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 HANDY_H /* Guard against nested #inclusion */
17 #if !defined(__STDC__)
25 # define Null(type) ((type)NULL)
30 =for apidoc AmU||Nullch
31 Null character pointer. (No longer available when C<PERL_CORE> is
34 =for apidoc AmU||Nullsv
35 Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
40 # define Nullch Null(char*)
41 # define Nullfp Null(PerlIO*)
42 # define Nullsv Null(SV*)
54 /* The MUTABLE_*() macros cast pointers to the types shown, in such a way
55 * (compiler permitting) that casting away const-ness will give a warning;
59 * AV *av1 = (AV*)sv; <== BAD: the const has been silently cast away
60 * AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
63 #if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
64 # define MUTABLE_PTR(p) ({ void *_p = (p); _p; })
66 # define MUTABLE_PTR(p) ((void *) (p))
69 #define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
70 #define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
71 #define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
72 #define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
73 #define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
74 #define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
76 #if defined(I_STDBOOL) && !defined(PERL_BOOL_AS_CHAR)
83 /* bool is built-in for g++-2.6.3 and later, which might be used
84 for extensions. <_G_config.h> defines _G_HAVE_BOOL, but we can't
85 be sure _G_config.h will be included before this file. _G_config.h
86 also defines _G_HAVE_BOOL for both gcc and g++, but only g++
87 actually has bool. Hence, _G_HAVE_BOOL is pretty useless for us.
88 g++ can be identified by __GNUG__.
89 Andy Dougherty February 2000
91 #ifdef __GNUG__ /* GNU g++ has bool built-in */
92 # ifndef PERL_BOOL_AS_CHAR
107 /* cast-to-bool. A simple (bool) cast may not do the right thing: if bool is
108 * defined as char for example, then the cast from int is
109 * implementation-defined (bool)!!(cbool) in a ternary triggers a bug in xlc on
111 #define cBOOL(cbool) ((cbool) ? (bool)1 : (bool)0)
113 /* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
114 * XXX Should really be a Configure probe, with HAS__FUNCTION__
115 * and FUNCTION__ as results.
116 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
117 #if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
118 # define FUNCTION__ __func__
120 # if (defined(USING_MSVC6)) || /* MSVC6 has neither __func__ nor __FUNCTION and no good workarounds, either. */ \
121 (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
122 # define FUNCTION__ ""
124 # define FUNCTION__ __FUNCTION__ /* Common extension. */
128 /* XXX A note on the perl source internal type system. The
129 original intent was that I32 be *exactly* 32 bits.
131 Currently, we only guarantee that I32 is *at least* 32 bits.
132 Specifically, if int is 64 bits, then so is I32. (This is the case
133 for the Cray.) This has the advantage of meshing nicely with
134 standard library calls (where we pass an I32 and the library is
135 expecting an int), but the disadvantage that an I32 is not 32 bits.
136 Andy Dougherty August 1996
138 There is no guarantee that there is *any* integral type with
139 exactly 32 bits. It is perfectly legal for a system to have
140 sizeof(short) == sizeof(int) == sizeof(long) == 8.
142 Similarly, there is no guarantee that I16 and U16 have exactly 16
145 For dealing with issues that may arise from various 32/64-bit
146 systems, we will ask Configure to check out
148 SHORTSIZE == sizeof(short)
149 INTSIZE == sizeof(int)
150 LONGSIZE == sizeof(long)
151 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
152 PTRSIZE == sizeof(void *)
153 DOUBLESIZE == sizeof(double)
154 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
158 #ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
159 # include <inttypes.h>
160 # ifdef INT32_MIN_BROKEN
162 # define INT32_MIN (-2147483647-1)
164 # ifdef INT64_MIN_BROKEN
166 # define INT64_MIN (-9223372036854775807LL-1)
182 /* INT64_C/UINT64_C are C99 from <stdint.h> (so they will not be
183 * available in strict C89 mode), but they are nice, so let's define
184 * them if necessary. */
185 #if defined(HAS_QUAD)
187 # undef PeRl_UINT64_C
188 /* Prefer the native integer types (int and long) over long long
189 * (which is not C89) and Win32-specific __int64. */
190 # if QUADKIND == QUAD_IS_INT && INTSIZE == 8
191 # define PeRl_INT64_C(c) (c)
192 # define PeRl_UINT64_C(c) CAT2(c,U)
194 # if QUADKIND == QUAD_IS_LONG && LONGSIZE == 8
195 # define PeRl_INT64_C(c) CAT2(c,L)
196 # define PeRl_UINT64_C(c) CAT2(c,UL)
198 # if QUADKIND == QUAD_IS_LONG_LONG && defined(HAS_LONG_LONG)
199 # define PeRl_INT64_C(c) CAT2(c,LL)
200 # define PeRl_UINT64_C(c) CAT2(c,ULL)
202 # if QUADKIND == QUAD_IS___INT64
203 # define PeRl_INT64_C(c) CAT2(c,I64)
204 # define PeRl_UINT64_C(c) CAT2(c,UI64)
206 # ifndef PeRl_INT64_C
207 # define PeRl_INT64_C(c) ((I64)(c)) /* last resort */
208 # define PeRl_UINT64_C(c) ((U64)(c))
210 /* In OS X the INT64_C/UINT64_C are defined with LL/ULL, which will
211 * not fly with C89-pedantic gcc, so let's undefine them first so that
212 * we can redefine them with our native integer preferring versions. */
213 # if defined(PERL_DARWIN) && defined(PERL_GCC_PEDANTIC)
218 # define INT64_C(c) PeRl_INT64_C(c)
221 # define UINT64_C(c) PeRl_UINT64_C(c)
225 #if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
227 /* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
228 Please search CHAR_MAX in perl.h for further details. */
229 #define U8_MAX UINT8_MAX
230 #define U8_MIN UINT8_MIN
232 #define I16_MAX INT16_MAX
233 #define I16_MIN INT16_MIN
234 #define U16_MAX UINT16_MAX
235 #define U16_MIN UINT16_MIN
237 #define I32_MAX INT32_MAX
238 #define I32_MIN INT32_MIN
239 #ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
240 # define U32_MAX UINT32_MAX
242 # define U32_MAX 4294967295U
244 #define U32_MIN UINT32_MIN
248 /* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
249 Please search CHAR_MAX in perl.h for further details. */
250 #define U8_MAX PERL_UCHAR_MAX
251 #define U8_MIN PERL_UCHAR_MIN
253 #define I16_MAX PERL_SHORT_MAX
254 #define I16_MIN PERL_SHORT_MIN
255 #define U16_MAX PERL_USHORT_MAX
256 #define U16_MIN PERL_USHORT_MIN
259 # define I32_MAX PERL_INT_MAX
260 # define I32_MIN PERL_INT_MIN
261 # define U32_MAX PERL_UINT_MAX
262 # define U32_MIN PERL_UINT_MIN
264 # define I32_MAX PERL_LONG_MAX
265 # define I32_MIN PERL_LONG_MIN
266 # define U32_MAX PERL_ULONG_MAX
267 # define U32_MIN PERL_ULONG_MIN
272 /* log(2) is pretty close to 0.30103, just in case anyone is grepping for it */
273 #define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log2(10) =~ 146/485 */
274 #define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
275 #define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
277 /* Unused by core; should be deprecated */
278 #define Ctl(ch) ((ch) & 037)
280 /* This is a helper macro to avoid preprocessor issues, replaced by nothing
281 * unless under DEBUGGING, where it expands to an assert of its argument,
282 * followed by a comma (hence the comma operator). If we just used a straight
283 * assert(), we would get a comma with nothing before it when not DEBUGGING.
285 * We also use empty definition under Coverity since the __ASSERT__
286 * checks often check for things that Really Cannot Happen, and Coverity
287 * detects that and gets all excited. */
289 #if defined(DEBUGGING) && !defined(__COVERITY__)
290 # define __ASSERT_(statement) assert(statement),
292 # define __ASSERT_(statement)
296 =head1 SV-Body Allocation
298 =for apidoc Ama|SV*|newSVpvs|const char* s
299 Like C<newSVpvn>, but takes a literal C<NUL>-terminated string instead of a
302 =for apidoc Ama|SV*|newSVpvs_flags|const char* s|U32 flags
303 Like C<newSVpvn_flags>, but takes a literal C<NUL>-terminated string instead of
304 a string/length pair.
306 =for apidoc Ama|SV*|newSVpvs_share|const char* s
307 Like C<newSVpvn_share>, but takes a literal C<NUL>-terminated string instead of
308 a string/length pair and omits the hash parameter.
310 =for apidoc Am|void|sv_catpvs_flags|SV* sv|const char* s|I32 flags
311 Like C<sv_catpvn_flags>, but takes a literal C<NUL>-terminated string instead
312 of a string/length pair.
314 =for apidoc Am|void|sv_catpvs_nomg|SV* sv|const char* s
315 Like C<sv_catpvn_nomg>, but takes a literal string instead of a
318 =for apidoc Am|void|sv_catpvs|SV* sv|const char* s
319 Like C<sv_catpvn>, but takes a literal string instead of a string/length pair.
321 =for apidoc Am|void|sv_catpvs_mg|SV* sv|const char* s
322 Like C<sv_catpvn_mg>, but takes a literal string instead of a
325 =for apidoc Am|void|sv_setpvs|SV* sv|const char* s
326 Like C<sv_setpvn>, but takes a literal string instead of a string/length pair.
328 =for apidoc Am|void|sv_setpvs_mg|SV* sv|const char* s
329 Like C<sv_setpvn_mg>, but takes a literal string instead of a
332 =for apidoc Am|SV *|sv_setref_pvs|const char* s
333 Like C<sv_setref_pvn>, but takes a literal string instead of a
336 =head1 Memory Management
338 =for apidoc Ama|char*|savepvs|const char* s
339 Like C<savepvn>, but takes a literal C<NUL>-terminated string instead of a
342 =for apidoc Ama|char*|savesharedpvs|const char* s
343 A version of C<savepvs()> which allocates the duplicate string in memory
344 which is shared between threads.
348 =for apidoc Am|HV*|gv_stashpvs|const char* name|I32 create
349 Like C<gv_stashpvn>, but takes a literal string instead of a string/length pair.
351 =head1 Hash Manipulation Functions
353 =for apidoc Am|SV**|hv_fetchs|HV* tb|const char* key|I32 lval
354 Like C<hv_fetch>, but takes a literal string instead of a string/length pair.
356 =for apidoc Am|SV**|hv_stores|HV* tb|const char* key|NULLOK SV* val
357 Like C<hv_store>, but takes a literal string instead of a string/length pair
358 and omits the hash parameter.
360 =head1 Lexer interface
362 =for apidoc Amx|void|lex_stuff_pvs|const char *pv|U32 flags
364 Like L</lex_stuff_pvn>, but takes a literal string instead of a
370 /* concatenating with "" ensures that only literal strings are accepted as
372 #define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
374 /* note that STR_WITH_LEN() can't be used as argument to macros or functions
375 * that under some configurations might be macros, which means that it requires
376 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
379 /* STR_WITH_LEN() shortcuts */
380 #define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
381 #define newSVpvs_flags(str,flags) \
382 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
383 #define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
384 #define sv_catpvs_flags(sv, str, flags) \
385 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
386 #define sv_catpvs_nomg(sv, str) \
387 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
388 #define sv_catpvs(sv, str) \
389 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
390 #define sv_catpvs_mg(sv, str) \
391 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
392 #define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
393 #define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
394 #define sv_setref_pvs(rv, classname, str) \
395 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
396 #define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
397 #define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
398 #define gv_stashpvs(str, create) \
399 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
400 #define gv_fetchpvs(namebeg, add, sv_type) \
401 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
402 #define gv_fetchpvn(namebeg, len, add, sv_type) \
403 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
404 #define sv_catxmlpvs(dsv, str, utf8) \
405 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
406 #define hv_fetchs(hv,key,lval) \
407 ((SV **)Perl_hv_common(aTHX_ (hv), NULL, STR_WITH_LEN(key), 0, \
408 (lval) ? (HV_FETCH_JUST_SV | HV_FETCH_LVALUE) \
409 : HV_FETCH_JUST_SV, NULL, 0))
411 #define hv_stores(hv,key,val) \
412 ((SV **)Perl_hv_common(aTHX_ (hv), NULL, STR_WITH_LEN(key), 0, \
413 (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), (val), 0))
415 #define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
417 #define get_cvs(str, flags) \
418 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
421 =head1 Miscellaneous Functions
423 =for apidoc Am|bool|strNE|char* s1|char* s2
424 Test two strings to see if they are different. Returns true or
427 =for apidoc Am|bool|strEQ|char* s1|char* s2
428 Test two strings to see if they are equal. Returns true or false.
430 =for apidoc Am|bool|strLT|char* s1|char* s2
431 Test two strings to see if the first, C<s1>, is less than the second,
432 C<s2>. Returns true or false.
434 =for apidoc Am|bool|strLE|char* s1|char* s2
435 Test two strings to see if the first, C<s1>, is less than or equal to the
436 second, C<s2>. Returns true or false.
438 =for apidoc Am|bool|strGT|char* s1|char* s2
439 Test two strings to see if the first, C<s1>, is greater than the second,
440 C<s2>. Returns true or false.
442 =for apidoc Am|bool|strGE|char* s1|char* s2
443 Test two strings to see if the first, C<s1>, is greater than or equal to
444 the second, C<s2>. Returns true or false.
446 =for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
447 Test two strings to see if they are different. The C<len> parameter
448 indicates the number of bytes to compare. Returns true or false. (A
449 wrapper for C<strncmp>).
451 =for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
452 Test two strings to see if they are equal. The C<len> parameter indicates
453 the number of bytes to compare. Returns true or false. (A wrapper for
456 =for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
457 Test two buffers (which may contain embedded C<NUL> characters, to see if they
458 are equal. The C<len> parameter indicates the number of bytes to compare.
459 Returns zero if equal, or non-zero if non-equal.
461 =for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
462 Test two buffers (which may contain embedded C<NUL> characters, to see if they
463 are not equal. The C<len> parameter indicates the number of bytes to compare.
464 Returns zero if non-equal, or non-zero if equal.
469 #define strNE(s1,s2) (strcmp(s1,s2))
470 #define strEQ(s1,s2) (!strcmp(s1,s2))
471 #define strLT(s1,s2) (strcmp(s1,s2) < 0)
472 #define strLE(s1,s2) (strcmp(s1,s2) <= 0)
473 #define strGT(s1,s2) (strcmp(s1,s2) > 0)
474 #define strGE(s1,s2) (strcmp(s1,s2) >= 0)
475 #define strnNE(s1,s2,l) (strncmp(s1,s2,l))
476 #define strnEQ(s1,s2,l) (!strncmp(s1,s2,l))
479 # define memNE(s1,s2,l) (memcmp(s1,s2,l))
480 # define memEQ(s1,s2,l) (!memcmp(s1,s2,l))
482 # define memNE(s1,s2,l) (bcmp(s1,s2,l))
483 # define memEQ(s1,s2,l) (!bcmp(s1,s2,l))
486 #define memEQs(s1, l, s2) \
487 (sizeof(s2)-1 == l && memEQ(s1, ("" s2 ""), (sizeof(s2)-1)))
488 #define memNEs(s1, l, s2) !memEQs(s1, l, s2)
493 * Unfortunately, the introduction of locales means that we
494 * can't trust isupper(), etc. to tell the truth. And when
495 * it comes to /\w+/ with tainting enabled, we *must* be able
496 * to trust our character classes.
498 * Therefore, the default tests in the text of Perl will be
499 * independent of locale. Any code that wants to depend on
500 * the current locale will use the tests that begin with "lc".
503 #ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
511 =head1 Character classification
512 This section is about functions (really macros) that classify characters
513 into types, such as punctuation versus alphabetic, etc. Most of these are
514 analogous to regular expression character classes. (See
515 L<perlrecharclass/POSIX Character Classes>.) There are several variants for
516 each class. (Not all macros have all variants; each item below lists the
517 ones valid for it.) None are affected by C<use bytes>, and only the ones
518 with C<LC> in the name are affected by the current locale.
520 The base function, e.g., C<isALPHA()>, takes an octet (either a C<char> or a
521 C<U8>) as input and returns a boolean as to whether or not the character
522 represented by that octet is (or on non-ASCII platforms, corresponds to) an
523 ASCII character in the named class based on platform, Unicode, and Perl rules.
524 If the input is a number that doesn't fit in an octet, FALSE is returned.
526 Variant C<isFOO_A> (e.g., C<isALPHA_A()>) is identical to the base function
527 with no suffix C<"_A">.
529 Variant C<isFOO_L1> imposes the Latin-1 (or EBCDIC equivlalent) character set
530 onto the platform. That is, the code points that are ASCII are unaffected,
531 since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
532 as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
533 true when called with the code point 0xDF, which is a word character in both
534 ASCII and EBCDIC (though it represents different characters in each).
536 Variant C<isFOO_uni> is like the C<isFOO_L1> variant, but accepts any UV code
537 point as input. If the code point is larger than 255, Unicode rules are used
538 to determine if it is in the character class. For example,
539 C<isWORDCHAR_uni(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
540 WITH MACRON in Unicode, and is a word character.
542 Variant C<isFOO_utf8> is like C<isFOO_uni>, but the input is a pointer to a
543 (known to be well-formed) UTF-8 encoded string (C<U8*> or C<char*>). The
544 classification of just the first (possibly multi-byte) character in the string
547 Variant C<isFOO_LC> is like the C<isFOO_A> and C<isFOO_L1> variants, but the
548 result is based on the current locale, which is what C<LC> in the name stands
549 for. If Perl can determine that the current locale is a UTF-8 locale, it uses
550 the published Unicode rules; otherwise, it uses the C library function that
551 gives the named classification. For example, C<isDIGIT_LC()> when not in a
552 UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
553 returned if the input won't fit into an octet. On some platforms where the C
554 library function is known to be defective, Perl changes its result to follow
555 the POSIX standard's rules.
557 Variant C<isFOO_LC_uvchr> is like C<isFOO_LC>, but is defined on any UV. It
558 returns the same as C<isFOO_LC> for input code points less than 256, and
559 returns the hard-coded, not-affected-by-locale, Unicode results for larger ones.
561 Variant C<isFOO_LC_utf8> is like C<isFOO_LC_uvchr>, but the input is a pointer
562 to a (known to be well-formed) UTF-8 encoded string (C<U8*> or C<char*>). The
563 classification of just the first (possibly multi-byte) character in the string
566 =for apidoc Am|bool|isALPHA|char ch
567 Returns a boolean indicating whether the specified character is an
568 alphabetic character, analogous to C<m/[[:alpha:]]/>.
569 See the L<top of this section|/Character classification> for an explanation of
571 C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uni>, C<isALPHA_utf8>, C<isALPHA_LC>,
572 C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8>.
574 =for apidoc Am|bool|isALPHANUMERIC|char ch
575 Returns a boolean indicating whether the specified character is a either an
576 alphabetic character or decimal digit, analogous to C<m/[[:alnum:]]/>.
577 See the L<top of this section|/Character classification> for an explanation of
579 C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uni>,
580 C<isALPHANUMERIC_utf8>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>, and
581 C<isALPHANUMERIC_LC_utf8>.
583 =for apidoc Am|bool|isASCII|char ch
584 Returns a boolean indicating whether the specified character is one of the 128
585 characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
586 On non-ASCII platforms, it returns TRUE iff this
587 character corresponds to an ASCII character. Variants C<isASCII_A()> and
588 C<isASCII_L1()> are identical to C<isASCII()>.
589 See the L<top of this section|/Character classification> for an explanation of
591 C<isASCII_uni>, C<isASCII_utf8>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
592 C<isASCII_LC_utf8>. Note, however, that some platforms do not have the C
593 library routine C<isascii()>. In these cases, the variants whose names contain
594 C<LC> are the same as the corresponding ones without.
596 Also note, that because all ASCII characters are UTF-8 invariant (meaning they
597 have the exact same representation (always a single byte) whether encoded in
598 UTF-8 or not), C<isASCII> will give the correct results when called with any
599 byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8> will
600 work properly on any string encoded or not in UTF-8.
602 =for apidoc Am|bool|isBLANK|char ch
603 Returns a boolean indicating whether the specified character is a
604 character considered to be a blank, analogous to C<m/[[:blank:]]/>.
605 See the L<top of this section|/Character classification> for an explanation of
607 C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uni>, C<isBLANK_utf8>, C<isBLANK_LC>,
608 C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8>. Note, however, that some
609 platforms do not have the C library routine C<isblank()>. In these cases, the
610 variants whose names contain C<LC> are the same as the corresponding ones
613 =for apidoc Am|bool|isCNTRL|char ch
614 Returns a boolean indicating whether the specified character is a
615 control character, analogous to C<m/[[:cntrl:]]/>.
616 See the L<top of this section|/Character classification> for an explanation of
618 C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uni>, C<isCNTRL_utf8>, C<isCNTRL_LC>,
619 C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8>
620 On EBCDIC platforms, you almost always want to use the C<isCNTRL_L1> variant.
622 =for apidoc Am|bool|isDIGIT|char ch
623 Returns a boolean indicating whether the specified character is a
624 digit, analogous to C<m/[[:digit:]]/>.
625 Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
626 See the L<top of this section|/Character classification> for an explanation of
628 C<isDIGIT_uni>, C<isDIGIT_utf8>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
631 =for apidoc Am|bool|isGRAPH|char ch
632 Returns a boolean indicating whether the specified character is a
633 graphic character, analogous to C<m/[[:graph:]]/>.
634 See the L<top of this section|/Character classification> for an explanation of
636 C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uni>, C<isGRAPH_utf8>, C<isGRAPH_LC>,
637 C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8>.
639 =for apidoc Am|bool|isLOWER|char ch
640 Returns a boolean indicating whether the specified character is a
641 lowercase character, analogous to C<m/[[:lower:]]/>.
642 See the L<top of this section|/Character classification> for an explanation of
644 C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uni>, C<isLOWER_utf8>, C<isLOWER_LC>,
645 C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8>.
647 =for apidoc Am|bool|isOCTAL|char ch
648 Returns a boolean indicating whether the specified character is an
650 The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
653 =for apidoc Am|bool|isPUNCT|char ch
654 Returns a boolean indicating whether the specified character is a
655 punctuation character, analogous to C<m/[[:punct:]]/>.
656 Note that the definition of what is punctuation isn't as
657 straightforward as one might desire. See L<perlrecharclass/POSIX Character
658 Classes> for details.
659 See the L<top of this section|/Character classification> for an explanation of
661 C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uni>, C<isPUNCT_utf8>, C<isPUNCT_LC>,
662 C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8>.
664 =for apidoc Am|bool|isSPACE|char ch
665 Returns a boolean indicating whether the specified character is a
666 whitespace character. This is analogous
667 to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
668 this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
669 locale forms of this macro (the ones with C<LC> in their names) matched
670 precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
671 in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
672 (See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
673 See the L<top of this section|/Character classification> for an explanation of
675 C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uni>, C<isSPACE_utf8>, C<isSPACE_LC>,
676 C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8>.
678 =for apidoc Am|bool|isPSXSPC|char ch
679 (short for Posix Space)
680 Starting in 5.18, this is identical in all its forms to the
681 corresponding C<isSPACE()> macros.
682 The locale forms of this macro are identical to their corresponding
683 C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
684 non-locale forms differ from their C<isSPACE()> forms only in that the
685 C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
686 Otherwise they are identical. Thus this macro is analogous to what
687 C<m/[[:space:]]/> matches in a regular expression.
688 See the L<top of this section|/Character classification> for an explanation of
689 variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uni>, C<isPSXSPC_utf8>,
690 C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8>.
692 =for apidoc Am|bool|isUPPER|char ch
693 Returns a boolean indicating whether the specified character is an
694 uppercase character, analogous to C<m/[[:upper:]]/>.
695 See the L<top of this section|/Character classification> for an explanation of
697 C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uni>, C<isUPPER_utf8>, C<isUPPER_LC>,
698 C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8>.
700 =for apidoc Am|bool|isPRINT|char ch
701 Returns a boolean indicating whether the specified character is a
702 printable character, analogous to C<m/[[:print:]]/>.
703 See the L<top of this section|/Character classification> for an explanation of
705 C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uni>, C<isPRINT_utf8>, C<isPRINT_LC>,
706 C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8>.
708 =for apidoc Am|bool|isWORDCHAR|char ch
709 Returns a boolean indicating whether the specified character is a character
710 that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
711 in a regular expression. A word character is an alphabetic character, a
712 decimal digit, a connecting punctuation character (such as an underscore), or
713 a "mark" character that attaches to one of those (like some sort of accent).
714 C<isALNUM()> is a synonym provided for backward compatibility, even though a
715 word character includes more than the standard C language meaning of
717 See the L<top of this section|/Character classification> for an explanation of
719 C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uni>, and C<isWORDCHAR_utf8>.
720 C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and C<isWORDCHAR_LC_utf8> are also as
721 described there, but additionally include the platform's native underscore.
723 =for apidoc Am|bool|isXDIGIT|char ch
724 Returns a boolean indicating whether the specified character is a hexadecimal
725 digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
726 and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
727 See the L<top of this section|/Character classification> for an explanation of
729 C<isXDIGIT_uni>, C<isXDIGIT_utf8>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>, and
732 =for apidoc Am|bool|isIDFIRST|char ch
733 Returns a boolean indicating whether the specified character can be the first
734 character of an identifier. This is very close to, but not quite the same as
735 the official Unicode property C<XID_Start>. The difference is that this
736 returns true only if the input character also matches L</isWORDCHAR>.
737 See the L<top of this section|/Character classification> for an explanation of
739 C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uni>, C<isIDFIRST_utf8>,
740 C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8>.
742 =for apidoc Am|bool|isIDCONT|char ch
743 Returns a boolean indicating whether the specified character can be the
744 second or succeeding character of an identifier. This is very close to, but
745 not quite the same as the official Unicode property C<XID_Continue>. The
746 difference is that this returns true only if the input character also matches
747 L</isWORDCHAR>. See the L<top of this section|/Character classification> for
749 explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uni>,
750 C<isIDCONT_utf8>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
753 =head1 Miscellaneous Functions
755 =for apidoc Am|U8|READ_XDIGIT|char str*
756 Returns the value of an ASCII-range hex digit and advances the string pointer.
757 Behaviour is only well defined when isXDIGIT(*str) is true.
759 =head1 Character case changing
761 =for apidoc Am|U8|toUPPER|U8 ch
762 Converts the specified character to uppercase. If the input is anything but an
763 ASCII lowercase character, that input character itself is returned. Variant
764 C<toUPPER_A> is equivalent.
766 =for apidoc Am|UV|toUPPER_uni|UV cp|U8* s|STRLEN* lenp
767 Converts the Unicode code point C<cp> to its uppercase version, and
768 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
769 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
770 bytes since the uppercase version may be longer than the original character.
772 The first code point of the uppercased version is returned
773 (but note, as explained just above, that there may be more.)
775 =for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
776 Converts the UTF-8 encoded character at C<p> to its uppercase version, and
777 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
778 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
779 bytes since the uppercase version may be longer than the original character.
781 The first code point of the uppercased version is returned
782 (but note, as explained just above, that there may be more.)
784 The input character at C<p> is assumed to be well-formed.
786 =for apidoc Am|U8|toFOLD|U8 ch
787 Converts the specified character to foldcase. If the input is anything but an
788 ASCII uppercase character, that input character itself is returned. Variant
789 C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
790 Latin1 range, as the full generality of L</toFOLD_uni> is needed there.)
792 =for apidoc Am|UV|toFOLD_uni|UV cp|U8* s|STRLEN* lenp
793 Converts the Unicode code point C<cp> to its foldcase version, and
794 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
795 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
796 bytes since the foldcase version may be longer than the original character.
798 The first code point of the foldcased version is returned
799 (but note, as explained just above, that there may be more.)
801 =for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
802 Converts the UTF-8 encoded character at C<p> to its foldcase version, and
803 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
804 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
805 bytes since the foldcase version may be longer than the original character.
807 The first code point of the foldcased version is returned
808 (but note, as explained just above, that there may be more.)
810 The input character at C<p> is assumed to be well-formed.
812 =for apidoc Am|U8|toLOWER|U8 ch
813 Converts the specified character to lowercase. If the input is anything but an
814 ASCII uppercase character, that input character itself is returned. Variant
815 C<toLOWER_A> is equivalent.
817 =for apidoc Am|U8|toLOWER_L1|U8 ch
818 Converts the specified Latin1 character to lowercase. The results are
819 undefined if the input doesn't fit in a byte.
821 =for apidoc Am|U8|toLOWER_LC|U8 ch
822 Converts the specified character to lowercase using the current locale's rules,
823 if possible; otherwise returns the input character itself.
825 =for apidoc Am|UV|toLOWER_uni|UV cp|U8* s|STRLEN* lenp
826 Converts the Unicode code point C<cp> to its lowercase version, and
827 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
828 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
829 bytes since the lowercase version may be longer than the original character.
831 The first code point of the lowercased version is returned
832 (but note, as explained just above, that there may be more.)
834 =for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
835 Converts the UTF-8 encoded character at C<p> to its lowercase version, and
836 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
837 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
838 bytes since the lowercase version may be longer than the original character.
840 The first code point of the lowercased version is returned
841 (but note, as explained just above, that there may be more.)
843 The input character at C<p> is assumed to be well-formed.
845 =for apidoc Am|U8|toTITLE|U8 ch
846 Converts the specified character to titlecase. If the input is anything but an
847 ASCII lowercase character, that input character itself is returned. Variant
848 C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
849 range, as the full generality of L</toTITLE_uni> is needed there. Titlecase is
850 not a concept used in locale handling, so there is no functionality for that.)
852 =for apidoc Am|UV|toTITLE_uni|UV cp|U8* s|STRLEN* lenp
853 Converts the Unicode code point C<cp> to its titlecase version, and
854 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
855 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
856 bytes since the titlecase version may be longer than the original character.
858 The first code point of the titlecased version is returned
859 (but note, as explained just above, that there may be more.)
861 =for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
862 Converts the UTF-8 encoded character at C<p> to its titlecase version, and
863 stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
864 that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
865 bytes since the titlecase version may be longer than the original character.
867 The first code point of the titlecased version is returned
868 (but note, as explained just above, that there may be more.)
870 The input character at C<p> is assumed to be well-formed.
874 XXX Still undocumented isVERTWS_uni and _utf8; it's unclear what their names
875 really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change.
877 Note that these macros are repeated in Devel::PPPort, so should also be
878 patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
882 /* Specify the widest unsigned type on the platform. Use U64TYPE because U64
883 * is known only in the perl core, and this macro can be called from outside
886 # define WIDEST_UTYPE U64TYPE
888 # define WIDEST_UTYPE U32
891 /* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
892 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
893 * bits of information are lost even on a 64-bit machine, but to get the
894 * compiler to optimize it out if possible. This is because Configure makes
895 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
896 * sizeof() guarantees that this evaluates to a constant true at compile time.
898 #define FITS_IN_8_BITS(c) ((sizeof(c) == 1) || !(((WIDEST_UTYPE)(c)) & ~0xFF))
902 /* The native libc isascii() et.al. functions return the wrong results
903 * on at least z/OS unless this is defined. */
904 # error _ALL_SOURCE should probably be defined
907 /* There is a simple definition of ASCII for ASCII platforms. But the
908 * EBCDIC one isn't so simple, so is defined using table look-up like the
909 * other macros below */
910 # define isASCII(c) ((WIDEST_UTYPE)(c) < 128)
913 /* The lower 3 bits in both the ASCII and EBCDIC representations of '0' are 0,
914 * and the 8 possible permutations of those bits exactly comprise the 8 octal
916 #define isOCTAL_A(c) cBOOL(FITS_IN_8_BITS(c) && (0xF8 & (c)) == '0')
918 #ifdef H_PERL /* If have access to perl.h, lookup in its table */
920 /* Character class numbers. For internal core Perl use only. The ones less
921 * than 32 are used in PL_charclass[] and the ones up through the one that
922 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
923 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
924 * their actual definitions are here. If that file has a name not used here,
927 * The first group of these is ordered in what I (khw) estimate to be the
928 * frequency of their use. This gives a slight edge to exiting a loop earlier
929 * (in reginclass() in regexec.c) */
930 # define _CC_WORDCHAR 0 /* \w and [:word:] */
931 # define _CC_DIGIT 1 /* \d and [:digit:] */
932 # define _CC_ALPHA 2 /* [:alpha:] */
933 # define _CC_LOWER 3 /* [:lower:] */
934 # define _CC_UPPER 4 /* [:upper:] */
935 # define _CC_PUNCT 5 /* [:punct:] */
936 # define _CC_PRINT 6 /* [:print:] */
937 # define _CC_ALPHANUMERIC 7 /* [:alnum:] */
938 # define _CC_GRAPH 8 /* [:graph:] */
939 # define _CC_CASED 9 /* [:lower:] and [:upper:] under /i */
941 #define _FIRST_NON_SWASH_CC 10
942 /* The character classes above are implemented with swashes. The second group
943 * (just below) contains the ones implemented without. These are also sorted
944 * in rough order of the frequency of their use, except that \v should be last,
945 * as it isn't a real Posix character class, and some (small) inefficiencies in
946 * regular expression handling would be introduced by putting it in the middle
947 * of those that are. Also, cntrl and ascii come after the others as it may be
948 * useful to group these which have no members that match above Latin1, (or
949 * above ASCII in the latter case) */
951 # define _CC_SPACE 10 /* \s, [:space:] */
952 # define _CC_BLANK 11 /* [:blank:] */
953 # define _CC_XDIGIT 12 /* [:xdigit:] */
954 # define _CC_CNTRL 13 /* [:cntrl:] */
955 # define _CC_ASCII 14 /* [:ascii:] */
956 # define _CC_VERTSPACE 15 /* \v */
958 # define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
960 /* The members of the third group below do not need to be coordinated with data
961 * structures in regcomp.[ch] and regexec.c. */
962 # define _CC_IDFIRST 16
963 # define _CC_CHARNAME_CONT 17
964 # define _CC_NONLATIN1_FOLD 18
965 # define _CC_NONLATIN1_SIMPLE_FOLD 19
966 # define _CC_QUOTEMETA 20
967 # define _CC_NON_FINAL_FOLD 21
968 # define _CC_IS_IN_SOME_FOLD 22
969 # define _CC_MNEMONIC_CNTRL 23
971 /* This next group is only used on EBCDIC platforms, so theoretically could be
972 * shared with something entirely different that's only on ASCII platforms */
973 # define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
974 # define _CC_UTF8_IS_START 29
975 # define _CC_UTF8_IS_DOWNGRADEABLE_START 30
976 # define _CC_UTF8_IS_CONTINUATION 31
978 * If more bits are needed, one could add a second word for non-64bit
979 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
980 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
981 * is used only for optimization (as of this writing), and differs in the
982 * Latin1 range from the ALPHA bit only in two relatively unimportant
983 * characters: the masculine and feminine ordinal indicators, so removing it
984 * would just cause /i regexes which match them to run less efficiently.
985 * Similarly the EBCDIC-only bits are used just for speed, and could be
986 * replaced by other means */
988 #if defined(PERL_CORE) || defined(PERL_EXT)
989 /* An enum version of the character class numbers, to help compilers
992 _CC_ENUM_ALPHA = _CC_ALPHA,
993 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
994 _CC_ENUM_ASCII = _CC_ASCII,
995 _CC_ENUM_BLANK = _CC_BLANK,
996 _CC_ENUM_CASED = _CC_CASED,
997 _CC_ENUM_CNTRL = _CC_CNTRL,
998 _CC_ENUM_DIGIT = _CC_DIGIT,
999 _CC_ENUM_GRAPH = _CC_GRAPH,
1000 _CC_ENUM_LOWER = _CC_LOWER,
1001 _CC_ENUM_PRINT = _CC_PRINT,
1002 _CC_ENUM_PUNCT = _CC_PUNCT,
1003 _CC_ENUM_SPACE = _CC_SPACE,
1004 _CC_ENUM_UPPER = _CC_UPPER,
1005 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
1006 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
1007 _CC_ENUM_XDIGIT = _CC_XDIGIT
1008 } _char_class_number;
1011 #define POSIX_SWASH_COUNT _FIRST_NON_SWASH_CC
1012 #define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
1014 #if defined(PERL_IN_UTF8_C) \
1015 || defined(PERL_IN_REGCOMP_C) \
1016 || defined(PERL_IN_REGEXEC_C)
1017 # if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 || _CC_LOWER != 3 \
1018 || _CC_UPPER != 4 || _CC_PUNCT != 5 || _CC_PRINT != 6 \
1019 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 || _CC_CASED != 9
1020 #error Need to adjust order of swash_property_names[]
1023 /* This is declared static in each of the few files that this is #defined for
1024 * to keep them from being publicly accessible. Hence there is a small amount
1025 * of wasted space */
1027 static const char* const swash_property_names[] = {
1043 EXTCONST U32 PL_charclass[] = {
1044 # include "l1_char_class_tab.h"
1047 # else /* ! DOINIT */
1048 EXTCONST U32 PL_charclass[];
1052 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
1053 # define _CC_mask(classnum) (1U << (classnum))
1055 /* For internal core Perl use only: the base macro for defining macros like
1057 # define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
1058 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
1060 /* The mask for the _A versions of the macros; it just adds in the bit for
1062 # define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1064 /* For internal core Perl use only: the base macro for defining macros like
1065 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1066 * the ASCII bit are present */
1067 # define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1068 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1069 == _CC_mask_A(classnum)))
1071 # define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
1072 # define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
1073 # define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1074 # define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
1075 # define isDIGIT_A(c) _generic_isCC(c, _CC_DIGIT) /* No non-ASCII digits */
1076 # define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1077 # define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1078 # define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
1079 # define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1080 # define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
1081 # define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1082 # define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
1083 # define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1085 # define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
1086 # define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1087 # define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1088 # define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1090 /* continuation character for legal NAME in \N{NAME} */
1091 # define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1093 # define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1094 # define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1095 # define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1096 # define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
1097 # define isPSXSPC_L1(c) isSPACE_L1(c)
1098 # define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1099 # define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1100 # define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1101 # define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1102 # define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
1105 # define isASCII(c) _generic_isCC(c, _CC_ASCII)
1108 /* Participates in a single-character fold with a character above 255 */
1109 # 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)))
1111 /* Like the above, but also can be part of a multi-char fold */
1112 # 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)))
1114 # define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
1115 # define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1116 _generic_isCC(c, _CC_NON_FINAL_FOLD)
1117 # define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1118 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
1119 # define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1120 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
1121 #else /* else we don't have perl.h H_PERL */
1123 /* If we don't have perl.h, we are compiling a utility program. Below we
1124 * hard-code various macro definitions that wouldn't otherwise be available
1125 * to it. Most are coded based on first principals. First some ones common
1126 * to both ASCII and EBCDIC */
1127 # define isDIGIT_A(c) ((c) <= '9' && (c) >= '0')
1128 # define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
1129 # define isSPACE_A(c) (isBLANK_A(c) \
1134 # ifdef EBCDIC /* There are gaps between 'i' and 'j'; 'r' and 's'. Same
1135 for uppercase. This is ordered to exclude most things
1137 # define isLOWER_A(c) ((c) >= 'a' && (c) <= 'z' \
1139 || ((c) >= 'j' && (c) <= 'r') \
1141 # define isUPPER_A(c) ((c) >= 'A' && (c) <= 'Z' \
1143 || ((c) >= 'J' && (c) <= 'R') \
1145 # else /* ASCII platform. */
1146 # define isLOWER_A(c) ((c) >= 'a' && (c) <= 'z')
1147 # define isUPPER_A(c) ((c) <= 'Z' && (c) >= 'A')
1150 /* Some more ASCII, non-ASCII common definitions */
1151 # define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1152 # define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
1153 # define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
1154 # define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
1155 # define isXDIGIT_A(c) (isDIGIT_A(c) \
1156 || ((c) >= 'a' && (c) <= 'f') \
1157 || ((c) <= 'F' && (c) >= 'A'))
1160 # define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1161 || (c) == '#' || (c) == '$' || (c) == '%' \
1162 || (c) == '&' || (c) == '\'' || (c) == '(' \
1163 || (c) == ')' || (c) == '*' || (c) == '+' \
1164 || (c) == ',' || (c) == '.' || (c) == '/' \
1165 || (c) == ':' || (c) == ';' || (c) == '<' \
1166 || (c) == '=' || (c) == '>' || (c) == '?' \
1167 || (c) == '@' || (c) == '[' || (c) == '\\' \
1168 || (c) == ']' || (c) == '^' || (c) == '_' \
1169 || (c) == '`' || (c) == '{' || (c) == '|' \
1170 || (c) == '}' || (c) == '~')
1171 # define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1172 # define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
1174 # ifdef QUESTION_MARK_CTRL
1175 # define _isQMC(c) ((c) == QUESTION_MARK_CTRL)
1177 # define _isQMC(c) 0
1180 /* I (khw) can't think of a way to define all the ASCII controls
1181 * without resorting to a libc (locale-sensitive) call. But we know
1182 * that all controls but the question-mark one are in the range 0-0x3f.
1183 * This makes sure that all the controls that have names are included,
1184 * and all controls that are also considered ASCII in the locale. This
1185 * may include more or fewer than what it actually should, but the
1186 * wrong ones are less-important controls, so likely won't impact
1187 * things (keep in mind that this is compiled only if perl.h isn't
1188 * available). The question mark control is included if available */
1189 # define isCNTRL_A(c) (((c) < 0x40 && isascii(c)) \
1190 || (c) == '\0' || (c) == '\a' || (c) == '\b' \
1191 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1192 || (c) == '\t' || (c) == '\v' || _isQMC(c))
1194 # define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
1195 # else /* ASCII platform; things are simpler, and isASCII has already
1197 # define isGRAPH_A(c) (((c) > ' ' && (c) < 127))
1198 # define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
1199 # define isPUNCT_A(c) (isGRAPH_A(c) && (! isALPHANUMERIC_A(c)))
1200 # define isCNTRL_A(c) (isASCII(c) && (! isPRINT_A(c)))
1203 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
1204 * during debugging, and it seems best to keep them. We may be called
1205 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1206 * do anything anyway, so make it not a problem */
1207 # if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1208 # define NATIVE_TO_LATIN1(ch) (ch)
1210 # define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1211 # define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1212 # define isBLANK_L1(c) (isBLANK_A(c) \
1213 || (FITS_IN_8_BITS(c) \
1214 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1215 # define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1216 # define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1217 # define isLOWER_L1(c) (isLOWER_A(c) \
1218 || (FITS_IN_8_BITS(c) \
1219 && ((NATIVE_TO_LATIN1((U8) c) >= 0xDF \
1220 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1221 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1222 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1223 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1224 # define isPRINT_L1(c) (isPRINT_A(c) \
1225 || (FITS_IN_8_BITS(c) \
1226 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
1227 # define isPUNCT_L1(c) (isPUNCT_A(c) \
1228 || (FITS_IN_8_BITS(c) \
1229 && (NATIVE_TO_LATIN1((U8) c) == 0xA1 \
1230 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1231 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1232 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1233 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1234 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1235 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1236 # define isSPACE_L1(c) (isSPACE_A(c) \
1237 || (FITS_IN_8_BITS(c) \
1238 && (NATIVE_TO_LATIN1((U8) c) == 0x85 \
1239 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1240 # define isUPPER_L1(c) (isUPPER_A(c) \
1241 || (FITS_IN_8_BITS(c) \
1242 && (NATIVE_TO_LATIN1((U8) c) >= 0xC0 \
1243 && NATIVE_TO_LATIN1((U8) c) <= 0xDE \
1244 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1245 # define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1246 # define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1247 # define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1252 /* The following are not fully accurate in the above-ASCII range. I (khw)
1253 * don't think it's necessary to be so for the purposes where this gets
1255 # define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1256 # define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1258 /* And these aren't accurate at all. They are useful only for above
1259 * Latin1, which utilities and bootstrapping don't deal with */
1260 # define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
1261 # define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1262 # define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1264 /* Many of the macros later in this file are defined in terms of these. By
1265 * implementing them with a function, which converts the class number into
1266 * a call to the desired macro, all of the later ones work. However, that
1267 * function won't be actually defined when building a utility program (no
1268 * perl.h), and so a compiler error will be generated if one is attempted
1269 * to be used. And the above-Latin1 code points require Unicode tables to
1270 * be present, something unlikely to be the case when bootstrapping */
1271 # define _generic_isCC(c, classnum) \
1272 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1273 # define _generic_isCC_A(c, classnum) \
1274 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
1275 #endif /* End of no perl.h H_PERL */
1277 #define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1278 #define isALPHA(c) isALPHA_A(c)
1279 #define isASCII_A(c) isASCII(c)
1280 #define isASCII_L1(c) isASCII(c)
1281 #define isBLANK(c) isBLANK_A(c)
1282 #define isCNTRL(c) isCNTRL_A(c)
1283 #define isDIGIT(c) isDIGIT_A(c)
1284 #define isGRAPH(c) isGRAPH_A(c)
1285 #define isIDFIRST(c) isIDFIRST_A(c)
1286 #define isLOWER(c) isLOWER_A(c)
1287 #define isPRINT(c) isPRINT_A(c)
1288 #define isPSXSPC_A(c) isSPACE_A(c)
1289 #define isPSXSPC(c) isPSXSPC_A(c)
1290 #define isPSXSPC_L1(c) isSPACE_L1(c)
1291 #define isPUNCT(c) isPUNCT_A(c)
1292 #define isSPACE(c) isSPACE_A(c)
1293 #define isUPPER(c) isUPPER_A(c)
1294 #define isWORDCHAR(c) isWORDCHAR_A(c)
1295 #define isXDIGIT(c) isXDIGIT_A(c)
1297 /* ASCII casing. These could also be written as
1298 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1299 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1300 which uses table lookup and mask instead of subtraction. (This would
1301 work because the _MOD does not apply in the ASCII range) */
1302 #define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1303 #define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
1305 /* In the ASCII range, these are equivalent to what they're here defined to be.
1306 * But by creating these definitions, other code doesn't have to be aware of
1308 #define toFOLD(c) toLOWER(c)
1309 #define toTITLE(c) toUPPER(c)
1311 #define toLOWER_A(c) toLOWER(c)
1312 #define toUPPER_A(c) toUPPER(c)
1313 #define toFOLD_A(c) toFOLD(c)
1314 #define toTITLE_A(c) toTITLE(c)
1316 /* Use table lookup for speed; returns the input itself if is out-of-range */
1317 #define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
1319 : PL_latin1_lc[ (U8) (c) ])
1320 #define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1322 /* Modified uc. Is correct uc except for three non-ascii chars which are
1323 * all mapped to one of them, and these need special handling; returns the
1324 * input itself if is out-of-range */
1325 #define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
1327 : PL_mod_latin1_uc[ (U8) (c) ])
1328 #define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
1330 /* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1332 /* For internal core Perl use only: the base macro for defining macros like
1333 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
1334 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1335 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1336 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1337 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1338 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1339 * code points under locale rules, use the macros based on _generic_LC_uvchr
1340 * instead of this. */
1341 #define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1342 (! FITS_IN_8_BITS(c) \
1344 : IN_UTF8_CTYPE_LOCALE \
1345 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
1348 /* For internal core Perl use only: a helper macro for defining macros like
1349 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1350 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1351 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1352 * _generic_LC_base for more info */
1353 #define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1354 _generic_LC_base(c,utf8_locale_classnum, \
1355 non_utf8_func( (_LC_CAST) (c)))
1357 /* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1358 * 'c' is the platform's native underscore character */
1359 #define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1360 _generic_LC_base(c, utf8_locale_classnum, \
1361 (non_utf8_func( (_LC_CAST) (c)) \
1362 || (char)(c) == '_'))
1364 /* These next three are also for internal core Perl use only: case-change
1366 #define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1368 : (IN_UTF8_CTYPE_LOCALE) \
1369 ? PL_latin1_lc[ (U8) (c) ] \
1370 : (cast)function((cast)(c)))
1372 /* Note that the result can be larger than a byte in a UTF-8 locale. It
1373 * returns a single value, so can't adequately return the upper case of LATIN
1374 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1375 * values "SS"); instead it asserts against that under DEBUGGING, and
1376 * otherwise returns its input */
1377 #define _generic_toUPPER_LC(c, function, cast) \
1378 (! FITS_IN_8_BITS(c) \
1380 : ((! IN_UTF8_CTYPE_LOCALE) \
1381 ? (cast)function((cast)(c)) \
1382 : ((((U8)(c)) == MICRO_SIGN) \
1383 ? GREEK_CAPITAL_LETTER_MU \
1384 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1385 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1386 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1387 ? (__ASSERT_(0) (c)) \
1388 : PL_mod_latin1_uc[ (U8) (c) ])))))
1390 /* Note that the result can be larger than a byte in a UTF-8 locale. It
1391 * returns a single value, so can't adequately return the fold case of LATIN
1392 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1393 * values "ss"); instead it asserts against that under DEBUGGING, and
1394 * otherwise returns its input */
1395 #define _generic_toFOLD_LC(c, function, cast) \
1396 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
1397 ? GREEK_SMALL_LETTER_MU \
1398 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1399 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1400 _generic_toLOWER_LC(c, function, cast)))
1402 /* Use the libc versions for these if available. */
1403 #if defined(HAS_ISASCII)
1404 # define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1406 # define isASCII_LC(c) isASCII(c)
1409 #if defined(HAS_ISBLANK)
1410 # define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
1411 #else /* Unlike isASCII, varies if in a UTF-8 locale */
1412 # define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
1418 /* The Windows functions don't bother to follow the POSIX standard, which
1419 * for example says that something can't both be a printable and a control.
1420 * But Windows treats the \t control as a printable, and does such things
1421 * as making superscripts into both digits and punctuation. This tames
1422 * these flaws by assuming that the definitions of both controls and space
1423 * are correct, and then making sure that other definitions don't have
1424 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1425 * Not all possible weirdnesses are checked for, just the ones that were
1426 * detected on actual Microsoft code pages */
1428 # define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1429 # define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1431 # define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1432 && isALPHANUMERIC_LC(c))
1433 # define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1435 # define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1436 isALPHANUMERIC_LC(c))
1437 # define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1438 # define isIDFIRST_LC(c) (((c) == '_') \
1439 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1440 # define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1441 # define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1442 # define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1443 # define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
1444 # define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
1445 # define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1446 && isALPHANUMERIC_LC(c))
1448 # define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1449 # define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1450 # define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1452 #elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
1453 /* For most other platforms */
1455 # define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1456 # define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1457 # define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1458 # define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1459 # define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1460 # define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1461 # define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1462 # define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1463 # define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1464 # define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1465 # define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1466 # define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1467 # define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1470 # define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1471 # define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1472 # define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1474 #else /* The final fallback position */
1476 # define isALPHA_LC(c) (isascii(c) && isalpha(c))
1477 # define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1478 # define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1479 # define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1480 # define isGRAPH_LC(c) (isascii(c) && isgraph(c))
1481 # define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
1482 # define isLOWER_LC(c) (isascii(c) && islower(c))
1483 # define isPRINT_LC(c) (isascii(c) && isprint(c))
1484 # define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1485 # define isSPACE_LC(c) (isascii(c) && isspace(c))
1486 # define isUPPER_LC(c) (isascii(c) && isupper(c))
1487 # define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
1488 # define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
1490 # define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1491 # define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1492 # define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
1496 #define isIDCONT(c) isWORDCHAR(c)
1497 #define isIDCONT_A(c) isWORDCHAR_A(c)
1498 #define isIDCONT_L1(c) isWORDCHAR_L1(c)
1499 #define isIDCONT_LC(c) isWORDCHAR_LC(c)
1500 #define isPSXSPC_LC(c) isSPACE_LC(c)
1502 /* For internal core Perl use only: the base macros for defining macros like
1503 * isALPHA_uni. 'c' is the code point to check. 'classnum' is the POSIX class
1504 * number defined earlier in this file. _generic_uni() is used for POSIX
1505 * classes where there is a macro or function 'above_latin1' that takes the
1506 * single argument 'c' and returns the desired value. These exist for those
1507 * classes which have simple definitions, avoiding the overhead of a hash
1508 * lookup or inversion list binary search. _generic_swash_uni() can be used
1509 * for classes where that overhead is faster than a direct lookup.
1510 * _generic_uni() won't compile if 'c' isn't unsigned, as it won't match the
1511 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1512 * have duplicate checks here, so could create versions of the macros that
1513 * don't, but experiments show that gcc optimizes them out anyway. */
1515 /* Note that all ignore 'use bytes' */
1516 #define _generic_uni(classnum, above_latin1, c) ((c) < 256 \
1517 ? _generic_isCC(c, classnum) \
1519 #define _generic_swash_uni(classnum, c) ((c) < 256 \
1520 ? _generic_isCC(c, classnum) \
1521 : _is_uni_FOO(classnum, c))
1522 #define isALPHA_uni(c) _generic_swash_uni(_CC_ALPHA, c)
1523 #define isALPHANUMERIC_uni(c) _generic_swash_uni(_CC_ALPHANUMERIC, c)
1524 #define isASCII_uni(c) isASCII(c)
1525 #define isBLANK_uni(c) _generic_uni(_CC_BLANK, is_HORIZWS_cp_high, c)
1526 #define isCNTRL_uni(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1527 #define isDIGIT_uni(c) _generic_swash_uni(_CC_DIGIT, c)
1528 #define isGRAPH_uni(c) _generic_swash_uni(_CC_GRAPH, c)
1529 #define isIDCONT_uni(c) _generic_uni(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1530 #define isIDFIRST_uni(c) _generic_uni(_CC_IDFIRST, _is_uni_perl_idstart, c)
1531 #define isLOWER_uni(c) _generic_swash_uni(_CC_LOWER, c)
1532 #define isPRINT_uni(c) _generic_swash_uni(_CC_PRINT, c)
1534 #define isPUNCT_uni(c) _generic_swash_uni(_CC_PUNCT, c)
1535 #define isSPACE_uni(c) _generic_uni(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1536 #define isPSXSPC_uni(c) isSPACE_uni(c)
1538 #define isUPPER_uni(c) _generic_swash_uni(_CC_UPPER, c)
1539 #define isVERTWS_uni(c) _generic_uni(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1540 #define isWORDCHAR_uni(c) _generic_swash_uni(_CC_WORDCHAR, c)
1541 #define isXDIGIT_uni(c) _generic_uni(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1543 #define toFOLD_uni(c,s,l) to_uni_fold(c,s,l)
1544 #define toLOWER_uni(c,s,l) to_uni_lower(c,s,l)
1545 #define toTITLE_uni(c,s,l) to_uni_title(c,s,l)
1546 #define toUPPER_uni(c,s,l) to_uni_upper(c,s,l)
1548 /* For internal core Perl use only: the base macros for defining macros like
1549 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
1550 * point, not just 0-255. Like _generic_uni, there are two versions, one for
1551 * simple class definitions; the other for more complex. These are like
1552 * _generic_uni, so see it for more info. */
1553 #define _generic_LC_uvchr(latin1, above_latin1, c) \
1554 (c < 256 ? latin1(c) : above_latin1(c))
1555 #define _generic_LC_swash_uvchr(latin1, classnum, c) \
1556 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1558 #define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1559 #define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1560 _CC_ALPHANUMERIC, c)
1561 #define isASCII_LC_uvchr(c) isASCII_LC(c)
1562 #define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1563 is_HORIZWS_cp_high, c)
1564 #define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
1565 #define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1566 #define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
1567 #define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
1568 _is_uni_perl_idcont, c)
1569 #define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
1570 _is_uni_perl_idstart, c)
1571 #define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1572 #define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
1573 #define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
1574 #define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
1575 #define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
1576 is_XPERLSPACE_cp_high, c)
1577 #define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
1578 #define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
1580 #define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1581 is_XDIGIT_cp_high, c)
1583 #define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
1585 /* For internal core Perl use only: the base macros for defining macros like
1586 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1587 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1588 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1589 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1590 * same representation whether utf8 or not. Note that it assumes that the utf8
1591 * has been validated, and ignores 'use bytes' */
1592 #define _generic_utf8(classnum, p, utf8) (UTF8_IS_INVARIANT(*(p)) \
1593 ? _generic_isCC(*(p), classnum) \
1594 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1596 EIGHT_BIT_UTF8_TO_NATIVE(*(p), \
1600 /* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1601 * 'above_latin1' can be a macro */
1602 #define _generic_func_utf8(classnum, above_latin1, p) \
1603 _generic_utf8(classnum, p, above_latin1(p))
1604 /* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
1605 * 'above_latin1' parameter */
1606 #define _generic_swash_utf8(classnum, p) \
1607 _generic_utf8(classnum, p, _is_utf8_FOO(classnum, p))
1609 /* Like the above, but should be used only when it is known that there are no
1610 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1611 * class is TRUE for. Hence it can skip the tests for this range.
1612 * 'above_latin1' should include its arguments */
1613 #define _generic_utf8_no_upper_latin1(classnum, p, above_latin1) \
1614 (UTF8_IS_INVARIANT(*(p)) \
1615 ? _generic_isCC(*(p), classnum) \
1616 : (UTF8_IS_ABOVE_LATIN1(*(p))) \
1620 /* NOTE that some of these macros have very similar ones in regcharclass.h.
1621 * For example, there is (at the time of this writing) an 'is_SPACE_utf8()'
1622 * there, differing in name only by an underscore from the one here
1623 * 'isSPACE_utf8(). The difference is that the ones here are probably more
1624 * efficient and smaller, using an O(1) array lookup for Latin1-range code
1625 * points; the regcharclass.h ones are implemented as a series of
1626 * "if-else-if-else ..." */
1628 #define isALPHA_utf8(p) _generic_swash_utf8(_CC_ALPHA, p)
1629 #define isALPHANUMERIC_utf8(p) _generic_swash_utf8(_CC_ALPHANUMERIC, p)
1630 #define isASCII_utf8(p) isASCII(*p) /* Because ASCII is invariant under
1631 utf8, the non-utf8 macro works
1633 #define isBLANK_utf8(p) _generic_func_utf8(_CC_BLANK, is_HORIZWS_high, p)
1636 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1637 * more efficient macro instead of the more general one */
1638 # define isCNTRL_utf8(p) isCNTRL_L1(*(p))
1640 # define isCNTRL_utf8(p) _generic_utf8(_CC_CNTRL, p, 0)
1643 #define isDIGIT_utf8(p) _generic_utf8_no_upper_latin1(_CC_DIGIT, p, \
1644 _is_utf8_FOO(_CC_DIGIT, p))
1645 #define isGRAPH_utf8(p) _generic_swash_utf8(_CC_GRAPH, p)
1646 #define isIDCONT_utf8(p) _generic_func_utf8(_CC_WORDCHAR, \
1647 _is_utf8_perl_idcont, p)
1649 /* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1650 * IDFIRST is an alnum. See
1651 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
1652 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1653 * This used to be not the XID version, but we decided to go with the more
1654 * modern Unicode definition */
1655 #define isIDFIRST_utf8(p) _generic_func_utf8(_CC_IDFIRST, \
1656 _is_utf8_perl_idstart, p)
1658 #define isLOWER_utf8(p) _generic_swash_utf8(_CC_LOWER, p)
1659 #define isPRINT_utf8(p) _generic_swash_utf8(_CC_PRINT, p)
1660 #define isPSXSPC_utf8(p) isSPACE_utf8(p)
1661 #define isPUNCT_utf8(p) _generic_swash_utf8(_CC_PUNCT, p)
1662 #define isSPACE_utf8(p) _generic_func_utf8(_CC_SPACE, is_XPERLSPACE_high, p)
1663 #define isUPPER_utf8(p) _generic_swash_utf8(_CC_UPPER, p)
1664 #define isVERTWS_utf8(p) _generic_func_utf8(_CC_VERTSPACE, is_VERTWS_high, p)
1665 #define isWORDCHAR_utf8(p) _generic_swash_utf8(_CC_WORDCHAR, p)
1666 #define isXDIGIT_utf8(p) _generic_utf8_no_upper_latin1(_CC_XDIGIT, p, \
1669 #define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
1670 #define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
1671 #define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
1672 #define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
1674 /* For internal core Perl use only: the base macros for defining macros like
1675 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
1676 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
1677 * 'macro' parameter */
1678 #define _generic_LC_utf8(macro, p, utf8) \
1679 (UTF8_IS_INVARIANT(*(p)) \
1681 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1682 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1)))\
1685 #define _generic_LC_swash_utf8(macro, classnum, p) \
1686 _generic_LC_utf8(macro, p, _is_utf8_FOO(classnum, p))
1687 #define _generic_LC_func_utf8(macro, above_latin1, p) \
1688 _generic_LC_utf8(macro, p, above_latin1(p))
1690 #define isALPHANUMERIC_LC_utf8(p) _generic_LC_swash_utf8(isALPHANUMERIC_LC, \
1691 _CC_ALPHANUMERIC, p)
1692 #define isALPHA_LC_utf8(p) _generic_LC_swash_utf8(isALPHA_LC, _CC_ALPHA, p)
1693 #define isASCII_LC_utf8(p) isASCII_LC(*p)
1694 #define isBLANK_LC_utf8(p) _generic_LC_func_utf8(isBLANK_LC, \
1696 #define isCNTRL_LC_utf8(p) _generic_LC_utf8(isCNTRL_LC, p, 0)
1697 #define isDIGIT_LC_utf8(p) _generic_LC_swash_utf8(isDIGIT_LC, _CC_DIGIT, p)
1698 #define isGRAPH_LC_utf8(p) _generic_LC_swash_utf8(isGRAPH_LC, _CC_GRAPH, p)
1699 #define isIDCONT_LC_utf8(p) _generic_LC_func_utf8(isIDCONT_LC, \
1700 _is_utf8_perl_idcont, p)
1701 #define isIDFIRST_LC_utf8(p) _generic_LC_func_utf8(isIDFIRST_LC, \
1702 _is_utf8_perl_idstart, p)
1703 #define isLOWER_LC_utf8(p) _generic_LC_swash_utf8(isLOWER_LC, _CC_LOWER, p)
1704 #define isPRINT_LC_utf8(p) _generic_LC_swash_utf8(isPRINT_LC, _CC_PRINT, p)
1705 #define isPSXSPC_LC_utf8(p) isSPACE_LC_utf8(p)
1706 #define isPUNCT_LC_utf8(p) _generic_LC_swash_utf8(isPUNCT_LC, _CC_PUNCT, p)
1707 #define isSPACE_LC_utf8(p) _generic_LC_func_utf8(isSPACE_LC, \
1708 is_XPERLSPACE_high, p)
1709 #define isUPPER_LC_utf8(p) _generic_LC_swash_utf8(isUPPER_LC, _CC_UPPER, p)
1710 #define isWORDCHAR_LC_utf8(p) _generic_LC_swash_utf8(isWORDCHAR_LC, \
1712 #define isXDIGIT_LC_utf8(p) _generic_LC_func_utf8(isXDIGIT_LC, \
1715 /* Macros for backwards compatibility and for completeness when the ASCII and
1716 * Latin1 values are identical */
1717 #define isALPHAU(c) isALPHA_L1(c)
1718 #define isDIGIT_L1(c) isDIGIT_A(c)
1719 #define isOCTAL(c) isOCTAL_A(c)
1720 #define isOCTAL_L1(c) isOCTAL_A(c)
1721 #define isXDIGIT_L1(c) isXDIGIT_A(c)
1722 #define isALNUM(c) isWORDCHAR(c)
1723 #define isALNUMU(c) isWORDCHAR_L1(c)
1724 #define isALNUM_LC(c) isWORDCHAR_LC(c)
1725 #define isALNUM_uni(c) isWORDCHAR_uni(c)
1726 #define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
1727 #define isALNUM_utf8(p) isWORDCHAR_utf8(p)
1728 #define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
1729 #define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
1730 #define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
1731 #define isALNUMC(c) isALPHANUMERIC(c)
1732 #define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
1733 #define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
1734 #define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
1735 #define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
1736 #define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
1738 /* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
1739 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
1740 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
1741 * '?' is special-cased on EBCDIC to APC, which is the control there that is
1742 * the outlier from the block that contains the other controls, just like
1743 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
1744 * block. If it weren't special cased, it would yield a non-control.
1745 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
1748 # define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
1750 # define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
1752 ? (UNLIKELY((c) == '?') \
1753 ? QUESTION_MARK_CTRL \
1754 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
1755 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
1757 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
1760 /* Line numbers are unsigned, 32 bits. */
1762 #define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
1764 /* Helpful alias for version prescan */
1765 #define is_LAX_VERSION(a,b) \
1766 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
1768 #define is_STRICT_VERSION(a,b) \
1769 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
1771 #define BADVERSION(a,b,c) \
1777 /* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
1778 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
1779 * which is advanced. The input is validated only by an assert() in DEBUGGING
1780 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
1781 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
1782 #define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
1785 #define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
1789 /* Converts a character known to represent an octal digit (0-7) to its numeric
1790 * value. The input is validated only by an assert() in DEBUGGING builds. In
1791 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
1792 #define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
1794 /* Efficiently returns a boolean as to if two native characters are equivalent
1795 * case-insenstively. At least one of the characters must be one of [A-Za-z];
1796 * the ALPHA in the name is to remind you of that. This is asserted() in
1797 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
1798 * works (on valid input) for both non- and UTF-8-encoded bytes.
1800 * When one of the inputs is a compile-time constant and gets folded by the
1801 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
1802 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
1803 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
1804 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
1805 * exclusive 'or' to find that bit and then inverts it to form a mask, with
1806 * just a single 0, in the bit position where the upper- and lowercase differ.
1808 #define isALPHA_FOLD_EQ(c1, c2) \
1809 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
1810 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
1811 #define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
1814 =head1 Memory Management
1816 =for apidoc Am|void|Newx|void* ptr|int nitems|type
1817 The XSUB-writer's interface to the C C<malloc> function.
1819 Memory obtained by this should B<ONLY> be freed with L<"Safefree">.
1821 In 5.9.3, Newx() and friends replace the older New() API, and drops
1822 the first parameter, I<x>, a debug aid which allowed callers to identify
1823 themselves. This aid has been superseded by a new build option,
1824 PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
1825 there for use in XS modules supporting older perls.
1827 =for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
1828 The XSUB-writer's interface to the C C<malloc> function, with
1829 cast. See also C<L</Newx>>.
1831 Memory obtained by this should B<ONLY> be freed with L<"Safefree">.
1833 =for apidoc Am|void|Newxz|void* ptr|int nitems|type
1834 The XSUB-writer's interface to the C C<malloc> function. The allocated
1835 memory is zeroed with C<memzero>. See also C<L</Newx>>.
1837 Memory obtained by this should B<ONLY> be freed with L<"Safefree">.
1839 =for apidoc Am|void|Renew|void* ptr|int nitems|type
1840 The XSUB-writer's interface to the C C<realloc> function.
1842 Memory obtained by this should B<ONLY> be freed with L<"Safefree">.
1844 =for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
1845 The XSUB-writer's interface to the C C<realloc> function, with
1848 Memory obtained by this should B<ONLY> be freed with L<"Safefree">.
1850 =for apidoc Am|void|Safefree|void* ptr
1851 The XSUB-writer's interface to the C C<free> function.
1853 This should B<ONLY> be used on memory obtained using L<"Newx"> and friends.
1855 =for apidoc Am|void|Move|void* src|void* dest|int nitems|type
1856 The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
1857 source, C<dest> is the destination, C<nitems> is the number of items, and
1858 C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
1860 =for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
1861 Like C<Move> but returns C<dest>. Useful
1862 for encouraging compilers to tail-call
1865 =for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
1866 The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
1867 source, C<dest> is the destination, C<nitems> is the number of items, and
1868 C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
1870 =for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
1872 Like C<Copy> but returns C<dest>. Useful
1873 for encouraging compilers to tail-call
1876 =for apidoc Am|void|Zero|void* dest|int nitems|type
1878 The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
1879 destination, C<nitems> is the number of items, and C<type> is the type.
1881 =for apidoc Am|void *|ZeroD|void* dest|int nitems|type
1883 Like C<Zero> but returns dest. Useful
1884 for encouraging compilers to tail-call
1887 =for apidoc Am|void|StructCopy|type *src|type *dest|type
1888 This is an architecture-independent macro to copy one structure to another.
1890 =for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
1892 Fill up memory with a byte pattern (a byte repeated over and over
1893 again) that hopefully catches attempts to access uninitialized memory.
1895 =for apidoc Am|void|PoisonNew|void* dest|int nitems|type
1897 PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
1899 =for apidoc Am|void|PoisonFree|void* dest|int nitems|type
1901 PoisonWith(0xEF) for catching access to freed memory.
1903 =for apidoc Am|void|Poison|void* dest|int nitems|type
1905 PoisonWith(0xEF) for catching access to freed memory.
1909 /* Maintained for backwards-compatibility only. Use newSV() instead. */
1911 #define NEWSV(x,len) newSV(len)
1914 #define MEM_SIZE_MAX ((MEM_SIZE)~0)
1917 #ifdef PERL_MALLOC_WRAP
1919 /* This expression will be constant-folded at compile time. It checks
1920 * whether or not the type of the count n is so small (e.g. U8 or U16, or
1921 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
1922 * As well as avoiding the need for a run-time check in some cases, it's
1923 * designed to avoid compiler warnings like:
1924 * comparison is always false due to limited range of data type
1925 * It's mathematically equivalent to
1926 * max(n) * sizeof(t) > MEM_SIZE_MAX
1929 # define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
1930 ( sizeof(MEM_SIZE) < sizeof(n) \
1931 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
1933 /* This is written in a slightly odd way to avoid various spurious
1934 * compiler warnings. We *want* to write the expression as
1935 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
1936 * (for some compile-time constant C), but even when the LHS
1937 * constant-folds to false at compile-time, g++ insists on emitting
1938 * warnings about the RHS (e.g. "comparison is always false"), so instead
1941 * (cond ? n : X) > C
1943 * where X is a constant with X > C always false. Choosing a value for X
1944 * is tricky. If 0, some compilers will complain about 0 > C always being
1945 * false; if 1, Coverity complains when n happens to be the constant value
1946 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
1947 * for X and hope that nothing else whines.
1950 # define _MEM_WRAP_WILL_WRAP(n,t) \
1951 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
1952 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
1954 # define MEM_WRAP_CHECK(n,t) \
1955 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
1956 && (croak_memory_wrap(),0))
1958 # define MEM_WRAP_CHECK_1(n,t,a) \
1959 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
1960 && (Perl_croak_nocontext("%s",(a)),0))
1962 #define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
1964 #define PERL_STRLEN_ROUNDUP(n) ((void)(((n) > MEM_SIZE_MAX - 2 * PERL_STRLEN_ROUNDUP_QUANTUM) ? (croak_memory_wrap(),0):0),((n-1+PERL_STRLEN_ROUNDUP_QUANTUM)&~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM-1)))
1967 #define MEM_WRAP_CHECK(n,t)
1968 #define MEM_WRAP_CHECK_1(n,t,a)
1969 #define MEM_WRAP_CHECK_2(n,t,a,b)
1970 #define MEM_WRAP_CHECK_(n,t)
1972 #define PERL_STRLEN_ROUNDUP(n) (((n-1+PERL_STRLEN_ROUNDUP_QUANTUM)&~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM-1)))
1978 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
1979 * go through functions, which are handy for debugging breakpoints, but
1980 * which more importantly get the immediate calling environment (file and
1981 * line number, and C function name if available) passed in. This info can
1982 * then be used for logging the calls, for which one gets a sample
1983 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
1986 * - not all memory allocs get logged, only those
1987 * that go through Newx() and derivatives (while all
1988 * Safefrees do get logged)
1989 * - __FILE__ and __LINE__ do not work everywhere
1990 * - __func__ or __FUNCTION__ even less so
1991 * - I think more goes on after the perlio frees but
1992 * the thing is that STDERR gets closed (as do all
1993 * the file descriptors)
1994 * - no deeper calling stack than the caller of the Newx()
1995 * or the kind, but do I look like a C reflection/introspection
1997 * - the function prototypes for the logging functions
1998 * probably should maybe be somewhere else than handy.h
1999 * - one could consider inlining (macrofying) the logging
2000 * for speed, but I am too lazy
2001 * - one could imagine recording the allocations in a hash,
2002 * (keyed by the allocation address?), and maintain that
2003 * through reallocs and frees, but how to do that without
2004 * any News() happening...?
2005 * - lots of -Ddefines to get useful/controllable output
2006 * - lots of ENV reads
2009 PERL_CALLCONV Malloc_t Perl_mem_log_alloc(const UV n, const UV typesize, const char *type_name, Malloc_t newalloc, const char *filename, const int linenumber, const char *funcname);
2011 PERL_CALLCONV Malloc_t Perl_mem_log_realloc(const UV n, const UV typesize, const char *type_name, Malloc_t oldalloc, Malloc_t newalloc, const char *filename, const int linenumber, const char *funcname);
2013 PERL_CALLCONV Malloc_t Perl_mem_log_free(Malloc_t oldalloc, const char *filename, const int linenumber, const char *funcname);
2016 # ifndef PERL_MEM_LOG_NOIMPL
2025 # if defined(PERL_IN_SV_C) /* those are only used in sv.c */
2026 void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2027 void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2034 #define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2035 #define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
2036 #define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
2039 #ifndef MEM_LOG_ALLOC
2040 #define MEM_LOG_ALLOC(n,t,a) (a)
2042 #ifndef MEM_LOG_REALLOC
2043 #define MEM_LOG_REALLOC(n,t,v,a) (a)
2045 #ifndef MEM_LOG_FREE
2046 #define MEM_LOG_FREE(a) (a)
2049 #define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2050 #define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2051 #define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
2054 /* pre 5.9.x compatibility */
2055 #define New(x,v,n,t) Newx(v,n,t)
2056 #define Newc(x,v,n,t,c) Newxc(v,n,t,c)
2057 #define Newz(x,v,n,t) Newxz(v,n,t)
2060 #define Renew(v,n,t) \
2061 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
2062 #define Renewc(v,n,t,c) \
2063 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
2066 #define Safefree(d) \
2067 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
2069 #define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
2072 #define Move(s,d,n,t) (MEM_WRAP_CHECK_(n,t) (void)memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2073 #define Copy(s,d,n,t) (MEM_WRAP_CHECK_(n,t) (void)memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2074 #define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) (void)memzero((char*)(d), (n) * sizeof(t)))
2076 #define MoveD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) memmove((char*)(d),(const char*)(s), (n) * sizeof(t)))
2077 #define CopyD(s,d,n,t) (MEM_WRAP_CHECK_(n,t) memcpy((char*)(d),(const char*)(s), (n) * sizeof(t)))
2079 #define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) memzero((char*)(d), (n) * sizeof(t)))
2081 /* Using bzero(), which returns void. */
2082 #define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) memzero((char*)(d), (n) * sizeof(t)),d)
2085 #define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2086 #define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2087 #define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2088 #define Poison(d,n,t) PoisonFree(d,n,t)
2091 # define PERL_POISON_EXPR(x) x
2093 # define PERL_POISON_EXPR(x)
2096 #ifdef USE_STRUCT_COPY
2097 #define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2099 #define StructCopy(s,d,t) Copy(s,d,1,t)
2102 /* C_ARRAY_LENGTH is the number of elements in the C array (so you
2103 * want your zero-based indices to be less than but not equal to).
2105 * C_ARRAY_END is one past the last: half-open/half-closed range,
2106 * not last-inclusive range. */
2107 #define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
2108 #define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
2112 # define Perl_va_copy(s, d) va_copy(d, s)
2114 # if defined(__va_copy)
2115 # define Perl_va_copy(s, d) __va_copy(d, s)
2117 # define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2122 /* convenience debug macros */
2124 #define pTHX_FORMAT "Perl interpreter: 0x%p"
2125 #define pTHX__FORMAT ", Perl interpreter: 0x%p"
2126 #define pTHX_VALUE_ (void *)my_perl,
2127 #define pTHX_VALUE (void *)my_perl
2128 #define pTHX__VALUE_ ,(void *)my_perl,
2129 #define pTHX__VALUE ,(void *)my_perl
2132 #define pTHX__FORMAT
2135 #define pTHX__VALUE_
2137 #endif /* USE_ITHREADS */
2139 /* Perl_deprecate was not part of the public API, and did not have a deprecate()
2140 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2141 CPAN::Unpack show any users outside the core. */
2143 # define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2144 "Use of " s " is deprecated")
2147 /* Internal macros to deal with gids and uids */
2150 # if Uid_t_size > IVSIZE
2151 # define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2152 # define SvUID(sv) SvNV(sv)
2154 # if Uid_t_sign <= 0
2155 # define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2156 # define SvUID(sv) SvIV(sv)
2158 # define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2159 # define SvUID(sv) SvUV(sv)
2161 # endif /* Uid_t_size */
2163 # if Gid_t_size > IVSIZE
2164 # define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2165 # define SvGID(sv) SvNV(sv)
2167 # if Gid_t_sign <= 0
2168 # define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2169 # define SvGID(sv) SvIV(sv)
2171 # define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2172 # define SvGID(sv) SvUV(sv)
2174 # endif /* Gid_t_size */
2178 #endif /* HANDY_H */
2181 * ex: set ts=8 sts=4 sw=4 et: