This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perlapi: Document XST_mUV
[perl5.git] / handy.h
CommitLineData
a0d0e21e 1/* handy.h
a687059c 2 *
1129b882 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1999, 2000,
da5d8dbb 4 * 2001, 2002, 2004, 2005, 2006, 2007, 2008, 2012 by Larry Wall and others
a687059c 5 *
6e21c824
LW
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.
8d063cd8 8 *
8d063cd8
LW
9 */
10
4650c663
KW
11/* IMPORTANT NOTE: Everything whose name begins with an underscore is for
12 * internal core Perl use only. */
13
6a5bc5ac
KW
14#ifndef PERL_HANDY_H_ /* Guard against nested #inclusion */
15#define PERL_HANDY_H_
9d745869 16
24792b8d
NC
17#ifndef PERL_CORE
18# define Null(type) ((type)NULL)
954c1994
GS
19
20/*
ccfc67b7 21=head1 Handy Values
954c1994 22
78342678 23=for apidoc AmnU||Nullch
72d33970
FC
24Null character pointer. (No longer available when C<PERL_CORE> is
25defined.)
2307c6d0 26
78342678 27=for apidoc AmnU||Nullsv
72d33970 28Null SV pointer. (No longer available when C<PERL_CORE> is defined.)
954c1994
GS
29
30=cut
31*/
32
24792b8d
NC
33# define Nullch Null(char*)
34# define Nullfp Null(PerlIO*)
35# define Nullsv Null(SV*)
36#endif
8d063cd8 37
641d3f0b
PP
38#ifdef TRUE
39#undef TRUE
40#endif
41#ifdef FALSE
42#undef FALSE
43#endif
44#define TRUE (1)
45#define FALSE (0)
46
cf3f0ffb
DM
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;
49 * e.g.:
50 *
51 * const SV *sv = ...;
52 * AV *av1 = (AV*)sv; <== BAD: the const has been silently cast away
53 * AV *av2 = MUTABLE_AV(sv); <== GOOD: it may warn
54 */
55
b1bc3f34 56#if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
f065df04 57# define MUTABLE_PTR(p) ({ void *_p = (p); _p; })
b1bc3f34
NC
58#else
59# define MUTABLE_PTR(p) ((void *) (p))
60#endif
61
a062e10d 62#define MUTABLE_AV(p) ((AV *)MUTABLE_PTR(p))
ea726b52 63#define MUTABLE_CV(p) ((CV *)MUTABLE_PTR(p))
159b6efe 64#define MUTABLE_GV(p) ((GV *)MUTABLE_PTR(p))
dbebbdb4 65#define MUTABLE_HV(p) ((HV *)MUTABLE_PTR(p))
a45c7426 66#define MUTABLE_IO(p) ((IO *)MUTABLE_PTR(p))
b1bc3f34 67#define MUTABLE_SV(p) ((SV *)MUTABLE_PTR(p))
27d4fb96 68
f789f6a4 69#if defined(I_STDBOOL) && !defined(PERL_BOOL_AS_CHAR)
bd31be4b
NC
70# include <stdbool.h>
71# ifndef HAS_BOOL
72# define HAS_BOOL 1
73# endif
74#endif
75
8e84507e 76/* bool is built-in for g++-2.6.3 and later, which might be used
c1d22f6b
GS
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
8e84507e 79 also defines _G_HAVE_BOOL for both gcc and g++, but only g++
c1d22f6b
GS
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
5d94fbed 83*/
3609ea0d 84#ifdef __GNUG__ /* GNU g++ has bool built-in */
f789f6a4 85# ifndef PERL_BOOL_AS_CHAR
5d94fbed 86# ifndef HAS_BOOL
c1d22f6b 87# define HAS_BOOL 1
5d94fbed 88# endif
f789f6a4 89# endif
5d94fbed
AD
90#endif
91
92#ifndef HAS_BOOL
f789f6a4
FC
93# ifdef bool
94# undef bool
95# endif
70d5cb32 96# define bool char
c1d22f6b 97# define HAS_BOOL 1
a687059c 98#endif
0d3e774c 99
236415e2
KW
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
103 * AIX */
18f5643b 104#define cBOOL(cbool) ((cbool) ? (bool)1 : (bool)0)
f2338a2e 105
46c6c7e2 106/* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
e352bcff
JH
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. */
46c6c7e2
JH
110#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
111# define FUNCTION__ __func__
07798b17
AC
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__ ""
46c6c7e2 115#else
07798b17 116# define FUNCTION__ __FUNCTION__ /* Common extension. */
46c6c7e2
JH
117#endif
118
27d4fb96
PP
119/* XXX A note on the perl source internal type system. The
120 original intent was that I32 be *exactly* 32 bits.
121
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
24fef2a7 128
dc45a647
MB
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.
693762b4 132
dc45a647
MB
133 Similarly, there is no guarantee that I16 and U16 have exactly 16
134 bits.
693762b4 135
8e84507e
NIS
136 For dealing with issues that may arise from various 32/64-bit
137 systems, we will ask Configure to check out
8175356b 138
3609ea0d
JH
139 SHORTSIZE == sizeof(short)
140 INTSIZE == sizeof(int)
141 LONGSIZE == sizeof(long)
dc45a647 142 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
3609ea0d 143 PTRSIZE == sizeof(void *)
dc45a647
MB
144 DOUBLESIZE == sizeof(double)
145 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
8175356b 146
27d4fb96
PP
147*/
148
69512466
JH
149#ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
150# include <inttypes.h>
dd0eed91
JH
151# ifdef INT32_MIN_BROKEN
152# undef INT32_MIN
153# define INT32_MIN (-2147483647-1)
154# endif
155# ifdef INT64_MIN_BROKEN
156# undef INT64_MIN
157# define INT64_MIN (-9223372036854775807LL-1)
158# endif
69512466
JH
159#endif
160
8175356b
JH
161typedef I8TYPE I8;
162typedef U8TYPE U8;
163typedef I16TYPE I16;
164typedef U16TYPE U16;
165typedef I32TYPE I32;
166typedef U32TYPE U32;
16d89be8 167
74b807c7 168#ifdef QUADKIND
8175356b
JH
169typedef I64TYPE I64;
170typedef U64TYPE U64;
16d89be8 171#endif
8175356b 172
d8668976 173#if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
5ff3f7a4 174
5ff3f7a4
GS
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
179
5ff3f7a4
GS
180#define I16_MAX INT16_MAX
181#define I16_MIN INT16_MIN
182#define U16_MAX UINT16_MAX
183#define U16_MIN UINT16_MIN
184
5ff3f7a4
GS
185#define I32_MAX INT32_MAX
186#define I32_MIN INT32_MIN
0e983133
GS
187#ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
188# define U32_MAX UINT32_MAX
189#else
190# define U32_MAX 4294967295U
191#endif
5ff3f7a4
GS
192#define U32_MIN UINT32_MIN
193
194#else
195
5c9fa16e
KA
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. */
27d4fb96
PP
198#define U8_MAX PERL_UCHAR_MAX
199#define U8_MIN PERL_UCHAR_MIN
79072805 200
27d4fb96
PP
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
79072805 205
c4f23d77 206#if LONGSIZE > 4
27d4fb96
PP
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
79072805 211#else
27d4fb96
PP
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
79072805
LW
216#endif
217
5ff3f7a4
GS
218#endif
219
247cee9f
KW
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.
223 *
224 * Restrict these to core for now until we are more certain this is a good
225 * idea. */
226#if defined(PERL_CORE) || defined(PERL_EXT)
227# ifdef I_STDINT
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;
232# else
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;
237# endif
238#endif
239
464decb6
KW
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 */
fc36a67e
PP
243#define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
244#define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
245
88794300 246/* Unused by core; should be deprecated */
ff68c719 247#define Ctl(ch) ((ch) & 037)
8d063cd8 248
98fce2a4
KW
249#if defined(PERL_CORE) || defined(PERL_EXT)
250# ifndef MIN
251# define MIN(a,b) ((a) < (b) ? (a) : (b))
252# endif
253# ifndef MAX
254# define MAX(a,b) ((a) > (b) ? (a) : (b))
255# endif
256#endif
257
84ff4fa9
KW
258/* Returns a boolean as to whether the input unsigned number is a power of 2
259 * (2**0, 2**1, etc). In other words if it has just a single bit set.
260 * If not, subtracting 1 would leave the uppermost bit set, so the & would
261 * yield non-zero */
262#if defined(PERL_CORE) || defined(PERL_EXT)
011b1419 263# define isPOWER_OF_2(n) ((n) && ((n) & ((n)-1)) == 0)
84ff4fa9
KW
264#endif
265
4650c663
KW
266/* This is a helper macro to avoid preprocessor issues, replaced by nothing
267 * unless under DEBUGGING, where it expands to an assert of its argument,
268 * followed by a comma (hence the comma operator). If we just used a straight
3e94db23
JH
269 * assert(), we would get a comma with nothing before it when not DEBUGGING.
270 *
271 * We also use empty definition under Coverity since the __ASSERT__
272 * checks often check for things that Really Cannot Happen, and Coverity
273 * detects that and gets all excited. */
274
275#if defined(DEBUGGING) && !defined(__COVERITY__)
0f092d08
KW
276# define __ASSERT_(statement) assert(statement),
277#else
278# define __ASSERT_(statement)
279#endif
280
3fe05580 281/*
a4ee4fb5 282=head1 SV Manipulation Functions
3fe05580 283
1568d13a
Z
284=for apidoc Ama|SV*|newSVpvs|"literal string" s
285Like C<newSVpvn>, but takes a literal string instead of a
30a15352 286string/length pair.
3fe05580 287
1568d13a
Z
288=for apidoc Ama|SV*|newSVpvs_flags|"literal string" s|U32 flags
289Like C<newSVpvn_flags>, but takes a literal string instead of
30a15352 290a string/length pair.
84bafc02 291
1568d13a
Z
292=for apidoc Ama|SV*|newSVpvs_share|"literal string" s
293Like C<newSVpvn_share>, but takes a literal string instead of
30a15352 294a string/length pair and omits the hash parameter.
3fe05580 295
1568d13a
Z
296=for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string" s|I32 flags
297Like C<sv_catpvn_flags>, but takes a literal string instead
30a15352 298of a string/length pair.
9dcc53ea 299
1568d13a
Z
300=for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string" s
301Like C<sv_catpvn_nomg>, but takes a literal string instead of
0c395ea5 302a string/length pair.
9dcc53ea 303
1568d13a
Z
304=for apidoc Am|void|sv_catpvs|SV* sv|"literal string" s
305Like C<sv_catpvn>, but takes a literal string instead of a
0c395ea5 306string/length pair.
3fe05580 307
1568d13a
Z
308=for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string" s
309Like C<sv_catpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
310string/length pair.
311
1568d13a
Z
312=for apidoc Am|void|sv_setpvs|SV* sv|"literal string" s
313Like C<sv_setpvn>, but takes a literal string instead of a
0c395ea5 314string/length pair.
3fe05580 315
1568d13a
Z
316=for apidoc Am|void|sv_setpvs_mg|SV* sv|"literal string" s
317Like C<sv_setpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
318string/length pair.
319
78342678 320=for apidoc Am|SV *|sv_setref_pvs|SV *const rv|const char *const classname|"literal string" s
1568d13a 321Like C<sv_setref_pvn>, but takes a literal string instead of
0c395ea5 322a string/length pair.
9dcc53ea 323
3fe05580
MHM
324=head1 Memory Management
325
1568d13a
Z
326=for apidoc Ama|char*|savepvs|"literal string" s
327Like C<savepvn>, but takes a literal string instead of a
30a15352 328string/length pair.
3fe05580 329
1568d13a 330=for apidoc Ama|char*|savesharedpvs|"literal string" s
9dcc53ea
Z
331A version of C<savepvs()> which allocates the duplicate string in memory
332which is shared between threads.
333
3fe05580
MHM
334=head1 GV Functions
335
1568d13a
Z
336=for apidoc Am|HV*|gv_stashpvs|"literal string" name|I32 create
337Like C<gv_stashpvn>, but takes a literal string instead of a
0c395ea5 338string/length pair.
3fe05580
MHM
339
340=head1 Hash Manipulation Functions
341
1568d13a
Z
342=for apidoc Am|SV**|hv_fetchs|HV* tb|"literal string" key|I32 lval
343Like C<hv_fetch>, but takes a literal string instead of a
0c395ea5 344string/length pair.
3fe05580 345
d106cb46 346=for apidoc Am|SV**|hv_stores|HV* tb|"literal string" key|SV* val
1568d13a 347Like C<hv_store>, but takes a literal string instead of a
0c395ea5 348string/length pair
3fe05580
MHM
349and omits the hash parameter.
350
510966aa
Z
351=head1 Lexer interface
352
1568d13a 353=for apidoc Amx|void|lex_stuff_pvs|"literal string" pv|U32 flags
510966aa 354
1568d13a 355Like L</lex_stuff_pvn>, but takes a literal string instead of
0c395ea5 356a string/length pair.
510966aa 357
3fe05580
MHM
358=cut
359*/
360
926bb54c
KW
361/* concatenating with "" ensures that only literal strings are accepted as
362 * argument */
26bfd924 363#define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
2efa8cc7 364
926bb54c
KW
365/* note that STR_WITH_LEN() can't be used as argument to macros or functions
366 * that under some configurations might be macros, which means that it requires
367 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
ba3a79e7
GA
368 */
369
370/* STR_WITH_LEN() shortcuts */
371#define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
84bafc02
NC
372#define newSVpvs_flags(str,flags) \
373 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
ba3a79e7 374#define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
9dcc53ea
Z
375#define sv_catpvs_flags(sv, str, flags) \
376 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
377#define sv_catpvs_nomg(sv, str) \
378 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
379#define sv_catpvs(sv, str) \
380 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
381#define sv_catpvs_mg(sv, str) \
382 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
3fe05580 383#define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
9dcc53ea
Z
384#define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
385#define sv_setref_pvs(rv, classname, str) \
386 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
ba3a79e7 387#define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
9dcc53ea
Z
388#define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
389#define gv_stashpvs(str, create) \
390 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
391#define gv_fetchpvs(namebeg, add, sv_type) \
392 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
393#define gv_fetchpvn(namebeg, len, add, sv_type) \
394 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
395#define sv_catxmlpvs(dsv, str, utf8) \
396 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
4ac46235 397
ba3a79e7 398
510966aa
Z
399#define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
400
b96d8cd9
NC
401#define get_cvs(str, flags) \
402 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
5c1737d1 403
954c1994 404/*
ccfc67b7
JH
405=head1 Miscellaneous Functions
406
954c1994 407=for apidoc Am|bool|strNE|char* s1|char* s2
dc6b0978
KW
408Test two C<NUL>-terminated strings to see if they are different. Returns true
409or false.
954c1994
GS
410
411=for apidoc Am|bool|strEQ|char* s1|char* s2
dc6b0978
KW
412Test two C<NUL>-terminated strings to see if they are equal. Returns true or
413false.
954c1994
GS
414
415=for apidoc Am|bool|strLT|char* s1|char* s2
dc6b0978
KW
416Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
417second, C<s2>. Returns true or false.
954c1994
GS
418
419=for apidoc Am|bool|strLE|char* s1|char* s2
dc6b0978
KW
420Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
421equal to the second, C<s2>. Returns true or false.
954c1994
GS
422
423=for apidoc Am|bool|strGT|char* s1|char* s2
dc6b0978
KW
424Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
425the second, C<s2>. Returns true or false.
954c1994
GS
426
427=for apidoc Am|bool|strGE|char* s1|char* s2
dc6b0978
KW
428Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
429or equal to the second, C<s2>. Returns true or false.
954c1994
GS
430
431=for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
dc6b0978
KW
432Test two C<NUL>-terminated strings to see if they are different. The C<len>
433parameter indicates the number of bytes to compare. Returns true or false. (A
954c1994
GS
434wrapper for C<strncmp>).
435
436=for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
dc6b0978
KW
437Test two C<NUL>-terminated strings to see if they are equal. The C<len>
438parameter indicates the number of bytes to compare. Returns true or false. (A
439wrapper for C<strncmp>).
954c1994 440
bd18bd40
KW
441=for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
442Test two buffers (which may contain embedded C<NUL> characters, to see if they
443are equal. The C<len> parameter indicates the number of bytes to compare.
444Returns zero if equal, or non-zero if non-equal.
445
446=for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
447Test two buffers (which may contain embedded C<NUL> characters, to see if they
448are not equal. The C<len> parameter indicates the number of bytes to compare.
449Returns zero if non-equal, or non-zero if equal.
450
954c1994 451=cut
fc169e00
KW
452
453New macros should use the following conventions for their names (which are
454based on the underlying C library functions):
455
456 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
457
458 Each has two main parameters, string-like operands that are compared
459 against each other, as specified by the macro name. Some macros may
460 additionally have one or potentially even two length parameters. If a length
461 parameter applies to both string parameters, it will be positioned third;
462 otherwise any length parameter immediately follows the string parameter it
463 applies to.
464
465 If the prefix to the name is 'str', the string parameter is a pointer to a C
466 language string. Such a string does not contain embedded NUL bytes; its
467 length may be unknown, but can be calculated by C<strlen()>, since it is
468 terminated by a NUL, which isn't included in its length.
469
470 The optional 'n' following 'str' means that that there is a third parameter,
471 giving the maximum number of bytes to look at in each string. Even if both
472 strings are longer than the length parameter, those extra bytes will be
473 unexamined.
474
475 The 's' suffix means that the 2nd byte string parameter is a literal C
476 double-quoted string. Its length will automatically be calculated by the
477 macro, so no length parameter will ever be needed for it.
478
479 If the prefix is 'mem', the string parameters don't have to be C strings;
480 they may contain embedded NUL bytes, do not necessarily have a terminating
481 NUL, and their lengths can be known only through other means, which in
482 practice are additional parameter(s) passed to the function. All 'mem'
483 functions have at least one length parameter. Barring any 'l' or 's' suffix,
484 there is a single length parameter, in position 3, which applies to both
485 string parameters. The 's' suffix means, as described above, that the 2nd
486 string is a literal double-quoted C string (hence its length is calculated by
487 the macro, and the length parameter to the function applies just to the first
488 string parameter, and hence is positioned just after it). An 'l' suffix
489 means that the 2nd string parameter has its own length parameter, and the
490 signature will look like memFOOl(s1, l1, s2, l2).
491
492 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
493 substring of the 1st string. 'P' if present indicates that the substring
494 must be a "proper" one in tha mathematical sense that the first one must be
495 strictly larger than the 2nd.
496
954c1994
GS
497*/
498
62946e08 499
75400963
KW
500#define strNE(s1,s2) (strcmp(s1,s2) != 0)
501#define strEQ(s1,s2) (strcmp(s1,s2) == 0)
8d063cd8
LW
502#define strLT(s1,s2) (strcmp(s1,s2) < 0)
503#define strLE(s1,s2) (strcmp(s1,s2) <= 0)
504#define strGT(s1,s2) (strcmp(s1,s2) > 0)
505#define strGE(s1,s2) (strcmp(s1,s2) >= 0)
62946e08 506
75400963
KW
507#define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
508#define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
378cc40b 509
9d3980bc
KW
510#define memEQ(s1,s2,l) (memcmp(((const void *) (s1)), ((const void *) (s2)), l) == 0)
511#define memNE(s1,s2,l) (! memEQ(s1,s2,l))
36477c24 512
085b7534 513/* memEQ and memNE where second comparand is a string constant */
568a785a 514#define memEQs(s1, l, s2) \
777fa2cb 515 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
5f50c6c9 516#define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
568a785a 517
fdbb9a7c
KW
518/* Keep these private until we decide it was a good idea */
519#if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
520
521#define strBEGINs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1) == 0)
522
bdb7e3f0 523#define memBEGINs(s1, l, s2) \
30a6480c 524 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
bdb7e3f0 525 && memEQ(s1, "" s2 "", sizeof(s2)-1))
de627158 526#define memBEGINPs(s1, l, s2) \
30a6480c 527 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) - 1 \
de627158 528 && memEQ(s1, "" s2 "", sizeof(s2)-1))
bdb7e3f0 529#define memENDs(s1, l, s2) \
30a6480c 530 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
bdb7e3f0 531 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
b80f8424 532#define memENDPs(s1, l, s2) \
30a6480c 533 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) \
b80f8424 534 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
fdbb9a7c 535#endif /* End of making macros private */
bdb7e3f0 536
062b6850
KW
537#define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
538#define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
539#define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
540#define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
541
bbce6d69
PP
542/*
543 * Character classes.
544 *
545 * Unfortunately, the introduction of locales means that we
546 * can't trust isupper(), etc. to tell the truth. And when
547 * it comes to /\w+/ with tainting enabled, we *must* be able
548 * to trust our character classes.
549 *
550 * Therefore, the default tests in the text of Perl will be
551 * independent of locale. Any code that wants to depend on
552 * the current locale will use the tests that begin with "lc".
553 */
554
2304df62
AD
555#ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
556# ifndef CTYPE256
557# define CTYPE256
558# endif
559#endif
560
954c1994 561/*
ccfc67b7 562
dcccc8ff 563=head1 Character classification
243effed
KW
564This section is about functions (really macros) that classify characters
565into types, such as punctuation versus alphabetic, etc. Most of these are
566analogous to regular expression character classes. (See
567L<perlrecharclass/POSIX Character Classes>.) There are several variants for
568each class. (Not all macros have all variants; each item below lists the
569ones valid for it.) None are affected by C<use bytes>, and only the ones
570with C<LC> in the name are affected by the current locale.
571
d713f9d9
KW
572The base function, e.g., C<isALPHA()>, takes any signed or unsigned value,
573treating it as a code point, and returns a boolean as to whether or not the
574character represented by it is (or on non-ASCII platforms, corresponds to) an
6aff1f14
KW
575ASCII character in the named class based on platform, Unicode, and Perl rules.
576If the input is a number that doesn't fit in an octet, FALSE is returned.
243effed 577
c98722a4 578Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
550da823
KW
579with no suffix C<"_A">. This variant is used to emphasize by its name that
580only ASCII-range characters can return TRUE.
4b9734bf 581
d60679e1 582Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
4b9734bf
KW
583onto the platform. That is, the code points that are ASCII are unaffected,
584since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
585as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
586true when called with the code point 0xDF, which is a word character in both
4650c663 587ASCII and EBCDIC (though it represents different characters in each).
d713f9d9
KW
588If the input is a number that doesn't fit in an octet, FALSE is returned.
589(Perl's documentation uses a colloquial definition of Latin-1, to include all
590code points below 256.)
243effed 591
d713f9d9
KW
592Variant C<isI<FOO>_uvchr> is exactly like the C<isI<FOO>_L1> variant, for
593inputs below 256, but if the code point is larger than 255, Unicode rules are
594used to determine if it is in the character class. For example,
d0da05db 595C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
6aff1f14 596WITH MACRON in Unicode, and is a word character.
243effed 597
c98722a4 598Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
d713f9d9
KW
599encoded strings. Each call classifies the first character of the string. This
600variant takes two parameters. The first, C<p>, is a
da8c1a98
KW
601pointer to the first byte of the character to be classified. (Recall that it
602may take more than one byte to represent a character in UTF-8 strings.) The
603second parameter, C<e>, points to anywhere in the string beyond the first
604character, up to one byte past the end of the entire string. The suffix
605C<_safe> in the function's name indicates that it will not attempt to read
606beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
607is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
608character is malformed in some way, the program may croak, or the function may
609return FALSE, at the discretion of the implementation, and subject to change in
610future releases.
611
c98722a4 612Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
da8c1a98 613parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 614in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
8b4b30c5 615beyond the end of the string. Starting in Perl v5.32, it will take a second
c98722a4 616parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
34aeb2e9 617program that uses it will have to be changed to successfully compile. In the
c98722a4 618meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
34aeb2e9 619program will raise a deprecation warning, enabled by default. You can convert
c98722a4 620your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 621extra measure of protection, or you can wait until v5.32, when you'll be forced
34aeb2e9 622to add the C<e> parameter.
243effed 623
d713f9d9
KW
624Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants,
625but the result is based on the current locale, which is what C<LC> in the name
626stands for. If Perl can determine that the current locale is a UTF-8 locale,
627it uses the published Unicode rules; otherwise, it uses the C library function
628that gives the named classification. For example, C<isDIGIT_LC()> when not in
629a UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
1a83413c
KW
630returned if the input won't fit into an octet. On some platforms where the C
631library function is known to be defective, Perl changes its result to follow
632the POSIX standard's rules.
243effed 633
d713f9d9
KW
634Variant C<isI<FOO>_LC_uvchr> acts exactly like C<isI<FOO>_LC> for inputs less
635than 256, but for larger ones it returns the Unicode classification of the code
636point.
243effed 637
c98722a4 638Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
d713f9d9
KW
639encoded strings. Each call classifies the first character of the string. This
640variant takes two parameters. The first, C<p>, is a pointer to the first byte
641of the character to be classified. (Recall that it may take more than one byte
642to represent a character in UTF-8 strings.) The second parameter, C<e>,
643points to anywhere in the string beyond the first character, up to one byte
644past the end of the entire string. The suffix C<_safe> in the function's name
645indicates that it will not attempt to read beyond S<C<e - 1>>, provided that
646the constraint S<C<s E<lt> e>> is true (this is asserted for in C<-DDEBUGGING>
647builds). If the UTF-8 for the input character is malformed in some way, the
648program may croak, or the function may return FALSE, at the discretion of the
649implementation, and subject to change in future releases.
da8c1a98 650
c98722a4 651Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
da8c1a98 652parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 653in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
8b4b30c5 654beyond the end of the string. Starting in Perl v5.32, it will take a second
c98722a4 655parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
34aeb2e9 656program that uses it will have to be changed to successfully compile. In the
c98722a4 657meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
34aeb2e9 658the program will raise a deprecation warning, enabled by default. You can
c98722a4 659convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
8b4b30c5 660and get an extra measure of protection, or you can wait until v5.32, when
34aeb2e9 661you'll be forced to add the C<e> parameter.
ccfc67b7 662
d713f9d9
KW
663=for apidoc Am|bool|isALPHA|int ch
664Returns a boolean indicating whether the specified input is one of C<[A-Za-z]>,
665analogous to C<m/[[:alpha:]]/>.
dcccc8ff
KW
666See the L<top of this section|/Character classification> for an explanation of
667variants
da8c1a98
KW
668C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
669C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
8a58bdcf 670
d713f9d9
KW
671=for apidoc Am|bool|isALPHANUMERIC|int ch
672Returns a boolean indicating whether the specified character is one of
673C<[A-Za-z0-9]>, analogous to C<m/[[:alnum:]]/>.
dcccc8ff
KW
674See the L<top of this section|/Character classification> for an explanation of
675variants
d0da05db 676C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
da8c1a98
KW
677C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
678and C<isALPHANUMERIC_LC_utf8_safe>.
15861f94 679
255b632a
KW
680A (discouraged from use) synonym is C<isALNUMC> (where the C<C> suffix means
681this corresponds to the C language alphanumeric definition). Also
682there are the variants
683C<isALNUMC_A>, C<isALNUMC_L1>
684C<isALNUMC_LC>, and C<isALNUMC_LC_uvchr>.
685
d713f9d9 686=for apidoc Am|bool|isASCII|int ch
8a58bdcf 687Returns a boolean indicating whether the specified character is one of the 128
243effed 688characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
e5ad6aba 689On non-ASCII platforms, it returns TRUE iff this
8a58bdcf
KW
690character corresponds to an ASCII character. Variants C<isASCII_A()> and
691C<isASCII_L1()> are identical to C<isASCII()>.
dcccc8ff
KW
692See the L<top of this section|/Character classification> for an explanation of
693variants
da8c1a98
KW
694C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
695C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
243effed
KW
696library routine C<isascii()>. In these cases, the variants whose names contain
697C<LC> are the same as the corresponding ones without.
698
d98532ea
KW
699Also note, that because all ASCII characters are UTF-8 invariant (meaning they
700have the exact same representation (always a single byte) whether encoded in
701UTF-8 or not), C<isASCII> will give the correct results when called with any
da8c1a98
KW
702byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
703will work properly on any string encoded or not in UTF-8.
d98532ea 704
243effed
KW
705=for apidoc Am|bool|isBLANK|char ch
706Returns a boolean indicating whether the specified character is a
6aff1f14 707character considered to be a blank, analogous to C<m/[[:blank:]]/>.
dcccc8ff
KW
708See the L<top of this section|/Character classification> for an explanation of
709variants
da8c1a98
KW
710C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
711C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
712however, that some platforms do not have the C library routine
713C<isblank()>. In these cases, the variants whose names contain C<LC> are
714the same as the corresponding ones without.
243effed
KW
715
716=for apidoc Am|bool|isCNTRL|char ch
717Returns a boolean indicating whether the specified character is a
6aff1f14 718control character, analogous to C<m/[[:cntrl:]]/>.
dcccc8ff
KW
719See the L<top of this section|/Character classification> for an explanation of
720variants
da8c1a98
KW
721C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
722C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
723platforms, you almost always want to use the C<isCNTRL_L1> variant.
954c1994 724
954c1994 725=for apidoc Am|bool|isDIGIT|char ch
2787a470 726Returns a boolean indicating whether the specified character is a
6aff1f14 727digit, analogous to C<m/[[:digit:]]/>.
8a58bdcf 728Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
dcccc8ff
KW
729See the L<top of this section|/Character classification> for an explanation of
730variants
da8c1a98
KW
731C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
732C<isDIGIT_LC_utf8_safe>.
243effed
KW
733
734=for apidoc Am|bool|isGRAPH|char ch
735Returns a boolean indicating whether the specified character is a
6aff1f14 736graphic character, analogous to C<m/[[:graph:]]/>.
dcccc8ff 737See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
738variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
739C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
954c1994 740
0c82b6df 741=for apidoc Am|bool|isLOWER|char ch
2787a470 742Returns a boolean indicating whether the specified character is a
6aff1f14 743lowercase character, analogous to C<m/[[:lower:]]/>.
dcccc8ff
KW
744See the L<top of this section|/Character classification> for an explanation of
745variants
da8c1a98
KW
746C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
747C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
0c82b6df 748
c99e91e9 749=for apidoc Am|bool|isOCTAL|char ch
2787a470 750Returns a boolean indicating whether the specified character is an
6aff1f14 751octal digit, [0-7].
243effed
KW
752The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
753C<isOCTAL>.
754
755=for apidoc Am|bool|isPUNCT|char ch
756Returns a boolean indicating whether the specified character is a
6aff1f14
KW
757punctuation character, analogous to C<m/[[:punct:]]/>.
758Note that the definition of what is punctuation isn't as
243effed
KW
759straightforward as one might desire. See L<perlrecharclass/POSIX Character
760Classes> for details.
dcccc8ff 761See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
762variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
763C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
c99e91e9 764
0c82b6df 765=for apidoc Am|bool|isSPACE|char ch
2787a470 766Returns a boolean indicating whether the specified character is a
6aff1f14 767whitespace character. This is analogous
398d098a 768to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
779cf272 769this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
398d098a
KW
770locale forms of this macro (the ones with C<LC> in their names) matched
771precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
772in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
773(See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
dcccc8ff
KW
774See the L<top of this section|/Character classification> for an explanation of
775variants
da8c1a98
KW
776C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
777C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
0c82b6df 778
398d098a
KW
779=for apidoc Am|bool|isPSXSPC|char ch
780(short for Posix Space)
779cf272
KW
781Starting in 5.18, this is identical in all its forms to the
782corresponding C<isSPACE()> macros.
398d098a
KW
783The locale forms of this macro are identical to their corresponding
784C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
785non-locale forms differ from their C<isSPACE()> forms only in that the
786C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
787Otherwise they are identical. Thus this macro is analogous to what
788C<m/[[:space:]]/> matches in a regular expression.
dcccc8ff 789See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
790variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
791C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
398d098a 792
954c1994 793=for apidoc Am|bool|isUPPER|char ch
2787a470 794Returns a boolean indicating whether the specified character is an
6aff1f14 795uppercase character, analogous to C<m/[[:upper:]]/>.
dcccc8ff 796See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
797variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
798C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
954c1994 799
243effed 800=for apidoc Am|bool|isPRINT|char ch
8eea39dd 801Returns a boolean indicating whether the specified character is a
6aff1f14 802printable character, analogous to C<m/[[:print:]]/>.
dcccc8ff
KW
803See the L<top of this section|/Character classification> for an explanation of
804variants
da8c1a98
KW
805C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
806C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
243effed
KW
807
808=for apidoc Am|bool|isWORDCHAR|char ch
809Returns a boolean indicating whether the specified character is a character
810that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
811in a regular expression. A word character is an alphabetic character, a
812decimal digit, a connecting punctuation character (such as an underscore), or
813a "mark" character that attaches to one of those (like some sort of accent).
814C<isALNUM()> is a synonym provided for backward compatibility, even though a
815word character includes more than the standard C language meaning of
816alphanumeric.
dcccc8ff 817See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
818variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
819C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
820C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
821include the platform's native underscore.
8a58bdcf
KW
822
823=for apidoc Am|bool|isXDIGIT|char ch
824Returns a boolean indicating whether the specified character is a hexadecimal
243effed
KW
825digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
826and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
dcccc8ff
KW
827See the L<top of this section|/Character classification> for an explanation of
828variants
da8c1a98
KW
829C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
830and C<isXDIGIT_LC_utf8_safe>.
243effed 831
3c3ecf18
KW
832=for apidoc Am|bool|isIDFIRST|char ch
833Returns a boolean indicating whether the specified character can be the first
834character of an identifier. This is very close to, but not quite the same as
835the official Unicode property C<XID_Start>. The difference is that this
836returns true only if the input character also matches L</isWORDCHAR>.
dcccc8ff
KW
837See the L<top of this section|/Character classification> for an explanation of
838variants
da8c1a98
KW
839C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
840C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
3c3ecf18
KW
841
842=for apidoc Am|bool|isIDCONT|char ch
843Returns a boolean indicating whether the specified character can be the
844second or succeeding character of an identifier. This is very close to, but
845not quite the same as the official Unicode property C<XID_Continue>. The
846difference is that this returns true only if the input character also matches
dcccc8ff
KW
847L</isWORDCHAR>. See the L<top of this section|/Character classification> for
848an
d0da05db 849explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
da8c1a98
KW
850C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
851C<isIDCONT_LC_utf8_safe>.
3c3ecf18 852
243effed 853=head1 Miscellaneous Functions
8eea39dd 854
95a59cab 855=for apidoc Am|U8|READ_XDIGIT|char str*
243effed 856Returns the value of an ASCII-range hex digit and advances the string pointer.
95a59cab
YO
857Behaviour is only well defined when isXDIGIT(*str) is true.
858
e7c1e6c1 859=head1 Character case changing
21da7284
KW
860Perl uses "full" Unicode case mappings. This means that converting a single
861character to another case may result in a sequence of more than one character.
862For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
863character sequence C<SS>. This presents some complications The lowercase of
864all characters in the range 0..255 is a single character, and thus
865C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
866return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
867an API that does allow every possible legal result to be returned.) Likewise
868no other function that is crippled by not being able to give the correct
869results for the full range of possible inputs has been implemented here.
e7c1e6c1 870
d713f9d9 871=for apidoc Am|U8|toUPPER|int ch
1f607577
KW
872Converts the specified character to uppercase. If the input is anything but an
873ASCII lowercase character, that input character itself is returned. Variant
c753c8d3 874C<toUPPER_A> is equivalent.
954c1994 875
d0da05db
KW
876=for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
877Converts the code point C<cp> to its uppercase version, and
878stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
879point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
880that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
881bytes since the uppercase version may be longer than the original character.
882
883The first code point of the uppercased version is returned
21da7284
KW
884(but note, as explained at L<the top of this section|/Character case
885changing>, that there may be more.)
1f607577 886
a239b1e2
KW
887=for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
888Converts the first UTF-8 encoded character in the sequence starting at C<p> and
889extending no further than S<C<e - 1>> to its uppercase version, and
1f607577
KW
890stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
891that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
892bytes since the uppercase version may be longer than the original character.
893
894The first code point of the uppercased version is returned
21da7284
KW
895(but note, as explained at L<the top of this section|/Character case
896changing>, that there may be more).
1f607577 897
a239b1e2
KW
898The suffix C<_safe> in the function's name indicates that it will not attempt
899to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
900true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
901input character is malformed in some way, the program may croak, or the
902function may return the REPLACEMENT CHARACTER, at the discretion of the
903implementation, and subject to change in future releases.
904
905=for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
906This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
907parameter The function therefore can't check if it is reading
8b4b30c5 908beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
909parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
910program that uses it will have to be changed to successfully compile. In the
911meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
912program will raise a deprecation warning, enabled by default. You can convert
913your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 914extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 915to add the C<e> parameter.
1f607577 916
25200305
KW
917=for apidoc Am|U8|toFOLD|U8 ch
918Converts the specified character to foldcase. If the input is anything but an
919ASCII uppercase character, that input character itself is returned. Variant
920C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
d0da05db 921Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
25200305 922
d0da05db
KW
923=for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
924Converts the code point C<cp> to its foldcase version, and
925stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
926point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
927that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
928bytes since the foldcase version may be longer than the original character.
929
930The first code point of the foldcased version is returned
21da7284
KW
931(but note, as explained at L<the top of this section|/Character case
932changing>, that there may be more).
1f607577 933
a239b1e2
KW
934=for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
935Converts the first UTF-8 encoded character in the sequence starting at C<p> and
936extending no further than S<C<e - 1>> to its foldcase version, and
1f607577
KW
937stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
938that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
939bytes since the foldcase version may be longer than the original character.
940
941The first code point of the foldcased version is returned
21da7284
KW
942(but note, as explained at L<the top of this section|/Character case
943changing>, that there may be more).
1f607577 944
a239b1e2
KW
945The suffix C<_safe> in the function's name indicates that it will not attempt
946to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
947true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
948input character is malformed in some way, the program may croak, or the
949function may return the REPLACEMENT CHARACTER, at the discretion of the
950implementation, and subject to change in future releases.
951
952=for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
953This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
954parameter The function therefore can't check if it is reading
8b4b30c5 955beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
956parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
957program that uses it will have to be changed to successfully compile. In the
958meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
959program will raise a deprecation warning, enabled by default. You can convert
960your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 961extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 962to add the C<e> parameter.
1f607577
KW
963
964=for apidoc Am|U8|toLOWER|U8 ch
965Converts the specified character to lowercase. If the input is anything but an
966ASCII uppercase character, that input character itself is returned. Variant
c753c8d3 967C<toLOWER_A> is equivalent.
954c1994 968
1f607577 969=for apidoc Am|U8|toLOWER_L1|U8 ch
b7d90381
KW
970Converts the specified Latin1 character to lowercase. The results are
971undefined if the input doesn't fit in a byte.
1f607577
KW
972
973=for apidoc Am|U8|toLOWER_LC|U8 ch
974Converts the specified character to lowercase using the current locale's rules,
975if possible; otherwise returns the input character itself.
976
d0da05db
KW
977=for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
978Converts the code point C<cp> to its lowercase version, and
979stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
980point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
981that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
982bytes since the lowercase version may be longer than the original character.
983
984The first code point of the lowercased version is returned
21da7284
KW
985(but note, as explained at L<the top of this section|/Character case
986changing>, that there may be more).
1f607577 987
a239b1e2
KW
988
989=for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
990Converts the first UTF-8 encoded character in the sequence starting at C<p> and
991extending no further than S<C<e - 1>> to its lowercase version, and
1f607577
KW
992stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
993that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
994bytes since the lowercase version may be longer than the original character.
995
996The first code point of the lowercased version is returned
21da7284
KW
997(but note, as explained at L<the top of this section|/Character case
998changing>, that there may be more).
1f607577 999
a239b1e2
KW
1000The suffix C<_safe> in the function's name indicates that it will not attempt
1001to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1002true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1003input character is malformed in some way, the program may croak, or the
1004function may return the REPLACEMENT CHARACTER, at the discretion of the
1005implementation, and subject to change in future releases.
1006
1007=for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
1008This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1009parameter The function therefore can't check if it is reading
8b4b30c5 1010beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
1011parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
1012program that uses it will have to be changed to successfully compile. In the
1013meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
1014program will raise a deprecation warning, enabled by default. You can convert
1015your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 1016extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 1017to add the C<e> parameter.
1f607577 1018
25200305
KW
1019=for apidoc Am|U8|toTITLE|U8 ch
1020Converts the specified character to titlecase. If the input is anything but an
1021ASCII lowercase character, that input character itself is returned. Variant
b7d90381 1022C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
d0da05db 1023range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
b7d90381 1024not a concept used in locale handling, so there is no functionality for that.)
25200305 1025
d0da05db
KW
1026=for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1027Converts the code point C<cp> to its titlecase version, and
1028stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
1029point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
1030that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1031bytes since the titlecase version may be longer than the original character.
1032
1033The first code point of the titlecased version is returned
21da7284
KW
1034(but note, as explained at L<the top of this section|/Character case
1035changing>, that there may be more).
1f607577 1036
a239b1e2
KW
1037=for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1038Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1039extending no further than S<C<e - 1>> to its titlecase version, and
1f607577
KW
1040stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1041that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1042bytes since the titlecase version may be longer than the original character.
1043
1044The first code point of the titlecased version is returned
21da7284
KW
1045(but note, as explained at L<the top of this section|/Character case
1046changing>, that there may be more).
1f607577 1047
a239b1e2
KW
1048The suffix C<_safe> in the function's name indicates that it will not attempt
1049to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1050true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1051input character is malformed in some way, the program may croak, or the
1052function may return the REPLACEMENT CHARACTER, at the discretion of the
1053implementation, and subject to change in future releases.
1054
1055=for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1056This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1057parameter The function therefore can't check if it is reading
8b4b30c5 1058beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
1059parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1060program that uses it will have to be changed to successfully compile. In the
1061meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1062program will raise a deprecation warning, enabled by default. You can convert
1063your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 1064extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 1065to add the C<e> parameter.
1f607577 1066
954c1994 1067=cut
353c9b6f 1068
d0da05db 1069XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1e222e4f
KW
1070really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1071and aren't general purpose as they don't work on U+DF, and assert against that.
243effed 1072
8a58bdcf 1073Note that these macros are repeated in Devel::PPPort, so should also be
62fa66b6
KW
1074patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1075
954c1994
GS
1076*/
1077
de9e2639
KW
1078/* Specify the widest unsigned type on the platform. */
1079#ifdef QUADKIND
1080# define WIDEST_UTYPE U64
7c062697
KW
1081#else
1082# define WIDEST_UTYPE U32
1083#endif
1084
3912bc88
KW
1085/* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1086 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1087 * bits of information are lost even on a 64-bit machine, but to get the
1088 * compiler to optimize it out if possible. This is because Configure makes
1089 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1090 * sizeof() guarantees that this evaluates to a constant true at compile time.
7e75d1a1
JH
1091 *
1092 * For Coverity, be always true, because otherwise Coverity thinks
1093 * it finds several expressions that are always true, independent
1094 * of operands. Well, they are, but that is kind of the point.
220c71bf 1095 */
7e75d1a1 1096#ifndef __COVERITY__
6c5b02ac
KW
1097 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1098 * pointer) */
1099#define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
ace3ad0f 1100 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
7e75d1a1
JH
1101#else
1102#define FITS_IN_8_BITS(c) (1)
1103#endif
cf301eb7 1104
94250c4f
KW
1105/* Returns true if c is in the range l..u, where 'l' is non-negative
1106 * Written this way so that after optimization, only one conditional test is
1107 * needed.
1108 *
1109 * This isn't fully general, except for the special cased 'signed char' (which
1110 * should be resolved at compile time): It won't work if 'c' is negative, and
1111 * 'l' is larger than the max for that signed type. Thus if 'c' is a negative
1112 * int, and 'l' is larger than INT_MAX, it will fail. To protect agains this
1113 * happening, there is an assert that will generate a warning if c is larger
1114 * than e.g. INT_MAX if it is an 'unsigned int'. This could be a false
1115 * positive, but khw couldn't figure out a way to make it better. It's good
1116 * enough so far */
1117#define inRANGE(c, l, u) (__ASSERT_((l) >= 0) __ASSERT_((u) >= (l)) \
1118 ((sizeof(c) == 1) \
1119 ? (((WIDEST_UTYPE) ((((U8) (c))|0) - (l))) <= ((WIDEST_UTYPE) ((u) - (l)))) \
1120 : (__ASSERT_( (((WIDEST_UTYPE) 1) << (CHARBITS * sizeof(c) - 1) & (c)) \
1121 /* sign bit of c is 0 */ == 0 \
1122 || (((~ ((WIDEST_UTYPE) 1) << ((CHARBITS * sizeof(c) - 1) - 1))\
1123 /* l not larger than largest value in c's signed type */ \
1124 & ~ ((WIDEST_UTYPE) 0)) & (l)) == 0) \
1125 ((WIDEST_UTYPE) (((c) - (l)) | 0) <= ((WIDEST_UTYPE) ((u) - (l)))))))
305fe86e 1126
41f43cc2 1127#ifdef EBCDIC
b6340bd0 1128# ifndef _ALL_SOURCE
0852beac
KW
1129 /* The native libc isascii() et.al. functions return the wrong results
1130 * on at least z/OS unless this is defined. */
b6340bd0
KW
1131# error _ALL_SOURCE should probably be defined
1132# endif
41f43cc2 1133#else
0852beac
KW
1134 /* There is a simple definition of ASCII for ASCII platforms. But the
1135 * EBCDIC one isn't so simple, so is defined using table look-up like the
9c903d59 1136 * other macros below.
3f3c579d
KW
1137 *
1138 * The cast here is used instead of '(c) >= 0', because some compilers emit
1139 * a warning that that test is always true when the parameter is an
1140 * unsigned type. khw supposes that it could be written as
1141 * && ((c) == '\0' || (c) > 0)
1142 * to avoid the message, but the cast will likely avoid extra branches even
1143 * with stupid compilers.
1144 *
1145 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1146 * a pointer) */
9c903d59 1147# define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
41f43cc2
KW
1148#endif
1149
38694112
KW
1150/* Take the eight possible bit patterns of the lower 3 bits and you get the
1151 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1152 * can be ignored. If the rest match '0', we have an octal */
1153#define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
c2da0b36 1154
9fb1bf9d 1155#ifdef H_PERL /* If have access to perl.h, lookup in its table */
f4cdb42c 1156
a500dc72
KW
1157/* Character class numbers. For internal core Perl use only. The ones less
1158 * than 32 are used in PL_charclass[] and the ones up through the one that
1159 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1160 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1161 * their actual definitions are here. If that file has a name not used here,
1162 * it won't compile.
1709d539
KW
1163 *
1164 * The first group of these is ordered in what I (khw) estimate to be the
31c7f561 1165 * frequency of their use. This gives a slight edge to exiting a loop earlier
58a3ba2c
KW
1166 * (in reginclass() in regexec.c). Except \v should be last, as it isn't a
1167 * real Posix character class, and some (small) inefficiencies in regular
1168 * expression handling would be introduced by putting it in the middle of those
1169 * that are. Also, cntrl and ascii come after the others as it may be useful
1170 * to group these which have no members that match above Latin1, (or above
1171 * ASCII in the latter case) */
1172
1709d539
KW
1173# define _CC_WORDCHAR 0 /* \w and [:word:] */
1174# define _CC_DIGIT 1 /* \d and [:digit:] */
1175# define _CC_ALPHA 2 /* [:alpha:] */
1176# define _CC_LOWER 3 /* [:lower:] */
1177# define _CC_UPPER 4 /* [:upper:] */
1178# define _CC_PUNCT 5 /* [:punct:] */
1179# define _CC_PRINT 6 /* [:print:] */
15861f94 1180# define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1709d539 1181# define _CC_GRAPH 8 /* [:graph:] */
359b005e 1182# define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
779cf272 1183# define _CC_SPACE 10 /* \s, [:space:] */
34aeb2e9
KW
1184# define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1185 when the deprecated isFOO_utf8()
1186 functions are removed */
b0d691b2
KW
1187# define _CC_BLANK 11 /* [:blank:] */
1188# define _CC_XDIGIT 12 /* [:xdigit:] */
779cf272
KW
1189# define _CC_CNTRL 13 /* [:cntrl:] */
1190# define _CC_ASCII 14 /* [:ascii:] */
1191# define _CC_VERTSPACE 15 /* \v */
1709d539 1192
a0947d7b
KW
1193# define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1194
1709d539 1195/* The members of the third group below do not need to be coordinated with data
3ffc8c70 1196 * structures in regcomp.[ch] and regexec.c. */
779cf272
KW
1197# define _CC_IDFIRST 16
1198# define _CC_CHARNAME_CONT 17
1199# define _CC_NONLATIN1_FOLD 18
1200# define _CC_NONLATIN1_SIMPLE_FOLD 19
1201# define _CC_QUOTEMETA 20
1202# define _CC_NON_FINAL_FOLD 21
1203# define _CC_IS_IN_SOME_FOLD 22
1204# define _CC_MNEMONIC_CNTRL 23
073c22b3 1205
34aeb2e9
KW
1206# define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1207 isFOO_utf8() functions are removed */
1208
073c22b3
KW
1209/* This next group is only used on EBCDIC platforms, so theoretically could be
1210 * shared with something entirely different that's only on ASCII platforms */
5d5376e2
KW
1211# define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1212# define _CC_UTF8_IS_START 29
1213# define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1214# define _CC_UTF8_IS_CONTINUATION 31
1215/* Unused: 24-27
f4cdb42c
KW
1216 * If more bits are needed, one could add a second word for non-64bit
1217 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
37ede926
KW
1218 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1219 * is used only for optimization (as of this writing), and differs in the
1220 * Latin1 range from the ALPHA bit only in two relatively unimportant
a500dc72 1221 * characters: the masculine and feminine ordinal indicators, so removing it
073c22b3
KW
1222 * would just cause /i regexes which match them to run less efficiently.
1223 * Similarly the EBCDIC-only bits are used just for speed, and could be
1224 * replaced by other means */
96ac0975 1225
3a371f2f
KW
1226#if defined(PERL_CORE) || defined(PERL_EXT)
1227/* An enum version of the character class numbers, to help compilers
1228 * optimize */
1229typedef enum {
3a371f2f 1230 _CC_ENUM_ALPHA = _CC_ALPHA,
e8d596e0
KW
1231 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1232 _CC_ENUM_ASCII = _CC_ASCII,
1233 _CC_ENUM_BLANK = _CC_BLANK,
b0d691b2 1234 _CC_ENUM_CASED = _CC_CASED,
e8d596e0 1235 _CC_ENUM_CNTRL = _CC_CNTRL,
3a371f2f
KW
1236 _CC_ENUM_DIGIT = _CC_DIGIT,
1237 _CC_ENUM_GRAPH = _CC_GRAPH,
1238 _CC_ENUM_LOWER = _CC_LOWER,
1239 _CC_ENUM_PRINT = _CC_PRINT,
1240 _CC_ENUM_PUNCT = _CC_PUNCT,
e8d596e0 1241 _CC_ENUM_SPACE = _CC_SPACE,
3a371f2f 1242 _CC_ENUM_UPPER = _CC_UPPER,
e8d596e0 1243 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
3a371f2f 1244 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
e8d596e0 1245 _CC_ENUM_XDIGIT = _CC_XDIGIT
3a371f2f
KW
1246} _char_class_number;
1247#endif
1248
86f72d56 1249#define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
63c61c3f 1250
6635f04f 1251START_EXTERN_C
96ac0975
NC
1252# ifdef DOINIT
1253EXTCONST U32 PL_charclass[] = {
1254# include "l1_char_class_tab.h"
1255};
1256
1257# else /* ! DOINIT */
1258EXTCONST U32 PL_charclass[];
1259# endif
6635f04f 1260END_EXTERN_C
96ac0975 1261
265c1f46 1262 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
430b7c70 1263# define _CC_mask(classnum) (1U << (classnum))
4650c663
KW
1264
1265 /* For internal core Perl use only: the base macro for defining macros like
1266 * isALPHA */
ff7ecfc3 1267# define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
f4cd282c 1268 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
4eeeb416 1269
f4cdb42c
KW
1270 /* The mask for the _A versions of the macros; it just adds in the bit for
1271 * ASCII. */
1272# define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1273
4650c663
KW
1274 /* For internal core Perl use only: the base macro for defining macros like
1275 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1276 * the ASCII bit are present */
b7d90381
KW
1277# define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1278 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1279 == _CC_mask_A(classnum)))
f4cdb42c 1280
26c1d9d8
KW
1281/* On ASCII platforms certain classes form a single range. It's faster to
1282 * special case these. isDIGIT is a single range on all platforms */
b877c1ff
KW
1283# ifdef EBCDIC
1284# define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
1285# define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1286# define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1287# define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
1288# define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1289# else
26c1d9d8 1290 /* By folding the upper and lowercase, we can use a single range */
b877c1ff 1291# define isALPHA_A(c) inRANGE((~('A' ^ 'a') & (c)), 'A', 'Z')
26c1d9d8 1292# define isGRAPH_A(c) inRANGE(c, ' ' + 1, 0x7e)
b877c1ff
KW
1293# define isLOWER_A(c) inRANGE(c, 'a', 'z')
1294# define isPRINT_A(c) inRANGE(c, ' ', 0x7e)
1295# define isUPPER_A(c) inRANGE(c, 'A', 'Z')
1296# endif
15861f94 1297# define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
f4cdb42c
KW
1298# define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1299# define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
b877c1ff 1300# define isDIGIT_A(c) inRANGE(c, '0', '9')
f4cdb42c
KW
1301# define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1302# define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
f4cdb42c 1303# define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
b7d90381
KW
1304# define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1305 */
d95f8b6a 1306# define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
3ded5eb0
KW
1307# define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1308# define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1309# define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1310
1311 /* continuation character for legal NAME in \N{NAME} */
1312# define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1313
1314# define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1315# define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1316# define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1317# define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
b7d90381 1318# define isPSXSPC_L1(c) isSPACE_L1(c)
3ded5eb0
KW
1319# define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1320# define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1321# define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1322# define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1323# define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
f4cdb42c 1324
0852beac
KW
1325# ifdef EBCDIC
1326# define isASCII(c) _generic_isCC(c, _CC_ASCII)
1327# endif
1328
f12c0118
KW
1329 /* Participates in a single-character fold with a character above 255 */
1330# 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)))
1331
1332 /* Like the above, but also can be part of a multi-char fold */
f4cd282c 1333# 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)))
430b7c70 1334
4eeeb416 1335# define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
26faadbd 1336# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1337 _generic_isCC(c, _CC_NON_FINAL_FOLD)
37ede926 1338# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1339 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
8e35b056
KW
1340# define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1341 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
687c8d01 1342#else /* else we don't have perl.h H_PERL */
3ded5eb0
KW
1343
1344 /* If we don't have perl.h, we are compiling a utility program. Below we
1345 * hard-code various macro definitions that wouldn't otherwise be available
fc273927 1346 * to it. Most are coded based on first principles. These are written to
74665a89 1347 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
182c4ace 1348# define isDIGIT_A(c) inRANGE(c, '0', '9')
0852beac 1349# define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
74665a89
KW
1350# define isSPACE_A(c) (isBLANK_A(c) \
1351 || (c) == '\n' \
1352 || (c) == '\r' \
1353 || (c) == '\v' \
0852beac 1354 || (c) == '\f')
74665a89
KW
1355 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1356 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1357 * performance (if optimization isn't on), and allow the same code to be
1358 * used for both platform types */
182c4ace
KW
1359# define isLOWER_A(c) inRANGE((c), 'a', 'i') \
1360 || inRANGE((c), 'j', 'r') \
1361 || inRANGE((c), 's', 'z')
1362# define isUPPER_A(c) inRANGE((c), 'A', 'I') \
1363 || inRANGE((c), 'J', 'R') \
1364 || inRANGE((c), 'S', 'Z')
a4d7a999
KW
1365# define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1366# define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
3ded5eb0 1367# define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
0852beac 1368# define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
182c4ace
KW
1369# define isXDIGIT_A(c) ( isDIGIT_A(c) \
1370 || inRANGE((c), 'a', 'f') \
1371 || inRANGE((c), 'A', 'F')
74665a89
KW
1372# define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1373 || (c) == '#' || (c) == '$' || (c) == '%' \
1374 || (c) == '&' || (c) == '\'' || (c) == '(' \
1375 || (c) == ')' || (c) == '*' || (c) == '+' \
1376 || (c) == ',' || (c) == '.' || (c) == '/' \
1377 || (c) == ':' || (c) == ';' || (c) == '<' \
1378 || (c) == '=' || (c) == '>' || (c) == '?' \
1379 || (c) == '@' || (c) == '[' || (c) == '\\' \
1380 || (c) == ']' || (c) == '^' || (c) == '_' \
1381 || (c) == '`' || (c) == '{' || (c) == '|' \
1382 || (c) == '}' || (c) == '~')
1383# define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1384# define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
3ded5eb0 1385
0852beac 1386# ifdef EBCDIC
74665a89
KW
1387 /* The below is accurate for the 3 EBCDIC code pages traditionally
1388 * supported by perl. The only difference between them in the controls
1389 * is the position of \n, and that is represented symbolically below */
1390# define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1391 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1392 || (c) == '\t' || (c) == '\v' \
182c4ace 1393 || inRANGE((c), 1, 3) /* SOH, STX, ETX */ \
74665a89 1394 || (c) == 7 /* U+7F DEL */ \
182c4ace
KW
1395 || inRANGE((c), 0x0E, 0x13) /* SO SI DLE \
1396 DC[1-3] */ \
74665a89
KW
1397 || (c) == 0x18 /* U+18 CAN */ \
1398 || (c) == 0x19 /* U+19 EOM */ \
182c4ace 1399 || inRANGE((c), 0x1C, 0x1F) /* [FGRU]S */ \
74665a89
KW
1400 || (c) == 0x26 /* U+17 ETB */ \
1401 || (c) == 0x27 /* U+1B ESC */ \
1402 || (c) == 0x2D /* U+05 ENQ */ \
1403 || (c) == 0x2E /* U+06 ACK */ \
1404 || (c) == 0x32 /* U+16 SYN */ \
1405 || (c) == 0x37 /* U+04 EOT */ \
1406 || (c) == 0x3C /* U+14 DC4 */ \
1407 || (c) == 0x3D /* U+15 NAK */ \
1408 || (c) == 0x3F)/* U+1A SUB */
0852beac 1409# define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
74665a89
KW
1410# else /* isASCII is already defined for ASCII platforms, so can use that to
1411 define isCNTRL */
1412# define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
0852beac
KW
1413# endif
1414
3ffc8c70 1415 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
caa94d35
KW
1416 * during debugging, and it seems best to keep them. We may be called
1417 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1418 * do anything anyway, so make it not a problem */
1419# if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1420# define NATIVE_TO_LATIN1(ch) (ch)
1421# endif
3ded5eb0
KW
1422# define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1423# define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1424# define isBLANK_L1(c) (isBLANK_A(c) \
1425 || (FITS_IN_8_BITS(c) \
1426 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1427# define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1428# define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1429# define isLOWER_L1(c) (isLOWER_A(c) \
1430 || (FITS_IN_8_BITS(c) \
ae683a5f 1431 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
3ded5eb0
KW
1432 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1433 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1434 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1435 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1436# define isPRINT_L1(c) (isPRINT_A(c) \
1437 || (FITS_IN_8_BITS(c) \
1438 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
3ded5eb0
KW
1439# define isPUNCT_L1(c) (isPUNCT_A(c) \
1440 || (FITS_IN_8_BITS(c) \
ae683a5f 1441 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
3ded5eb0
KW
1442 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1443 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1444 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1445 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1446 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1447 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1448# define isSPACE_L1(c) (isSPACE_A(c) \
1449 || (FITS_IN_8_BITS(c) \
ae683a5f 1450 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
3ded5eb0
KW
1451 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1452# define isUPPER_L1(c) (isUPPER_A(c) \
1453 || (FITS_IN_8_BITS(c) \
182c4ace
KW
1454 && ( IN_RANGE(NATIVE_TO_LATIN1((U8) c), \
1455 0xC0, 0xDE) \
3ded5eb0
KW
1456 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1457# define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1458# define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1459# define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1460 || isBLANK_L1(c) \
1461 || (c) == '-' \
1462 || (c) == '(' \
1463 || (c) == ')')
1464 /* The following are not fully accurate in the above-ASCII range. I (khw)
1465 * don't think it's necessary to be so for the purposes where this gets
1466 * compiled */
1467# define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1468# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1469
1470 /* And these aren't accurate at all. They are useful only for above
1471 * Latin1, which utilities and bootstrapping don't deal with */
1472# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
6838b41e 1473# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
3ded5eb0
KW
1474# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1475
1476 /* Many of the macros later in this file are defined in terms of these. By
1477 * implementing them with a function, which converts the class number into
1478 * a call to the desired macro, all of the later ones work. However, that
1479 * function won't be actually defined when building a utility program (no
1480 * perl.h), and so a compiler error will be generated if one is attempted
1481 * to be used. And the above-Latin1 code points require Unicode tables to
1482 * be present, something unlikely to be the case when bootstrapping */
1483# define _generic_isCC(c, classnum) \
1484 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1485# define _generic_isCC_A(c, classnum) \
1486 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
687c8d01 1487#endif /* End of no perl.h H_PERL */
8a58bdcf 1488
e66b99e9
KW
1489#define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1490#define isALPHA(c) isALPHA_A(c)
0852beac
KW
1491#define isASCII_A(c) isASCII(c)
1492#define isASCII_L1(c) isASCII(c)
e66b99e9
KW
1493#define isBLANK(c) isBLANK_A(c)
1494#define isCNTRL(c) isCNTRL_A(c)
1495#define isDIGIT(c) isDIGIT_A(c)
1496#define isGRAPH(c) isGRAPH_A(c)
1497#define isIDFIRST(c) isIDFIRST_A(c)
1498#define isLOWER(c) isLOWER_A(c)
1499#define isPRINT(c) isPRINT_A(c)
779cf272 1500#define isPSXSPC_A(c) isSPACE_A(c)
e66b99e9 1501#define isPSXSPC(c) isPSXSPC_A(c)
779cf272 1502#define isPSXSPC_L1(c) isSPACE_L1(c)
e66b99e9
KW
1503#define isPUNCT(c) isPUNCT_A(c)
1504#define isSPACE(c) isSPACE_A(c)
1505#define isUPPER(c) isUPPER_A(c)
1506#define isWORDCHAR(c) isWORDCHAR_A(c)
1507#define isXDIGIT(c) isXDIGIT_A(c)
1508
1509/* ASCII casing. These could also be written as
1510 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1511 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1512 which uses table lookup and mask instead of subtraction. (This would
c5e9991e
KW
1513 work because the _MOD does not apply in the ASCII range).
1514
1515 These actually are UTF-8 invariant casing, not just ASCII, as any non-ASCII
1516 UTF-8 invariants are neither upper nor lower. (Only on EBCDIC platforms are
1517 there non-ASCII invariants, and all of them are controls.) */
68067e4e
DM
1518#define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1519#define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
bbce6d69 1520
25200305
KW
1521/* In the ASCII range, these are equivalent to what they're here defined to be.
1522 * But by creating these definitions, other code doesn't have to be aware of
c5e9991e
KW
1523 * this detail. Actually this works for all UTF-8 invariants, not just the
1524 * ASCII range. (EBCDIC platforms can have non-ASCII invariants.) */
25200305 1525#define toFOLD(c) toLOWER(c)
25200305
KW
1526#define toTITLE(c) toUPPER(c)
1527
c753c8d3
KW
1528#define toLOWER_A(c) toLOWER(c)
1529#define toUPPER_A(c) toUPPER(c)
25200305
KW
1530#define toFOLD_A(c) toFOLD(c)
1531#define toTITLE_A(c) toTITLE(c)
1a0901db 1532
4650c663 1533/* Use table lookup for speed; returns the input itself if is out-of-range */
b2bf251f 1534#define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1535 ? (c) \
f4cd282c 1536 : PL_latin1_lc[ (U8) (c) ])
c753c8d3
KW
1537#define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1538
1a0901db 1539/* Modified uc. Is correct uc except for three non-ascii chars which are
4650c663
KW
1540 * all mapped to one of them, and these need special handling; returns the
1541 * input itself if is out-of-range */
b2bf251f 1542#define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1543 ? (c) \
f4cd282c 1544 : PL_mod_latin1_uc[ (U8) (c) ])
31f05a37 1545#define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
84061b6a 1546
beab9ebe
KW
1547/* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1548
1549/* For internal core Perl use only: the base macro for defining macros like
1550 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
31f05a37
KW
1551 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1552 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1553 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1554 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1555 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1556 * code points under locale rules, use the macros based on _generic_LC_uvchr
1557 * instead of this. */
beab9ebe
KW
1558#define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1559 (! FITS_IN_8_BITS(c) \
1560 ? 0 \
31f05a37
KW
1561 : IN_UTF8_CTYPE_LOCALE \
1562 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
beab9ebe
KW
1563 : cBOOL(non_utf8))
1564
1565/* For internal core Perl use only: a helper macro for defining macros like
1566 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1567 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1568 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1569 * _generic_LC_base for more info */
1570#define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1571 _generic_LC_base(c,utf8_locale_classnum, \
1572 non_utf8_func( (_LC_CAST) (c)))
1573
1574/* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1575 * 'c' is the platform's native underscore character */
1576#define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1577 _generic_LC_base(c, utf8_locale_classnum, \
1578 (non_utf8_func( (_LC_CAST) (c)) \
1579 || (char)(c) == '_'))
1580
1581/* These next three are also for internal core Perl use only: case-change
247985d4
KW
1582 * helper macros. The reason for using the PL_latin arrays is in case the
1583 * system function is defective; it ensures uniform results that conform to the
b257a28c 1584 * Unicod standard. It does not handle the anomalies in UTF-8 Turkic locales */
beab9ebe
KW
1585#define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1586 ? (c) \
31f05a37
KW
1587 : (IN_UTF8_CTYPE_LOCALE) \
1588 ? PL_latin1_lc[ (U8) (c) ] \
5a10328c 1589 : (cast)function((cast)(c)))
beab9ebe 1590
31f05a37
KW
1591/* Note that the result can be larger than a byte in a UTF-8 locale. It
1592 * returns a single value, so can't adequately return the upper case of LATIN
1593 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1594 * values "SS"); instead it asserts against that under DEBUGGING, and
b257a28c
KW
1595 * otherwise returns its input. It does not handle the anomalies in UTF-8
1596 * Turkic locales. */
beab9ebe
KW
1597#define _generic_toUPPER_LC(c, function, cast) \
1598 (! FITS_IN_8_BITS(c) \
1599 ? (c) \
31f05a37 1600 : ((! IN_UTF8_CTYPE_LOCALE) \
b7d90381 1601 ? (cast)function((cast)(c)) \
31f05a37
KW
1602 : ((((U8)(c)) == MICRO_SIGN) \
1603 ? GREEK_CAPITAL_LETTER_MU \
1604 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1605 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1606 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1607 ? (__ASSERT_(0) (c)) \
1608 : PL_mod_latin1_uc[ (U8) (c) ])))))
1609
1610/* Note that the result can be larger than a byte in a UTF-8 locale. It
1611 * returns a single value, so can't adequately return the fold case of LATIN
1612 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1613 * values "ss"); instead it asserts against that under DEBUGGING, and
b257a28c
KW
1614 * otherwise returns its input. It does not handle the anomalies in UTF-8
1615 * Turkic locales */
beab9ebe 1616#define _generic_toFOLD_LC(c, function, cast) \
4d7db1b9 1617 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
31f05a37 1618 ? GREEK_SMALL_LETTER_MU \
4d7db1b9
KW
1619 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1620 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1621 _generic_toLOWER_LC(c, function, cast)))
beab9ebe 1622
84061b6a 1623/* Use the libc versions for these if available. */
f05550c0 1624#if defined(HAS_ISASCII)
84061b6a
KW
1625# define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1626#else
1627# define isASCII_LC(c) isASCII(c)
1628#endif
1629
f05550c0 1630#if defined(HAS_ISBLANK)
84061b6a 1631# define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
95f34b6f 1632#else /* Unlike isASCII, varies if in a UTF-8 locale */
278b2b56 1633# define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
84061b6a
KW
1634#endif
1635
f05550c0 1636#define _LC_CAST U8
bbce6d69 1637
f05550c0 1638#ifdef WIN32
375f5f06
KW
1639 /* The Windows functions don't bother to follow the POSIX standard, which
1640 * for example says that something can't both be a printable and a control.
1641 * But Windows treats the \t control as a printable, and does such things
1642 * as making superscripts into both digits and punctuation. This tames
1643 * these flaws by assuming that the definitions of both controls and space
1644 * are correct, and then making sure that other definitions don't have
1645 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1646 * Not all possible weirdnesses are checked for, just the ones that were
1647 * detected on actual Microsoft code pages */
1648
b7d90381
KW
1649# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1650# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1651
1652# define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1653 && isALPHANUMERIC_LC(c))
1654# define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1655 ! isPUNCT_LC(c))
1656# define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1657 isALPHANUMERIC_LC(c))
1658# define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1659# define isIDFIRST_LC(c) (((c) == '_') \
1660 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1661# define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1662# define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1663# define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1664# define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
f05550c0 1665# define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
b7d90381
KW
1666# define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1667 && isALPHANUMERIC_LC(c))
f05550c0
BF
1668
1669# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1670# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1671# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1672
1673#elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
4650c663 1674 /* For most other platforms */
beab9ebe 1675
f05550c0
BF
1676# define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1677# define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1678# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1679# define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1680# define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1681# define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1682# define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1683# define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1684# define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1685# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1686# define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1687# define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1688# define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1689
1690
1691# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1692# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1693# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1694
1695#else /* The final fallback position */
1696
b7d90381
KW
1697# define isALPHA_LC(c) (isascii(c) && isalpha(c))
1698# define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1699# define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1700# define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1701# define isGRAPH_LC(c) (isascii(c) && isgraph(c))
f05550c0 1702# define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
b7d90381
KW
1703# define isLOWER_LC(c) (isascii(c) && islower(c))
1704# define isPRINT_LC(c) (isascii(c) && isprint(c))
1705# define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1706# define isSPACE_LC(c) (isascii(c) && isspace(c))
1707# define isUPPER_LC(c) (isascii(c) && isupper(c))
f05550c0 1708# define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
b7d90381 1709# define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
f05550c0
BF
1710
1711# define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1712# define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1713# define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
bbce6d69 1714
f05550c0 1715#endif
55204971 1716
eba68aa0
KW
1717#define isIDCONT(c) isWORDCHAR(c)
1718#define isIDCONT_A(c) isWORDCHAR_A(c)
1719#define isIDCONT_L1(c) isWORDCHAR_L1(c)
1720#define isIDCONT_LC(c) isWORDCHAR_LC(c)
13380643 1721#define isPSXSPC_LC(c) isSPACE_LC(c)
aaa51d5e 1722
4650c663 1723/* For internal core Perl use only: the base macros for defining macros like
d0da05db
KW
1724 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1725 * number defined earlier in this file. _generic_uvchr() is used for POSIX
4650c663
KW
1726 * classes where there is a macro or function 'above_latin1' that takes the
1727 * single argument 'c' and returns the desired value. These exist for those
1728 * classes which have simple definitions, avoiding the overhead of a hash
d0da05db 1729 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
4650c663 1730 * for classes where that overhead is faster than a direct lookup.
d0da05db 1731 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
4650c663
KW
1732 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1733 * have duplicate checks here, so could create versions of the macros that
1734 * don't, but experiments show that gcc optimizes them out anyway. */
66c17564
KW
1735
1736/* Note that all ignore 'use bytes' */
1e222e4f
KW
1737#define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1738 ? _generic_isCC(c, classnum) \
cd500f2f 1739 : above_latin1(c))
1e222e4f
KW
1740#define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1741 ? _generic_isCC(c, classnum) \
922e8cb4 1742 : _is_uni_FOO(classnum, c))
d0da05db
KW
1743#define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1744#define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1745#define isASCII_uvchr(c) isASCII(c)
1746#define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1747#define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1748#define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1749#define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1e222e4f
KW
1750#define isIDCONT_uvchr(c) \
1751 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1752#define isIDFIRST_uvchr(c) \
1753 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
d0da05db
KW
1754#define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1755#define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1756
1757#define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1758#define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1759#define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1760
1761#define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1762#define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1763#define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1764#define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1765
1766#define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1767#define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1768#define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1769#define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1770
1771/* For backwards compatibility, even though '_uni' should mean official Unicode
1772 * code points, in Perl it means native for those below 256 */
1773#define isALPHA_uni(c) isALPHA_uvchr(c)
1774#define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1775#define isASCII_uni(c) isASCII_uvchr(c)
1776#define isBLANK_uni(c) isBLANK_uvchr(c)
1777#define isCNTRL_uni(c) isCNTRL_uvchr(c)
1778#define isDIGIT_uni(c) isDIGIT_uvchr(c)
1779#define isGRAPH_uni(c) isGRAPH_uvchr(c)
1780#define isIDCONT_uni(c) isIDCONT_uvchr(c)
1781#define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1782#define isLOWER_uni(c) isLOWER_uvchr(c)
1783#define isPRINT_uni(c) isPRINT_uvchr(c)
1784#define isPUNCT_uni(c) isPUNCT_uvchr(c)
1785#define isSPACE_uni(c) isSPACE_uvchr(c)
1786#define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1787#define isUPPER_uni(c) isUPPER_uvchr(c)
1788#define isVERTWS_uni(c) isVERTWS_uvchr(c)
1789#define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1790#define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1791#define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1792#define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1793#define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1794#define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
a0ed51b3 1795
4650c663
KW
1796/* For internal core Perl use only: the base macros for defining macros like
1797 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
d0da05db 1798 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
4650c663 1799 * simple class definitions; the other for more complex. These are like
d0da05db 1800 * _generic_uvchr, so see it for more info. */
cd500f2f
KW
1801#define _generic_LC_uvchr(latin1, above_latin1, c) \
1802 (c < 256 ? latin1(c) : above_latin1(c))
1803#define _generic_LC_swash_uvchr(latin1, classnum, c) \
1804 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1805
1806#define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1807#define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1808 _CC_ALPHANUMERIC, c)
b7d90381
KW
1809#define isASCII_LC_uvchr(c) isASCII_LC(c)
1810#define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1811 is_HORIZWS_cp_high, c)
feeab5a9 1812#define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
cd500f2f
KW
1813#define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1814#define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
b7d90381 1815#define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
eba68aa0 1816 _is_uni_perl_idcont, c)
b7d90381 1817#define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
cd500f2f
KW
1818 _is_uni_perl_idstart, c)
1819#define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1820#define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
b7d90381 1821#define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
cd500f2f 1822#define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
b7d90381 1823#define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
509fb054 1824 is_XPERLSPACE_cp_high, c)
cd500f2f 1825#define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
b7d90381 1826#define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
cd500f2f 1827 _CC_WORDCHAR, c)
b7d90381
KW
1828#define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1829 is_XDIGIT_cp_high, c)
e712593e 1830
b7d90381 1831#define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
aaa51d5e 1832
4650c663
KW
1833/* For internal core Perl use only: the base macros for defining macros like
1834 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1835 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1836 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1837 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1838 * same representation whether utf8 or not. Note that it assumes that the utf8
1839 * has been validated, and ignores 'use bytes' */
34aeb2e9
KW
1840#define _base_generic_utf8(enum_name, name, p, use_locale ) \
1841 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1842 (const U8 *) p, \
1843 "is" STRINGIFY(name) "_utf8", \
1844 "is" STRINGIFY(name) "_utf8_safe", \
1845 1, use_locale, __FILE__,__LINE__)
1846
1847#define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
da8c1a98
KW
1848
1849/* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1850 * they don't otherwise go out of their way to look for malformed UTF-8. If
1851 * they can return accurate results without knowing if the input is otherwise
1852 * malformed, they do so. For example isASCII is accurate in spite of any
1853 * non-length malformations because it looks only at a single byte. Likewise
1854 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1855 * variant ones return FALSE. But, if the input has to be well-formed in order
1856 * for the results to be accurate, the macros will test and if malformed will
1857 * call a routine to die
1858 *
1859 * Except for toke.c, the macros do assume that e > p, asserting that on
1860 * DEBUGGING builds. Much code that calls these depends on this being true,
1861 * for other reasons. toke.c is treated specially as using the regular
1862 * assertion breaks it in many ways. All strings that these operate on there
1863 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1864 * bunch of code in toke.c assumes that this is true, so the assertion allows
1865 * for that */
1866#ifdef PERL_IN_TOKE_C
1867# define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1868#else
1869# define _utf8_safe_assert(p,e) ((e) > (p))
1870#endif
1871
1872#define _generic_utf8_safe(classnum, p, e, above_latin1) \
1873 (__ASSERT_(_utf8_safe_assert(p, e)) \
1874 (UTF8_IS_INVARIANT(*(p))) \
1875 ? _generic_isCC(*(p), classnum) \
1876 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1877 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1878 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1879 classnum) \
1880 : (_force_out_malformed_utf8_message( \
1881 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1882 : above_latin1))
b7d90381
KW
1883/* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1884 * 'above_latin1' can be a macro */
da8c1a98
KW
1885#define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1886 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1887#define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1888 _generic_utf8_safe(classnum, p, e, \
1889 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1890 ? (_force_out_malformed_utf8_message( \
1891 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1892 : above_latin1(p)))
4650c663 1893/* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
cd500f2f 1894 * 'above_latin1' parameter */
da8c1a98
KW
1895#define _generic_swash_utf8_safe(classnum, p, e) \
1896_generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
922e8cb4 1897
cc8ab7c0 1898/* Like the above, but should be used only when it is known that there are no
ff7ecfc3
KW
1899 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1900 * class is TRUE for. Hence it can skip the tests for this range.
1901 * 'above_latin1' should include its arguments */
da8c1a98
KW
1902#define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1903 (__ASSERT_(_utf8_safe_assert(p, e)) \
1904 (UTF8_IS_INVARIANT(*(p))) \
1905 ? _generic_isCC(*(p), classnum) \
1906 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1907 ? 0 /* Note that doesn't check validity for latin1 */ \
1908 : above_latin1)
1909
84238efa 1910
34aeb2e9
KW
1911#define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1912#define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1913#define isASCII_utf8(p) _generic_utf8(ASCII, p)
1914#define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1915#define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1916#define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1917#define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1918#define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1919#define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1920#define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1921#define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1922#define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1923#define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1924#define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1925#define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1926#define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1927#define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1928#define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
e8fa43e2 1929
da8c1a98
KW
1930#define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1931#define isALPHANUMERIC_utf8_safe(p, e) \
1932 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1933#define isASCII_utf8_safe(p, e) \
1934 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1935 * works */ \
1936 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1937#define isBLANK_utf8_safe(p, e) \
1938 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1939
e8fa43e2
KW
1940#ifdef EBCDIC
1941 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1942 * more efficient macro instead of the more general one */
da8c1a98 1943# define isCNTRL_utf8_safe(p, e) \
56d02b8c 1944 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
e8fa43e2 1945#else
da8c1a98 1946# define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
e8fa43e2
KW
1947#endif
1948
da8c1a98
KW
1949#define isDIGIT_utf8_safe(p, e) \
1950 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1951 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1952#define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1953#define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1954 _is_utf8_perl_idcont_with_len, p, e)
e5dcd934 1955
c11ff943
KW
1956/* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1957 * IDFIRST is an alnum. See
c8362b00 1958 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
f91dcd13
KW
1959 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1960 * This used to be not the XID version, but we decided to go with the more
1961 * modern Unicode definition */
da8c1a98
KW
1962#define isIDFIRST_utf8_safe(p, e) \
1963 _generic_func_utf8_safe(_CC_IDFIRST, \
1964 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1965
1966#define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1967#define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1968#define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1969#define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1970#define isSPACE_utf8_safe(p, e) \
1971 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1972#define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1973#define isVERTWS_utf8_safe(p, e) \
1974 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1975#define isWORDCHAR_utf8_safe(p, e) \
1976 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1977#define isXDIGIT_utf8_safe(p, e) \
1978 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
1979 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1980 ? (_force_out_malformed_utf8_message( \
1981 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1982 : is_XDIGIT_high(p)))
a0ed51b3 1983
7d2b61dd 1984#define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
509fb054
KW
1985#define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
1986#define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
1987#define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
2e8adce6 1988
567b353c 1989/* For internal core use only, subject to change */
607313a1
KW
1990#define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
1991#define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
1992#define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
1993#define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
a1a5ec35
KW
1994
1995#define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
1996#define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
1997#define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
1998#define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
567b353c 1999
4650c663
KW
2000/* For internal core Perl use only: the base macros for defining macros like
2001 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
2002 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
2003 * 'macro' parameter */
34aeb2e9
KW
2004#define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
2005
2006#define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
2007#define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
2008#define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
2009#define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
2010#define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
2011#define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
2012#define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
2013#define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
2014#define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
2015#define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
2016#define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
2017#define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
2018#define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
2019#define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
2020#define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
2021#define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
2022#define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
2023
da8c1a98
KW
2024/* For internal core Perl use only: the base macros for defining macros like
2025 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
2026 * point in 'p' is within the 0-255 range, it uses locale rules from the
2027 * passed-in 'macro' parameter */
2028#define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
2029 (__ASSERT_(_utf8_safe_assert(p, e)) \
2030 (UTF8_IS_INVARIANT(*(p))) \
2031 ? macro(*(p)) \
2032 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2033 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2034 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2035 : (_force_out_malformed_utf8_message( \
2036 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2037 : above_latin1))
2038
2039#define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2040 _generic_LC_utf8_safe(macro, p, e, \
2041 _is_utf8_FOO_with_len(classnum, p, e))
2042
2043#define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2044 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2045
2046#define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2047 _generic_LC_utf8_safe(classnum, p, e, \
2048 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2049 ? (_force_out_malformed_utf8_message( \
2050 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2051 : above_latin1(p)))
2052
2053#define isALPHANUMERIC_LC_utf8_safe(p, e) \
2054 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2055 _CC_ALPHANUMERIC, p, e)
2056#define isALPHA_LC_utf8_safe(p, e) \
2057 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2058#define isASCII_LC_utf8_safe(p, e) \
2059 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2060#define isBLANK_LC_utf8_safe(p, e) \
2061 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2062#define isCNTRL_LC_utf8_safe(p, e) \
2063 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2064#define isDIGIT_LC_utf8_safe(p, e) \
2065 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2066#define isGRAPH_LC_utf8_safe(p, e) \
2067 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2068#define isIDCONT_LC_utf8_safe(p, e) \
2069 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2070 _is_utf8_perl_idcont_with_len, p, e)
2071#define isIDFIRST_LC_utf8_safe(p, e) \
2072 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2073 _is_utf8_perl_idstart_with_len, p, e)
2074#define isLOWER_LC_utf8_safe(p, e) \
2075 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2076#define isPRINT_LC_utf8_safe(p, e) \
2077 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2078#define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2079#define isPUNCT_LC_utf8_safe(p, e) \
2080 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2081#define isSPACE_LC_utf8_safe(p, e) \
2082 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2083#define isUPPER_LC_utf8_safe(p, e) \
2084 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2085#define isWORDCHAR_LC_utf8_safe(p, e) \
2086 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2087#define isXDIGIT_LC_utf8_safe(p, e) \
2088 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
aaa51d5e 2089
fbc19f27
KW
2090/* Macros for backwards compatibility and for completeness when the ASCII and
2091 * Latin1 values are identical */
b7d90381
KW
2092#define isALPHAU(c) isALPHA_L1(c)
2093#define isDIGIT_L1(c) isDIGIT_A(c)
2094#define isOCTAL(c) isOCTAL_A(c)
2095#define isOCTAL_L1(c) isOCTAL_A(c)
2096#define isXDIGIT_L1(c) isXDIGIT_A(c)
2097#define isALNUM(c) isWORDCHAR(c)
a377c856 2098#define isALNUM_A(c) isALNUM(c)
b7d90381
KW
2099#define isALNUMU(c) isWORDCHAR_L1(c)
2100#define isALNUM_LC(c) isWORDCHAR_LC(c)
2101#define isALNUM_uni(c) isWORDCHAR_uni(c)
2e28f0b9 2102#define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
b7d90381
KW
2103#define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2104#define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2105#define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2106#define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2107#define isALNUMC(c) isALPHANUMERIC(c)
2108#define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2109#define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
15861f94 2110#define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
b7d90381 2111#define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
15861f94 2112#define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
fbc19f27 2113
2bd1cbf6
KW
2114/* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2115 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2116 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2117 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2118 * the outlier from the block that contains the other controls, just like
2119 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2120 * block. If it weren't special cased, it would yield a non-control.
88794300
KW
2121 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2122 * etc. */
2bd1cbf6 2123#ifndef EBCDIC
75763b3a 2124# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2bd1cbf6 2125#else
75763b3a
KW
2126# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2127 ((isPRINT_A(c)) \
2128 ? (UNLIKELY((c) == '?') \
2129 ? QUESTION_MARK_CTRL \
2130 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2131 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2132 ? '?' \
2133 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2bd1cbf6 2134#endif
bbce6d69 2135
dea28490
JJ
2136/* Line numbers are unsigned, 32 bits. */
2137typedef U32 line_t;
e5dcd934 2138#define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
378cc40b 2139
91152fc1
DG
2140/* Helpful alias for version prescan */
2141#define is_LAX_VERSION(a,b) \
2142 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2143
2144#define is_STRICT_VERSION(a,b) \
2145 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2146
2147#define BADVERSION(a,b,c) \
2148 if (b) { \
2149 *b = c; \
2150 } \
2151 return a;
8c52afec 2152
cb27eebd
KW
2153/* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2154 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2155 * which is advanced. The input is validated only by an assert() in DEBUGGING
2156 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2157 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2158#define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2159 ? (c) \
2160 : ((c) + 9))))
902008b8
KW
2161#define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2162 ? (*(s)++) \
2163 : (*(s)++ + 9))))
95a59cab 2164
cb27eebd
KW
2165/* Converts a character known to represent an octal digit (0-7) to its numeric
2166 * value. The input is validated only by an assert() in DEBUGGING builds. In
2167 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2168#define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2169
305b8651
KW
2170/* Efficiently returns a boolean as to if two native characters are equivalent
2171 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2172 * the ALPHA in the name is to remind you of that. This is asserted() in
2173 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2174 * works (on valid input) for both non- and UTF-8-encoded bytes.
2175 *
2176 * When one of the inputs is a compile-time constant and gets folded by the
2177 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2178 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2179 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
96ca48da
KW
2180 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2181 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2182 * just a single 0, in the bit position where the upper- and lowercase differ.
2183 * */
305b8651
KW
2184#define isALPHA_FOLD_EQ(c1, c2) \
2185 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2186 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2187#define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2188
8e84507e 2189/*
ccfc67b7
JH
2190=head1 Memory Management
2191
a02a5408 2192=for apidoc Am|void|Newx|void* ptr|int nitems|type
954c1994
GS
2193The XSUB-writer's interface to the C C<malloc> function.
2194
596f7718 2195Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2196
c5008215
JC
2197In 5.9.3, Newx() and friends replace the older New() API, and drops
2198the first parameter, I<x>, a debug aid which allowed callers to identify
37b8b4c9 2199themselves. This aid has been superseded by a new build option,
d10b4965 2200PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
c5008215
JC
2201there for use in XS modules supporting older perls.
2202
a02a5408 2203=for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
954c1994 2204The XSUB-writer's interface to the C C<malloc> function, with
fbe13c60 2205cast. See also C<L</Newx>>.
954c1994 2206
596f7718 2207Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2208
a02a5408 2209=for apidoc Am|void|Newxz|void* ptr|int nitems|type
954c1994 2210The XSUB-writer's interface to the C C<malloc> function. The allocated
fbe13c60 2211memory is zeroed with C<memzero>. See also C<L</Newx>>.
a02a5408 2212
596f7718 2213Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2214
954c1994
GS
2215=for apidoc Am|void|Renew|void* ptr|int nitems|type
2216The XSUB-writer's interface to the C C<realloc> function.
2217
596f7718 2218Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2219
954c1994
GS
2220=for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2221The XSUB-writer's interface to the C C<realloc> function, with
2222cast.
2223
596f7718 2224Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2225
49b8b560 2226=for apidoc Am|void|Safefree|void* ptr
954c1994
GS
2227The XSUB-writer's interface to the C C<free> function.
2228
596f7718 2229This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
0d7b2759 2230
954c1994
GS
2231=for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2232The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
926bb54c 2233source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2234C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
954c1994 2235
e90e2364 2236=for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
796b6530 2237Like C<Move> but returns C<dest>. Useful
72d33970 2238for encouraging compilers to tail-call
e90e2364
NC
2239optimise.
2240
954c1994
GS
2241=for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2242The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
926bb54c 2243source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2244C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
954c1994 2245
e90e2364
NC
2246=for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2247
796b6530 2248Like C<Copy> but returns C<dest>. Useful
72d33970 2249for encouraging compilers to tail-call
e90e2364
NC
2250optimise.
2251
954c1994
GS
2252=for apidoc Am|void|Zero|void* dest|int nitems|type
2253
2254The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2255destination, C<nitems> is the number of items, and C<type> is the type.
2256
e90e2364
NC
2257=for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2258
72d33970
FC
2259Like C<Zero> but returns dest. Useful
2260for encouraging compilers to tail-call
e90e2364
NC
2261optimise.
2262
da5d8dbb 2263=for apidoc Am|void|StructCopy|type *src|type *dest|type
4375e838 2264This is an architecture-independent macro to copy one structure to another.
954c1994 2265
7e337ee0
JH
2266=for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2267
2268Fill up memory with a byte pattern (a byte repeated over and over
2269again) that hopefully catches attempts to access uninitialized memory.
2270
2271=for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2272
2273PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2274
1c12ffb4 2275=for apidoc Am|void|PoisonFree|void* dest|int nitems|type
7e337ee0
JH
2276
2277PoisonWith(0xEF) for catching access to freed memory.
2278
9965345d
JH
2279=for apidoc Am|void|Poison|void* dest|int nitems|type
2280
7e337ee0 2281PoisonWith(0xEF) for catching access to freed memory.
9965345d
JH
2282
2283=cut */
954c1994 2284
561b68a9
SH
2285/* Maintained for backwards-compatibility only. Use newSV() instead. */
2286#ifndef PERL_CORE
ff06c60c 2287#define NEWSV(x,len) newSV(len)
561b68a9 2288#endif
ff06c60c 2289
b7112dce 2290#define MEM_SIZE_MAX ((MEM_SIZE)-1)
19a94d75 2291
a500027b 2292#define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
e6bdf523 2293
27d5b266 2294#ifdef PERL_MALLOC_WRAP
e6bdf523
DM
2295
2296/* This expression will be constant-folded at compile time. It checks
2297 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2298 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2299 * As well as avoiding the need for a run-time check in some cases, it's
2300 * designed to avoid compiler warnings like:
2301 * comparison is always false due to limited range of data type
73e8ff00
DM
2302 * It's mathematically equivalent to
2303 * max(n) * sizeof(t) > MEM_SIZE_MAX
e6bdf523
DM
2304 */
2305
2306# define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
445198b9
LM
2307 ( sizeof(MEM_SIZE) < sizeof(n) \
2308 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
e6bdf523 2309
88f9f128 2310/* This is written in a slightly odd way to avoid various spurious
d98e5cde
DM
2311 * compiler warnings. We *want* to write the expression as
2312 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2313 * (for some compile-time constant C), but even when the LHS
2314 * constant-folds to false at compile-time, g++ insists on emitting
2315 * warnings about the RHS (e.g. "comparison is always false"), so instead
2316 * we write it as
e6bdf523 2317 *
d98e5cde 2318 * (cond ? n : X) > C
88f9f128 2319 *
d98e5cde
DM
2320 * where X is a constant with X > C always false. Choosing a value for X
2321 * is tricky. If 0, some compilers will complain about 0 > C always being
2322 * false; if 1, Coverity complains when n happens to be the constant value
2323 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2324 * for X and hope that nothing else whines.
e6bdf523
DM
2325 */
2326
2327# define _MEM_WRAP_WILL_WRAP(n,t) \
88f9f128
DM
2328 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2329 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
e6bdf523
DM
2330
2331# define MEM_WRAP_CHECK(n,t) \
2332 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2333 && (croak_memory_wrap(),0))
2334
2335# define MEM_WRAP_CHECK_1(n,t,a) \
2336 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2337 && (Perl_croak_nocontext("%s",(a)),0))
2338
814eedc8
DD
2339/* "a" arg must be a string literal */
2340# define MEM_WRAP_CHECK_s(n,t,a) \
2341 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2342 && (Perl_croak_nocontext("" a ""),0))
2343
8b44ba4c 2344#define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
27d5b266 2345
a500027b 2346#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))
27d5b266
JH
2347#else
2348
410319be
NC
2349#define MEM_WRAP_CHECK(n,t)
2350#define MEM_WRAP_CHECK_1(n,t,a)
814eedc8 2351#define MEM_WRAP_CHECK_s(n,t,a)
8b44ba4c
NC
2352#define MEM_WRAP_CHECK_(n,t)
2353
a500027b 2354#define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
27d5b266 2355
1936d2a7 2356#endif
8b44ba4c 2357
fe4f188c 2358#ifdef PERL_MEM_LOG
46c6c7e2 2359/*
9f653bb5 2360 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
46c6c7e2
JH
2361 * go through functions, which are handy for debugging breakpoints, but
2362 * which more importantly get the immediate calling environment (file and
e352bcff
JH
2363 * line number, and C function name if available) passed in. This info can
2364 * then be used for logging the calls, for which one gets a sample
73d1d973 2365 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
3609ea0d 2366 *
46c6c7e2 2367 * Known problems:
94e892a6 2368 * - not all memory allocs get logged, only those
46c6c7e2 2369 * that go through Newx() and derivatives (while all
94e892a6 2370 * Safefrees do get logged)
46c6c7e2
JH
2371 * - __FILE__ and __LINE__ do not work everywhere
2372 * - __func__ or __FUNCTION__ even less so
2373 * - I think more goes on after the perlio frees but
2374 * the thing is that STDERR gets closed (as do all
2375 * the file descriptors)
2376 * - no deeper calling stack than the caller of the Newx()
2377 * or the kind, but do I look like a C reflection/introspection
2378 * utility to you?
2379 * - the function prototypes for the logging functions
2380 * probably should maybe be somewhere else than handy.h
2381 * - one could consider inlining (macrofying) the logging
2382 * for speed, but I am too lazy
2383 * - one could imagine recording the allocations in a hash,
2384 * (keyed by the allocation address?), and maintain that
2385 * through reallocs and frees, but how to do that without
2386 * any News() happening...?
73d1d973 2387 * - lots of -Ddefines to get useful/controllable output
b953482e 2388 * - lots of ENV reads
46c6c7e2
JH
2389 */
2390
0b0ab801 2391# ifdef PERL_CORE
73d1d973 2392# ifndef PERL_MEM_LOG_NOIMPL
0b0ab801
MHM
2393enum mem_log_type {
2394 MLT_ALLOC,
2395 MLT_REALLOC,
d7a2c63c
MHM
2396 MLT_FREE,
2397 MLT_NEW_SV,
2398 MLT_DEL_SV
0b0ab801
MHM
2399};
2400# endif
12754f92 2401# if defined(PERL_IN_SV_C) /* those are only used in sv.c */
d7a2c63c
MHM
2402void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2403void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
12754f92 2404# endif
0b0ab801
MHM
2405# endif
2406
fe4f188c
JH
2407#endif
2408
2409#ifdef PERL_MEM_LOG
d1401ee9
MHM
2410#define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2411#define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
46c6c7e2 2412#define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
fe4f188c
JH
2413#endif
2414
2415#ifndef MEM_LOG_ALLOC
2416#define MEM_LOG_ALLOC(n,t,a) (a)
2417#endif
2418#ifndef MEM_LOG_REALLOC
2419#define MEM_LOG_REALLOC(n,t,v,a) (a)
2420#endif
2421#ifndef MEM_LOG_FREE
2422#define MEM_LOG_FREE(a) (a)
2423#endif
2424
d1401ee9
MHM
2425#define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2426#define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2427#define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
a6f6820f
NC
2428
2429#ifndef PERL_CORE
a02a5408
JC
2430/* pre 5.9.x compatibility */
2431#define New(x,v,n,t) Newx(v,n,t)
2432#define Newc(x,v,n,t,c) Newxc(v,n,t,c)
4541904d 2433#define Newz(x,v,n,t) Newxz(v,n,t)
a6f6820f 2434#endif
a02a5408 2435
ff68c719 2436#define Renew(v,n,t) \
d1401ee9 2437 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
ff68c719 2438#define Renewc(v,n,t,c) \
d1401ee9 2439 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
94010e71
NC
2440
2441#ifdef PERL_POISON
2442#define Safefree(d) \
06c0cc96 2443 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
94010e71 2444#else
fe4f188c 2445#define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
94010e71 2446#endif
55497cff 2447
dbb57106
YO
2448/* assert that a valid ptr has been supplied - use this instead of assert(ptr) *
2449 * as it handles cases like constant string arguments without throwing warnings *
2450 * the cast is required, as is the inequality check, to avoid warnings */
45908e4d 2451#define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
55497cff 2452
45908e4d
YO
2453
2454#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)))
2455#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)))
2456#define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2457
bdd1531d 2458/* Like above, but returns a pointer to 'd' */
45908e4d
YO
2459#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)))
2460#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)))
45908e4d 2461#define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
e90e2364 2462
7e337ee0
JH
2463#define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2464#define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2465#define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2466#define Poison(d,n,t) PoisonFree(d,n,t)
27d5b266 2467
caa674f3
DD
2468#ifdef PERL_POISON
2469# define PERL_POISON_EXPR(x) x
2470#else
2471# define PERL_POISON_EXPR(x)
2472#endif
2473
ff68c719 2474#define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2cc61e15 2475
c3caa5c3
JH
2476/* C_ARRAY_LENGTH is the number of elements in the C array (so you
2477 * want your zero-based indices to be less than but not equal to).
2478 *
2479 * C_ARRAY_END is one past the last: half-open/half-closed range,
2480 * not last-inclusive range. */
622913ab 2481#define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
c3caa5c3 2482#define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
622913ab 2483
2cc61e15
DD
2484#ifdef NEED_VA_COPY
2485# ifdef va_copy
2486# define Perl_va_copy(s, d) va_copy(d, s)
07798b17
AC
2487# elif defined(__va_copy)
2488# define Perl_va_copy(s, d) __va_copy(d, s)
2cc61e15 2489# else
07798b17 2490# define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2cc61e15
DD
2491# endif
2492#endif
2493
472d47bc
SB
2494/* convenience debug macros */
2495#ifdef USE_ITHREADS
2496#define pTHX_FORMAT "Perl interpreter: 0x%p"
2497#define pTHX__FORMAT ", Perl interpreter: 0x%p"
f54cb97a
AL
2498#define pTHX_VALUE_ (void *)my_perl,
2499#define pTHX_VALUE (void *)my_perl
2500#define pTHX__VALUE_ ,(void *)my_perl,
2501#define pTHX__VALUE ,(void *)my_perl
472d47bc 2502#else
3609ea0d 2503#define pTHX_FORMAT
472d47bc 2504#define pTHX__FORMAT
3609ea0d 2505#define pTHX_VALUE_
472d47bc 2506#define pTHX_VALUE
3609ea0d 2507#define pTHX__VALUE_
472d47bc
SB
2508#define pTHX__VALUE
2509#endif /* USE_ITHREADS */
3609ea0d 2510
2acdbac1
NC
2511/* Perl_deprecate was not part of the public API, and did not have a deprecate()
2512 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2513 CPAN::Unpack show any users outside the core. */
2514#ifdef PERL_CORE
dc6e8de0
A
2515# define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2516 "Use of " s " is deprecated")
c9680906
A
2517# define deprecate_disappears_in(when,message) \
2518 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2519 message ", and will disappear in Perl " when)
ac641426
A
2520# define deprecate_fatal_in(when,message) \
2521 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2522 message ". Its use will be fatal in Perl " when)
2acdbac1
NC
2523#endif
2524
dfff4baf
BF
2525/* Internal macros to deal with gids and uids */
2526#ifdef PERL_CORE
2527
2528# if Uid_t_size > IVSIZE
2529# define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2530# define SvUID(sv) SvNV(sv)
07798b17
AC
2531# elif Uid_t_sign <= 0
2532# define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2533# define SvUID(sv) SvIV(sv)
dfff4baf 2534# else
07798b17
AC
2535# define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2536# define SvUID(sv) SvUV(sv)
dfff4baf
BF
2537# endif /* Uid_t_size */
2538
2539# if Gid_t_size > IVSIZE
2540# define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2541# define SvGID(sv) SvNV(sv)
07798b17
AC
2542# elif Gid_t_sign <= 0
2543# define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2544# define SvGID(sv) SvIV(sv)
dfff4baf 2545# else
07798b17
AC
2546# define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2547# define SvGID(sv) SvUV(sv)
dfff4baf
BF
2548# endif /* Gid_t_size */
2549
2550#endif
2551
6a5bc5ac 2552#endif /* PERL_HANDY_H_ */
9d745869 2553
e9a8c099 2554/*
14d04a33 2555 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 2556 */