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 /* These C99 typedefs are useful sometimes for, say, loop variables whose
221 * maximum values are small, but for which speed trumps size. If we have a C99
222 * compiler, use that. Otherwise, a plain 'int' should be good enough.
224 * Restrict these to core for now until we are more certain this is a good
226 #if defined(PERL_CORE) || defined(PERL_EXT)
228 typedef int_fast8_t PERL_INT_FAST8_T;
229 typedef uint_fast8_t PERL_UINT_FAST8_T;
230 typedef int_fast16_t PERL_INT_FAST16_T;
231 typedef uint_fast16_t PERL_UINT_FAST16_T;
233 typedef int PERL_INT_FAST8_T;
234 typedef unsigned int PERL_UINT_FAST8_T;
235 typedef int PERL_INT_FAST16_T;
236 typedef unsigned int PERL_UINT_FAST16_T;
240 /* log(2) (i.e., log base 10 of 2) is pretty close to 0.30103, just in case
241 * anyone is grepping for it */
242 #define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log10(2) =~ 146/485 */
243 #define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
244 #define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
246 /* Unused by core; should be deprecated */
247 #define Ctl(ch) ((ch) & 037)
249 #if defined(PERL_CORE) || defined(PERL_EXT)
251 # define MIN(a,b) ((a) < (b) ? (a) : (b))
254 # define MAX(a,b) ((a) > (b) ? (a) : (b))
258 /* This is a helper macro to avoid preprocessor issues, replaced by nothing
259 * unless under DEBUGGING, where it expands to an assert of its argument,
260 * followed by a comma (hence the comma operator). If we just used a straight
261 * assert(), we would get a comma with nothing before it when not DEBUGGING.
263 * We also use empty definition under Coverity since the __ASSERT__
264 * checks often check for things that Really Cannot Happen, and Coverity
265 * detects that and gets all excited. */
267 #if defined(DEBUGGING) && !defined(__COVERITY__)
268 # define __ASSERT_(statement) assert(statement),
270 # define __ASSERT_(statement)
274 =head1 SV-Body Allocation
276 =for apidoc Ama|SV*|newSVpvs|"literal string" s
277 Like C<newSVpvn>, but takes a literal string instead of a
280 =for apidoc Ama|SV*|newSVpvs_flags|"literal string" s|U32 flags
281 Like C<newSVpvn_flags>, but takes a literal string instead of
282 a string/length pair.
284 =for apidoc Ama|SV*|newSVpvs_share|"literal string" s
285 Like C<newSVpvn_share>, but takes a literal string instead of
286 a string/length pair and omits the hash parameter.
288 =for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string" s|I32 flags
289 Like C<sv_catpvn_flags>, but takes a literal string instead
290 of a string/length pair.
292 =for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string" s
293 Like C<sv_catpvn_nomg>, but takes a literal string instead of
294 a string/length pair.
296 =for apidoc Am|void|sv_catpvs|SV* sv|"literal string" s
297 Like C<sv_catpvn>, but takes a literal string instead of a
300 =for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string" s
301 Like C<sv_catpvn_mg>, but takes a literal string instead of a
304 =for apidoc Am|void|sv_setpvs|SV* sv|"literal string" s
305 Like C<sv_setpvn>, but takes a literal string instead of a
308 =for apidoc Am|void|sv_setpvs_mg|SV* sv|"literal string" s
309 Like C<sv_setpvn_mg>, but takes a literal string instead of a
312 =for apidoc Am|SV *|sv_setref_pvs|"literal string" s
313 Like C<sv_setref_pvn>, but takes a literal string instead of
314 a string/length pair.
316 =head1 Memory Management
318 =for apidoc Ama|char*|savepvs|"literal string" s
319 Like C<savepvn>, but takes a literal string instead of a
322 =for apidoc Ama|char*|savesharedpvs|"literal string" s
323 A version of C<savepvs()> which allocates the duplicate string in memory
324 which is shared between threads.
328 =for apidoc Am|HV*|gv_stashpvs|"literal string" name|I32 create
329 Like C<gv_stashpvn>, but takes a literal string instead of a
332 =head1 Hash Manipulation Functions
334 =for apidoc Am|SV**|hv_fetchs|HV* tb|"literal string" key|I32 lval
335 Like C<hv_fetch>, but takes a literal string instead of a
338 =for apidoc Am|SV**|hv_stores|HV* tb|"literal string" key|SV* val
339 Like C<hv_store>, but takes a literal string instead of a
341 and omits the hash parameter.
343 =head1 Lexer interface
345 =for apidoc Amx|void|lex_stuff_pvs|"literal string" pv|U32 flags
347 Like L</lex_stuff_pvn>, but takes a literal string instead of
348 a string/length pair.
353 /* concatenating with "" ensures that only literal strings are accepted as
355 #define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
357 /* note that STR_WITH_LEN() can't be used as argument to macros or functions
358 * that under some configurations might be macros, which means that it requires
359 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
362 /* STR_WITH_LEN() shortcuts */
363 #define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
364 #define newSVpvs_flags(str,flags) \
365 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
366 #define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
367 #define sv_catpvs_flags(sv, str, flags) \
368 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
369 #define sv_catpvs_nomg(sv, str) \
370 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
371 #define sv_catpvs(sv, str) \
372 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
373 #define sv_catpvs_mg(sv, str) \
374 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
375 #define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
376 #define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
377 #define sv_setref_pvs(rv, classname, str) \
378 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
379 #define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
380 #define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
381 #define gv_stashpvs(str, create) \
382 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
383 #define gv_fetchpvs(namebeg, add, sv_type) \
384 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
385 #define gv_fetchpvn(namebeg, len, add, sv_type) \
386 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
387 #define sv_catxmlpvs(dsv, str, utf8) \
388 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
391 #define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
393 #define get_cvs(str, flags) \
394 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
397 =head1 Miscellaneous Functions
399 =for apidoc Am|bool|strNE|char* s1|char* s2
400 Test two C<NUL>-terminated strings to see if they are different. Returns true
403 =for apidoc Am|bool|strEQ|char* s1|char* s2
404 Test two C<NUL>-terminated strings to see if they are equal. Returns true or
407 =for apidoc Am|bool|strLT|char* s1|char* s2
408 Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
409 second, C<s2>. Returns true or false.
411 =for apidoc Am|bool|strLE|char* s1|char* s2
412 Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
413 equal to the second, C<s2>. Returns true or false.
415 =for apidoc Am|bool|strGT|char* s1|char* s2
416 Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
417 the second, C<s2>. Returns true or false.
419 =for apidoc Am|bool|strGE|char* s1|char* s2
420 Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
421 or equal to the second, C<s2>. Returns true or false.
423 =for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
424 Test two C<NUL>-terminated strings to see if they are different. The C<len>
425 parameter indicates the number of bytes to compare. Returns true or false. (A
426 wrapper for C<strncmp>).
428 =for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
429 Test two C<NUL>-terminated strings to see if they are equal. The C<len>
430 parameter indicates the number of bytes to compare. Returns true or false. (A
431 wrapper for C<strncmp>).
433 =for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
434 Test two buffers (which may contain embedded C<NUL> characters, to see if they
435 are equal. The C<len> parameter indicates the number of bytes to compare.
436 Returns zero if equal, or non-zero if non-equal.
438 =for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
439 Test two buffers (which may contain embedded C<NUL> characters, to see if they
440 are not equal. The C<len> parameter indicates the number of bytes to compare.
441 Returns zero if non-equal, or non-zero if equal.
445 New macros should use the following conventions for their names (which are
446 based on the underlying C library functions):
448 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
450 Each has two main parameters, string-like operands that are compared
451 against each other, as specified by the macro name. Some macros may
452 additionally have one or potentially even two length parameters. If a length
453 parameter applies to both string parameters, it will be positioned third;
454 otherwise any length parameter immediately follows the string parameter it
457 If the prefix to the name is 'str', the string parameter is a pointer to a C
458 language string. Such a string does not contain embedded NUL bytes; its
459 length may be unknown, but can be calculated by C<strlen()>, since it is
460 terminated by a NUL, which isn't included in its length.
462 The optional 'n' following 'str' means that that there is a third parameter,
463 giving the maximum number of bytes to look at in each string. Even if both
464 strings are longer than the length parameter, those extra bytes will be
467 The 's' suffix means that the 2nd byte string parameter is a literal C
468 double-quoted string. Its length will automatically be calculated by the
469 macro, so no length parameter will ever be needed for it.
471 If the prefix is 'mem', the string parameters don't have to be C strings;
472 they may contain embedded NUL bytes, do not necessarily have a terminating
473 NUL, and their lengths can be known only through other means, which in
474 practice are additional parameter(s) passed to the function. All 'mem'
475 functions have at least one length parameter. Barring any 'l' or 's' suffix,
476 there is a single length parameter, in position 3, which applies to both
477 string parameters. The 's' suffix means, as described above, that the 2nd
478 string is a literal double-quoted C string (hence its length is calculated by
479 the macro, and the length parameter to the function applies just to the first
480 string parameter, and hence is positioned just after it). An 'l' suffix
481 means that the 2nd string parameter has its own length parameter, and the
482 signature will look like memFOOl(s1, l1, s2, l2).
484 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
485 substring of the 1st string. 'P' if present indicates that the substring
486 must be a "proper" one in tha mathematical sense that the first one must be
487 strictly larger than the 2nd.
492 #define strNE(s1,s2) (strcmp(s1,s2) != 0)
493 #define strEQ(s1,s2) (strcmp(s1,s2) == 0)
494 #define strLT(s1,s2) (strcmp(s1,s2) < 0)
495 #define strLE(s1,s2) (strcmp(s1,s2) <= 0)
496 #define strGT(s1,s2) (strcmp(s1,s2) > 0)
497 #define strGE(s1,s2) (strcmp(s1,s2) >= 0)
499 #define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
500 #define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
502 #define memNE(s1,s2,l) (memcmp(s1,s2,l) != 0)
503 #define memEQ(s1,s2,l) (memcmp(s1,s2,l) == 0)
505 /* memEQ and memNE where second comparand is a string constant */
506 #define memEQs(s1, l, s2) \
507 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
508 #define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
510 /* Keep these private until we decide it was a good idea */
511 #if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
513 #define strBEGINs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1) == 0)
515 #define memBEGINs(s1, l, s2) \
516 ( (l) >= sizeof(s2) - 1 \
517 && memEQ(s1, "" s2 "", sizeof(s2)-1))
518 #define memBEGINPs(s1, l, s2) \
519 ( (l) > sizeof(s2) - 1 \
520 && memEQ(s1, "" s2 "", sizeof(s2)-1))
521 #define memENDs(s1, l, s2) \
522 ( (l) >= sizeof(s2) - 1 \
523 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
524 #define memENDPs(s1, l, s2) \
526 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
527 #endif /* End of making macros private */
529 #define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
530 #define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
531 #define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
532 #define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
537 * Unfortunately, the introduction of locales means that we
538 * can't trust isupper(), etc. to tell the truth. And when
539 * it comes to /\w+/ with tainting enabled, we *must* be able
540 * to trust our character classes.
542 * Therefore, the default tests in the text of Perl will be
543 * independent of locale. Any code that wants to depend on
544 * the current locale will use the tests that begin with "lc".
547 #ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
555 =head1 Character classification
556 This section is about functions (really macros) that classify characters
557 into types, such as punctuation versus alphabetic, etc. Most of these are
558 analogous to regular expression character classes. (See
559 L<perlrecharclass/POSIX Character Classes>.) There are several variants for
560 each class. (Not all macros have all variants; each item below lists the
561 ones valid for it.) None are affected by C<use bytes>, and only the ones
562 with C<LC> in the name are affected by the current locale.
564 The base function, e.g., C<isALPHA()>, takes an octet (either a C<char> or a
565 C<U8>) as input and returns a boolean as to whether or not the character
566 represented by that octet is (or on non-ASCII platforms, corresponds to) an
567 ASCII character in the named class based on platform, Unicode, and Perl rules.
568 If the input is a number that doesn't fit in an octet, FALSE is returned.
570 Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
571 with no suffix C<"_A">. This variant is used to emphasize by its name that
572 only ASCII-range characters can return TRUE.
574 Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
575 onto the platform. That is, the code points that are ASCII are unaffected,
576 since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
577 as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
578 true when called with the code point 0xDF, which is a word character in both
579 ASCII and EBCDIC (though it represents different characters in each).
581 Variant C<isI<FOO>_uvchr> is like the C<isI<FOO>_L1> variant, but accepts any UV code
582 point as input. If the code point is larger than 255, Unicode rules are used
583 to determine if it is in the character class. For example,
584 C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
585 WITH MACRON in Unicode, and is a word character.
587 Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
588 encoded strings. Each call classifies one character, even if the string
589 contains many. This variant takes two parameters. The first, C<p>, is a
590 pointer to the first byte of the character to be classified. (Recall that it
591 may take more than one byte to represent a character in UTF-8 strings.) The
592 second parameter, C<e>, points to anywhere in the string beyond the first
593 character, up to one byte past the end of the entire string. The suffix
594 C<_safe> in the function's name indicates that it will not attempt to read
595 beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
596 is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
597 character is malformed in some way, the program may croak, or the function may
598 return FALSE, at the discretion of the implementation, and subject to change in
601 Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
602 parameter, C<p>, which has the same meaning as the corresponding parameter does
603 in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
604 beyond the end of the string. Starting in Perl v5.30, it will take a second
605 parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
606 program that uses it will have to be changed to successfully compile. In the
607 meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
608 program will raise a deprecation warning, enabled by default. You can convert
609 your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
610 extra measure of protection, or you can wait until v5.30, when you'll be forced
611 to add the C<e> parameter.
613 Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants, but the
614 result is based on the current locale, which is what C<LC> in the name stands
615 for. If Perl can determine that the current locale is a UTF-8 locale, it uses
616 the published Unicode rules; otherwise, it uses the C library function that
617 gives the named classification. For example, C<isDIGIT_LC()> when not in a
618 UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
619 returned if the input won't fit into an octet. On some platforms where the C
620 library function is known to be defective, Perl changes its result to follow
621 the POSIX standard's rules.
623 Variant C<isI<FOO>_LC_uvchr> is like C<isI<FOO>_LC>, but is defined on any UV. It
624 returns the same as C<isI<FOO>_LC> for input code points less than 256, and
625 returns the hard-coded, not-affected-by-locale, Unicode results for larger ones.
627 Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
628 encoded strings. Each call classifies one character, even if the string
629 contains many. This variant takes two parameters. The first, C<p>, is a
630 pointer to the first byte of the character to be classified. (Recall that it
631 may take more than one byte to represent a character in UTF-8 strings.) The
632 second parameter, C<e>, points to anywhere in the string beyond the first
633 character, up to one byte past the end of the entire string. The suffix
634 C<_safe> in the function's name indicates that it will not attempt to read
635 beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
636 is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
637 character is malformed in some way, the program may croak, or the function may
638 return FALSE, at the discretion of the implementation, and subject to change in
641 Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
642 parameter, C<p>, which has the same meaning as the corresponding parameter does
643 in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
644 beyond the end of the string. Starting in Perl v5.30, it will take a second
645 parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
646 program that uses it will have to be changed to successfully compile. In the
647 meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
648 the program will raise a deprecation warning, enabled by default. You can
649 convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
650 and get an extra measure of protection, or you can wait until v5.30, when
651 you'll be forced to add the C<e> parameter.
653 =for apidoc Am|bool|isALPHA|char ch
654 Returns a boolean indicating whether the specified character is an
655 alphabetic character, analogous to C<m/[[:alpha:]]/>.
656 See the L<top of this section|/Character classification> for an explanation of
658 C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
659 C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
661 =for apidoc Am|bool|isALPHANUMERIC|char ch
662 Returns a boolean indicating whether the specified character is a either an
663 alphabetic character or decimal digit, analogous to C<m/[[:alnum:]]/>.
664 See the L<top of this section|/Character classification> for an explanation of
666 C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
667 C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
668 and C<isALPHANUMERIC_LC_utf8_safe>.
670 =for apidoc Am|bool|isASCII|char ch
671 Returns a boolean indicating whether the specified character is one of the 128
672 characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
673 On non-ASCII platforms, it returns TRUE iff this
674 character corresponds to an ASCII character. Variants C<isASCII_A()> and
675 C<isASCII_L1()> are identical to C<isASCII()>.
676 See the L<top of this section|/Character classification> for an explanation of
678 C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
679 C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
680 library routine C<isascii()>. In these cases, the variants whose names contain
681 C<LC> are the same as the corresponding ones without.
683 Also note, that because all ASCII characters are UTF-8 invariant (meaning they
684 have the exact same representation (always a single byte) whether encoded in
685 UTF-8 or not), C<isASCII> will give the correct results when called with any
686 byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
687 will work properly on any string encoded or not in UTF-8.
689 =for apidoc Am|bool|isBLANK|char ch
690 Returns a boolean indicating whether the specified character is a
691 character considered to be a blank, analogous to C<m/[[:blank:]]/>.
692 See the L<top of this section|/Character classification> for an explanation of
694 C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
695 C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
696 however, that some platforms do not have the C library routine
697 C<isblank()>. In these cases, the variants whose names contain C<LC> are
698 the same as the corresponding ones without.
700 =for apidoc Am|bool|isCNTRL|char ch
701 Returns a boolean indicating whether the specified character is a
702 control character, analogous to C<m/[[:cntrl:]]/>.
703 See the L<top of this section|/Character classification> for an explanation of
705 C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
706 C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
707 platforms, you almost always want to use the C<isCNTRL_L1> variant.
709 =for apidoc Am|bool|isDIGIT|char ch
710 Returns a boolean indicating whether the specified character is a
711 digit, analogous to C<m/[[:digit:]]/>.
712 Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
713 See the L<top of this section|/Character classification> for an explanation of
715 C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
716 C<isDIGIT_LC_utf8_safe>.
718 =for apidoc Am|bool|isGRAPH|char ch
719 Returns a boolean indicating whether the specified character is a
720 graphic character, analogous to C<m/[[:graph:]]/>.
721 See the L<top of this section|/Character classification> for an explanation of
722 variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
723 C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
725 =for apidoc Am|bool|isLOWER|char ch
726 Returns a boolean indicating whether the specified character is a
727 lowercase character, analogous to C<m/[[:lower:]]/>.
728 See the L<top of this section|/Character classification> for an explanation of
730 C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
731 C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
733 =for apidoc Am|bool|isOCTAL|char ch
734 Returns a boolean indicating whether the specified character is an
736 The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
739 =for apidoc Am|bool|isPUNCT|char ch
740 Returns a boolean indicating whether the specified character is a
741 punctuation character, analogous to C<m/[[:punct:]]/>.
742 Note that the definition of what is punctuation isn't as
743 straightforward as one might desire. See L<perlrecharclass/POSIX Character
744 Classes> for details.
745 See the L<top of this section|/Character classification> for an explanation of
746 variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
747 C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
749 =for apidoc Am|bool|isSPACE|char ch
750 Returns a boolean indicating whether the specified character is a
751 whitespace character. This is analogous
752 to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
753 this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
754 locale forms of this macro (the ones with C<LC> in their names) matched
755 precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
756 in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
757 (See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
758 See the L<top of this section|/Character classification> for an explanation of
760 C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
761 C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
763 =for apidoc Am|bool|isPSXSPC|char ch
764 (short for Posix Space)
765 Starting in 5.18, this is identical in all its forms to the
766 corresponding C<isSPACE()> macros.
767 The locale forms of this macro are identical to their corresponding
768 C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
769 non-locale forms differ from their C<isSPACE()> forms only in that the
770 C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
771 Otherwise they are identical. Thus this macro is analogous to what
772 C<m/[[:space:]]/> matches in a regular expression.
773 See the L<top of this section|/Character classification> for an explanation of
774 variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
775 C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
777 =for apidoc Am|bool|isUPPER|char ch
778 Returns a boolean indicating whether the specified character is an
779 uppercase character, analogous to C<m/[[:upper:]]/>.
780 See the L<top of this section|/Character classification> for an explanation of
781 variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
782 C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
784 =for apidoc Am|bool|isPRINT|char ch
785 Returns a boolean indicating whether the specified character is a
786 printable character, analogous to C<m/[[:print:]]/>.
787 See the L<top of this section|/Character classification> for an explanation of
789 C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
790 C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
792 =for apidoc Am|bool|isWORDCHAR|char ch
793 Returns a boolean indicating whether the specified character is a character
794 that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
795 in a regular expression. A word character is an alphabetic character, a
796 decimal digit, a connecting punctuation character (such as an underscore), or
797 a "mark" character that attaches to one of those (like some sort of accent).
798 C<isALNUM()> is a synonym provided for backward compatibility, even though a
799 word character includes more than the standard C language meaning of
801 See the L<top of this section|/Character classification> for an explanation of
802 variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
803 C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
804 C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
805 include the platform's native underscore.
807 =for apidoc Am|bool|isXDIGIT|char ch
808 Returns a boolean indicating whether the specified character is a hexadecimal
809 digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
810 and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
811 See the L<top of this section|/Character classification> for an explanation of
813 C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
814 and C<isXDIGIT_LC_utf8_safe>.
816 =for apidoc Am|bool|isIDFIRST|char ch
817 Returns a boolean indicating whether the specified character can be the first
818 character of an identifier. This is very close to, but not quite the same as
819 the official Unicode property C<XID_Start>. The difference is that this
820 returns true only if the input character also matches L</isWORDCHAR>.
821 See the L<top of this section|/Character classification> for an explanation of
823 C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
824 C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
826 =for apidoc Am|bool|isIDCONT|char ch
827 Returns a boolean indicating whether the specified character can be the
828 second or succeeding character of an identifier. This is very close to, but
829 not quite the same as the official Unicode property C<XID_Continue>. The
830 difference is that this returns true only if the input character also matches
831 L</isWORDCHAR>. See the L<top of this section|/Character classification> for
833 explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
834 C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
835 C<isIDCONT_LC_utf8_safe>.
837 =head1 Miscellaneous Functions
839 =for apidoc Am|U8|READ_XDIGIT|char str*
840 Returns the value of an ASCII-range hex digit and advances the string pointer.
841 Behaviour is only well defined when isXDIGIT(*str) is true.
843 =head1 Character case changing
844 Perl uses "full" Unicode case mappings. This means that converting a single
845 character to another case may result in a sequence of more than one character.
846 For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
847 character sequence C<SS>. This presents some complications The lowercase of
848 all characters in the range 0..255 is a single character, and thus
849 C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
850 return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
851 an API that does allow every possible legal result to be returned.) Likewise
852 no other function that is crippled by not being able to give the correct
853 results for the full range of possible inputs has been implemented here.
855 =for apidoc Am|U8|toUPPER|U8 ch
856 Converts the specified character to uppercase. If the input is anything but an
857 ASCII lowercase character, that input character itself is returned. Variant
858 C<toUPPER_A> is equivalent.
860 =for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
861 Converts the code point C<cp> to its uppercase version, and
862 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
863 point is interpreted as native if less than 256; otherwise as Unicode. Note
864 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
865 bytes since the uppercase version may be longer than the original character.
867 The first code point of the uppercased version is returned
868 (but note, as explained at L<the top of this section|/Character case
869 changing>, that there may be more.)
871 =for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
872 Converts the first UTF-8 encoded character in the sequence starting at C<p> and
873 extending no further than S<C<e - 1>> to its uppercase version, and
874 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
875 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
876 bytes since the uppercase version may be longer than the original character.
878 The first code point of the uppercased version is returned
879 (but note, as explained at L<the top of this section|/Character case
880 changing>, that there may be more).
882 The suffix C<_safe> in the function's name indicates that it will not attempt
883 to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
884 true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
885 input character is malformed in some way, the program may croak, or the
886 function may return the REPLACEMENT CHARACTER, at the discretion of the
887 implementation, and subject to change in future releases.
889 =for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
890 This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
891 parameter The function therefore can't check if it is reading
892 beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
893 parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
894 program that uses it will have to be changed to successfully compile. In the
895 meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
896 program will raise a deprecation warning, enabled by default. You can convert
897 your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
898 extra measure of protection, or you can wait until v5.30, when you'll be forced
899 to add the C<e> parameter.
901 =for apidoc Am|U8|toFOLD|U8 ch
902 Converts the specified character to foldcase. If the input is anything but an
903 ASCII uppercase character, that input character itself is returned. Variant
904 C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
905 Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
907 =for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
908 Converts the code point C<cp> to its foldcase version, and
909 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
910 point is interpreted as native if less than 256; otherwise as Unicode. Note
911 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
912 bytes since the foldcase version may be longer than the original character.
914 The first code point of the foldcased version is returned
915 (but note, as explained at L<the top of this section|/Character case
916 changing>, that there may be more).
918 =for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
919 Converts the first UTF-8 encoded character in the sequence starting at C<p> and
920 extending no further than S<C<e - 1>> to its foldcase version, and
921 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
922 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
923 bytes since the foldcase version may be longer than the original character.
925 The first code point of the foldcased version is returned
926 (but note, as explained at L<the top of this section|/Character case
927 changing>, that there may be more).
929 The suffix C<_safe> in the function's name indicates that it will not attempt
930 to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
931 true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
932 input character is malformed in some way, the program may croak, or the
933 function may return the REPLACEMENT CHARACTER, at the discretion of the
934 implementation, and subject to change in future releases.
936 =for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
937 This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
938 parameter The function therefore can't check if it is reading
939 beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
940 parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
941 program that uses it will have to be changed to successfully compile. In the
942 meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
943 program will raise a deprecation warning, enabled by default. You can convert
944 your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
945 extra measure of protection, or you can wait until v5.30, when you'll be forced
946 to add the C<e> parameter.
948 =for apidoc Am|U8|toLOWER|U8 ch
949 Converts the specified character to lowercase. If the input is anything but an
950 ASCII uppercase character, that input character itself is returned. Variant
951 C<toLOWER_A> is equivalent.
953 =for apidoc Am|U8|toLOWER_L1|U8 ch
954 Converts the specified Latin1 character to lowercase. The results are
955 undefined if the input doesn't fit in a byte.
957 =for apidoc Am|U8|toLOWER_LC|U8 ch
958 Converts the specified character to lowercase using the current locale's rules,
959 if possible; otherwise returns the input character itself.
961 =for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
962 Converts the code point C<cp> to its lowercase version, and
963 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
964 point is interpreted as native if less than 256; otherwise as Unicode. Note
965 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
966 bytes since the lowercase version may be longer than the original character.
968 The first code point of the lowercased version is returned
969 (but note, as explained at L<the top of this section|/Character case
970 changing>, that there may be more).
973 =for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
974 Converts the first UTF-8 encoded character in the sequence starting at C<p> and
975 extending no further than S<C<e - 1>> to its lowercase version, and
976 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
977 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
978 bytes since the lowercase version may be longer than the original character.
980 The first code point of the lowercased version is returned
981 (but note, as explained at L<the top of this section|/Character case
982 changing>, that there may be more).
984 The suffix C<_safe> in the function's name indicates that it will not attempt
985 to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
986 true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
987 input character is malformed in some way, the program may croak, or the
988 function may return the REPLACEMENT CHARACTER, at the discretion of the
989 implementation, and subject to change in future releases.
991 =for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
992 This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
993 parameter The function therefore can't check if it is reading
994 beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
995 parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
996 program that uses it will have to be changed to successfully compile. In the
997 meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
998 program will raise a deprecation warning, enabled by default. You can convert
999 your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
1000 extra measure of protection, or you can wait until v5.30, when you'll be forced
1001 to add the C<e> parameter.
1003 =for apidoc Am|U8|toTITLE|U8 ch
1004 Converts the specified character to titlecase. If the input is anything but an
1005 ASCII lowercase character, that input character itself is returned. Variant
1006 C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
1007 range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
1008 not a concept used in locale handling, so there is no functionality for that.)
1010 =for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1011 Converts the code point C<cp> to its titlecase version, and
1012 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
1013 point is interpreted as native if less than 256; otherwise as Unicode. Note
1014 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1015 bytes since the titlecase version may be longer than the original character.
1017 The first code point of the titlecased version is returned
1018 (but note, as explained at L<the top of this section|/Character case
1019 changing>, that there may be more).
1021 =for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1022 Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1023 extending no further than S<C<e - 1>> to its titlecase version, and
1024 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1025 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1026 bytes since the titlecase version may be longer than the original character.
1028 The first code point of the titlecased version is returned
1029 (but note, as explained at L<the top of this section|/Character case
1030 changing>, that there may be more).
1032 The suffix C<_safe> in the function's name indicates that it will not attempt
1033 to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1034 true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1035 input character is malformed in some way, the program may croak, or the
1036 function may return the REPLACEMENT CHARACTER, at the discretion of the
1037 implementation, and subject to change in future releases.
1039 =for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1040 This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1041 parameter The function therefore can't check if it is reading
1042 beyond the end of the string. Starting in Perl v5.30, it will take the C<e>
1043 parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1044 program that uses it will have to be changed to successfully compile. In the
1045 meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1046 program will raise a deprecation warning, enabled by default. You can convert
1047 your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
1048 extra measure of protection, or you can wait until v5.30, when you'll be forced
1049 to add the C<e> parameter.
1053 XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1054 really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1055 and aren't general purpose as they don't work on U+DF, and assert against that.
1057 Note that these macros are repeated in Devel::PPPort, so should also be
1058 patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1062 /* Specify the widest unsigned type on the platform. */
1064 # define WIDEST_UTYPE U64
1066 # define WIDEST_UTYPE U32
1069 /* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1070 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1071 * bits of information are lost even on a 64-bit machine, but to get the
1072 * compiler to optimize it out if possible. This is because Configure makes
1073 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1074 * sizeof() guarantees that this evaluates to a constant true at compile time.
1076 * For Coverity, be always true, because otherwise Coverity thinks
1077 * it finds several expressions that are always true, independent
1078 * of operands. Well, they are, but that is kind of the point.
1080 #ifndef __COVERITY__
1081 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1083 #define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
1084 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
1086 #define FITS_IN_8_BITS(c) (1)
1090 # ifndef _ALL_SOURCE
1091 /* The native libc isascii() et.al. functions return the wrong results
1092 * on at least z/OS unless this is defined. */
1093 # error _ALL_SOURCE should probably be defined
1096 /* There is a simple definition of ASCII for ASCII platforms. But the
1097 * EBCDIC one isn't so simple, so is defined using table look-up like the
1098 * other macros below.
1100 * The cast here is used instead of '(c) >= 0', because some compilers emit
1101 * a warning that that test is always true when the parameter is an
1102 * unsigned type. khw supposes that it could be written as
1103 * && ((c) == '\0' || (c) > 0)
1104 * to avoid the message, but the cast will likely avoid extra branches even
1105 * with stupid compilers.
1107 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1109 # define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
1112 /* Take the eight possible bit patterns of the lower 3 bits and you get the
1113 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1114 * can be ignored. If the rest match '0', we have an octal */
1115 #define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
1117 #ifdef H_PERL /* If have access to perl.h, lookup in its table */
1119 /* Character class numbers. For internal core Perl use only. The ones less
1120 * than 32 are used in PL_charclass[] and the ones up through the one that
1121 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1122 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1123 * their actual definitions are here. If that file has a name not used here,
1126 * The first group of these is ordered in what I (khw) estimate to be the
1127 * frequency of their use. This gives a slight edge to exiting a loop earlier
1128 * (in reginclass() in regexec.c) */
1129 # define _CC_WORDCHAR 0 /* \w and [:word:] */
1130 # define _CC_DIGIT 1 /* \d and [:digit:] */
1131 # define _CC_ALPHA 2 /* [:alpha:] */
1132 # define _CC_LOWER 3 /* [:lower:] */
1133 # define _CC_UPPER 4 /* [:upper:] */
1134 # define _CC_PUNCT 5 /* [:punct:] */
1135 # define _CC_PRINT 6 /* [:print:] */
1136 # define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1137 # define _CC_GRAPH 8 /* [:graph:] */
1138 # define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
1140 #define _FIRST_NON_SWASH_CC 10
1141 /* The character classes above are implemented with swashes. The second group
1142 * (just below) contains the ones implemented without. These are also sorted
1143 * in rough order of the frequency of their use, except that \v should be last,
1144 * as it isn't a real Posix character class, and some (small) inefficiencies in
1145 * regular expression handling would be introduced by putting it in the middle
1146 * of those that are. Also, cntrl and ascii come after the others as it may be
1147 * useful to group these which have no members that match above Latin1, (or
1148 * above ASCII in the latter case) */
1150 # define _CC_SPACE 10 /* \s, [:space:] */
1151 # define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1152 when the deprecated isFOO_utf8()
1153 functions are removed */
1154 # define _CC_BLANK 11 /* [:blank:] */
1155 # define _CC_XDIGIT 12 /* [:xdigit:] */
1156 # define _CC_CNTRL 13 /* [:cntrl:] */
1157 # define _CC_ASCII 14 /* [:ascii:] */
1158 # define _CC_VERTSPACE 15 /* \v */
1160 # define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1162 /* The members of the third group below do not need to be coordinated with data
1163 * structures in regcomp.[ch] and regexec.c. */
1164 # define _CC_IDFIRST 16
1165 # define _CC_CHARNAME_CONT 17
1166 # define _CC_NONLATIN1_FOLD 18
1167 # define _CC_NONLATIN1_SIMPLE_FOLD 19
1168 # define _CC_QUOTEMETA 20
1169 # define _CC_NON_FINAL_FOLD 21
1170 # define _CC_IS_IN_SOME_FOLD 22
1171 # define _CC_MNEMONIC_CNTRL 23
1173 # define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1174 isFOO_utf8() functions are removed */
1176 /* This next group is only used on EBCDIC platforms, so theoretically could be
1177 * shared with something entirely different that's only on ASCII platforms */
1178 # define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1179 # define _CC_UTF8_IS_START 29
1180 # define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1181 # define _CC_UTF8_IS_CONTINUATION 31
1183 * If more bits are needed, one could add a second word for non-64bit
1184 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
1185 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1186 * is used only for optimization (as of this writing), and differs in the
1187 * Latin1 range from the ALPHA bit only in two relatively unimportant
1188 * characters: the masculine and feminine ordinal indicators, so removing it
1189 * would just cause /i regexes which match them to run less efficiently.
1190 * Similarly the EBCDIC-only bits are used just for speed, and could be
1191 * replaced by other means */
1193 #if defined(PERL_CORE) || defined(PERL_EXT)
1194 /* An enum version of the character class numbers, to help compilers
1197 _CC_ENUM_ALPHA = _CC_ALPHA,
1198 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1199 _CC_ENUM_ASCII = _CC_ASCII,
1200 _CC_ENUM_BLANK = _CC_BLANK,
1201 _CC_ENUM_CASED = _CC_CASED,
1202 _CC_ENUM_CNTRL = _CC_CNTRL,
1203 _CC_ENUM_DIGIT = _CC_DIGIT,
1204 _CC_ENUM_GRAPH = _CC_GRAPH,
1205 _CC_ENUM_LOWER = _CC_LOWER,
1206 _CC_ENUM_PRINT = _CC_PRINT,
1207 _CC_ENUM_PUNCT = _CC_PUNCT,
1208 _CC_ENUM_SPACE = _CC_SPACE,
1209 _CC_ENUM_UPPER = _CC_UPPER,
1210 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
1211 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
1212 _CC_ENUM_XDIGIT = _CC_XDIGIT
1213 } _char_class_number;
1216 #define POSIX_SWASH_COUNT _FIRST_NON_SWASH_CC
1217 #define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
1219 #if defined(PERL_IN_UTF8_C) \
1220 || defined(PERL_IN_REGCOMP_C) \
1221 || defined(PERL_IN_REGEXEC_C)
1222 # if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 || _CC_LOWER != 3 \
1223 || _CC_UPPER != 4 || _CC_PUNCT != 5 || _CC_PRINT != 6 \
1224 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 || _CC_CASED != 9
1225 #error Need to adjust order of swash_property_names[]
1228 /* This is declared static in each of the few files that this is #defined for
1229 * to keep them from being publicly accessible. Hence there is a small amount
1230 * of wasted space */
1232 static const char* const swash_property_names[] = {
1248 EXTCONST U32 PL_charclass[] = {
1249 # include "l1_char_class_tab.h"
1252 # else /* ! DOINIT */
1253 EXTCONST U32 PL_charclass[];
1257 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
1258 # define _CC_mask(classnum) (1U << (classnum))
1260 /* For internal core Perl use only: the base macro for defining macros like
1262 # define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
1263 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
1265 /* The mask for the _A versions of the macros; it just adds in the bit for
1267 # define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1269 /* For internal core Perl use only: the base macro for defining macros like
1270 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1271 * the ASCII bit are present */
1272 # define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1273 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1274 == _CC_mask_A(classnum)))
1276 # define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
1277 # define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
1278 # define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1279 # define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
1280 # define isDIGIT_A(c) _generic_isCC(c, _CC_DIGIT) /* No non-ASCII digits */
1281 # define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1282 # define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1283 # define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
1284 # define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1285 # define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
1286 # define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1287 # define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
1288 # define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1290 # define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
1291 # define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1292 # define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1293 # define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1295 /* continuation character for legal NAME in \N{NAME} */
1296 # define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1298 # define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1299 # define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1300 # define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1301 # define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
1302 # define isPSXSPC_L1(c) isSPACE_L1(c)
1303 # define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1304 # define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1305 # define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1306 # define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1307 # define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
1310 # define isASCII(c) _generic_isCC(c, _CC_ASCII)
1313 /* Participates in a single-character fold with a character above 255 */
1314 # 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)))
1316 /* Like the above, but also can be part of a multi-char fold */
1317 # 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)))
1319 # define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
1320 # define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1321 _generic_isCC(c, _CC_NON_FINAL_FOLD)
1322 # define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1323 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
1324 # define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1325 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
1326 #else /* else we don't have perl.h H_PERL */
1328 /* If we don't have perl.h, we are compiling a utility program. Below we
1329 * hard-code various macro definitions that wouldn't otherwise be available
1330 * to it. Most are coded based on first principles. These are written to
1331 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
1332 # define isDIGIT_A(c) ((c) <= '9' && (c) >= '0')
1333 # define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
1334 # define isSPACE_A(c) (isBLANK_A(c) \
1339 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1340 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1341 * performance (if optimization isn't on), and allow the same code to be
1342 * used for both platform types */
1343 # define isLOWER_A(c) ((c) >= 'a' && (c) <= 'z' \
1345 || ((c) >= 'j' && (c) <= 'r') \
1347 # define isUPPER_A(c) ((c) >= 'A' && (c) <= 'Z' \
1349 || ((c) >= 'J' && (c) <= 'R') \
1351 # define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1352 # define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
1353 # define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
1354 # define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
1355 # define isXDIGIT_A(c) (isDIGIT_A(c) \
1356 || ((c) >= 'a' && (c) <= 'f') \
1357 || ((c) <= 'F' && (c) >= 'A'))
1358 # define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1359 || (c) == '#' || (c) == '$' || (c) == '%' \
1360 || (c) == '&' || (c) == '\'' || (c) == '(' \
1361 || (c) == ')' || (c) == '*' || (c) == '+' \
1362 || (c) == ',' || (c) == '.' || (c) == '/' \
1363 || (c) == ':' || (c) == ';' || (c) == '<' \
1364 || (c) == '=' || (c) == '>' || (c) == '?' \
1365 || (c) == '@' || (c) == '[' || (c) == '\\' \
1366 || (c) == ']' || (c) == '^' || (c) == '_' \
1367 || (c) == '`' || (c) == '{' || (c) == '|' \
1368 || (c) == '}' || (c) == '~')
1369 # define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1370 # define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
1373 /* The below is accurate for the 3 EBCDIC code pages traditionally
1374 * supported by perl. The only difference between them in the controls
1375 * is the position of \n, and that is represented symbolically below */
1376 # define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1377 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1378 || (c) == '\t' || (c) == '\v' \
1379 || ((c) <= 3 && (c) >= 1) /* SOH, STX, ETX */ \
1380 || (c) == 7 /* U+7F DEL */ \
1381 || ((c) <= 0x13 && (c) >= 0x0E) /* SO, SI */ \
1382 /* DLE, DC[1-3] */ \
1383 || (c) == 0x18 /* U+18 CAN */ \
1384 || (c) == 0x19 /* U+19 EOM */ \
1385 || ((c) <= 0x1F && (c) >= 0x1C) /* [FGRU]S */ \
1386 || (c) == 0x26 /* U+17 ETB */ \
1387 || (c) == 0x27 /* U+1B ESC */ \
1388 || (c) == 0x2D /* U+05 ENQ */ \
1389 || (c) == 0x2E /* U+06 ACK */ \
1390 || (c) == 0x32 /* U+16 SYN */ \
1391 || (c) == 0x37 /* U+04 EOT */ \
1392 || (c) == 0x3C /* U+14 DC4 */ \
1393 || (c) == 0x3D /* U+15 NAK */ \
1394 || (c) == 0x3F)/* U+1A SUB */
1395 # define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
1396 # else /* isASCII is already defined for ASCII platforms, so can use that to
1398 # define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
1401 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
1402 * during debugging, and it seems best to keep them. We may be called
1403 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1404 * do anything anyway, so make it not a problem */
1405 # if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1406 # define NATIVE_TO_LATIN1(ch) (ch)
1408 # define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1409 # define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1410 # define isBLANK_L1(c) (isBLANK_A(c) \
1411 || (FITS_IN_8_BITS(c) \
1412 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1413 # define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1414 # define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1415 # define isLOWER_L1(c) (isLOWER_A(c) \
1416 || (FITS_IN_8_BITS(c) \
1417 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
1418 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1419 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1420 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1421 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1422 # define isPRINT_L1(c) (isPRINT_A(c) \
1423 || (FITS_IN_8_BITS(c) \
1424 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
1425 # define isPUNCT_L1(c) (isPUNCT_A(c) \
1426 || (FITS_IN_8_BITS(c) \
1427 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
1428 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1429 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1430 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1431 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1432 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1433 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1434 # define isSPACE_L1(c) (isSPACE_A(c) \
1435 || (FITS_IN_8_BITS(c) \
1436 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
1437 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1438 # define isUPPER_L1(c) (isUPPER_A(c) \
1439 || (FITS_IN_8_BITS(c) \
1440 && ( NATIVE_TO_LATIN1((U8) c) >= 0xC0 \
1441 && NATIVE_TO_LATIN1((U8) c) <= 0xDE \
1442 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1443 # define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1444 # define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1445 # define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1450 /* The following are not fully accurate in the above-ASCII range. I (khw)
1451 * don't think it's necessary to be so for the purposes where this gets
1453 # define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1454 # define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1456 /* And these aren't accurate at all. They are useful only for above
1457 * Latin1, which utilities and bootstrapping don't deal with */
1458 # define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
1459 # define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1460 # define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1462 /* Many of the macros later in this file are defined in terms of these. By
1463 * implementing them with a function, which converts the class number into
1464 * a call to the desired macro, all of the later ones work. However, that
1465 * function won't be actually defined when building a utility program (no
1466 * perl.h), and so a compiler error will be generated if one is attempted
1467 * to be used. And the above-Latin1 code points require Unicode tables to
1468 * be present, something unlikely to be the case when bootstrapping */
1469 # define _generic_isCC(c, classnum) \
1470 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1471 # define _generic_isCC_A(c, classnum) \
1472 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
1473 #endif /* End of no perl.h H_PERL */
1475 #define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1476 #define isALPHA(c) isALPHA_A(c)
1477 #define isASCII_A(c) isASCII(c)
1478 #define isASCII_L1(c) isASCII(c)
1479 #define isBLANK(c) isBLANK_A(c)
1480 #define isCNTRL(c) isCNTRL_A(c)
1481 #define isDIGIT(c) isDIGIT_A(c)
1482 #define isGRAPH(c) isGRAPH_A(c)
1483 #define isIDFIRST(c) isIDFIRST_A(c)
1484 #define isLOWER(c) isLOWER_A(c)
1485 #define isPRINT(c) isPRINT_A(c)
1486 #define isPSXSPC_A(c) isSPACE_A(c)
1487 #define isPSXSPC(c) isPSXSPC_A(c)
1488 #define isPSXSPC_L1(c) isSPACE_L1(c)
1489 #define isPUNCT(c) isPUNCT_A(c)
1490 #define isSPACE(c) isSPACE_A(c)
1491 #define isUPPER(c) isUPPER_A(c)
1492 #define isWORDCHAR(c) isWORDCHAR_A(c)
1493 #define isXDIGIT(c) isXDIGIT_A(c)
1495 /* ASCII casing. These could also be written as
1496 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1497 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1498 which uses table lookup and mask instead of subtraction. (This would
1499 work because the _MOD does not apply in the ASCII range) */
1500 #define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1501 #define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
1503 /* In the ASCII range, these are equivalent to what they're here defined to be.
1504 * But by creating these definitions, other code doesn't have to be aware of
1506 #define toFOLD(c) toLOWER(c)
1507 #define toTITLE(c) toUPPER(c)
1509 #define toLOWER_A(c) toLOWER(c)
1510 #define toUPPER_A(c) toUPPER(c)
1511 #define toFOLD_A(c) toFOLD(c)
1512 #define toTITLE_A(c) toTITLE(c)
1514 /* Use table lookup for speed; returns the input itself if is out-of-range */
1515 #define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
1517 : PL_latin1_lc[ (U8) (c) ])
1518 #define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1520 /* Modified uc. Is correct uc except for three non-ascii chars which are
1521 * all mapped to one of them, and these need special handling; returns the
1522 * input itself if is out-of-range */
1523 #define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
1525 : PL_mod_latin1_uc[ (U8) (c) ])
1526 #define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
1528 /* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1530 /* For internal core Perl use only: the base macro for defining macros like
1531 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
1532 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1533 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1534 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1535 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1536 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1537 * code points under locale rules, use the macros based on _generic_LC_uvchr
1538 * instead of this. */
1539 #define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1540 (! FITS_IN_8_BITS(c) \
1542 : IN_UTF8_CTYPE_LOCALE \
1543 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
1546 /* For internal core Perl use only: a helper macro for defining macros like
1547 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1548 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1549 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1550 * _generic_LC_base for more info */
1551 #define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1552 _generic_LC_base(c,utf8_locale_classnum, \
1553 non_utf8_func( (_LC_CAST) (c)))
1555 /* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1556 * 'c' is the platform's native underscore character */
1557 #define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1558 _generic_LC_base(c, utf8_locale_classnum, \
1559 (non_utf8_func( (_LC_CAST) (c)) \
1560 || (char)(c) == '_'))
1562 /* These next three are also for internal core Perl use only: case-change
1564 #define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1566 : (IN_UTF8_CTYPE_LOCALE) \
1567 ? PL_latin1_lc[ (U8) (c) ] \
1568 : (cast)function((cast)(c)))
1570 /* Note that the result can be larger than a byte in a UTF-8 locale. It
1571 * returns a single value, so can't adequately return the upper case of LATIN
1572 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1573 * values "SS"); instead it asserts against that under DEBUGGING, and
1574 * otherwise returns its input */
1575 #define _generic_toUPPER_LC(c, function, cast) \
1576 (! FITS_IN_8_BITS(c) \
1578 : ((! IN_UTF8_CTYPE_LOCALE) \
1579 ? (cast)function((cast)(c)) \
1580 : ((((U8)(c)) == MICRO_SIGN) \
1581 ? GREEK_CAPITAL_LETTER_MU \
1582 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1583 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1584 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1585 ? (__ASSERT_(0) (c)) \
1586 : PL_mod_latin1_uc[ (U8) (c) ])))))
1588 /* Note that the result can be larger than a byte in a UTF-8 locale. It
1589 * returns a single value, so can't adequately return the fold case of LATIN
1590 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1591 * values "ss"); instead it asserts against that under DEBUGGING, and
1592 * otherwise returns its input */
1593 #define _generic_toFOLD_LC(c, function, cast) \
1594 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
1595 ? GREEK_SMALL_LETTER_MU \
1596 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1597 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1598 _generic_toLOWER_LC(c, function, cast)))
1600 /* Use the libc versions for these if available. */
1601 #if defined(HAS_ISASCII)
1602 # define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1604 # define isASCII_LC(c) isASCII(c)
1607 #if defined(HAS_ISBLANK)
1608 # define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
1609 #else /* Unlike isASCII, varies if in a UTF-8 locale */
1610 # define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
1616 /* The Windows functions don't bother to follow the POSIX standard, which
1617 * for example says that something can't both be a printable and a control.
1618 * But Windows treats the \t control as a printable, and does such things
1619 * as making superscripts into both digits and punctuation. This tames
1620 * these flaws by assuming that the definitions of both controls and space
1621 * are correct, and then making sure that other definitions don't have
1622 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1623 * Not all possible weirdnesses are checked for, just the ones that were
1624 * detected on actual Microsoft code pages */
1626 # define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1627 # define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1629 # define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1630 && isALPHANUMERIC_LC(c))
1631 # define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1633 # define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1634 isALPHANUMERIC_LC(c))
1635 # define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1636 # define isIDFIRST_LC(c) (((c) == '_') \
1637 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1638 # define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1639 # define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1640 # define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1641 # define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
1642 # define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
1643 # define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1644 && isALPHANUMERIC_LC(c))
1646 # define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1647 # define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1648 # define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1650 #elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
1651 /* For most other platforms */
1653 # define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1654 # define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1655 # define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1656 # define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1657 # define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1658 # define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1659 # define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1660 # define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1661 # define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1662 # define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1663 # define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1664 # define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1665 # define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1668 # define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1669 # define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1670 # define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1672 #else /* The final fallback position */
1674 # define isALPHA_LC(c) (isascii(c) && isalpha(c))
1675 # define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1676 # define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1677 # define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1678 # define isGRAPH_LC(c) (isascii(c) && isgraph(c))
1679 # define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
1680 # define isLOWER_LC(c) (isascii(c) && islower(c))
1681 # define isPRINT_LC(c) (isascii(c) && isprint(c))
1682 # define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1683 # define isSPACE_LC(c) (isascii(c) && isspace(c))
1684 # define isUPPER_LC(c) (isascii(c) && isupper(c))
1685 # define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
1686 # define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
1688 # define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1689 # define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1690 # define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
1694 #define isIDCONT(c) isWORDCHAR(c)
1695 #define isIDCONT_A(c) isWORDCHAR_A(c)
1696 #define isIDCONT_L1(c) isWORDCHAR_L1(c)
1697 #define isIDCONT_LC(c) isWORDCHAR_LC(c)
1698 #define isPSXSPC_LC(c) isSPACE_LC(c)
1700 /* For internal core Perl use only: the base macros for defining macros like
1701 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1702 * number defined earlier in this file. _generic_uvchr() is used for POSIX
1703 * classes where there is a macro or function 'above_latin1' that takes the
1704 * single argument 'c' and returns the desired value. These exist for those
1705 * classes which have simple definitions, avoiding the overhead of a hash
1706 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
1707 * for classes where that overhead is faster than a direct lookup.
1708 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
1709 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1710 * have duplicate checks here, so could create versions of the macros that
1711 * don't, but experiments show that gcc optimizes them out anyway. */
1713 /* Note that all ignore 'use bytes' */
1714 #define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1715 ? _generic_isCC(c, classnum) \
1717 #define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1718 ? _generic_isCC(c, classnum) \
1719 : _is_uni_FOO(classnum, c))
1720 #define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1721 #define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1722 #define isASCII_uvchr(c) isASCII(c)
1723 #define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1724 #define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1725 #define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1726 #define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1727 #define isIDCONT_uvchr(c) \
1728 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1729 #define isIDFIRST_uvchr(c) \
1730 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
1731 #define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1732 #define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1734 #define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1735 #define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1736 #define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1738 #define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1739 #define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1740 #define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1741 #define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1743 #define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1744 #define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1745 #define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1746 #define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1748 /* For backwards compatibility, even though '_uni' should mean official Unicode
1749 * code points, in Perl it means native for those below 256 */
1750 #define isALPHA_uni(c) isALPHA_uvchr(c)
1751 #define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1752 #define isASCII_uni(c) isASCII_uvchr(c)
1753 #define isBLANK_uni(c) isBLANK_uvchr(c)
1754 #define isCNTRL_uni(c) isCNTRL_uvchr(c)
1755 #define isDIGIT_uni(c) isDIGIT_uvchr(c)
1756 #define isGRAPH_uni(c) isGRAPH_uvchr(c)
1757 #define isIDCONT_uni(c) isIDCONT_uvchr(c)
1758 #define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1759 #define isLOWER_uni(c) isLOWER_uvchr(c)
1760 #define isPRINT_uni(c) isPRINT_uvchr(c)
1761 #define isPUNCT_uni(c) isPUNCT_uvchr(c)
1762 #define isSPACE_uni(c) isSPACE_uvchr(c)
1763 #define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1764 #define isUPPER_uni(c) isUPPER_uvchr(c)
1765 #define isVERTWS_uni(c) isVERTWS_uvchr(c)
1766 #define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1767 #define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1768 #define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1769 #define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1770 #define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1771 #define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
1773 /* For internal core Perl use only: the base macros for defining macros like
1774 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
1775 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
1776 * simple class definitions; the other for more complex. These are like
1777 * _generic_uvchr, so see it for more info. */
1778 #define _generic_LC_uvchr(latin1, above_latin1, c) \
1779 (c < 256 ? latin1(c) : above_latin1(c))
1780 #define _generic_LC_swash_uvchr(latin1, classnum, c) \
1781 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1783 #define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1784 #define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1785 _CC_ALPHANUMERIC, c)
1786 #define isASCII_LC_uvchr(c) isASCII_LC(c)
1787 #define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1788 is_HORIZWS_cp_high, c)
1789 #define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
1790 #define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1791 #define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
1792 #define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
1793 _is_uni_perl_idcont, c)
1794 #define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
1795 _is_uni_perl_idstart, c)
1796 #define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1797 #define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
1798 #define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
1799 #define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
1800 #define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
1801 is_XPERLSPACE_cp_high, c)
1802 #define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
1803 #define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
1805 #define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1806 is_XDIGIT_cp_high, c)
1808 #define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
1810 /* For internal core Perl use only: the base macros for defining macros like
1811 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1812 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1813 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1814 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1815 * same representation whether utf8 or not. Note that it assumes that the utf8
1816 * has been validated, and ignores 'use bytes' */
1817 #define _base_generic_utf8(enum_name, name, p, use_locale ) \
1818 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1820 "is" STRINGIFY(name) "_utf8", \
1821 "is" STRINGIFY(name) "_utf8_safe", \
1822 1, use_locale, __FILE__,__LINE__)
1824 #define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
1826 /* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1827 * they don't otherwise go out of their way to look for malformed UTF-8. If
1828 * they can return accurate results without knowing if the input is otherwise
1829 * malformed, they do so. For example isASCII is accurate in spite of any
1830 * non-length malformations because it looks only at a single byte. Likewise
1831 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1832 * variant ones return FALSE. But, if the input has to be well-formed in order
1833 * for the results to be accurate, the macros will test and if malformed will
1834 * call a routine to die
1836 * Except for toke.c, the macros do assume that e > p, asserting that on
1837 * DEBUGGING builds. Much code that calls these depends on this being true,
1838 * for other reasons. toke.c is treated specially as using the regular
1839 * assertion breaks it in many ways. All strings that these operate on there
1840 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1841 * bunch of code in toke.c assumes that this is true, so the assertion allows
1843 #ifdef PERL_IN_TOKE_C
1844 # define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1846 # define _utf8_safe_assert(p,e) ((e) > (p))
1849 #define _generic_utf8_safe(classnum, p, e, above_latin1) \
1850 (__ASSERT_(_utf8_safe_assert(p, e)) \
1851 (UTF8_IS_INVARIANT(*(p))) \
1852 ? _generic_isCC(*(p), classnum) \
1853 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1854 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1855 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1857 : (_force_out_malformed_utf8_message( \
1858 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1860 /* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1861 * 'above_latin1' can be a macro */
1862 #define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1863 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1864 #define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1865 _generic_utf8_safe(classnum, p, e, \
1866 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1867 ? (_force_out_malformed_utf8_message( \
1868 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1870 /* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
1871 * 'above_latin1' parameter */
1872 #define _generic_swash_utf8_safe(classnum, p, e) \
1873 _generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
1875 /* Like the above, but should be used only when it is known that there are no
1876 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1877 * class is TRUE for. Hence it can skip the tests for this range.
1878 * 'above_latin1' should include its arguments */
1879 #define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1880 (__ASSERT_(_utf8_safe_assert(p, e)) \
1881 (UTF8_IS_INVARIANT(*(p))) \
1882 ? _generic_isCC(*(p), classnum) \
1883 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1884 ? 0 /* Note that doesn't check validity for latin1 */ \
1887 /* NOTE that some of these macros have very similar ones in regcharclass.h.
1888 * For example, there is (at the time of this writing) an 'is_SPACE_utf8()'
1889 * there, differing in name only by an underscore from the one here
1890 * 'isSPACE_utf8(). The difference is that the ones here are probably more
1891 * efficient and smaller, using an O(1) array lookup for Latin1-range code
1892 * points; the regcharclass.h ones are implemented as a series of
1893 * "if-else-if-else ..." */
1895 #define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1896 #define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1897 #define isASCII_utf8(p) _generic_utf8(ASCII, p)
1898 #define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1899 #define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1900 #define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1901 #define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1902 #define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1903 #define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1904 #define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1905 #define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1906 #define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1907 #define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1908 #define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1909 #define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1910 #define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1911 #define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1912 #define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
1914 #define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1915 #define isALPHANUMERIC_utf8_safe(p, e) \
1916 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1917 #define isASCII_utf8_safe(p, e) \
1918 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1920 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1921 #define isBLANK_utf8_safe(p, e) \
1922 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1925 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1926 * more efficient macro instead of the more general one */
1927 # define isCNTRL_utf8_safe(p, e) \
1928 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
1930 # define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
1933 #define isDIGIT_utf8_safe(p, e) \
1934 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1935 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1936 #define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1937 #define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1938 _is_utf8_perl_idcont_with_len, p, e)
1940 /* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1941 * IDFIRST is an alnum. See
1942 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
1943 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1944 * This used to be not the XID version, but we decided to go with the more
1945 * modern Unicode definition */
1946 #define isIDFIRST_utf8_safe(p, e) \
1947 _generic_func_utf8_safe(_CC_IDFIRST, \
1948 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1950 #define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1951 #define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1952 #define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1953 #define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1954 #define isSPACE_utf8_safe(p, e) \
1955 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1956 #define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1957 #define isVERTWS_utf8_safe(p, e) \
1958 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1959 #define isWORDCHAR_utf8_safe(p, e) \
1960 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1961 #define isXDIGIT_utf8_safe(p, e) \
1962 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
1963 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1964 ? (_force_out_malformed_utf8_message( \
1965 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1966 : is_XDIGIT_high(p)))
1968 #define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
1969 #define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
1970 #define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
1971 #define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
1973 /* For internal core use only, subject to change */
1974 #define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
1975 #define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
1976 #define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
1977 #define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
1979 #define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
1980 #define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
1981 #define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
1982 #define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
1984 /* For internal core Perl use only: the base macros for defining macros like
1985 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
1986 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
1987 * 'macro' parameter */
1988 #define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
1990 #define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
1991 #define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
1992 #define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
1993 #define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
1994 #define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
1995 #define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
1996 #define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
1997 #define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
1998 #define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
1999 #define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
2000 #define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
2001 #define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
2002 #define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
2003 #define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
2004 #define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
2005 #define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
2006 #define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
2008 /* For internal core Perl use only: the base macros for defining macros like
2009 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
2010 * point in 'p' is within the 0-255 range, it uses locale rules from the
2011 * passed-in 'macro' parameter */
2012 #define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
2013 (__ASSERT_(_utf8_safe_assert(p, e)) \
2014 (UTF8_IS_INVARIANT(*(p))) \
2016 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2017 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2018 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2019 : (_force_out_malformed_utf8_message( \
2020 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2023 #define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2024 _generic_LC_utf8_safe(macro, p, e, \
2025 _is_utf8_FOO_with_len(classnum, p, e))
2027 #define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2028 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2030 #define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2031 _generic_LC_utf8_safe(classnum, p, e, \
2032 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2033 ? (_force_out_malformed_utf8_message( \
2034 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2037 #define isALPHANUMERIC_LC_utf8_safe(p, e) \
2038 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2039 _CC_ALPHANUMERIC, p, e)
2040 #define isALPHA_LC_utf8_safe(p, e) \
2041 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2042 #define isASCII_LC_utf8_safe(p, e) \
2043 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2044 #define isBLANK_LC_utf8_safe(p, e) \
2045 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2046 #define isCNTRL_LC_utf8_safe(p, e) \
2047 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2048 #define isDIGIT_LC_utf8_safe(p, e) \
2049 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2050 #define isGRAPH_LC_utf8_safe(p, e) \
2051 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2052 #define isIDCONT_LC_utf8_safe(p, e) \
2053 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2054 _is_utf8_perl_idcont_with_len, p, e)
2055 #define isIDFIRST_LC_utf8_safe(p, e) \
2056 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2057 _is_utf8_perl_idstart_with_len, p, e)
2058 #define isLOWER_LC_utf8_safe(p, e) \
2059 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2060 #define isPRINT_LC_utf8_safe(p, e) \
2061 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2062 #define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2063 #define isPUNCT_LC_utf8_safe(p, e) \
2064 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2065 #define isSPACE_LC_utf8_safe(p, e) \
2066 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2067 #define isUPPER_LC_utf8_safe(p, e) \
2068 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2069 #define isWORDCHAR_LC_utf8_safe(p, e) \
2070 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2071 #define isXDIGIT_LC_utf8_safe(p, e) \
2072 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
2074 /* Macros for backwards compatibility and for completeness when the ASCII and
2075 * Latin1 values are identical */
2076 #define isALPHAU(c) isALPHA_L1(c)
2077 #define isDIGIT_L1(c) isDIGIT_A(c)
2078 #define isOCTAL(c) isOCTAL_A(c)
2079 #define isOCTAL_L1(c) isOCTAL_A(c)
2080 #define isXDIGIT_L1(c) isXDIGIT_A(c)
2081 #define isALNUM(c) isWORDCHAR(c)
2082 #define isALNUMU(c) isWORDCHAR_L1(c)
2083 #define isALNUM_LC(c) isWORDCHAR_LC(c)
2084 #define isALNUM_uni(c) isWORDCHAR_uni(c)
2085 #define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
2086 #define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2087 #define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2088 #define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2089 #define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2090 #define isALNUMC(c) isALPHANUMERIC(c)
2091 #define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2092 #define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
2093 #define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
2094 #define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
2095 #define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
2097 /* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2098 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2099 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2100 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2101 * the outlier from the block that contains the other controls, just like
2102 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2103 * block. If it weren't special cased, it would yield a non-control.
2104 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2107 # define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2109 # define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2111 ? (UNLIKELY((c) == '?') \
2112 ? QUESTION_MARK_CTRL \
2113 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2114 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2116 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2119 /* Line numbers are unsigned, 32 bits. */
2121 #define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
2123 /* Helpful alias for version prescan */
2124 #define is_LAX_VERSION(a,b) \
2125 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2127 #define is_STRICT_VERSION(a,b) \
2128 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2130 #define BADVERSION(a,b,c) \
2136 /* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2137 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2138 * which is advanced. The input is validated only by an assert() in DEBUGGING
2139 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2140 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2141 #define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2144 #define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2148 /* Converts a character known to represent an octal digit (0-7) to its numeric
2149 * value. The input is validated only by an assert() in DEBUGGING builds. In
2150 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2151 #define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2153 /* Efficiently returns a boolean as to if two native characters are equivalent
2154 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2155 * the ALPHA in the name is to remind you of that. This is asserted() in
2156 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2157 * works (on valid input) for both non- and UTF-8-encoded bytes.
2159 * When one of the inputs is a compile-time constant and gets folded by the
2160 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2161 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2162 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
2163 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2164 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2165 * just a single 0, in the bit position where the upper- and lowercase differ.
2167 #define isALPHA_FOLD_EQ(c1, c2) \
2168 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2169 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2170 #define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2173 =head1 Memory Management
2175 =for apidoc Am|void|Newx|void* ptr|int nitems|type
2176 The XSUB-writer's interface to the C C<malloc> function.
2178 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2180 In 5.9.3, Newx() and friends replace the older New() API, and drops
2181 the first parameter, I<x>, a debug aid which allowed callers to identify
2182 themselves. This aid has been superseded by a new build option,
2183 PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
2184 there for use in XS modules supporting older perls.
2186 =for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
2187 The XSUB-writer's interface to the C C<malloc> function, with
2188 cast. See also C<L</Newx>>.
2190 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2192 =for apidoc Am|void|Newxz|void* ptr|int nitems|type
2193 The XSUB-writer's interface to the C C<malloc> function. The allocated
2194 memory is zeroed with C<memzero>. See also C<L</Newx>>.
2196 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2198 =for apidoc Am|void|Renew|void* ptr|int nitems|type
2199 The XSUB-writer's interface to the C C<realloc> function.
2201 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2203 =for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2204 The XSUB-writer's interface to the C C<realloc> function, with
2207 Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
2209 =for apidoc Am|void|Safefree|void* ptr
2210 The XSUB-writer's interface to the C C<free> function.
2212 This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
2214 =for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2215 The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
2216 source, C<dest> is the destination, C<nitems> is the number of items, and
2217 C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
2219 =for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
2220 Like C<Move> but returns C<dest>. Useful
2221 for encouraging compilers to tail-call
2224 =for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2225 The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
2226 source, C<dest> is the destination, C<nitems> is the number of items, and
2227 C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
2229 =for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2231 Like C<Copy> but returns C<dest>. Useful
2232 for encouraging compilers to tail-call
2235 =for apidoc Am|void|Zero|void* dest|int nitems|type
2237 The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2238 destination, C<nitems> is the number of items, and C<type> is the type.
2240 =for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2242 Like C<Zero> but returns dest. Useful
2243 for encouraging compilers to tail-call
2246 =for apidoc Am|void|StructCopy|type *src|type *dest|type
2247 This is an architecture-independent macro to copy one structure to another.
2249 =for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2251 Fill up memory with a byte pattern (a byte repeated over and over
2252 again) that hopefully catches attempts to access uninitialized memory.
2254 =for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2256 PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2258 =for apidoc Am|void|PoisonFree|void* dest|int nitems|type
2260 PoisonWith(0xEF) for catching access to freed memory.
2262 =for apidoc Am|void|Poison|void* dest|int nitems|type
2264 PoisonWith(0xEF) for catching access to freed memory.
2268 /* Maintained for backwards-compatibility only. Use newSV() instead. */
2270 #define NEWSV(x,len) newSV(len)
2273 #define MEM_SIZE_MAX ((MEM_SIZE)-1)
2275 #define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
2277 #ifdef PERL_MALLOC_WRAP
2279 /* This expression will be constant-folded at compile time. It checks
2280 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2281 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2282 * As well as avoiding the need for a run-time check in some cases, it's
2283 * designed to avoid compiler warnings like:
2284 * comparison is always false due to limited range of data type
2285 * It's mathematically equivalent to
2286 * max(n) * sizeof(t) > MEM_SIZE_MAX
2289 # define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
2290 ( sizeof(MEM_SIZE) < sizeof(n) \
2291 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
2293 /* This is written in a slightly odd way to avoid various spurious
2294 * compiler warnings. We *want* to write the expression as
2295 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2296 * (for some compile-time constant C), but even when the LHS
2297 * constant-folds to false at compile-time, g++ insists on emitting
2298 * warnings about the RHS (e.g. "comparison is always false"), so instead
2301 * (cond ? n : X) > C
2303 * where X is a constant with X > C always false. Choosing a value for X
2304 * is tricky. If 0, some compilers will complain about 0 > C always being
2305 * false; if 1, Coverity complains when n happens to be the constant value
2306 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2307 * for X and hope that nothing else whines.
2310 # define _MEM_WRAP_WILL_WRAP(n,t) \
2311 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2312 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
2314 # define MEM_WRAP_CHECK(n,t) \
2315 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2316 && (croak_memory_wrap(),0))
2318 # define MEM_WRAP_CHECK_1(n,t,a) \
2319 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2320 && (Perl_croak_nocontext("%s",(a)),0))
2322 #define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
2324 #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))
2327 #define MEM_WRAP_CHECK(n,t)
2328 #define MEM_WRAP_CHECK_1(n,t,a)
2329 #define MEM_WRAP_CHECK_2(n,t,a,b)
2330 #define MEM_WRAP_CHECK_(n,t)
2332 #define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
2338 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
2339 * go through functions, which are handy for debugging breakpoints, but
2340 * which more importantly get the immediate calling environment (file and
2341 * line number, and C function name if available) passed in. This info can
2342 * then be used for logging the calls, for which one gets a sample
2343 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
2346 * - not all memory allocs get logged, only those
2347 * that go through Newx() and derivatives (while all
2348 * Safefrees do get logged)
2349 * - __FILE__ and __LINE__ do not work everywhere
2350 * - __func__ or __FUNCTION__ even less so
2351 * - I think more goes on after the perlio frees but
2352 * the thing is that STDERR gets closed (as do all
2353 * the file descriptors)
2354 * - no deeper calling stack than the caller of the Newx()
2355 * or the kind, but do I look like a C reflection/introspection
2357 * - the function prototypes for the logging functions
2358 * probably should maybe be somewhere else than handy.h
2359 * - one could consider inlining (macrofying) the logging
2360 * for speed, but I am too lazy
2361 * - one could imagine recording the allocations in a hash,
2362 * (keyed by the allocation address?), and maintain that
2363 * through reallocs and frees, but how to do that without
2364 * any News() happening...?
2365 * - lots of -Ddefines to get useful/controllable output
2366 * - lots of ENV reads
2370 # ifndef PERL_MEM_LOG_NOIMPL
2379 # if defined(PERL_IN_SV_C) /* those are only used in sv.c */
2380 void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2381 void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2388 #define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2389 #define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
2390 #define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
2393 #ifndef MEM_LOG_ALLOC
2394 #define MEM_LOG_ALLOC(n,t,a) (a)
2396 #ifndef MEM_LOG_REALLOC
2397 #define MEM_LOG_REALLOC(n,t,v,a) (a)
2399 #ifndef MEM_LOG_FREE
2400 #define MEM_LOG_FREE(a) (a)
2403 #define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2404 #define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2405 #define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
2408 /* pre 5.9.x compatibility */
2409 #define New(x,v,n,t) Newx(v,n,t)
2410 #define Newc(x,v,n,t,c) Newxc(v,n,t,c)
2411 #define Newz(x,v,n,t) Newxz(v,n,t)
2414 #define Renew(v,n,t) \
2415 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
2416 #define Renewc(v,n,t,c) \
2417 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
2420 #define Safefree(d) \
2421 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
2423 #define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
2426 /* assert that a valid ptr has been supplied - use this instead of assert(ptr) *
2427 * as it handles cases like constant string arguments without throwing warnings *
2428 * the cast is required, as is the inequality check, to avoid warnings */
2429 #define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
2432 #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)))
2433 #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)))
2434 #define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2436 /* Like above, but returns a pointer to 'd' */
2437 #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)))
2438 #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)))
2439 #define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
2441 #define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2442 #define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2443 #define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2444 #define Poison(d,n,t) PoisonFree(d,n,t)
2447 # define PERL_POISON_EXPR(x) x
2449 # define PERL_POISON_EXPR(x)
2452 #define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2454 /* C_ARRAY_LENGTH is the number of elements in the C array (so you
2455 * want your zero-based indices to be less than but not equal to).
2457 * C_ARRAY_END is one past the last: half-open/half-closed range,
2458 * not last-inclusive range. */
2459 #define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
2460 #define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
2464 # define Perl_va_copy(s, d) va_copy(d, s)
2465 # elif defined(__va_copy)
2466 # define Perl_va_copy(s, d) __va_copy(d, s)
2468 # define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2472 /* convenience debug macros */
2474 #define pTHX_FORMAT "Perl interpreter: 0x%p"
2475 #define pTHX__FORMAT ", Perl interpreter: 0x%p"
2476 #define pTHX_VALUE_ (void *)my_perl,
2477 #define pTHX_VALUE (void *)my_perl
2478 #define pTHX__VALUE_ ,(void *)my_perl,
2479 #define pTHX__VALUE ,(void *)my_perl
2482 #define pTHX__FORMAT
2485 #define pTHX__VALUE_
2487 #endif /* USE_ITHREADS */
2489 /* Perl_deprecate was not part of the public API, and did not have a deprecate()
2490 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2491 CPAN::Unpack show any users outside the core. */
2493 # define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2494 "Use of " s " is deprecated")
2495 # define deprecate_disappears_in(when,message) \
2496 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2497 message ", and will disappear in Perl " when)
2498 # define deprecate_fatal_in(when,message) \
2499 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2500 message ". Its use will be fatal in Perl " when)
2503 /* Internal macros to deal with gids and uids */
2506 # if Uid_t_size > IVSIZE
2507 # define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2508 # define SvUID(sv) SvNV(sv)
2509 # elif Uid_t_sign <= 0
2510 # define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2511 # define SvUID(sv) SvIV(sv)
2513 # define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2514 # define SvUID(sv) SvUV(sv)
2515 # endif /* Uid_t_size */
2517 # if Gid_t_size > IVSIZE
2518 # define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2519 # define SvGID(sv) SvNV(sv)
2520 # elif Gid_t_sign <= 0
2521 # define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2522 # define SvGID(sv) SvIV(sv)
2524 # define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2525 # define SvGID(sv) SvUV(sv)
2526 # endif /* Gid_t_size */
2530 #endif /* PERL_HANDY_H_ */
2533 * ex: set ts=8 sts=4 sw=4 et: