LHS
2277 * constant-folds to false at compile-time, g++ insists on emitting
2278 * warnings about the RHS (e.g. "comparison is always false"), so instead
2281 * (cond ? n : X) > C
2283 * where X is a constant with X > C always false. Choosing a value for X
2284 * is tricky. If 0, some compilers will complain about 0 > C always being
2285 * false; if 1, Coverity complains when n happens to be the constant value
2286 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2287 * for X and hope that nothing else whines.
2290 # define _MEM_WRAP_WILL_WRAP(n,t) \
2291 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2292 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
2294 # define MEM_WRAP_CHECK(n,t) \
2295 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2296 && (croak_memory_wrap(),0))
2298 # define MEM_WRAP_CHECK_1(n,t,a) \
2299 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2300 && (Perl_croak_nocontext("%s",(a)),0))
2302 #define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
2304 #define PERL_STRLEN_ROUNDUP(n) ((void)(((n) > MEM_SIZE_MAX - 2 * PERL_STRLEN_ROUNDUP_QUANTUM) ? (croak_memory_wrap(),0) : 0), _PERL_STRLEN_ROUNDUP_UNCHECKED(n))
2307 #define MEM_WRAP_CHECK(n,t)
2308 #define MEM_WRAP_CHECK_1(n,t,a)
2309 #define MEM_WRAP_CHECK_2(n,t,a,b)
2310 #define MEM_WRAP_CHECK_(n,t)
2312 #define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
2318 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
2319 * go through functions, which are handy for debugging breakpoints, but
2320 * which more importantly get the immediate calling environment (file and
2321 * line number, and C function name if available) passed in. This info can
2322 * then be used for logging the calls, for which one gets a sample
2323 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
2326 * - not all memory allocs get logged, only those
2327 * that go through Newx() and derivatives (while all
2328 * Safefrees do get logged)
2329 * - __FILE__ and __LINE__ do not work everywhere
2330 * - __func__ or __FUNCTION__ even less so
2331 * - I think more goes on after the perlio frees but
2332 * the thing is that STDERR gets closed (as do all
2333 * the file descriptors)
2334 * - no deeper calling stack than the caller of the Newx()
2335 * or the kind, but do I look like a C reflection/introspection
2337 * - the function prototypes for the logging functions
2338 * probably should maybe be somewhere else than handy.h
2339 * - one could consider inlining (macrofying) the logging
2340 * for speed, but I am too lazy
2341 * - one could imagine recording the allocations in a hash,
2342 * (keyed by the allocation address?), and maintain that
2343 * through reallocs and frees, but how to do that without
2344 * any News() happening...?
2345 * - lots of -Ddefines to get useful/controllable output
2346 * - lots of ENV reads
2350 # ifndef PERL_MEM_LOG_NOIMPL
2359 # if defined(PERL_IN_SV_C) /* those are only used in sv.c */
2360 void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2361 void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2368 #define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2369 #define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
2370 #define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
2373 #ifndef MEM_LOG_ALLOC
2374 #define MEM_LOG_ALLOC(n,t,a) (a)
2376 #ifndef MEM_LOG_REALLOC
2377 #define MEM_LOG_REALLOC(n,t,v,a) (a)
2379 #ifndef MEM_LOG_FREE
2380 #define MEM_LOG_FREE(a) (a)
2383 #define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2384 #define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2385 #define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
2388 /* pre 5.9.x compatibility */
2389 #define New(x,v,n,t) Newx(v,n,t)
2390 #define Newc(x,v,n,t,c) Newxc(v,n,t,c)
2391 #define Newz(x,v,n,t) Newxz(v,n,t)
2394 #define Renew(v,n,t) \
2395 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
2396 #define Renewc(v,n,t,c) \
2397 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
2400 #define Safefree(d) \
2401 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
2403 #define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
2406 /* assert that a valid ptr has been supplied - use this instead of assert(ptr) *
2407 * as it handles cases like constant string arguments without throwing warnings *
2408 * the cast is required, as is the inequality check, to avoid warnings */
2409 #define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
2412 #define Move(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), (void)memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2413 #define Copy(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), (void)memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2414 #define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2416 #define MoveD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2417 #define CopyD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), perl_assert_ptr(s), memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2418 #define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
2420 #define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2421 #define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2422 #define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2423 #define Poison(d,n,t) PoisonFree(d,n,t)
2426 # define PERL_POISON_EXPR(x) x
2428 # define PERL_POISON_EXPR(x)
2431 #define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2433 /* C_ARRAY_LENGTH is the number of elements in the C array (so you
2434 * want your zero-based indices to be less than but not equal to).
2436 * C_ARRAY_END is one past the last: half-open/half-closed range,
2437 * not last-inclusive range. */
2438 #define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
2439 #define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
2443 # define Perl_va_copy(s, d) va_copy(d, s)
2444 # elif defined(__va_copy)
2445 # define Perl_va_copy(s, d) __va_copy(d, s)
2447 # define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2451 /* convenience debug macros */
2453 #define pTHX_FORMAT "Perl interpreter: 0x%p"
2454 #define pTHX__FORMAT ", Perl interpreter: 0x%p"
2455 #define pTHX_VALUE_ (void *)my_perl,
2456 #define pTHX_VALUE (void *)my_perl
2457 #define pTHX__VALUE_ ,(void *)my_perl,
2458 #define pTHX__VALUE ,(void *)my_perl
2461 #define pTHX__FORMAT
2464 #define pTHX__VALUE_
2466 #endif /* USE_ITHREADS */
2468 /* Perl_deprecate was not part of the public API, and did not have a deprecate()
2469 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2470 CPAN::Unpack show any users outside the core. */
2472 # define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2473 "Use of " s " is deprecated")
2474 # define deprecate_disappears_in(when,message) \
2475 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2476 message ", and will disappear in Perl " when)
2477 # define deprecate_fatal_in(when,message) \
2478 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2479 message ". Its use will be fatal in Perl " when)
2482 /* Internal macros to deal with gids and uids */
2485 # if Uid_t_size > IVSIZE
2486 # define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2487 # define SvUID(sv) SvNV(sv)
2488 # elif Uid_t_sign <= 0
2489 # define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2490 # define SvUID(sv) SvIV(sv)
2492 # define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2493 # define SvUID(sv) SvUV(sv)
2494 # endif /* Uid_t_size */
2496 # if Gid_t_size > IVSIZE
2497 # define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2498 # define SvGID(sv) SvNV(sv)
2499 # elif Gid_t_sign <= 0
2500 # define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2501 # define SvGID(sv) SvIV(sv)
2503 # define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2504 # define SvGID(sv) SvUV(sv)
2505 # endif /* Gid_t_size */
2509 #endif /* PERL_HANDY_H_ */
2512 * ex: set ts=8 sts=4 sw=4 et:
https://perl5.git.perl.org / perl5.git / blob