This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perlapi: Document __ASSERT_
[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
25ba28ce
KW
100/*
101=for apidoc Am|bool|cBOOL|bool expr
102
103Cast-to-bool. A simple S<C<(bool) I<expr>>> cast may not do the right thing:
104if C<bool> is defined as C<char>, for example, then the cast from C<int> is
105implementation-defined.
106
107C<(bool)!!(cbool)> in a ternary triggers a bug in xlc on AIX
108
109=cut
110*/
18f5643b 111#define cBOOL(cbool) ((cbool) ? (bool)1 : (bool)0)
f2338a2e 112
46c6c7e2 113/* Try to figure out __func__ or __FUNCTION__ equivalent, if any.
e352bcff
JH
114 * XXX Should really be a Configure probe, with HAS__FUNCTION__
115 * and FUNCTION__ as results.
116 * XXX Similarly, a Configure probe for __FILE__ and __LINE__ is needed. */
46c6c7e2
JH
117#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || (defined(__SUNPRO_C)) /* C99 or close enough. */
118# define FUNCTION__ __func__
07798b17
AC
119#elif (defined(USING_MSVC6)) || /* MSVC6 has neither __func__ nor __FUNCTION and no good workarounds, either. */ \
120 (defined(__DECC_VER)) /* Tru64 or VMS, and strict C89 being used, but not modern enough cc (in Tur64, -c99 not known, only -std1). */
121# define FUNCTION__ ""
46c6c7e2 122#else
07798b17 123# define FUNCTION__ __FUNCTION__ /* Common extension. */
46c6c7e2
JH
124#endif
125
27d4fb96
PP
126/* XXX A note on the perl source internal type system. The
127 original intent was that I32 be *exactly* 32 bits.
128
129 Currently, we only guarantee that I32 is *at least* 32 bits.
130 Specifically, if int is 64 bits, then so is I32. (This is the case
131 for the Cray.) This has the advantage of meshing nicely with
132 standard library calls (where we pass an I32 and the library is
133 expecting an int), but the disadvantage that an I32 is not 32 bits.
134 Andy Dougherty August 1996
24fef2a7 135
dc45a647
MB
136 There is no guarantee that there is *any* integral type with
137 exactly 32 bits. It is perfectly legal for a system to have
138 sizeof(short) == sizeof(int) == sizeof(long) == 8.
693762b4 139
dc45a647
MB
140 Similarly, there is no guarantee that I16 and U16 have exactly 16
141 bits.
693762b4 142
8e84507e
NIS
143 For dealing with issues that may arise from various 32/64-bit
144 systems, we will ask Configure to check out
8175356b 145
3609ea0d
JH
146 SHORTSIZE == sizeof(short)
147 INTSIZE == sizeof(int)
148 LONGSIZE == sizeof(long)
dc45a647 149 LONGLONGSIZE == sizeof(long long) (if HAS_LONG_LONG)
3609ea0d 150 PTRSIZE == sizeof(void *)
dc45a647
MB
151 DOUBLESIZE == sizeof(double)
152 LONG_DOUBLESIZE == sizeof(long double) (if HAS_LONG_DOUBLE).
8175356b 153
27d4fb96
PP
154*/
155
69512466
JH
156#ifdef I_INTTYPES /* e.g. Linux has int64_t without <inttypes.h> */
157# include <inttypes.h>
dd0eed91
JH
158# ifdef INT32_MIN_BROKEN
159# undef INT32_MIN
160# define INT32_MIN (-2147483647-1)
161# endif
162# ifdef INT64_MIN_BROKEN
163# undef INT64_MIN
164# define INT64_MIN (-9223372036854775807LL-1)
165# endif
69512466
JH
166#endif
167
8175356b
JH
168typedef I8TYPE I8;
169typedef U8TYPE U8;
170typedef I16TYPE I16;
171typedef U16TYPE U16;
172typedef I32TYPE I32;
173typedef U32TYPE U32;
16d89be8 174
74b807c7 175#ifdef QUADKIND
8175356b
JH
176typedef I64TYPE I64;
177typedef U64TYPE U64;
16d89be8 178#endif
8175356b 179
d8668976 180#if defined(UINT8_MAX) && defined(INT16_MAX) && defined(INT32_MAX)
5ff3f7a4 181
5ff3f7a4
GS
182/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
183 Please search CHAR_MAX in perl.h for further details. */
184#define U8_MAX UINT8_MAX
185#define U8_MIN UINT8_MIN
186
5ff3f7a4
GS
187#define I16_MAX INT16_MAX
188#define I16_MIN INT16_MIN
189#define U16_MAX UINT16_MAX
190#define U16_MIN UINT16_MIN
191
5ff3f7a4
GS
192#define I32_MAX INT32_MAX
193#define I32_MIN INT32_MIN
0e983133
GS
194#ifndef UINT32_MAX_BROKEN /* e.g. HP-UX with gcc messes this up */
195# define U32_MAX UINT32_MAX
196#else
197# define U32_MAX 4294967295U
198#endif
5ff3f7a4
GS
199#define U32_MIN UINT32_MIN
200
201#else
202
5c9fa16e
KA
203/* I8_MAX and I8_MIN constants are not defined, as I8 is an ambiguous type.
204 Please search CHAR_MAX in perl.h for further details. */
27d4fb96
PP
205#define U8_MAX PERL_UCHAR_MAX
206#define U8_MIN PERL_UCHAR_MIN
79072805 207
27d4fb96
PP
208#define I16_MAX PERL_SHORT_MAX
209#define I16_MIN PERL_SHORT_MIN
210#define U16_MAX PERL_USHORT_MAX
211#define U16_MIN PERL_USHORT_MIN
79072805 212
c4f23d77 213#if LONGSIZE > 4
27d4fb96
PP
214# define I32_MAX PERL_INT_MAX
215# define I32_MIN PERL_INT_MIN
216# define U32_MAX PERL_UINT_MAX
217# define U32_MIN PERL_UINT_MIN
79072805 218#else
27d4fb96
PP
219# define I32_MAX PERL_LONG_MAX
220# define I32_MIN PERL_LONG_MIN
221# define U32_MAX PERL_ULONG_MAX
222# define U32_MIN PERL_ULONG_MIN
79072805
LW
223#endif
224
5ff3f7a4
GS
225#endif
226
247cee9f
KW
227/* These C99 typedefs are useful sometimes for, say, loop variables whose
228 * maximum values are small, but for which speed trumps size. If we have a C99
229 * compiler, use that. Otherwise, a plain 'int' should be good enough.
230 *
231 * Restrict these to core for now until we are more certain this is a good
232 * idea. */
233#if defined(PERL_CORE) || defined(PERL_EXT)
234# ifdef I_STDINT
235 typedef int_fast8_t PERL_INT_FAST8_T;
236 typedef uint_fast8_t PERL_UINT_FAST8_T;
237 typedef int_fast16_t PERL_INT_FAST16_T;
238 typedef uint_fast16_t PERL_UINT_FAST16_T;
239# else
240 typedef int PERL_INT_FAST8_T;
241 typedef unsigned int PERL_UINT_FAST8_T;
242 typedef int PERL_INT_FAST16_T;
243 typedef unsigned int PERL_UINT_FAST16_T;
244# endif
245#endif
246
464decb6
KW
247/* log(2) (i.e., log base 10 of 2) is pretty close to 0.30103, just in case
248 * anyone is grepping for it */
249#define BIT_DIGITS(N) (((N)*146)/485 + 1) /* log10(2) =~ 146/485 */
fc36a67e
PP
250#define TYPE_DIGITS(T) BIT_DIGITS(sizeof(T) * 8)
251#define TYPE_CHARS(T) (TYPE_DIGITS(T) + 2) /* sign, NUL */
252
88794300 253/* Unused by core; should be deprecated */
ff68c719 254#define Ctl(ch) ((ch) & 037)
8d063cd8 255
98fce2a4
KW
256#if defined(PERL_CORE) || defined(PERL_EXT)
257# ifndef MIN
258# define MIN(a,b) ((a) < (b) ? (a) : (b))
259# endif
260# ifndef MAX
261# define MAX(a,b) ((a) > (b) ? (a) : (b))
262# endif
263#endif
264
84ff4fa9
KW
265/* Returns a boolean as to whether the input unsigned number is a power of 2
266 * (2**0, 2**1, etc). In other words if it has just a single bit set.
267 * If not, subtracting 1 would leave the uppermost bit set, so the & would
268 * yield non-zero */
269#if defined(PERL_CORE) || defined(PERL_EXT)
011b1419 270# define isPOWER_OF_2(n) ((n) && ((n) & ((n)-1)) == 0)
84ff4fa9
KW
271#endif
272
8d9433eb
KW
273/*
274=for apidoc Am|void|__ASSERT_|bool expr
275
276This is a helper macro to avoid preprocessor issues, replaced by nothing
277unless under DEBUGGING, where it expands to an assert of its argument,
278followed by a comma (hence the comma operator). If we just used a straight
279assert(), we would get a comma with nothing before it when not DEBUGGING.
280
281=cut
282
283We also use empty definition under Coverity since the __ASSERT__
284checks often check for things that Really Cannot Happen, and Coverity
285detects that and gets all excited. */
3e94db23
JH
286
287#if defined(DEBUGGING) && !defined(__COVERITY__)
0f092d08
KW
288# define __ASSERT_(statement) assert(statement),
289#else
290# define __ASSERT_(statement)
291#endif
292
3fe05580 293/*
a4ee4fb5 294=head1 SV Manipulation Functions
3fe05580 295
1568d13a
Z
296=for apidoc Ama|SV*|newSVpvs|"literal string" s
297Like C<newSVpvn>, but takes a literal string instead of a
30a15352 298string/length pair.
3fe05580 299
1568d13a
Z
300=for apidoc Ama|SV*|newSVpvs_flags|"literal string" s|U32 flags
301Like C<newSVpvn_flags>, but takes a literal string instead of
30a15352 302a string/length pair.
84bafc02 303
1568d13a
Z
304=for apidoc Ama|SV*|newSVpvs_share|"literal string" s
305Like C<newSVpvn_share>, but takes a literal string instead of
30a15352 306a string/length pair and omits the hash parameter.
3fe05580 307
1568d13a
Z
308=for apidoc Am|void|sv_catpvs_flags|SV* sv|"literal string" s|I32 flags
309Like C<sv_catpvn_flags>, but takes a literal string instead
30a15352 310of a string/length pair.
9dcc53ea 311
1568d13a
Z
312=for apidoc Am|void|sv_catpvs_nomg|SV* sv|"literal string" s
313Like C<sv_catpvn_nomg>, but takes a literal string instead of
0c395ea5 314a string/length pair.
9dcc53ea 315
1568d13a
Z
316=for apidoc Am|void|sv_catpvs|SV* sv|"literal string" s
317Like C<sv_catpvn>, but takes a literal string instead of a
0c395ea5 318string/length pair.
3fe05580 319
1568d13a
Z
320=for apidoc Am|void|sv_catpvs_mg|SV* sv|"literal string" s
321Like C<sv_catpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
322string/length pair.
323
1568d13a
Z
324=for apidoc Am|void|sv_setpvs|SV* sv|"literal string" s
325Like C<sv_setpvn>, but takes a literal string instead of a
0c395ea5 326string/length pair.
3fe05580 327
1568d13a
Z
328=for apidoc Am|void|sv_setpvs_mg|SV* sv|"literal string" s
329Like C<sv_setpvn_mg>, but takes a literal string instead of a
9dcc53ea
Z
330string/length pair.
331
78342678 332=for apidoc Am|SV *|sv_setref_pvs|SV *const rv|const char *const classname|"literal string" s
1568d13a 333Like C<sv_setref_pvn>, but takes a literal string instead of
0c395ea5 334a string/length pair.
9dcc53ea 335
3fe05580
MHM
336=head1 Memory Management
337
1568d13a
Z
338=for apidoc Ama|char*|savepvs|"literal string" s
339Like C<savepvn>, but takes a literal string instead of a
30a15352 340string/length pair.
3fe05580 341
1568d13a 342=for apidoc Ama|char*|savesharedpvs|"literal string" s
9dcc53ea
Z
343A version of C<savepvs()> which allocates the duplicate string in memory
344which is shared between threads.
345
3fe05580
MHM
346=head1 GV Functions
347
1568d13a
Z
348=for apidoc Am|HV*|gv_stashpvs|"literal string" name|I32 create
349Like C<gv_stashpvn>, but takes a literal string instead of a
0c395ea5 350string/length pair.
3fe05580
MHM
351
352=head1 Hash Manipulation Functions
353
1568d13a
Z
354=for apidoc Am|SV**|hv_fetchs|HV* tb|"literal string" key|I32 lval
355Like C<hv_fetch>, but takes a literal string instead of a
0c395ea5 356string/length pair.
3fe05580 357
d106cb46 358=for apidoc Am|SV**|hv_stores|HV* tb|"literal string" key|SV* val
1568d13a 359Like C<hv_store>, but takes a literal string instead of a
0c395ea5 360string/length pair
3fe05580
MHM
361and omits the hash parameter.
362
510966aa
Z
363=head1 Lexer interface
364
1568d13a 365=for apidoc Amx|void|lex_stuff_pvs|"literal string" pv|U32 flags
510966aa 366
1568d13a 367Like L</lex_stuff_pvn>, but takes a literal string instead of
0c395ea5 368a string/length pair.
510966aa 369
3fe05580
MHM
370=cut
371*/
372
926bb54c
KW
373/* concatenating with "" ensures that only literal strings are accepted as
374 * argument */
26bfd924 375#define STR_WITH_LEN(s) ("" s ""), (sizeof(s)-1)
2efa8cc7 376
926bb54c
KW
377/* note that STR_WITH_LEN() can't be used as argument to macros or functions
378 * that under some configurations might be macros, which means that it requires
379 * the full Perl_xxx(aTHX_ ...) form for any API calls where it's used.
ba3a79e7
GA
380 */
381
382/* STR_WITH_LEN() shortcuts */
383#define newSVpvs(str) Perl_newSVpvn(aTHX_ STR_WITH_LEN(str))
84bafc02
NC
384#define newSVpvs_flags(str,flags) \
385 Perl_newSVpvn_flags(aTHX_ STR_WITH_LEN(str), flags)
ba3a79e7 386#define newSVpvs_share(str) Perl_newSVpvn_share(aTHX_ STR_WITH_LEN(str), 0)
9dcc53ea
Z
387#define sv_catpvs_flags(sv, str, flags) \
388 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), flags)
389#define sv_catpvs_nomg(sv, str) \
390 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), 0)
391#define sv_catpvs(sv, str) \
392 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC)
393#define sv_catpvs_mg(sv, str) \
394 Perl_sv_catpvn_flags(aTHX_ sv, STR_WITH_LEN(str), SV_GMAGIC|SV_SMAGIC)
3fe05580 395#define sv_setpvs(sv, str) Perl_sv_setpvn(aTHX_ sv, STR_WITH_LEN(str))
9dcc53ea
Z
396#define sv_setpvs_mg(sv, str) Perl_sv_setpvn_mg(aTHX_ sv, STR_WITH_LEN(str))
397#define sv_setref_pvs(rv, classname, str) \
398 Perl_sv_setref_pvn(aTHX_ rv, classname, STR_WITH_LEN(str))
ba3a79e7 399#define savepvs(str) Perl_savepvn(aTHX_ STR_WITH_LEN(str))
9dcc53ea
Z
400#define savesharedpvs(str) Perl_savesharedpvn(aTHX_ STR_WITH_LEN(str))
401#define gv_stashpvs(str, create) \
402 Perl_gv_stashpvn(aTHX_ STR_WITH_LEN(str), create)
403#define gv_fetchpvs(namebeg, add, sv_type) \
404 Perl_gv_fetchpvn_flags(aTHX_ STR_WITH_LEN(namebeg), add, sv_type)
405#define gv_fetchpvn(namebeg, len, add, sv_type) \
406 Perl_gv_fetchpvn_flags(aTHX_ namebeg, len, add, sv_type)
407#define sv_catxmlpvs(dsv, str, utf8) \
408 Perl_sv_catxmlpvn(aTHX_ dsv, STR_WITH_LEN(str), utf8)
4ac46235 409
ba3a79e7 410
510966aa
Z
411#define lex_stuff_pvs(pv,flags) Perl_lex_stuff_pvn(aTHX_ STR_WITH_LEN(pv), flags)
412
b96d8cd9
NC
413#define get_cvs(str, flags) \
414 Perl_get_cvn_flags(aTHX_ STR_WITH_LEN(str), (flags))
5c1737d1 415
954c1994 416/*
ccfc67b7
JH
417=head1 Miscellaneous Functions
418
954c1994 419=for apidoc Am|bool|strNE|char* s1|char* s2
dc6b0978
KW
420Test two C<NUL>-terminated strings to see if they are different. Returns true
421or false.
954c1994
GS
422
423=for apidoc Am|bool|strEQ|char* s1|char* s2
dc6b0978
KW
424Test two C<NUL>-terminated strings to see if they are equal. Returns true or
425false.
954c1994
GS
426
427=for apidoc Am|bool|strLT|char* s1|char* s2
dc6b0978
KW
428Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than the
429second, C<s2>. Returns true or false.
954c1994
GS
430
431=for apidoc Am|bool|strLE|char* s1|char* s2
dc6b0978
KW
432Test two C<NUL>-terminated strings to see if the first, C<s1>, is less than or
433equal to the second, C<s2>. Returns true or false.
954c1994
GS
434
435=for apidoc Am|bool|strGT|char* s1|char* s2
dc6b0978
KW
436Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
437the second, C<s2>. Returns true or false.
954c1994
GS
438
439=for apidoc Am|bool|strGE|char* s1|char* s2
dc6b0978
KW
440Test two C<NUL>-terminated strings to see if the first, C<s1>, is greater than
441or equal to the second, C<s2>. Returns true or false.
954c1994
GS
442
443=for apidoc Am|bool|strnNE|char* s1|char* s2|STRLEN len
dc6b0978
KW
444Test two C<NUL>-terminated strings to see if they are different. The C<len>
445parameter indicates the number of bytes to compare. Returns true or false. (A
954c1994
GS
446wrapper for C<strncmp>).
447
448=for apidoc Am|bool|strnEQ|char* s1|char* s2|STRLEN len
dc6b0978
KW
449Test two C<NUL>-terminated strings to see if they are equal. The C<len>
450parameter indicates the number of bytes to compare. Returns true or false. (A
451wrapper for C<strncmp>).
954c1994 452
bd18bd40
KW
453=for apidoc Am|bool|memEQ|char* s1|char* s2|STRLEN len
454Test two buffers (which may contain embedded C<NUL> characters, to see if they
455are equal. The C<len> parameter indicates the number of bytes to compare.
456Returns zero if equal, or non-zero if non-equal.
457
458=for apidoc Am|bool|memNE|char* s1|char* s2|STRLEN len
459Test two buffers (which may contain embedded C<NUL> characters, to see if they
460are not equal. The C<len> parameter indicates the number of bytes to compare.
461Returns zero if non-equal, or non-zero if equal.
462
954c1994 463=cut
fc169e00
KW
464
465New macros should use the following conventions for their names (which are
466based on the underlying C library functions):
467
468 (mem | str n? ) (EQ | NE | LT | GT | GE | (( BEGIN | END ) P? )) l? s?
469
470 Each has two main parameters, string-like operands that are compared
471 against each other, as specified by the macro name. Some macros may
472 additionally have one or potentially even two length parameters. If a length
473 parameter applies to both string parameters, it will be positioned third;
474 otherwise any length parameter immediately follows the string parameter it
475 applies to.
476
477 If the prefix to the name is 'str', the string parameter is a pointer to a C
478 language string. Such a string does not contain embedded NUL bytes; its
479 length may be unknown, but can be calculated by C<strlen()>, since it is
480 terminated by a NUL, which isn't included in its length.
481
482 The optional 'n' following 'str' means that that there is a third parameter,
483 giving the maximum number of bytes to look at in each string. Even if both
484 strings are longer than the length parameter, those extra bytes will be
485 unexamined.
486
487 The 's' suffix means that the 2nd byte string parameter is a literal C
488 double-quoted string. Its length will automatically be calculated by the
489 macro, so no length parameter will ever be needed for it.
490
491 If the prefix is 'mem', the string parameters don't have to be C strings;
492 they may contain embedded NUL bytes, do not necessarily have a terminating
493 NUL, and their lengths can be known only through other means, which in
494 practice are additional parameter(s) passed to the function. All 'mem'
495 functions have at least one length parameter. Barring any 'l' or 's' suffix,
496 there is a single length parameter, in position 3, which applies to both
497 string parameters. The 's' suffix means, as described above, that the 2nd
498 string is a literal double-quoted C string (hence its length is calculated by
499 the macro, and the length parameter to the function applies just to the first
500 string parameter, and hence is positioned just after it). An 'l' suffix
501 means that the 2nd string parameter has its own length parameter, and the
502 signature will look like memFOOl(s1, l1, s2, l2).
503
504 BEGIN (and END) are for testing if the 2nd string is an initial (or final)
505 substring of the 1st string. 'P' if present indicates that the substring
506 must be a "proper" one in tha mathematical sense that the first one must be
507 strictly larger than the 2nd.
508
954c1994
GS
509*/
510
62946e08 511
75400963
KW
512#define strNE(s1,s2) (strcmp(s1,s2) != 0)
513#define strEQ(s1,s2) (strcmp(s1,s2) == 0)
8d063cd8
LW
514#define strLT(s1,s2) (strcmp(s1,s2) < 0)
515#define strLE(s1,s2) (strcmp(s1,s2) <= 0)
516#define strGT(s1,s2) (strcmp(s1,s2) > 0)
517#define strGE(s1,s2) (strcmp(s1,s2) >= 0)
62946e08 518
75400963
KW
519#define strnNE(s1,s2,l) (strncmp(s1,s2,l) != 0)
520#define strnEQ(s1,s2,l) (strncmp(s1,s2,l) == 0)
378cc40b 521
9d3980bc
KW
522#define memEQ(s1,s2,l) (memcmp(((const void *) (s1)), ((const void *) (s2)), l) == 0)
523#define memNE(s1,s2,l) (! memEQ(s1,s2,l))
36477c24 524
085b7534 525/* memEQ and memNE where second comparand is a string constant */
568a785a 526#define memEQs(s1, l, s2) \
777fa2cb 527 (((sizeof(s2)-1) == (l)) && memEQ((s1), ("" s2 ""), (sizeof(s2)-1)))
5f50c6c9 528#define memNEs(s1, l, s2) (! memEQs(s1, l, s2))
568a785a 529
fdbb9a7c
KW
530/* Keep these private until we decide it was a good idea */
531#if defined(PERL_CORE) || defined(PERL_EXT) || defined(PERL_EXT_POSIX)
532
533#define strBEGINs(s1,s2) (strncmp(s1,"" s2 "", sizeof(s2)-1) == 0)
534
bdb7e3f0 535#define memBEGINs(s1, l, s2) \
30a6480c 536 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
bdb7e3f0 537 && memEQ(s1, "" s2 "", sizeof(s2)-1))
de627158 538#define memBEGINPs(s1, l, s2) \
30a6480c 539 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) - 1 \
de627158 540 && memEQ(s1, "" s2 "", sizeof(s2)-1))
bdb7e3f0 541#define memENDs(s1, l, s2) \
30a6480c 542 ( (Ptrdiff_t) (l) >= (Ptrdiff_t) sizeof(s2) - 1 \
bdb7e3f0 543 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
b80f8424 544#define memENDPs(s1, l, s2) \
30a6480c 545 ( (Ptrdiff_t) (l) > (Ptrdiff_t) sizeof(s2) \
b80f8424 546 && memEQ(s1 + (l) - (sizeof(s2) - 1), "" s2 "", sizeof(s2)-1))
fdbb9a7c 547#endif /* End of making macros private */
bdb7e3f0 548
062b6850
KW
549#define memLT(s1,s2,l) (memcmp(s1,s2,l) < 0)
550#define memLE(s1,s2,l) (memcmp(s1,s2,l) <= 0)
551#define memGT(s1,s2,l) (memcmp(s1,s2,l) > 0)
552#define memGE(s1,s2,l) (memcmp(s1,s2,l) >= 0)
553
bbce6d69
PP
554/*
555 * Character classes.
556 *
557 * Unfortunately, the introduction of locales means that we
558 * can't trust isupper(), etc. to tell the truth. And when
559 * it comes to /\w+/ with tainting enabled, we *must* be able
560 * to trust our character classes.
561 *
562 * Therefore, the default tests in the text of Perl will be
563 * independent of locale. Any code that wants to depend on
564 * the current locale will use the tests that begin with "lc".
565 */
566
2304df62
AD
567#ifdef HAS_SETLOCALE /* XXX Is there a better test for this? */
568# ifndef CTYPE256
569# define CTYPE256
570# endif
571#endif
572
954c1994 573/*
ccfc67b7 574
dcccc8ff 575=head1 Character classification
243effed
KW
576This section is about functions (really macros) that classify characters
577into types, such as punctuation versus alphabetic, etc. Most of these are
578analogous to regular expression character classes. (See
579L<perlrecharclass/POSIX Character Classes>.) There are several variants for
580each class. (Not all macros have all variants; each item below lists the
581ones valid for it.) None are affected by C<use bytes>, and only the ones
582with C<LC> in the name are affected by the current locale.
583
d713f9d9
KW
584The base function, e.g., C<isALPHA()>, takes any signed or unsigned value,
585treating it as a code point, and returns a boolean as to whether or not the
586character represented by it is (or on non-ASCII platforms, corresponds to) an
6aff1f14
KW
587ASCII character in the named class based on platform, Unicode, and Perl rules.
588If the input is a number that doesn't fit in an octet, FALSE is returned.
243effed 589
c98722a4 590Variant C<isI<FOO>_A> (e.g., C<isALPHA_A()>) is identical to the base function
550da823
KW
591with no suffix C<"_A">. This variant is used to emphasize by its name that
592only ASCII-range characters can return TRUE.
4b9734bf 593
d60679e1 594Variant C<isI<FOO>_L1> imposes the Latin-1 (or EBCDIC equivalent) character set
4b9734bf
KW
595onto the platform. That is, the code points that are ASCII are unaffected,
596since ASCII is a subset of Latin-1. But the non-ASCII code points are treated
597as if they are Latin-1 characters. For example, C<isWORDCHAR_L1()> will return
598true when called with the code point 0xDF, which is a word character in both
4650c663 599ASCII and EBCDIC (though it represents different characters in each).
d713f9d9
KW
600If the input is a number that doesn't fit in an octet, FALSE is returned.
601(Perl's documentation uses a colloquial definition of Latin-1, to include all
602code points below 256.)
243effed 603
d713f9d9
KW
604Variant C<isI<FOO>_uvchr> is exactly like the C<isI<FOO>_L1> variant, for
605inputs below 256, but if the code point is larger than 255, Unicode rules are
606used to determine if it is in the character class. For example,
d0da05db 607C<isWORDCHAR_uvchr(0x100)> returns TRUE, since 0x100 is LATIN CAPITAL LETTER A
6aff1f14 608WITH MACRON in Unicode, and is a word character.
243effed 609
c98722a4 610Variant C<isI<FOO>_utf8_safe> is like C<isI<FOO>_uvchr>, but is used for UTF-8
d713f9d9
KW
611encoded strings. Each call classifies the first character of the string. This
612variant takes two parameters. The first, C<p>, is a
da8c1a98
KW
613pointer to the first byte of the character to be classified. (Recall that it
614may take more than one byte to represent a character in UTF-8 strings.) The
615second parameter, C<e>, points to anywhere in the string beyond the first
616character, up to one byte past the end of the entire string. The suffix
617C<_safe> in the function's name indicates that it will not attempt to read
618beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is true (this
619is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the input
620character is malformed in some way, the program may croak, or the function may
621return FALSE, at the discretion of the implementation, and subject to change in
622future releases.
623
c98722a4 624Variant C<isI<FOO>_utf8> is like C<isI<FOO>_utf8_safe>, but takes just a single
da8c1a98 625parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 626in C<isI<FOO>_utf8_safe>. The function therefore can't check if it is reading
8b4b30c5 627beyond the end of the string. Starting in Perl v5.32, it will take a second
c98722a4 628parameter, becoming a synonym for C<isI<FOO>_utf8_safe>. At that time every
34aeb2e9 629program that uses it will have to be changed to successfully compile. In the
c98722a4 630meantime, the first runtime call to C<isI<FOO>_utf8> from each call point in the
34aeb2e9 631program will raise a deprecation warning, enabled by default. You can convert
c98722a4 632your program now to use C<isI<FOO>_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 633extra measure of protection, or you can wait until v5.32, when you'll be forced
34aeb2e9 634to add the C<e> parameter.
243effed 635
d713f9d9
KW
636Variant C<isI<FOO>_LC> is like the C<isI<FOO>_A> and C<isI<FOO>_L1> variants,
637but the result is based on the current locale, which is what C<LC> in the name
638stands for. If Perl can determine that the current locale is a UTF-8 locale,
639it uses the published Unicode rules; otherwise, it uses the C library function
640that gives the named classification. For example, C<isDIGIT_LC()> when not in
641a UTF-8 locale returns the result of calling C<isdigit()>. FALSE is always
1a83413c
KW
642returned if the input won't fit into an octet. On some platforms where the C
643library function is known to be defective, Perl changes its result to follow
644the POSIX standard's rules.
243effed 645
d713f9d9
KW
646Variant C<isI<FOO>_LC_uvchr> acts exactly like C<isI<FOO>_LC> for inputs less
647than 256, but for larger ones it returns the Unicode classification of the code
648point.
243effed 649
c98722a4 650Variant C<isI<FOO>_LC_utf8_safe> is like C<isI<FOO>_LC_uvchr>, but is used for UTF-8
d713f9d9
KW
651encoded strings. Each call classifies the first character of the string. This
652variant takes two parameters. The first, C<p>, is a pointer to the first byte
653of the character to be classified. (Recall that it may take more than one byte
654to represent a character in UTF-8 strings.) The second parameter, C<e>,
655points to anywhere in the string beyond the first character, up to one byte
656past the end of the entire string. The suffix C<_safe> in the function's name
657indicates that it will not attempt to read beyond S<C<e - 1>>, provided that
658the constraint S<C<s E<lt> e>> is true (this is asserted for in C<-DDEBUGGING>
659builds). If the UTF-8 for the input character is malformed in some way, the
660program may croak, or the function may return FALSE, at the discretion of the
661implementation, and subject to change in future releases.
da8c1a98 662
c98722a4 663Variant C<isI<FOO>_LC_utf8> is like C<isI<FOO>_LC_utf8_safe>, but takes just a single
da8c1a98 664parameter, C<p>, which has the same meaning as the corresponding parameter does
c98722a4 665in C<isI<FOO>_LC_utf8_safe>. The function therefore can't check if it is reading
8b4b30c5 666beyond the end of the string. Starting in Perl v5.32, it will take a second
c98722a4 667parameter, becoming a synonym for C<isI<FOO>_LC_utf8_safe>. At that time every
34aeb2e9 668program that uses it will have to be changed to successfully compile. In the
c98722a4 669meantime, the first runtime call to C<isI<FOO>_LC_utf8> from each call point in
34aeb2e9 670the program will raise a deprecation warning, enabled by default. You can
c98722a4 671convert your program now to use C<isI<FOO>_LC_utf8_safe>, and avoid the warnings,
8b4b30c5 672and get an extra measure of protection, or you can wait until v5.32, when
34aeb2e9 673you'll be forced to add the C<e> parameter.
ccfc67b7 674
d713f9d9
KW
675=for apidoc Am|bool|isALPHA|int ch
676Returns a boolean indicating whether the specified input is one of C<[A-Za-z]>,
677analogous to C<m/[[:alpha:]]/>.
dcccc8ff
KW
678See the L<top of this section|/Character classification> for an explanation of
679variants
da8c1a98
KW
680C<isALPHA_A>, C<isALPHA_L1>, C<isALPHA_uvchr>, C<isALPHA_utf8_safe>,
681C<isALPHA_LC>, C<isALPHA_LC_uvchr>, and C<isALPHA_LC_utf8_safe>.
8a58bdcf 682
d713f9d9
KW
683=for apidoc Am|bool|isALPHANUMERIC|int ch
684Returns a boolean indicating whether the specified character is one of
685C<[A-Za-z0-9]>, analogous to C<m/[[:alnum:]]/>.
dcccc8ff
KW
686See the L<top of this section|/Character classification> for an explanation of
687variants
d0da05db 688C<isALPHANUMERIC_A>, C<isALPHANUMERIC_L1>, C<isALPHANUMERIC_uvchr>,
da8c1a98
KW
689C<isALPHANUMERIC_utf8_safe>, C<isALPHANUMERIC_LC>, C<isALPHANUMERIC_LC_uvchr>,
690and C<isALPHANUMERIC_LC_utf8_safe>.
15861f94 691
255b632a
KW
692A (discouraged from use) synonym is C<isALNUMC> (where the C<C> suffix means
693this corresponds to the C language alphanumeric definition). Also
694there are the variants
695C<isALNUMC_A>, C<isALNUMC_L1>
696C<isALNUMC_LC>, and C<isALNUMC_LC_uvchr>.
697
d713f9d9 698=for apidoc Am|bool|isASCII|int ch
8a58bdcf 699Returns a boolean indicating whether the specified character is one of the 128
243effed 700characters in the ASCII character set, analogous to C<m/[[:ascii:]]/>.
e5ad6aba 701On non-ASCII platforms, it returns TRUE iff this
8a58bdcf
KW
702character corresponds to an ASCII character. Variants C<isASCII_A()> and
703C<isASCII_L1()> are identical to C<isASCII()>.
dcccc8ff
KW
704See the L<top of this section|/Character classification> for an explanation of
705variants
da8c1a98
KW
706C<isASCII_uvchr>, C<isASCII_utf8_safe>, C<isASCII_LC>, C<isASCII_LC_uvchr>, and
707C<isASCII_LC_utf8_safe>. Note, however, that some platforms do not have the C
243effed
KW
708library routine C<isascii()>. In these cases, the variants whose names contain
709C<LC> are the same as the corresponding ones without.
710
d98532ea
KW
711Also note, that because all ASCII characters are UTF-8 invariant (meaning they
712have the exact same representation (always a single byte) whether encoded in
713UTF-8 or not), C<isASCII> will give the correct results when called with any
da8c1a98
KW
714byte in any string encoded or not in UTF-8. And similarly C<isASCII_utf8_safe>
715will work properly on any string encoded or not in UTF-8.
d98532ea 716
243effed
KW
717=for apidoc Am|bool|isBLANK|char ch
718Returns a boolean indicating whether the specified character is a
6aff1f14 719character considered to be a blank, analogous to C<m/[[:blank:]]/>.
dcccc8ff
KW
720See the L<top of this section|/Character classification> for an explanation of
721variants
da8c1a98
KW
722C<isBLANK_A>, C<isBLANK_L1>, C<isBLANK_uvchr>, C<isBLANK_utf8_safe>,
723C<isBLANK_LC>, C<isBLANK_LC_uvchr>, and C<isBLANK_LC_utf8_safe>. Note,
724however, that some platforms do not have the C library routine
725C<isblank()>. In these cases, the variants whose names contain C<LC> are
726the same as the corresponding ones without.
243effed
KW
727
728=for apidoc Am|bool|isCNTRL|char ch
729Returns a boolean indicating whether the specified character is a
6aff1f14 730control character, analogous to C<m/[[:cntrl:]]/>.
dcccc8ff
KW
731See the L<top of this section|/Character classification> for an explanation of
732variants
da8c1a98
KW
733C<isCNTRL_A>, C<isCNTRL_L1>, C<isCNTRL_uvchr>, C<isCNTRL_utf8_safe>,
734C<isCNTRL_LC>, C<isCNTRL_LC_uvchr>, and C<isCNTRL_LC_utf8_safe> On EBCDIC
735platforms, you almost always want to use the C<isCNTRL_L1> variant.
954c1994 736
954c1994 737=for apidoc Am|bool|isDIGIT|char ch
2787a470 738Returns a boolean indicating whether the specified character is a
6aff1f14 739digit, analogous to C<m/[[:digit:]]/>.
8a58bdcf 740Variants C<isDIGIT_A> and C<isDIGIT_L1> are identical to C<isDIGIT>.
dcccc8ff
KW
741See the L<top of this section|/Character classification> for an explanation of
742variants
da8c1a98
KW
743C<isDIGIT_uvchr>, C<isDIGIT_utf8_safe>, C<isDIGIT_LC>, C<isDIGIT_LC_uvchr>, and
744C<isDIGIT_LC_utf8_safe>.
243effed
KW
745
746=for apidoc Am|bool|isGRAPH|char ch
747Returns a boolean indicating whether the specified character is a
6aff1f14 748graphic character, analogous to C<m/[[:graph:]]/>.
dcccc8ff 749See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
750variants C<isGRAPH_A>, C<isGRAPH_L1>, C<isGRAPH_uvchr>, C<isGRAPH_utf8_safe>,
751C<isGRAPH_LC>, C<isGRAPH_LC_uvchr>, and C<isGRAPH_LC_utf8_safe>.
954c1994 752
0c82b6df 753=for apidoc Am|bool|isLOWER|char ch
2787a470 754Returns a boolean indicating whether the specified character is a
6aff1f14 755lowercase character, analogous to C<m/[[:lower:]]/>.
dcccc8ff
KW
756See the L<top of this section|/Character classification> for an explanation of
757variants
da8c1a98
KW
758C<isLOWER_A>, C<isLOWER_L1>, C<isLOWER_uvchr>, C<isLOWER_utf8_safe>,
759C<isLOWER_LC>, C<isLOWER_LC_uvchr>, and C<isLOWER_LC_utf8_safe>.
0c82b6df 760
c99e91e9 761=for apidoc Am|bool|isOCTAL|char ch
2787a470 762Returns a boolean indicating whether the specified character is an
6aff1f14 763octal digit, [0-7].
243effed
KW
764The only two variants are C<isOCTAL_A> and C<isOCTAL_L1>; each is identical to
765C<isOCTAL>.
766
767=for apidoc Am|bool|isPUNCT|char ch
768Returns a boolean indicating whether the specified character is a
6aff1f14
KW
769punctuation character, analogous to C<m/[[:punct:]]/>.
770Note that the definition of what is punctuation isn't as
243effed
KW
771straightforward as one might desire. See L<perlrecharclass/POSIX Character
772Classes> for details.
dcccc8ff 773See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
774variants C<isPUNCT_A>, C<isPUNCT_L1>, C<isPUNCT_uvchr>, C<isPUNCT_utf8_safe>,
775C<isPUNCT_LC>, C<isPUNCT_LC_uvchr>, and C<isPUNCT_LC_utf8_safe>.
c99e91e9 776
0c82b6df 777=for apidoc Am|bool|isSPACE|char ch
2787a470 778Returns a boolean indicating whether the specified character is a
6aff1f14 779whitespace character. This is analogous
398d098a 780to what C<m/\s/> matches in a regular expression. Starting in Perl 5.18
779cf272 781this also matches what C<m/[[:space:]]/> does. Prior to 5.18, only the
398d098a
KW
782locale forms of this macro (the ones with C<LC> in their names) matched
783precisely what C<m/[[:space:]]/> does. In those releases, the only difference,
784in the non-locale variants, was that C<isSPACE()> did not match a vertical tab.
785(See L</isPSXSPC> for a macro that matches a vertical tab in all releases.)
dcccc8ff
KW
786See the L<top of this section|/Character classification> for an explanation of
787variants
da8c1a98
KW
788C<isSPACE_A>, C<isSPACE_L1>, C<isSPACE_uvchr>, C<isSPACE_utf8_safe>,
789C<isSPACE_LC>, C<isSPACE_LC_uvchr>, and C<isSPACE_LC_utf8_safe>.
0c82b6df 790
398d098a
KW
791=for apidoc Am|bool|isPSXSPC|char ch
792(short for Posix Space)
779cf272
KW
793Starting in 5.18, this is identical in all its forms to the
794corresponding C<isSPACE()> macros.
398d098a
KW
795The locale forms of this macro are identical to their corresponding
796C<isSPACE()> forms in all Perl releases. In releases prior to 5.18, the
797non-locale forms differ from their C<isSPACE()> forms only in that the
798C<isSPACE()> forms don't match a Vertical Tab, and the C<isPSXSPC()> forms do.
799Otherwise they are identical. Thus this macro is analogous to what
800C<m/[[:space:]]/> matches in a regular expression.
dcccc8ff 801See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
802variants C<isPSXSPC_A>, C<isPSXSPC_L1>, C<isPSXSPC_uvchr>, C<isPSXSPC_utf8_safe>,
803C<isPSXSPC_LC>, C<isPSXSPC_LC_uvchr>, and C<isPSXSPC_LC_utf8_safe>.
398d098a 804
954c1994 805=for apidoc Am|bool|isUPPER|char ch
2787a470 806Returns a boolean indicating whether the specified character is an
6aff1f14 807uppercase character, analogous to C<m/[[:upper:]]/>.
dcccc8ff 808See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
809variants C<isUPPER_A>, C<isUPPER_L1>, C<isUPPER_uvchr>, C<isUPPER_utf8_safe>,
810C<isUPPER_LC>, C<isUPPER_LC_uvchr>, and C<isUPPER_LC_utf8_safe>.
954c1994 811
243effed 812=for apidoc Am|bool|isPRINT|char ch
8eea39dd 813Returns a boolean indicating whether the specified character is a
6aff1f14 814printable character, analogous to C<m/[[:print:]]/>.
dcccc8ff
KW
815See the L<top of this section|/Character classification> for an explanation of
816variants
da8c1a98
KW
817C<isPRINT_A>, C<isPRINT_L1>, C<isPRINT_uvchr>, C<isPRINT_utf8_safe>,
818C<isPRINT_LC>, C<isPRINT_LC_uvchr>, and C<isPRINT_LC_utf8_safe>.
243effed
KW
819
820=for apidoc Am|bool|isWORDCHAR|char ch
821Returns a boolean indicating whether the specified character is a character
822that is a word character, analogous to what C<m/\w/> and C<m/[[:word:]]/> match
823in a regular expression. A word character is an alphabetic character, a
824decimal digit, a connecting punctuation character (such as an underscore), or
825a "mark" character that attaches to one of those (like some sort of accent).
826C<isALNUM()> is a synonym provided for backward compatibility, even though a
827word character includes more than the standard C language meaning of
828alphanumeric.
dcccc8ff 829See the L<top of this section|/Character classification> for an explanation of
da8c1a98
KW
830variants C<isWORDCHAR_A>, C<isWORDCHAR_L1>, C<isWORDCHAR_uvchr>, and
831C<isWORDCHAR_utf8_safe>. C<isWORDCHAR_LC>, C<isWORDCHAR_LC_uvchr>, and
832C<isWORDCHAR_LC_utf8_safe> are also as described there, but additionally
833include the platform's native underscore.
8a58bdcf
KW
834
835=for apidoc Am|bool|isXDIGIT|char ch
836Returns a boolean indicating whether the specified character is a hexadecimal
243effed
KW
837digit. In the ASCII range these are C<[0-9A-Fa-f]>. Variants C<isXDIGIT_A()>
838and C<isXDIGIT_L1()> are identical to C<isXDIGIT()>.
dcccc8ff
KW
839See the L<top of this section|/Character classification> for an explanation of
840variants
da8c1a98
KW
841C<isXDIGIT_uvchr>, C<isXDIGIT_utf8_safe>, C<isXDIGIT_LC>, C<isXDIGIT_LC_uvchr>,
842and C<isXDIGIT_LC_utf8_safe>.
243effed 843
3c3ecf18
KW
844=for apidoc Am|bool|isIDFIRST|char ch
845Returns a boolean indicating whether the specified character can be the first
846character of an identifier. This is very close to, but not quite the same as
847the official Unicode property C<XID_Start>. The difference is that this
848returns true only if the input character also matches L</isWORDCHAR>.
dcccc8ff
KW
849See the L<top of this section|/Character classification> for an explanation of
850variants
da8c1a98
KW
851C<isIDFIRST_A>, C<isIDFIRST_L1>, C<isIDFIRST_uvchr>, C<isIDFIRST_utf8_safe>,
852C<isIDFIRST_LC>, C<isIDFIRST_LC_uvchr>, and C<isIDFIRST_LC_utf8_safe>.
3c3ecf18
KW
853
854=for apidoc Am|bool|isIDCONT|char ch
855Returns a boolean indicating whether the specified character can be the
856second or succeeding character of an identifier. This is very close to, but
857not quite the same as the official Unicode property C<XID_Continue>. The
858difference is that this returns true only if the input character also matches
dcccc8ff
KW
859L</isWORDCHAR>. See the L<top of this section|/Character classification> for
860an
d0da05db 861explanation of variants C<isIDCONT_A>, C<isIDCONT_L1>, C<isIDCONT_uvchr>,
da8c1a98
KW
862C<isIDCONT_utf8_safe>, C<isIDCONT_LC>, C<isIDCONT_LC_uvchr>, and
863C<isIDCONT_LC_utf8_safe>.
3c3ecf18 864
243effed 865=head1 Miscellaneous Functions
8eea39dd 866
95a59cab 867=for apidoc Am|U8|READ_XDIGIT|char str*
243effed 868Returns the value of an ASCII-range hex digit and advances the string pointer.
95a59cab
YO
869Behaviour is only well defined when isXDIGIT(*str) is true.
870
e7c1e6c1 871=head1 Character case changing
21da7284
KW
872Perl uses "full" Unicode case mappings. This means that converting a single
873character to another case may result in a sequence of more than one character.
874For example, the uppercase of C<E<223>> (LATIN SMALL LETTER SHARP S) is the two
875character sequence C<SS>. This presents some complications The lowercase of
876all characters in the range 0..255 is a single character, and thus
877C<L</toLOWER_L1>> is furnished. But, C<toUPPER_L1> can't exist, as it couldn't
878return a valid result for all legal inputs. Instead C<L</toUPPER_uvchr>> has
879an API that does allow every possible legal result to be returned.) Likewise
880no other function that is crippled by not being able to give the correct
881results for the full range of possible inputs has been implemented here.
e7c1e6c1 882
d713f9d9 883=for apidoc Am|U8|toUPPER|int ch
1f607577
KW
884Converts the specified character to uppercase. If the input is anything but an
885ASCII lowercase character, that input character itself is returned. Variant
c753c8d3 886C<toUPPER_A> is equivalent.
954c1994 887
d0da05db
KW
888=for apidoc Am|UV|toUPPER_uvchr|UV cp|U8* s|STRLEN* lenp
889Converts the code point C<cp> to its uppercase version, and
890stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
891point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
892that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
893bytes since the uppercase version may be longer than the original character.
894
895The first code point of the uppercased version is returned
21da7284
KW
896(but note, as explained at L<the top of this section|/Character case
897changing>, that there may be more.)
1f607577 898
a239b1e2
KW
899=for apidoc Am|UV|toUPPER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
900Converts the first UTF-8 encoded character in the sequence starting at C<p> and
901extending no further than S<C<e - 1>> to its uppercase version, and
1f607577
KW
902stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
903that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
904bytes since the uppercase version may be longer than the original character.
905
906The first code point of the uppercased version is returned
21da7284
KW
907(but note, as explained at L<the top of this section|/Character case
908changing>, that there may be more).
1f607577 909
a239b1e2
KW
910The suffix C<_safe> in the function's name indicates that it will not attempt
911to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
912true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
913input character is malformed in some way, the program may croak, or the
914function may return the REPLACEMENT CHARACTER, at the discretion of the
915implementation, and subject to change in future releases.
916
917=for apidoc Am|UV|toUPPER_utf8|U8* p|U8* s|STRLEN* lenp
918This is like C<L</toUPPER_utf8_safe>>, but doesn't have the C<e>
919parameter The function therefore can't check if it is reading
8b4b30c5 920beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
921parameter, becoming a synonym for C<toUPPER_utf8_safe>. At that time every
922program that uses it will have to be changed to successfully compile. In the
923meantime, the first runtime call to C<toUPPER_utf8> from each call point in the
924program will raise a deprecation warning, enabled by default. You can convert
925your program now to use C<toUPPER_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 926extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 927to add the C<e> parameter.
1f607577 928
25200305
KW
929=for apidoc Am|U8|toFOLD|U8 ch
930Converts the specified character to foldcase. If the input is anything but an
931ASCII uppercase character, that input character itself is returned. Variant
932C<toFOLD_A> is equivalent. (There is no equivalent C<to_FOLD_L1> for the full
d0da05db 933Latin1 range, as the full generality of L</toFOLD_uvchr> is needed there.)
25200305 934
d0da05db
KW
935=for apidoc Am|UV|toFOLD_uvchr|UV cp|U8* s|STRLEN* lenp
936Converts the code point C<cp> to its foldcase version, and
937stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
938point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
939that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
940bytes since the foldcase version may be longer than the original character.
941
942The first code point of the foldcased version is returned
21da7284
KW
943(but note, as explained at L<the top of this section|/Character case
944changing>, that there may be more).
1f607577 945
a239b1e2
KW
946=for apidoc Am|UV|toFOLD_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
947Converts the first UTF-8 encoded character in the sequence starting at C<p> and
948extending no further than S<C<e - 1>> to its foldcase version, and
1f607577
KW
949stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
950that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
951bytes since the foldcase version may be longer than the original character.
952
953The first code point of the foldcased version is returned
21da7284
KW
954(but note, as explained at L<the top of this section|/Character case
955changing>, that there may be more).
1f607577 956
a239b1e2
KW
957The suffix C<_safe> in the function's name indicates that it will not attempt
958to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
959true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
960input character is malformed in some way, the program may croak, or the
961function may return the REPLACEMENT CHARACTER, at the discretion of the
962implementation, and subject to change in future releases.
963
964=for apidoc Am|UV|toFOLD_utf8|U8* p|U8* s|STRLEN* lenp
965This is like C<L</toFOLD_utf8_safe>>, but doesn't have the C<e>
966parameter The function therefore can't check if it is reading
8b4b30c5 967beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
968parameter, becoming a synonym for C<toFOLD_utf8_safe>. At that time every
969program that uses it will have to be changed to successfully compile. In the
970meantime, the first runtime call to C<toFOLD_utf8> from each call point in the
971program will raise a deprecation warning, enabled by default. You can convert
972your program now to use C<toFOLD_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 973extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 974to add the C<e> parameter.
1f607577
KW
975
976=for apidoc Am|U8|toLOWER|U8 ch
977Converts the specified character to lowercase. If the input is anything but an
978ASCII uppercase character, that input character itself is returned. Variant
c753c8d3 979C<toLOWER_A> is equivalent.
954c1994 980
1f607577 981=for apidoc Am|U8|toLOWER_L1|U8 ch
b7d90381
KW
982Converts the specified Latin1 character to lowercase. The results are
983undefined if the input doesn't fit in a byte.
1f607577
KW
984
985=for apidoc Am|U8|toLOWER_LC|U8 ch
986Converts the specified character to lowercase using the current locale's rules,
987if possible; otherwise returns the input character itself.
988
d0da05db
KW
989=for apidoc Am|UV|toLOWER_uvchr|UV cp|U8* s|STRLEN* lenp
990Converts the code point C<cp> to its lowercase version, and
991stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
992point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
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
1000
1001=for apidoc Am|UV|toLOWER_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1002Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1003extending no further than S<C<e - 1>> to its lowercase version, and
1f607577
KW
1004stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1005that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1006bytes since the lowercase version may be longer than the original character.
1007
1008The first code point of the lowercased version is returned
21da7284
KW
1009(but note, as explained at L<the top of this section|/Character case
1010changing>, that there may be more).
1f607577 1011
a239b1e2
KW
1012The suffix C<_safe> in the function's name indicates that it will not attempt
1013to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1014true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1015input character is malformed in some way, the program may croak, or the
1016function may return the REPLACEMENT CHARACTER, at the discretion of the
1017implementation, and subject to change in future releases.
1018
1019=for apidoc Am|UV|toLOWER_utf8|U8* p|U8* s|STRLEN* lenp
1020This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1021parameter The function therefore can't check if it is reading
8b4b30c5 1022beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
1023parameter, becoming a synonym for C<toLOWER_utf8_safe>. At that time every
1024program that uses it will have to be changed to successfully compile. In the
1025meantime, the first runtime call to C<toLOWER_utf8> from each call point in the
1026program will raise a deprecation warning, enabled by default. You can convert
1027your program now to use C<toLOWER_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 1028extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 1029to add the C<e> parameter.
1f607577 1030
25200305
KW
1031=for apidoc Am|U8|toTITLE|U8 ch
1032Converts the specified character to titlecase. If the input is anything but an
1033ASCII lowercase character, that input character itself is returned. Variant
b7d90381 1034C<toTITLE_A> is equivalent. (There is no C<toTITLE_L1> for the full Latin1
d0da05db 1035range, as the full generality of L</toTITLE_uvchr> is needed there. Titlecase is
b7d90381 1036not a concept used in locale handling, so there is no functionality for that.)
25200305 1037
d0da05db
KW
1038=for apidoc Am|UV|toTITLE_uvchr|UV cp|U8* s|STRLEN* lenp
1039Converts the code point C<cp> to its titlecase version, and
1040stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. The code
1041point is interpreted as native if less than 256; otherwise as Unicode. Note
1f607577
KW
1042that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1043bytes since the titlecase version may be longer than the original character.
1044
1045The first code point of the titlecased version is returned
21da7284
KW
1046(but note, as explained at L<the top of this section|/Character case
1047changing>, that there may be more).
1f607577 1048
a239b1e2
KW
1049=for apidoc Am|UV|toTITLE_utf8_safe|U8* p|U8* e|U8* s|STRLEN* lenp
1050Converts the first UTF-8 encoded character in the sequence starting at C<p> and
1051extending no further than S<C<e - 1>> to its titlecase version, and
1f607577
KW
1052stores that in UTF-8 in C<s>, and its length in bytes in C<lenp>. Note
1053that the buffer pointed to by C<s> needs to be at least C<UTF8_MAXBYTES_CASE+1>
1054bytes since the titlecase version may be longer than the original character.
1055
1056The first code point of the titlecased version is returned
21da7284
KW
1057(but note, as explained at L<the top of this section|/Character case
1058changing>, that there may be more).
1f607577 1059
a239b1e2
KW
1060The suffix C<_safe> in the function's name indicates that it will not attempt
1061to read beyond S<C<e - 1>>, provided that the constraint S<C<s E<lt> e>> is
1062true (this is asserted for in C<-DDEBUGGING> builds). If the UTF-8 for the
1063input character is malformed in some way, the program may croak, or the
1064function may return the REPLACEMENT CHARACTER, at the discretion of the
1065implementation, and subject to change in future releases.
1066
1067=for apidoc Am|UV|toTITLE_utf8|U8* p|U8* s|STRLEN* lenp
1068This is like C<L</toLOWER_utf8_safe>>, but doesn't have the C<e>
1069parameter The function therefore can't check if it is reading
8b4b30c5 1070beyond the end of the string. Starting in Perl v5.32, it will take the C<e>
607313a1
KW
1071parameter, becoming a synonym for C<toTITLE_utf8_safe>. At that time every
1072program that uses it will have to be changed to successfully compile. In the
1073meantime, the first runtime call to C<toTITLE_utf8> from each call point in the
1074program will raise a deprecation warning, enabled by default. You can convert
1075your program now to use C<toTITLE_utf8_safe>, and avoid the warnings, and get an
8b4b30c5 1076extra measure of protection, or you can wait until v5.32, when you'll be forced
607313a1 1077to add the C<e> parameter.
1f607577 1078
954c1994 1079=cut
353c9b6f 1080
d0da05db 1081XXX Still undocumented isVERTWS_uvchr and _utf8; it's unclear what their names
1e222e4f
KW
1082really should be. Also toUPPER_LC and toFOLD_LC, which are subject to change,
1083and aren't general purpose as they don't work on U+DF, and assert against that.
243effed 1084
8a58bdcf 1085Note that these macros are repeated in Devel::PPPort, so should also be
62fa66b6
KW
1086patched there. The file as of this writing is cpan/Devel-PPPort/parts/inc/misc
1087
954c1994
GS
1088*/
1089
de9e2639
KW
1090/* Specify the widest unsigned type on the platform. */
1091#ifdef QUADKIND
1092# define WIDEST_UTYPE U64
7c062697
KW
1093#else
1094# define WIDEST_UTYPE U32
1095#endif
1096
3912bc88
KW
1097/* FITS_IN_8_BITS(c) returns true if c doesn't have a bit set other than in
1098 * the lower 8. It is designed to be hopefully bomb-proof, making sure that no
1099 * bits of information are lost even on a 64-bit machine, but to get the
1100 * compiler to optimize it out if possible. This is because Configure makes
1101 * sure that the machine has an 8-bit byte, so if c is stored in a byte, the
1102 * sizeof() guarantees that this evaluates to a constant true at compile time.
7e75d1a1
JH
1103 *
1104 * For Coverity, be always true, because otherwise Coverity thinks
1105 * it finds several expressions that are always true, independent
1106 * of operands. Well, they are, but that is kind of the point.
220c71bf 1107 */
7e75d1a1 1108#ifndef __COVERITY__
6c5b02ac
KW
1109 /* The '| 0' part ensures a compiler error if c is not integer (like e.g., a
1110 * pointer) */
1111#define FITS_IN_8_BITS(c) ( (sizeof(c) == 1) \
ace3ad0f 1112 || !(((WIDEST_UTYPE)((c) | 0)) & ~0xFF))
7e75d1a1
JH
1113#else
1114#define FITS_IN_8_BITS(c) (1)
1115#endif
cf301eb7 1116
94250c4f
KW
1117/* Returns true if c is in the range l..u, where 'l' is non-negative
1118 * Written this way so that after optimization, only one conditional test is
1119 * needed.
1120 *
1121 * This isn't fully general, except for the special cased 'signed char' (which
1122 * should be resolved at compile time): It won't work if 'c' is negative, and
1123 * 'l' is larger than the max for that signed type. Thus if 'c' is a negative
1124 * int, and 'l' is larger than INT_MAX, it will fail. To protect agains this
1125 * happening, there is an assert that will generate a warning if c is larger
1126 * than e.g. INT_MAX if it is an 'unsigned int'. This could be a false
1127 * positive, but khw couldn't figure out a way to make it better. It's good
1128 * enough so far */
1129#define inRANGE(c, l, u) (__ASSERT_((l) >= 0) __ASSERT_((u) >= (l)) \
1130 ((sizeof(c) == 1) \
1131 ? (((WIDEST_UTYPE) ((((U8) (c))|0) - (l))) <= ((WIDEST_UTYPE) ((u) - (l)))) \
1132 : (__ASSERT_( (((WIDEST_UTYPE) 1) << (CHARBITS * sizeof(c) - 1) & (c)) \
1133 /* sign bit of c is 0 */ == 0 \
1134 || (((~ ((WIDEST_UTYPE) 1) << ((CHARBITS * sizeof(c) - 1) - 1))\
1135 /* l not larger than largest value in c's signed type */ \
1136 & ~ ((WIDEST_UTYPE) 0)) & (l)) == 0) \
1137 ((WIDEST_UTYPE) (((c) - (l)) | 0) <= ((WIDEST_UTYPE) ((u) - (l)))))))
305fe86e 1138
41f43cc2 1139#ifdef EBCDIC
b6340bd0 1140# ifndef _ALL_SOURCE
0852beac
KW
1141 /* The native libc isascii() et.al. functions return the wrong results
1142 * on at least z/OS unless this is defined. */
b6340bd0
KW
1143# error _ALL_SOURCE should probably be defined
1144# endif
41f43cc2 1145#else
0852beac
KW
1146 /* There is a simple definition of ASCII for ASCII platforms. But the
1147 * EBCDIC one isn't so simple, so is defined using table look-up like the
9c903d59 1148 * other macros below.
3f3c579d
KW
1149 *
1150 * The cast here is used instead of '(c) >= 0', because some compilers emit
1151 * a warning that that test is always true when the parameter is an
1152 * unsigned type. khw supposes that it could be written as
1153 * && ((c) == '\0' || (c) > 0)
1154 * to avoid the message, but the cast will likely avoid extra branches even
1155 * with stupid compilers.
1156 *
1157 * The '| 0' part ensures a compiler error if c is not integer (like e.g.,
1158 * a pointer) */
9c903d59 1159# define isASCII(c) ((WIDEST_UTYPE)((c) | 0) < 128)
41f43cc2
KW
1160#endif
1161
38694112
KW
1162/* Take the eight possible bit patterns of the lower 3 bits and you get the
1163 * lower 3 bits of the 8 octal digits, in both ASCII and EBCDIC, so those bits
1164 * can be ignored. If the rest match '0', we have an octal */
1165#define isOCTAL_A(c) (((WIDEST_UTYPE)((c) | 0) & ~7) == '0')
c2da0b36 1166
9fb1bf9d 1167#ifdef H_PERL /* If have access to perl.h, lookup in its table */
f4cdb42c 1168
a500dc72
KW
1169/* Character class numbers. For internal core Perl use only. The ones less
1170 * than 32 are used in PL_charclass[] and the ones up through the one that
1171 * corresponds to <_HIGHEST_REGCOMP_DOT_H_SYNC> are used by regcomp.h and
1172 * related files. PL_charclass ones use names used in l1_char_class_tab.h but
1173 * their actual definitions are here. If that file has a name not used here,
1174 * it won't compile.
1709d539
KW
1175 *
1176 * The first group of these is ordered in what I (khw) estimate to be the
31c7f561 1177 * frequency of their use. This gives a slight edge to exiting a loop earlier
58a3ba2c
KW
1178 * (in reginclass() in regexec.c). Except \v should be last, as it isn't a
1179 * real Posix character class, and some (small) inefficiencies in regular
1180 * expression handling would be introduced by putting it in the middle of those
1181 * that are. Also, cntrl and ascii come after the others as it may be useful
1182 * to group these which have no members that match above Latin1, (or above
1183 * ASCII in the latter case) */
1184
1709d539
KW
1185# define _CC_WORDCHAR 0 /* \w and [:word:] */
1186# define _CC_DIGIT 1 /* \d and [:digit:] */
1187# define _CC_ALPHA 2 /* [:alpha:] */
1188# define _CC_LOWER 3 /* [:lower:] */
1189# define _CC_UPPER 4 /* [:upper:] */
1190# define _CC_PUNCT 5 /* [:punct:] */
1191# define _CC_PRINT 6 /* [:print:] */
15861f94 1192# define _CC_ALPHANUMERIC 7 /* [:alnum:] */
1709d539 1193# define _CC_GRAPH 8 /* [:graph:] */
359b005e 1194# define _CC_CASED 9 /* [:lower:] or [:upper:] under /i */
779cf272 1195# define _CC_SPACE 10 /* \s, [:space:] */
34aeb2e9
KW
1196# define _CC_PSXSPC _CC_SPACE /* XXX Temporary, can be removed
1197 when the deprecated isFOO_utf8()
1198 functions are removed */
b0d691b2
KW
1199# define _CC_BLANK 11 /* [:blank:] */
1200# define _CC_XDIGIT 12 /* [:xdigit:] */
779cf272
KW
1201# define _CC_CNTRL 13 /* [:cntrl:] */
1202# define _CC_ASCII 14 /* [:ascii:] */
1203# define _CC_VERTSPACE 15 /* \v */
1709d539 1204
a0947d7b
KW
1205# define _HIGHEST_REGCOMP_DOT_H_SYNC _CC_VERTSPACE
1206
1709d539 1207/* The members of the third group below do not need to be coordinated with data
3ffc8c70 1208 * structures in regcomp.[ch] and regexec.c. */
779cf272
KW
1209# define _CC_IDFIRST 16
1210# define _CC_CHARNAME_CONT 17
1211# define _CC_NONLATIN1_FOLD 18
1212# define _CC_NONLATIN1_SIMPLE_FOLD 19
1213# define _CC_QUOTEMETA 20
1214# define _CC_NON_FINAL_FOLD 21
1215# define _CC_IS_IN_SOME_FOLD 22
1216# define _CC_MNEMONIC_CNTRL 23
073c22b3 1217
34aeb2e9
KW
1218# define _CC_IDCONT 24 /* XXX Temporary, can be removed when the deprecated
1219 isFOO_utf8() functions are removed */
1220
073c22b3
KW
1221/* This next group is only used on EBCDIC platforms, so theoretically could be
1222 * shared with something entirely different that's only on ASCII platforms */
5d5376e2
KW
1223# define _CC_UTF8_START_BYTE_IS_FOR_AT_LEAST_SURROGATE 28
1224# define _CC_UTF8_IS_START 29
1225# define _CC_UTF8_IS_DOWNGRADEABLE_START 30
1226# define _CC_UTF8_IS_CONTINUATION 31
1227/* Unused: 24-27
f4cdb42c
KW
1228 * If more bits are needed, one could add a second word for non-64bit
1229 * QUAD_IS_INT systems, using some #ifdefs to distinguish between having a 2nd
37ede926
KW
1230 * word or not. The IS_IN_SOME_FOLD bit is the most easily expendable, as it
1231 * is used only for optimization (as of this writing), and differs in the
1232 * Latin1 range from the ALPHA bit only in two relatively unimportant
a500dc72 1233 * characters: the masculine and feminine ordinal indicators, so removing it
073c22b3
KW
1234 * would just cause /i regexes which match them to run less efficiently.
1235 * Similarly the EBCDIC-only bits are used just for speed, and could be
1236 * replaced by other means */
96ac0975 1237
3a371f2f
KW
1238#if defined(PERL_CORE) || defined(PERL_EXT)
1239/* An enum version of the character class numbers, to help compilers
1240 * optimize */
1241typedef enum {
3a371f2f 1242 _CC_ENUM_ALPHA = _CC_ALPHA,
e8d596e0
KW
1243 _CC_ENUM_ALPHANUMERIC = _CC_ALPHANUMERIC,
1244 _CC_ENUM_ASCII = _CC_ASCII,
1245 _CC_ENUM_BLANK = _CC_BLANK,
b0d691b2 1246 _CC_ENUM_CASED = _CC_CASED,
e8d596e0 1247 _CC_ENUM_CNTRL = _CC_CNTRL,
3a371f2f
KW
1248 _CC_ENUM_DIGIT = _CC_DIGIT,
1249 _CC_ENUM_GRAPH = _CC_GRAPH,
1250 _CC_ENUM_LOWER = _CC_LOWER,
1251 _CC_ENUM_PRINT = _CC_PRINT,
1252 _CC_ENUM_PUNCT = _CC_PUNCT,
e8d596e0 1253 _CC_ENUM_SPACE = _CC_SPACE,
3a371f2f 1254 _CC_ENUM_UPPER = _CC_UPPER,
e8d596e0 1255 _CC_ENUM_VERTSPACE = _CC_VERTSPACE,
3a371f2f 1256 _CC_ENUM_WORDCHAR = _CC_WORDCHAR,
e8d596e0 1257 _CC_ENUM_XDIGIT = _CC_XDIGIT
3a371f2f
KW
1258} _char_class_number;
1259#endif
1260
86f72d56 1261#define POSIX_CC_COUNT (_HIGHEST_REGCOMP_DOT_H_SYNC + 1)
63c61c3f 1262
6635f04f 1263START_EXTERN_C
96ac0975
NC
1264# ifdef DOINIT
1265EXTCONST U32 PL_charclass[] = {
1266# include "l1_char_class_tab.h"
1267};
1268
1269# else /* ! DOINIT */
1270EXTCONST U32 PL_charclass[];
1271# endif
6635f04f 1272END_EXTERN_C
96ac0975 1273
265c1f46 1274 /* The 1U keeps Solaris from griping when shifting sets the uppermost bit */
430b7c70 1275# define _CC_mask(classnum) (1U << (classnum))
4650c663
KW
1276
1277 /* For internal core Perl use only: the base macro for defining macros like
1278 * isALPHA */
ff7ecfc3 1279# define _generic_isCC(c, classnum) cBOOL(FITS_IN_8_BITS(c) \
f4cd282c 1280 && (PL_charclass[(U8) (c)] & _CC_mask(classnum)))
4eeeb416 1281
f4cdb42c
KW
1282 /* The mask for the _A versions of the macros; it just adds in the bit for
1283 * ASCII. */
1284# define _CC_mask_A(classnum) (_CC_mask(classnum) | _CC_mask(_CC_ASCII))
1285
4650c663
KW
1286 /* For internal core Perl use only: the base macro for defining macros like
1287 * isALPHA_A. The foo_A version makes sure that both the desired bit and
1288 * the ASCII bit are present */
b7d90381
KW
1289# define _generic_isCC_A(c, classnum) (FITS_IN_8_BITS(c) \
1290 && ((PL_charclass[(U8) (c)] & _CC_mask_A(classnum)) \
1291 == _CC_mask_A(classnum)))
f4cdb42c 1292
26c1d9d8
KW
1293/* On ASCII platforms certain classes form a single range. It's faster to
1294 * special case these. isDIGIT is a single range on all platforms */
b877c1ff
KW
1295# ifdef EBCDIC
1296# define isALPHA_A(c) _generic_isCC_A(c, _CC_ALPHA)
1297# define isGRAPH_A(c) _generic_isCC_A(c, _CC_GRAPH)
1298# define isLOWER_A(c) _generic_isCC_A(c, _CC_LOWER)
1299# define isPRINT_A(c) _generic_isCC_A(c, _CC_PRINT)
1300# define isUPPER_A(c) _generic_isCC_A(c, _CC_UPPER)
1301# else
26c1d9d8 1302 /* By folding the upper and lowercase, we can use a single range */
b877c1ff 1303# define isALPHA_A(c) inRANGE((~('A' ^ 'a') & (c)), 'A', 'Z')
26c1d9d8 1304# define isGRAPH_A(c) inRANGE(c, ' ' + 1, 0x7e)
b877c1ff
KW
1305# define isLOWER_A(c) inRANGE(c, 'a', 'z')
1306# define isPRINT_A(c) inRANGE(c, ' ', 0x7e)
1307# define isUPPER_A(c) inRANGE(c, 'A', 'Z')
1308# endif
15861f94 1309# define isALPHANUMERIC_A(c) _generic_isCC_A(c, _CC_ALPHANUMERIC)
f4cdb42c
KW
1310# define isBLANK_A(c) _generic_isCC_A(c, _CC_BLANK)
1311# define isCNTRL_A(c) _generic_isCC_A(c, _CC_CNTRL)
b877c1ff 1312# define isDIGIT_A(c) inRANGE(c, '0', '9')
f4cdb42c
KW
1313# define isPUNCT_A(c) _generic_isCC_A(c, _CC_PUNCT)
1314# define isSPACE_A(c) _generic_isCC_A(c, _CC_SPACE)
f4cdb42c 1315# define isWORDCHAR_A(c) _generic_isCC_A(c, _CC_WORDCHAR)
b7d90381
KW
1316# define isXDIGIT_A(c) _generic_isCC(c, _CC_XDIGIT) /* No non-ASCII xdigits
1317 */
d95f8b6a 1318# define isIDFIRST_A(c) _generic_isCC_A(c, _CC_IDFIRST)
3ded5eb0
KW
1319# define isALPHA_L1(c) _generic_isCC(c, _CC_ALPHA)
1320# define isALPHANUMERIC_L1(c) _generic_isCC(c, _CC_ALPHANUMERIC)
1321# define isBLANK_L1(c) _generic_isCC(c, _CC_BLANK)
1322
1323 /* continuation character for legal NAME in \N{NAME} */
1324# define isCHARNAME_CONT(c) _generic_isCC(c, _CC_CHARNAME_CONT)
1325
1326# define isCNTRL_L1(c) _generic_isCC(c, _CC_CNTRL)
1327# define isGRAPH_L1(c) _generic_isCC(c, _CC_GRAPH)
1328# define isLOWER_L1(c) _generic_isCC(c, _CC_LOWER)
1329# define isPRINT_L1(c) _generic_isCC(c, _CC_PRINT)
b7d90381 1330# define isPSXSPC_L1(c) isSPACE_L1(c)
3ded5eb0
KW
1331# define isPUNCT_L1(c) _generic_isCC(c, _CC_PUNCT)
1332# define isSPACE_L1(c) _generic_isCC(c, _CC_SPACE)
1333# define isUPPER_L1(c) _generic_isCC(c, _CC_UPPER)
1334# define isWORDCHAR_L1(c) _generic_isCC(c, _CC_WORDCHAR)
1335# define isIDFIRST_L1(c) _generic_isCC(c, _CC_IDFIRST)
f4cdb42c 1336
0852beac
KW
1337# ifdef EBCDIC
1338# define isASCII(c) _generic_isCC(c, _CC_ASCII)
1339# endif
1340
f12c0118
KW
1341 /* Participates in a single-character fold with a character above 255 */
1342# 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)))
1343
1344 /* Like the above, but also can be part of a multi-char fold */
f4cd282c 1345# 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 1346
4eeeb416 1347# define _isQUOTEMETA(c) _generic_isCC(c, _CC_QUOTEMETA)
26faadbd 1348# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1349 _generic_isCC(c, _CC_NON_FINAL_FOLD)
37ede926 1350# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
b7d90381 1351 _generic_isCC(c, _CC_IS_IN_SOME_FOLD)
8e35b056
KW
1352# define _IS_MNEMONIC_CNTRL_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) \
1353 _generic_isCC(c, _CC_MNEMONIC_CNTRL)
687c8d01 1354#else /* else we don't have perl.h H_PERL */
3ded5eb0
KW
1355
1356 /* If we don't have perl.h, we are compiling a utility program. Below we
1357 * hard-code various macro definitions that wouldn't otherwise be available
fc273927 1358 * to it. Most are coded based on first principles. These are written to
74665a89 1359 * avoid EBCDIC vs. ASCII #ifdef's as much as possible. */
182c4ace 1360# define isDIGIT_A(c) inRANGE(c, '0', '9')
0852beac 1361# define isBLANK_A(c) ((c) == ' ' || (c) == '\t')
74665a89
KW
1362# define isSPACE_A(c) (isBLANK_A(c) \
1363 || (c) == '\n' \
1364 || (c) == '\r' \
1365 || (c) == '\v' \
0852beac 1366 || (c) == '\f')
74665a89
KW
1367 /* On EBCDIC, there are gaps between 'i' and 'j'; 'r' and 's'. Same for
1368 * uppercase. The tests for those aren't necessary on ASCII, but hurt only
1369 * performance (if optimization isn't on), and allow the same code to be
1370 * used for both platform types */
182c4ace
KW
1371# define isLOWER_A(c) inRANGE((c), 'a', 'i') \
1372 || inRANGE((c), 'j', 'r') \
1373 || inRANGE((c), 's', 'z')
1374# define isUPPER_A(c) inRANGE((c), 'A', 'I') \
1375 || inRANGE((c), 'J', 'R') \
1376 || inRANGE((c), 'S', 'Z')
a4d7a999
KW
1377# define isALPHA_A(c) (isUPPER_A(c) || isLOWER_A(c))
1378# define isALPHANUMERIC_A(c) (isALPHA_A(c) || isDIGIT_A(c))
3ded5eb0 1379# define isWORDCHAR_A(c) (isALPHANUMERIC_A(c) || (c) == '_')
0852beac 1380# define isIDFIRST_A(c) (isALPHA_A(c) || (c) == '_')
182c4ace
KW
1381# define isXDIGIT_A(c) ( isDIGIT_A(c) \
1382 || inRANGE((c), 'a', 'f') \
1383 || inRANGE((c), 'A', 'F')
74665a89
KW
1384# define isPUNCT_A(c) ((c) == '-' || (c) == '!' || (c) == '"' \
1385 || (c) == '#' || (c) == '$' || (c) == '%' \
1386 || (c) == '&' || (c) == '\'' || (c) == '(' \
1387 || (c) == ')' || (c) == '*' || (c) == '+' \
1388 || (c) == ',' || (c) == '.' || (c) == '/' \
1389 || (c) == ':' || (c) == ';' || (c) == '<' \
1390 || (c) == '=' || (c) == '>' || (c) == '?' \
1391 || (c) == '@' || (c) == '[' || (c) == '\\' \
1392 || (c) == ']' || (c) == '^' || (c) == '_' \
1393 || (c) == '`' || (c) == '{' || (c) == '|' \
1394 || (c) == '}' || (c) == '~')
1395# define isGRAPH_A(c) (isALPHANUMERIC_A(c) || isPUNCT_A(c))
1396# define isPRINT_A(c) (isGRAPH_A(c) || (c) == ' ')
3ded5eb0 1397
0852beac 1398# ifdef EBCDIC
74665a89
KW
1399 /* The below is accurate for the 3 EBCDIC code pages traditionally
1400 * supported by perl. The only difference between them in the controls
1401 * is the position of \n, and that is represented symbolically below */
1402# define isCNTRL_A(c) ((c) == '\0' || (c) == '\a' || (c) == '\b' \
1403 || (c) == '\f' || (c) == '\n' || (c) == '\r' \
1404 || (c) == '\t' || (c) == '\v' \
182c4ace 1405 || inRANGE((c), 1, 3) /* SOH, STX, ETX */ \
74665a89 1406 || (c) == 7 /* U+7F DEL */ \
182c4ace
KW
1407 || inRANGE((c), 0x0E, 0x13) /* SO SI DLE \
1408 DC[1-3] */ \
74665a89
KW
1409 || (c) == 0x18 /* U+18 CAN */ \
1410 || (c) == 0x19 /* U+19 EOM */ \
182c4ace 1411 || inRANGE((c), 0x1C, 0x1F) /* [FGRU]S */ \
74665a89
KW
1412 || (c) == 0x26 /* U+17 ETB */ \
1413 || (c) == 0x27 /* U+1B ESC */ \
1414 || (c) == 0x2D /* U+05 ENQ */ \
1415 || (c) == 0x2E /* U+06 ACK */ \
1416 || (c) == 0x32 /* U+16 SYN */ \
1417 || (c) == 0x37 /* U+04 EOT */ \
1418 || (c) == 0x3C /* U+14 DC4 */ \
1419 || (c) == 0x3D /* U+15 NAK */ \
1420 || (c) == 0x3F)/* U+1A SUB */
0852beac 1421# define isASCII(c) (isCNTRL_A(c) || isPRINT_A(c))
74665a89
KW
1422# else /* isASCII is already defined for ASCII platforms, so can use that to
1423 define isCNTRL */
1424# define isCNTRL_A(c) (isASCII(c) && ! isPRINT_A(c))
0852beac
KW
1425# endif
1426
3ffc8c70 1427 /* The _L1 macros may be unnecessary for the utilities; I (khw) added them
caa94d35
KW
1428 * during debugging, and it seems best to keep them. We may be called
1429 * without NATIVE_TO_LATIN1 being defined. On ASCII platforms, it doesn't
1430 * do anything anyway, so make it not a problem */
1431# if ! defined(EBCDIC) && ! defined(NATIVE_TO_LATIN1)
1432# define NATIVE_TO_LATIN1(ch) (ch)
1433# endif
3ded5eb0
KW
1434# define isALPHA_L1(c) (isUPPER_L1(c) || isLOWER_L1(c))
1435# define isALPHANUMERIC_L1(c) (isALPHA_L1(c) || isDIGIT_A(c))
1436# define isBLANK_L1(c) (isBLANK_A(c) \
1437 || (FITS_IN_8_BITS(c) \
1438 && NATIVE_TO_LATIN1((U8) c) == 0xA0))
1439# define isCNTRL_L1(c) (FITS_IN_8_BITS(c) && (! isPRINT_L1(c)))
1440# define isGRAPH_L1(c) (isPRINT_L1(c) && (! isBLANK_L1(c)))
1441# define isLOWER_L1(c) (isLOWER_A(c) \
1442 || (FITS_IN_8_BITS(c) \
ae683a5f 1443 && (( NATIVE_TO_LATIN1((U8) c) >= 0xDF \
3ded5eb0
KW
1444 && NATIVE_TO_LATIN1((U8) c) != 0xF7) \
1445 || NATIVE_TO_LATIN1((U8) c) == 0xAA \
1446 || NATIVE_TO_LATIN1((U8) c) == 0xBA \
1447 || NATIVE_TO_LATIN1((U8) c) == 0xB5)))
1448# define isPRINT_L1(c) (isPRINT_A(c) \
1449 || (FITS_IN_8_BITS(c) \
1450 && NATIVE_TO_LATIN1((U8) c) >= 0xA0))
3ded5eb0
KW
1451# define isPUNCT_L1(c) (isPUNCT_A(c) \
1452 || (FITS_IN_8_BITS(c) \
ae683a5f 1453 && ( NATIVE_TO_LATIN1((U8) c) == 0xA1 \
3ded5eb0
KW
1454 || NATIVE_TO_LATIN1((U8) c) == 0xA7 \
1455 || NATIVE_TO_LATIN1((U8) c) == 0xAB \
1456 || NATIVE_TO_LATIN1((U8) c) == 0xB6 \
1457 || NATIVE_TO_LATIN1((U8) c) == 0xB7 \
1458 || NATIVE_TO_LATIN1((U8) c) == 0xBB \
1459 || NATIVE_TO_LATIN1((U8) c) == 0xBF)))
1460# define isSPACE_L1(c) (isSPACE_A(c) \
1461 || (FITS_IN_8_BITS(c) \
ae683a5f 1462 && ( NATIVE_TO_LATIN1((U8) c) == 0x85 \
3ded5eb0
KW
1463 || NATIVE_TO_LATIN1((U8) c) == 0xA0)))
1464# define isUPPER_L1(c) (isUPPER_A(c) \
1465 || (FITS_IN_8_BITS(c) \
182c4ace
KW
1466 && ( IN_RANGE(NATIVE_TO_LATIN1((U8) c), \
1467 0xC0, 0xDE) \
3ded5eb0
KW
1468 && NATIVE_TO_LATIN1((U8) c) != 0xD7)))
1469# define isWORDCHAR_L1(c) (isIDFIRST_L1(c) || isDIGIT_A(c))
1470# define isIDFIRST_L1(c) (isALPHA_L1(c) || NATIVE_TO_LATIN1(c) == '_')
1471# define isCHARNAME_CONT(c) (isWORDCHAR_L1(c) \
1472 || isBLANK_L1(c) \
1473 || (c) == '-' \
1474 || (c) == '(' \
1475 || (c) == ')')
1476 /* The following are not fully accurate in the above-ASCII range. I (khw)
1477 * don't think it's necessary to be so for the purposes where this gets
1478 * compiled */
1479# define _isQUOTEMETA(c) (FITS_IN_8_BITS(c) && ! isWORDCHAR_L1(c))
1480# define _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) isALPHA_L1(c)
1481
1482 /* And these aren't accurate at all. They are useful only for above
1483 * Latin1, which utilities and bootstrapping don't deal with */
1484# define _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c) 0
6838b41e 1485# define _HAS_NONLATIN1_SIMPLE_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
3ded5eb0
KW
1486# define _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(c) 0
1487
1488 /* Many of the macros later in this file are defined in terms of these. By
1489 * implementing them with a function, which converts the class number into
1490 * a call to the desired macro, all of the later ones work. However, that
1491 * function won't be actually defined when building a utility program (no
1492 * perl.h), and so a compiler error will be generated if one is attempted
1493 * to be used. And the above-Latin1 code points require Unicode tables to
1494 * be present, something unlikely to be the case when bootstrapping */
1495# define _generic_isCC(c, classnum) \
1496 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), TRUE))
1497# define _generic_isCC_A(c, classnum) \
1498 (FITS_IN_8_BITS(c) && S_bootstrap_ctype((U8) (c), (classnum), FALSE))
687c8d01 1499#endif /* End of no perl.h H_PERL */
8a58bdcf 1500
e66b99e9
KW
1501#define isALPHANUMERIC(c) isALPHANUMERIC_A(c)
1502#define isALPHA(c) isALPHA_A(c)
0852beac
KW
1503#define isASCII_A(c) isASCII(c)
1504#define isASCII_L1(c) isASCII(c)
e66b99e9
KW
1505#define isBLANK(c) isBLANK_A(c)
1506#define isCNTRL(c) isCNTRL_A(c)
1507#define isDIGIT(c) isDIGIT_A(c)
1508#define isGRAPH(c) isGRAPH_A(c)
1509#define isIDFIRST(c) isIDFIRST_A(c)
1510#define isLOWER(c) isLOWER_A(c)
1511#define isPRINT(c) isPRINT_A(c)
779cf272 1512#define isPSXSPC_A(c) isSPACE_A(c)
e66b99e9 1513#define isPSXSPC(c) isPSXSPC_A(c)
779cf272 1514#define isPSXSPC_L1(c) isSPACE_L1(c)
e66b99e9
KW
1515#define isPUNCT(c) isPUNCT_A(c)
1516#define isSPACE(c) isSPACE_A(c)
1517#define isUPPER(c) isUPPER_A(c)
1518#define isWORDCHAR(c) isWORDCHAR_A(c)
1519#define isXDIGIT(c) isXDIGIT_A(c)
1520
1521/* ASCII casing. These could also be written as
1522 #define toLOWER(c) (isASCII(c) ? toLOWER_LATIN1(c) : (c))
1523 #define toUPPER(c) (isASCII(c) ? toUPPER_LATIN1_MOD(c) : (c))
1524 which uses table lookup and mask instead of subtraction. (This would
c5e9991e
KW
1525 work because the _MOD does not apply in the ASCII range).
1526
1527 These actually are UTF-8 invariant casing, not just ASCII, as any non-ASCII
1528 UTF-8 invariants are neither upper nor lower. (Only on EBCDIC platforms are
1529 there non-ASCII invariants, and all of them are controls.) */
68067e4e
DM
1530#define toLOWER(c) (isUPPER(c) ? (U8)((c) + ('a' - 'A')) : (c))
1531#define toUPPER(c) (isLOWER(c) ? (U8)((c) - ('a' - 'A')) : (c))
bbce6d69 1532
25200305
KW
1533/* In the ASCII range, these are equivalent to what they're here defined to be.
1534 * But by creating these definitions, other code doesn't have to be aware of
c5e9991e
KW
1535 * this detail. Actually this works for all UTF-8 invariants, not just the
1536 * ASCII range. (EBCDIC platforms can have non-ASCII invariants.) */
25200305 1537#define toFOLD(c) toLOWER(c)
25200305
KW
1538#define toTITLE(c) toUPPER(c)
1539
c753c8d3
KW
1540#define toLOWER_A(c) toLOWER(c)
1541#define toUPPER_A(c) toUPPER(c)
25200305
KW
1542#define toFOLD_A(c) toFOLD(c)
1543#define toTITLE_A(c) toTITLE(c)
1a0901db 1544
4650c663 1545/* Use table lookup for speed; returns the input itself if is out-of-range */
b2bf251f 1546#define toLOWER_LATIN1(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1547 ? (c) \
f4cd282c 1548 : PL_latin1_lc[ (U8) (c) ])
c753c8d3
KW
1549#define toLOWER_L1(c) toLOWER_LATIN1(c) /* Synonym for consistency */
1550
1a0901db 1551/* Modified uc. Is correct uc except for three non-ascii chars which are
4650c663
KW
1552 * all mapped to one of them, and these need special handling; returns the
1553 * input itself if is out-of-range */
b2bf251f 1554#define toUPPER_LATIN1_MOD(c) ((! FITS_IN_8_BITS(c)) \
8e7c6e7d 1555 ? (c) \
f4cd282c 1556 : PL_mod_latin1_uc[ (U8) (c) ])
31f05a37 1557#define IN_UTF8_CTYPE_LOCALE PL_in_utf8_CTYPE_locale
84061b6a 1558
beab9ebe
KW
1559/* Use foo_LC_uvchr() instead of these for beyond the Latin1 range */
1560
1561/* For internal core Perl use only: the base macro for defining macros like
1562 * isALPHA_LC, which uses the current LC_CTYPE locale. 'c' is the code point
31f05a37
KW
1563 * (0-255) to check. In a UTF-8 locale, the result is the same as calling
1564 * isFOO_L1(); the 'utf8_locale_classnum' parameter is something like
1565 * _CC_UPPER, which gives the class number for doing this. For non-UTF-8
1566 * locales, the code to actually do the test this is passed in 'non_utf8'. If
1567 * 'c' is above 255, 0 is returned. For accessing the full range of possible
1568 * code points under locale rules, use the macros based on _generic_LC_uvchr
1569 * instead of this. */
beab9ebe
KW
1570#define _generic_LC_base(c, utf8_locale_classnum, non_utf8) \
1571 (! FITS_IN_8_BITS(c) \
1572 ? 0 \
31f05a37
KW
1573 : IN_UTF8_CTYPE_LOCALE \
1574 ? cBOOL(PL_charclass[(U8) (c)] & _CC_mask(utf8_locale_classnum)) \
beab9ebe
KW
1575 : cBOOL(non_utf8))
1576
1577/* For internal core Perl use only: a helper macro for defining macros like
1578 * isALPHA_LC. 'c' is the code point (0-255) to check. The function name to
1579 * actually do this test is passed in 'non_utf8_func', which is called on 'c',
1580 * casting 'c' to the macro _LC_CAST, which should not be parenthesized. See
1581 * _generic_LC_base for more info */
1582#define _generic_LC(c, utf8_locale_classnum, non_utf8_func) \
1583 _generic_LC_base(c,utf8_locale_classnum, \
1584 non_utf8_func( (_LC_CAST) (c)))
1585
1586/* For internal core Perl use only: like _generic_LC, but also returns TRUE if
1587 * 'c' is the platform's native underscore character */
1588#define _generic_LC_underscore(c,utf8_locale_classnum,non_utf8_func) \
1589 _generic_LC_base(c, utf8_locale_classnum, \
1590 (non_utf8_func( (_LC_CAST) (c)) \
1591 || (char)(c) == '_'))
1592
1593/* These next three are also for internal core Perl use only: case-change
247985d4
KW
1594 * helper macros. The reason for using the PL_latin arrays is in case the
1595 * system function is defective; it ensures uniform results that conform to the
b257a28c 1596 * Unicod standard. It does not handle the anomalies in UTF-8 Turkic locales */
beab9ebe
KW
1597#define _generic_toLOWER_LC(c, function, cast) (! FITS_IN_8_BITS(c) \
1598 ? (c) \
31f05a37
KW
1599 : (IN_UTF8_CTYPE_LOCALE) \
1600 ? PL_latin1_lc[ (U8) (c) ] \
5a10328c 1601 : (cast)function((cast)(c)))
beab9ebe 1602
31f05a37
KW
1603/* Note that the result can be larger than a byte in a UTF-8 locale. It
1604 * returns a single value, so can't adequately return the upper case of LATIN
1605 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1606 * values "SS"); instead it asserts against that under DEBUGGING, and
b257a28c
KW
1607 * otherwise returns its input. It does not handle the anomalies in UTF-8
1608 * Turkic locales. */
beab9ebe
KW
1609#define _generic_toUPPER_LC(c, function, cast) \
1610 (! FITS_IN_8_BITS(c) \
1611 ? (c) \
31f05a37 1612 : ((! IN_UTF8_CTYPE_LOCALE) \
b7d90381 1613 ? (cast)function((cast)(c)) \
31f05a37
KW
1614 : ((((U8)(c)) == MICRO_SIGN) \
1615 ? GREEK_CAPITAL_LETTER_MU \
1616 : ((((U8)(c)) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) \
1617 ? LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS \
1618 : ((((U8)(c)) == LATIN_SMALL_LETTER_SHARP_S) \
1619 ? (__ASSERT_(0) (c)) \
1620 : PL_mod_latin1_uc[ (U8) (c) ])))))
1621
1622/* Note that the result can be larger than a byte in a UTF-8 locale. It
1623 * returns a single value, so can't adequately return the fold case of LATIN
1624 * SMALL LETTER SHARP S in a UTF-8 locale (which should be a string of two
1625 * values "ss"); instead it asserts against that under DEBUGGING, and
b257a28c
KW
1626 * otherwise returns its input. It does not handle the anomalies in UTF-8
1627 * Turkic locales */
beab9ebe 1628#define _generic_toFOLD_LC(c, function, cast) \
4d7db1b9 1629 ((UNLIKELY((c) == MICRO_SIGN) && IN_UTF8_CTYPE_LOCALE) \
31f05a37 1630 ? GREEK_SMALL_LETTER_MU \
4d7db1b9
KW
1631 : (__ASSERT_(! IN_UTF8_CTYPE_LOCALE \
1632 || (c) != LATIN_SMALL_LETTER_SHARP_S) \
1633 _generic_toLOWER_LC(c, function, cast)))
beab9ebe 1634
84061b6a 1635/* Use the libc versions for these if available. */
f05550c0 1636#if defined(HAS_ISASCII)
84061b6a
KW
1637# define isASCII_LC(c) (FITS_IN_8_BITS(c) && isascii( (U8) (c)))
1638#else
1639# define isASCII_LC(c) isASCII(c)
1640#endif
1641
f05550c0 1642#if defined(HAS_ISBLANK)
84061b6a 1643# define isBLANK_LC(c) _generic_LC(c, _CC_BLANK, isblank)
95f34b6f 1644#else /* Unlike isASCII, varies if in a UTF-8 locale */
278b2b56 1645# define isBLANK_LC(c) ((IN_UTF8_CTYPE_LOCALE) ? isBLANK_L1(c) : isBLANK(c))
84061b6a
KW
1646#endif
1647
f05550c0 1648#define _LC_CAST U8
bbce6d69 1649
f05550c0 1650#ifdef WIN32
375f5f06
KW
1651 /* The Windows functions don't bother to follow the POSIX standard, which
1652 * for example says that something can't both be a printable and a control.
1653 * But Windows treats the \t control as a printable, and does such things
1654 * as making superscripts into both digits and punctuation. This tames
1655 * these flaws by assuming that the definitions of both controls and space
1656 * are correct, and then making sure that other definitions don't have
1657 * weirdnesses, by making sure that isalnum() isn't also ispunct(), etc.
1658 * Not all possible weirdnesses are checked for, just the ones that were
1659 * detected on actual Microsoft code pages */
1660
b7d90381
KW
1661# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1662# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1663
1664# define isALPHA_LC(c) (_generic_LC(c, _CC_ALPHA, isalpha) \
1665 && isALPHANUMERIC_LC(c))
1666# define isALPHANUMERIC_LC(c) (_generic_LC(c, _CC_ALPHANUMERIC, isalnum) && \
1667 ! isPUNCT_LC(c))
1668# define isDIGIT_LC(c) (_generic_LC(c, _CC_DIGIT, isdigit) && \
1669 isALPHANUMERIC_LC(c))
1670# define isGRAPH_LC(c) (_generic_LC(c, _CC_GRAPH, isgraph) && isPRINT_LC(c))
1671# define isIDFIRST_LC(c) (((c) == '_') \
1672 || (_generic_LC(c, _CC_IDFIRST, isalpha) && ! isPUNCT_LC(c)))
1673# define isLOWER_LC(c) (_generic_LC(c, _CC_LOWER, islower) && isALPHA_LC(c))
1674# define isPRINT_LC(c) (_generic_LC(c, _CC_PRINT, isprint) && ! isCNTRL_LC(c))
1675# define isPUNCT_LC(c) (_generic_LC(c, _CC_PUNCT, ispunct) && ! isCNTRL_LC(c))
1676# define isUPPER_LC(c) (_generic_LC(c, _CC_UPPER, isupper) && isALPHA_LC(c))
f05550c0 1677# define isWORDCHAR_LC(c) (((c) == '_') || isALPHANUMERIC_LC(c))
b7d90381
KW
1678# define isXDIGIT_LC(c) (_generic_LC(c, _CC_XDIGIT, isxdigit) \
1679 && isALPHANUMERIC_LC(c))
f05550c0
BF
1680
1681# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1682# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1683# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1684
1685#elif defined(CTYPE256) || (!defined(isascii) && !defined(HAS_ISASCII))
4650c663 1686 /* For most other platforms */
beab9ebe 1687
f05550c0
BF
1688# define isALPHA_LC(c) _generic_LC(c, _CC_ALPHA, isalpha)
1689# define isALPHANUMERIC_LC(c) _generic_LC(c, _CC_ALPHANUMERIC, isalnum)
1690# define isCNTRL_LC(c) _generic_LC(c, _CC_CNTRL, iscntrl)
1691# define isDIGIT_LC(c) _generic_LC(c, _CC_DIGIT, isdigit)
1692# define isGRAPH_LC(c) _generic_LC(c, _CC_GRAPH, isgraph)
1693# define isIDFIRST_LC(c) _generic_LC_underscore(c, _CC_IDFIRST, isalpha)
1694# define isLOWER_LC(c) _generic_LC(c, _CC_LOWER, islower)
1695# define isPRINT_LC(c) _generic_LC(c, _CC_PRINT, isprint)
1696# define isPUNCT_LC(c) _generic_LC(c, _CC_PUNCT, ispunct)
1697# define isSPACE_LC(c) _generic_LC(c, _CC_SPACE, isspace)
1698# define isUPPER_LC(c) _generic_LC(c, _CC_UPPER, isupper)
1699# define isWORDCHAR_LC(c) _generic_LC_underscore(c, _CC_WORDCHAR, isalnum)
1700# define isXDIGIT_LC(c) _generic_LC(c, _CC_XDIGIT, isxdigit)
1701
1702
1703# define toLOWER_LC(c) _generic_toLOWER_LC((c), tolower, U8)
1704# define toUPPER_LC(c) _generic_toUPPER_LC((c), toupper, U8)
1705# define toFOLD_LC(c) _generic_toFOLD_LC((c), tolower, U8)
1706
1707#else /* The final fallback position */
1708
b7d90381
KW
1709# define isALPHA_LC(c) (isascii(c) && isalpha(c))
1710# define isALPHANUMERIC_LC(c) (isascii(c) && isalnum(c))
1711# define isCNTRL_LC(c) (isascii(c) && iscntrl(c))
1712# define isDIGIT_LC(c) (isascii(c) && isdigit(c))
1713# define isGRAPH_LC(c) (isascii(c) && isgraph(c))
f05550c0 1714# define isIDFIRST_LC(c) (isascii(c) && (isalpha(c) || (c) == '_'))
b7d90381
KW
1715# define isLOWER_LC(c) (isascii(c) && islower(c))
1716# define isPRINT_LC(c) (isascii(c) && isprint(c))
1717# define isPUNCT_LC(c) (isascii(c) && ispunct(c))
1718# define isSPACE_LC(c) (isascii(c) && isspace(c))
1719# define isUPPER_LC(c) (isascii(c) && isupper(c))
f05550c0 1720# define isWORDCHAR_LC(c) (isascii(c) && (isalnum(c) || (c) == '_'))
b7d90381 1721# define isXDIGIT_LC(c) (isascii(c) && isxdigit(c))
f05550c0
BF
1722
1723# define toLOWER_LC(c) (isascii(c) ? tolower(c) : (c))
1724# define toUPPER_LC(c) (isascii(c) ? toupper(c) : (c))
1725# define toFOLD_LC(c) (isascii(c) ? tolower(c) : (c))
bbce6d69 1726
f05550c0 1727#endif
55204971 1728
eba68aa0
KW
1729#define isIDCONT(c) isWORDCHAR(c)
1730#define isIDCONT_A(c) isWORDCHAR_A(c)
1731#define isIDCONT_L1(c) isWORDCHAR_L1(c)
1732#define isIDCONT_LC(c) isWORDCHAR_LC(c)
13380643 1733#define isPSXSPC_LC(c) isSPACE_LC(c)
aaa51d5e 1734
4650c663 1735/* For internal core Perl use only: the base macros for defining macros like
d0da05db
KW
1736 * isALPHA_uvchr. 'c' is the code point to check. 'classnum' is the POSIX class
1737 * number defined earlier in this file. _generic_uvchr() is used for POSIX
4650c663
KW
1738 * classes where there is a macro or function 'above_latin1' that takes the
1739 * single argument 'c' and returns the desired value. These exist for those
1740 * classes which have simple definitions, avoiding the overhead of a hash
d0da05db 1741 * lookup or inversion list binary search. _generic_swash_uvchr() can be used
4650c663 1742 * for classes where that overhead is faster than a direct lookup.
d0da05db 1743 * _generic_uvchr() won't compile if 'c' isn't unsigned, as it won't match the
4650c663
KW
1744 * 'above_latin1' prototype. _generic_isCC() macro does bounds checking, so
1745 * have duplicate checks here, so could create versions of the macros that
1746 * don't, but experiments show that gcc optimizes them out anyway. */
66c17564
KW
1747
1748/* Note that all ignore 'use bytes' */
1e222e4f
KW
1749#define _generic_uvchr(classnum, above_latin1, c) ((c) < 256 \
1750 ? _generic_isCC(c, classnum) \
cd500f2f 1751 : above_latin1(c))
1e222e4f
KW
1752#define _generic_swash_uvchr(classnum, c) ((c) < 256 \
1753 ? _generic_isCC(c, classnum) \
922e8cb4 1754 : _is_uni_FOO(classnum, c))
d0da05db
KW
1755#define isALPHA_uvchr(c) _generic_swash_uvchr(_CC_ALPHA, c)
1756#define isALPHANUMERIC_uvchr(c) _generic_swash_uvchr(_CC_ALPHANUMERIC, c)
1757#define isASCII_uvchr(c) isASCII(c)
1758#define isBLANK_uvchr(c) _generic_uvchr(_CC_BLANK, is_HORIZWS_cp_high, c)
1759#define isCNTRL_uvchr(c) isCNTRL_L1(c) /* All controls are in Latin1 */
1760#define isDIGIT_uvchr(c) _generic_swash_uvchr(_CC_DIGIT, c)
1761#define isGRAPH_uvchr(c) _generic_swash_uvchr(_CC_GRAPH, c)
1e222e4f
KW
1762#define isIDCONT_uvchr(c) \
1763 _generic_uvchr(_CC_WORDCHAR, _is_uni_perl_idcont, c)
1764#define isIDFIRST_uvchr(c) \
1765 _generic_uvchr(_CC_IDFIRST, _is_uni_perl_idstart, c)
d0da05db
KW
1766#define isLOWER_uvchr(c) _generic_swash_uvchr(_CC_LOWER, c)
1767#define isPRINT_uvchr(c) _generic_swash_uvchr(_CC_PRINT, c)
1768
1769#define isPUNCT_uvchr(c) _generic_swash_uvchr(_CC_PUNCT, c)
1770#define isSPACE_uvchr(c) _generic_uvchr(_CC_SPACE, is_XPERLSPACE_cp_high, c)
1771#define isPSXSPC_uvchr(c) isSPACE_uvchr(c)
1772
1773#define isUPPER_uvchr(c) _generic_swash_uvchr(_CC_UPPER, c)
1774#define isVERTWS_uvchr(c) _generic_uvchr(_CC_VERTSPACE, is_VERTWS_cp_high, c)
1775#define isWORDCHAR_uvchr(c) _generic_swash_uvchr(_CC_WORDCHAR, c)
1776#define isXDIGIT_uvchr(c) _generic_uvchr(_CC_XDIGIT, is_XDIGIT_cp_high, c)
1777
1778#define toFOLD_uvchr(c,s,l) to_uni_fold(c,s,l)
1779#define toLOWER_uvchr(c,s,l) to_uni_lower(c,s,l)
1780#define toTITLE_uvchr(c,s,l) to_uni_title(c,s,l)
1781#define toUPPER_uvchr(c,s,l) to_uni_upper(c,s,l)
1782
1783/* For backwards compatibility, even though '_uni' should mean official Unicode
1784 * code points, in Perl it means native for those below 256 */
1785#define isALPHA_uni(c) isALPHA_uvchr(c)
1786#define isALPHANUMERIC_uni(c) isALPHANUMERIC_uvchr(c)
1787#define isASCII_uni(c) isASCII_uvchr(c)
1788#define isBLANK_uni(c) isBLANK_uvchr(c)
1789#define isCNTRL_uni(c) isCNTRL_uvchr(c)
1790#define isDIGIT_uni(c) isDIGIT_uvchr(c)
1791#define isGRAPH_uni(c) isGRAPH_uvchr(c)
1792#define isIDCONT_uni(c) isIDCONT_uvchr(c)
1793#define isIDFIRST_uni(c) isIDFIRST_uvchr(c)
1794#define isLOWER_uni(c) isLOWER_uvchr(c)
1795#define isPRINT_uni(c) isPRINT_uvchr(c)
1796#define isPUNCT_uni(c) isPUNCT_uvchr(c)
1797#define isSPACE_uni(c) isSPACE_uvchr(c)
1798#define isPSXSPC_uni(c) isPSXSPC_uvchr(c)
1799#define isUPPER_uni(c) isUPPER_uvchr(c)
1800#define isVERTWS_uni(c) isVERTWS_uvchr(c)
1801#define isWORDCHAR_uni(c) isWORDCHAR_uvchr(c)
1802#define isXDIGIT_uni(c) isXDIGIT_uvchr(c)
1803#define toFOLD_uni(c,s,l) toFOLD_uvchr(c,s,l)
1804#define toLOWER_uni(c,s,l) toLOWER_uvchr(c,s,l)
1805#define toTITLE_uni(c,s,l) toTITLE_uvchr(c,s,l)
1806#define toUPPER_uni(c,s,l) toUPPER_uvchr(c,s,l)
a0ed51b3 1807
4650c663
KW
1808/* For internal core Perl use only: the base macros for defining macros like
1809 * isALPHA_LC_uvchr. These are like isALPHA_LC, but the input can be any code
d0da05db 1810 * point, not just 0-255. Like _generic_uvchr, there are two versions, one for
4650c663 1811 * simple class definitions; the other for more complex. These are like
d0da05db 1812 * _generic_uvchr, so see it for more info. */
cd500f2f
KW
1813#define _generic_LC_uvchr(latin1, above_latin1, c) \
1814 (c < 256 ? latin1(c) : above_latin1(c))
1815#define _generic_LC_swash_uvchr(latin1, classnum, c) \
1816 (c < 256 ? latin1(c) : _is_uni_FOO(classnum, c))
1817
1818#define isALPHA_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHA_LC, _CC_ALPHA, c)
1819#define isALPHANUMERIC_LC_uvchr(c) _generic_LC_swash_uvchr(isALPHANUMERIC_LC, \
1820 _CC_ALPHANUMERIC, c)
b7d90381
KW
1821#define isASCII_LC_uvchr(c) isASCII_LC(c)
1822#define isBLANK_LC_uvchr(c) _generic_LC_uvchr(isBLANK_LC, \
1823 is_HORIZWS_cp_high, c)
feeab5a9 1824#define isCNTRL_LC_uvchr(c) (c < 256 ? isCNTRL_LC(c) : 0)
cd500f2f
KW
1825#define isDIGIT_LC_uvchr(c) _generic_LC_swash_uvchr(isDIGIT_LC, _CC_DIGIT, c)
1826#define isGRAPH_LC_uvchr(c) _generic_LC_swash_uvchr(isGRAPH_LC, _CC_GRAPH, c)
b7d90381 1827#define isIDCONT_LC_uvchr(c) _generic_LC_uvchr(isIDCONT_LC, \
eba68aa0 1828 _is_uni_perl_idcont, c)
b7d90381 1829#define isIDFIRST_LC_uvchr(c) _generic_LC_uvchr(isIDFIRST_LC, \
cd500f2f
KW
1830 _is_uni_perl_idstart, c)
1831#define isLOWER_LC_uvchr(c) _generic_LC_swash_uvchr(isLOWER_LC, _CC_LOWER, c)
1832#define isPRINT_LC_uvchr(c) _generic_LC_swash_uvchr(isPRINT_LC, _CC_PRINT, c)
b7d90381 1833#define isPSXSPC_LC_uvchr(c) isSPACE_LC_uvchr(c)
cd500f2f 1834#define isPUNCT_LC_uvchr(c) _generic_LC_swash_uvchr(isPUNCT_LC, _CC_PUNCT, c)
b7d90381 1835#define isSPACE_LC_uvchr(c) _generic_LC_uvchr(isSPACE_LC, \
509fb054 1836 is_XPERLSPACE_cp_high, c)
cd500f2f 1837#define isUPPER_LC_uvchr(c) _generic_LC_swash_uvchr(isUPPER_LC, _CC_UPPER, c)
b7d90381 1838#define isWORDCHAR_LC_uvchr(c) _generic_LC_swash_uvchr(isWORDCHAR_LC, \
cd500f2f 1839 _CC_WORDCHAR, c)
b7d90381
KW
1840#define isXDIGIT_LC_uvchr(c) _generic_LC_uvchr(isXDIGIT_LC, \
1841 is_XDIGIT_cp_high, c)
e712593e 1842
b7d90381 1843#define isBLANK_LC_uni(c) isBLANK_LC_uvchr(UNI_TO_NATIVE(c))
aaa51d5e 1844
4650c663
KW
1845/* For internal core Perl use only: the base macros for defining macros like
1846 * isALPHA_utf8. These are like the earlier defined macros, but take an input
1847 * UTF-8 encoded string 'p'. If the input is in the Latin1 range, use
1848 * the Latin1 macro 'classnum' on 'p'. Otherwise use the value given by the
1849 * 'utf8' parameter. This relies on the fact that ASCII characters have the
1850 * same representation whether utf8 or not. Note that it assumes that the utf8
1851 * has been validated, and ignores 'use bytes' */
34aeb2e9
KW
1852#define _base_generic_utf8(enum_name, name, p, use_locale ) \
1853 _is_utf8_FOO(CAT2(_CC_, enum_name), \
1854 (const U8 *) p, \
1855 "is" STRINGIFY(name) "_utf8", \
1856 "is" STRINGIFY(name) "_utf8_safe", \
1857 1, use_locale, __FILE__,__LINE__)
1858
1859#define _generic_utf8(name, p) _base_generic_utf8(name, name, p, 0)
da8c1a98
KW
1860
1861/* The "_safe" macros make sure that we don't attempt to read beyond 'e', but
1862 * they don't otherwise go out of their way to look for malformed UTF-8. If
1863 * they can return accurate results without knowing if the input is otherwise
1864 * malformed, they do so. For example isASCII is accurate in spite of any
1865 * non-length malformations because it looks only at a single byte. Likewise
1866 * isDIGIT looks just at the first byte for code points 0-255, as all UTF-8
1867 * variant ones return FALSE. But, if the input has to be well-formed in order
1868 * for the results to be accurate, the macros will test and if malformed will
1869 * call a routine to die
1870 *
1871 * Except for toke.c, the macros do assume that e > p, asserting that on
1872 * DEBUGGING builds. Much code that calls these depends on this being true,
1873 * for other reasons. toke.c is treated specially as using the regular
1874 * assertion breaks it in many ways. All strings that these operate on there
1875 * are supposed to have an extra NUL character at the end, so that *e = \0. A
1876 * bunch of code in toke.c assumes that this is true, so the assertion allows
1877 * for that */
1878#ifdef PERL_IN_TOKE_C
1879# define _utf8_safe_assert(p,e) ((e) > (p) || ((e) == (p) && *(p) == '\0'))
1880#else
1881# define _utf8_safe_assert(p,e) ((e) > (p))
1882#endif
1883
1884#define _generic_utf8_safe(classnum, p, e, above_latin1) \
1885 (__ASSERT_(_utf8_safe_assert(p, e)) \
1886 (UTF8_IS_INVARIANT(*(p))) \
1887 ? _generic_isCC(*(p), classnum) \
1888 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
1889 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
1890 ? _generic_isCC(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1 )), \
1891 classnum) \
1892 : (_force_out_malformed_utf8_message( \
1893 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
1894 : above_latin1))
b7d90381
KW
1895/* Like the above, but calls 'above_latin1(p)' to get the utf8 value.
1896 * 'above_latin1' can be a macro */
da8c1a98
KW
1897#define _generic_func_utf8_safe(classnum, above_latin1, p, e) \
1898 _generic_utf8_safe(classnum, p, e, above_latin1(p, e))
1899#define _generic_non_swash_utf8_safe(classnum, above_latin1, p, e) \
1900 _generic_utf8_safe(classnum, p, e, \
1901 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1902 ? (_force_out_malformed_utf8_message( \
1903 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1904 : above_latin1(p)))
4650c663 1905/* Like the above, but passes classnum to _isFOO_utf8(), instead of having an
cd500f2f 1906 * 'above_latin1' parameter */
da8c1a98
KW
1907#define _generic_swash_utf8_safe(classnum, p, e) \
1908_generic_utf8_safe(classnum, p, e, _is_utf8_FOO_with_len(classnum, p, e))
922e8cb4 1909
cc8ab7c0 1910/* Like the above, but should be used only when it is known that there are no
ff7ecfc3
KW
1911 * characters in the upper-Latin1 range (128-255 on ASCII platforms) which the
1912 * class is TRUE for. Hence it can skip the tests for this range.
1913 * 'above_latin1' should include its arguments */
da8c1a98
KW
1914#define _generic_utf8_safe_no_upper_latin1(classnum, p, e, above_latin1) \
1915 (__ASSERT_(_utf8_safe_assert(p, e)) \
1916 (UTF8_IS_INVARIANT(*(p))) \
1917 ? _generic_isCC(*(p), classnum) \
1918 : (UTF8_IS_DOWNGRADEABLE_START(*(p))) \
1919 ? 0 /* Note that doesn't check validity for latin1 */ \
1920 : above_latin1)
1921
84238efa 1922
34aeb2e9
KW
1923#define isALPHA_utf8(p) _generic_utf8(ALPHA, p)
1924#define isALPHANUMERIC_utf8(p) _generic_utf8(ALPHANUMERIC, p)
1925#define isASCII_utf8(p) _generic_utf8(ASCII, p)
1926#define isBLANK_utf8(p) _generic_utf8(BLANK, p)
1927#define isCNTRL_utf8(p) _generic_utf8(CNTRL, p)
1928#define isDIGIT_utf8(p) _generic_utf8(DIGIT, p)
1929#define isGRAPH_utf8(p) _generic_utf8(GRAPH, p)
1930#define isIDCONT_utf8(p) _generic_utf8(IDCONT, p)
1931#define isIDFIRST_utf8(p) _generic_utf8(IDFIRST, p)
1932#define isLOWER_utf8(p) _generic_utf8(LOWER, p)
1933#define isPRINT_utf8(p) _generic_utf8(PRINT, p)
1934#define isPSXSPC_utf8(p) _generic_utf8(PSXSPC, p)
1935#define isPUNCT_utf8(p) _generic_utf8(PUNCT, p)
1936#define isSPACE_utf8(p) _generic_utf8(SPACE, p)
1937#define isUPPER_utf8(p) _generic_utf8(UPPER, p)
1938#define isVERTWS_utf8(p) _generic_utf8(VERTSPACE, p)
1939#define isWORDCHAR_utf8(p) _generic_utf8(WORDCHAR, p)
1940#define isXDIGIT_utf8(p) _generic_utf8(XDIGIT, p)
e8fa43e2 1941
da8c1a98
KW
1942#define isALPHA_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_ALPHA, p, e)
1943#define isALPHANUMERIC_utf8_safe(p, e) \
1944 _generic_swash_utf8_safe(_CC_ALPHANUMERIC, p, e)
1945#define isASCII_utf8_safe(p, e) \
1946 /* Because ASCII is invariant under utf8, the non-utf8 macro \
1947 * works */ \
1948 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII(*(p)))
1949#define isBLANK_utf8_safe(p, e) \
1950 _generic_non_swash_utf8_safe(_CC_BLANK, is_HORIZWS_high, p, e)
1951
e8fa43e2
KW
1952#ifdef EBCDIC
1953 /* Because all controls are UTF-8 invariants in EBCDIC, we can use this
1954 * more efficient macro instead of the more general one */
da8c1a98 1955# define isCNTRL_utf8_safe(p, e) \
56d02b8c 1956 (__ASSERT_(_utf8_safe_assert(p, e)) isCNTRL_L1(*(p)))
e8fa43e2 1957#else
da8c1a98 1958# define isCNTRL_utf8_safe(p, e) _generic_utf8_safe(_CC_CNTRL, p, e, 0)
e8fa43e2
KW
1959#endif
1960
da8c1a98
KW
1961#define isDIGIT_utf8_safe(p, e) \
1962 _generic_utf8_safe_no_upper_latin1(_CC_DIGIT, p, e, \
1963 _is_utf8_FOO_with_len(_CC_DIGIT, p, e))
1964#define isGRAPH_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_GRAPH, p, e)
1965#define isIDCONT_utf8_safe(p, e) _generic_func_utf8_safe(_CC_WORDCHAR, \
1966 _is_utf8_perl_idcont_with_len, p, e)
e5dcd934 1967
c11ff943
KW
1968/* To prevent S_scan_word in toke.c from hanging, we have to make sure that
1969 * IDFIRST is an alnum. See
c8362b00 1970 * http://rt.perl.org/rt3/Ticket/Display.html?id=74022 for more detail than you
f91dcd13
KW
1971 * ever wanted to know about. (In the ASCII range, there isn't a difference.)
1972 * This used to be not the XID version, but we decided to go with the more
1973 * modern Unicode definition */
da8c1a98
KW
1974#define isIDFIRST_utf8_safe(p, e) \
1975 _generic_func_utf8_safe(_CC_IDFIRST, \
1976 _is_utf8_perl_idstart_with_len, (U8 *) (p), (U8 *) (e))
1977
1978#define isLOWER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_LOWER, p, e)
1979#define isPRINT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PRINT, p, e)
1980#define isPSXSPC_utf8_safe(p, e) isSPACE_utf8_safe(p, e)
1981#define isPUNCT_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_PUNCT, p, e)
1982#define isSPACE_utf8_safe(p, e) \
1983 _generic_non_swash_utf8_safe(_CC_SPACE, is_XPERLSPACE_high, p, e)
1984#define isUPPER_utf8_safe(p, e) _generic_swash_utf8_safe(_CC_UPPER, p, e)
1985#define isVERTWS_utf8_safe(p, e) \
1986 _generic_non_swash_utf8_safe(_CC_VERTSPACE, is_VERTWS_high, p, e)
1987#define isWORDCHAR_utf8_safe(p, e) \
1988 _generic_swash_utf8_safe(_CC_WORDCHAR, p, e)
1989#define isXDIGIT_utf8_safe(p, e) \
1990 _generic_utf8_safe_no_upper_latin1(_CC_XDIGIT, p, e, \
1991 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
1992 ? (_force_out_malformed_utf8_message( \
1993 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
1994 : is_XDIGIT_high(p)))
a0ed51b3 1995
7d2b61dd 1996#define toFOLD_utf8(p,s,l) to_utf8_fold(p,s,l)
509fb054
KW
1997#define toLOWER_utf8(p,s,l) to_utf8_lower(p,s,l)
1998#define toTITLE_utf8(p,s,l) to_utf8_title(p,s,l)
1999#define toUPPER_utf8(p,s,l) to_utf8_upper(p,s,l)
2e8adce6 2000
567b353c 2001/* For internal core use only, subject to change */
607313a1
KW
2002#define _toFOLD_utf8_flags(p,e,s,l,f) _to_utf8_fold_flags (p,e,s,l,f, "", 0)
2003#define _toLOWER_utf8_flags(p,e,s,l,f) _to_utf8_lower_flags(p,e,s,l,f, "", 0)
2004#define _toTITLE_utf8_flags(p,e,s,l,f) _to_utf8_title_flags(p,e,s,l,f, "", 0)
2005#define _toUPPER_utf8_flags(p,e,s,l,f) _to_utf8_upper_flags(p,e,s,l,f, "", 0)
a1a5ec35
KW
2006
2007#define toFOLD_utf8_safe(p,e,s,l) _toFOLD_utf8_flags(p,e,s,l, FOLD_FLAGS_FULL)
2008#define toLOWER_utf8_safe(p,e,s,l) _toLOWER_utf8_flags(p,e,s,l, 0)
2009#define toTITLE_utf8_safe(p,e,s,l) _toTITLE_utf8_flags(p,e,s,l, 0)
2010#define toUPPER_utf8_safe(p,e,s,l) _toUPPER_utf8_flags(p,e,s,l, 0)
567b353c 2011
4650c663
KW
2012/* For internal core Perl use only: the base macros for defining macros like
2013 * isALPHA_LC_utf8. These are like _generic_utf8, but if the first code point
2014 * in 'p' is within the 0-255 range, it uses locale rules from the passed-in
2015 * 'macro' parameter */
34aeb2e9
KW
2016#define _generic_LC_utf8(name, p) _base_generic_utf8(name, name, p, 1)
2017
2018#define isALPHA_LC_utf8(p) _generic_LC_utf8(ALPHA, p)
2019#define isALPHANUMERIC_LC_utf8(p) _generic_LC_utf8(ALPHANUMERIC, p)
2020#define isASCII_LC_utf8(p) _generic_LC_utf8(ASCII, p)
2021#define isBLANK_LC_utf8(p) _generic_LC_utf8(BLANK, p)
2022#define isCNTRL_LC_utf8(p) _generic_LC_utf8(CNTRL, p)
2023#define isDIGIT_LC_utf8(p) _generic_LC_utf8(DIGIT, p)
2024#define isGRAPH_LC_utf8(p) _generic_LC_utf8(GRAPH, p)
2025#define isIDCONT_LC_utf8(p) _generic_LC_utf8(IDCONT, p)
2026#define isIDFIRST_LC_utf8(p) _generic_LC_utf8(IDFIRST, p)
2027#define isLOWER_LC_utf8(p) _generic_LC_utf8(LOWER, p)
2028#define isPRINT_LC_utf8(p) _generic_LC_utf8(PRINT, p)
2029#define isPSXSPC_LC_utf8(p) _generic_LC_utf8(PSXSPC, p)
2030#define isPUNCT_LC_utf8(p) _generic_LC_utf8(PUNCT, p)
2031#define isSPACE_LC_utf8(p) _generic_LC_utf8(SPACE, p)
2032#define isUPPER_LC_utf8(p) _generic_LC_utf8(UPPER, p)
2033#define isWORDCHAR_LC_utf8(p) _generic_LC_utf8(WORDCHAR, p)
2034#define isXDIGIT_LC_utf8(p) _generic_LC_utf8(XDIGIT, p)
2035
da8c1a98
KW
2036/* For internal core Perl use only: the base macros for defining macros like
2037 * isALPHA_LC_utf8_safe. These are like _generic_utf8, but if the first code
2038 * point in 'p' is within the 0-255 range, it uses locale rules from the
2039 * passed-in 'macro' parameter */
2040#define _generic_LC_utf8_safe(macro, p, e, above_latin1) \
2041 (__ASSERT_(_utf8_safe_assert(p, e)) \
2042 (UTF8_IS_INVARIANT(*(p))) \
2043 ? macro(*(p)) \
2044 : (UTF8_IS_DOWNGRADEABLE_START(*(p)) \
2045 ? ((LIKELY((e) - (p) > 1 && UTF8_IS_CONTINUATION(*((p)+1)))) \
2046 ? macro(EIGHT_BIT_UTF8_TO_NATIVE(*(p), *((p)+1))) \
2047 : (_force_out_malformed_utf8_message( \
2048 (U8 *) (p), (U8 *) (e), 0, 1), 0)) \
2049 : above_latin1))
2050
2051#define _generic_LC_swash_utf8_safe(macro, classnum, p, e) \
2052 _generic_LC_utf8_safe(macro, p, e, \
2053 _is_utf8_FOO_with_len(classnum, p, e))
2054
2055#define _generic_LC_func_utf8_safe(macro, above_latin1, p, e) \
2056 _generic_LC_utf8_safe(macro, p, e, above_latin1(p, e))
2057
2058#define _generic_LC_non_swash_utf8_safe(classnum, above_latin1, p, e) \
2059 _generic_LC_utf8_safe(classnum, p, e, \
2060 (UNLIKELY((e) - (p) < UTF8SKIP(p)) \
2061 ? (_force_out_malformed_utf8_message( \
2062 (U8 *) (p), (U8 *) (e), 0, 1), 0) \
2063 : above_latin1(p)))
2064
2065#define isALPHANUMERIC_LC_utf8_safe(p, e) \
2066 _generic_LC_swash_utf8_safe(isALPHANUMERIC_LC, \
2067 _CC_ALPHANUMERIC, p, e)
2068#define isALPHA_LC_utf8_safe(p, e) \
2069 _generic_LC_swash_utf8_safe(isALPHA_LC, _CC_ALPHA, p, e)
2070#define isASCII_LC_utf8_safe(p, e) \
2071 (__ASSERT_(_utf8_safe_assert(p, e)) isASCII_LC(*(p)))
2072#define isBLANK_LC_utf8_safe(p, e) \
2073 _generic_LC_non_swash_utf8_safe(isBLANK_LC, is_HORIZWS_high, p, e)
2074#define isCNTRL_LC_utf8_safe(p, e) \
2075 _generic_LC_utf8_safe(isCNTRL_LC, p, e, 0)
2076#define isDIGIT_LC_utf8_safe(p, e) \
2077 _generic_LC_swash_utf8_safe(isDIGIT_LC, _CC_DIGIT, p, e)
2078#define isGRAPH_LC_utf8_safe(p, e) \
2079 _generic_LC_swash_utf8_safe(isGRAPH_LC, _CC_GRAPH, p, e)
2080#define isIDCONT_LC_utf8_safe(p, e) \
2081 _generic_LC_func_utf8_safe(isIDCONT_LC, \
2082 _is_utf8_perl_idcont_with_len, p, e)
2083#define isIDFIRST_LC_utf8_safe(p, e) \
2084 _generic_LC_func_utf8_safe(isIDFIRST_LC, \
2085 _is_utf8_perl_idstart_with_len, p, e)
2086#define isLOWER_LC_utf8_safe(p, e) \
2087 _generic_LC_swash_utf8_safe(isLOWER_LC, _CC_LOWER, p, e)
2088#define isPRINT_LC_utf8_safe(p, e) \
2089 _generic_LC_swash_utf8_safe(isPRINT_LC, _CC_PRINT, p, e)
2090#define isPSXSPC_LC_utf8_safe(p, e) isSPACE_LC_utf8_safe(p, e)
2091#define isPUNCT_LC_utf8_safe(p, e) \
2092 _generic_LC_swash_utf8_safe(isPUNCT_LC, _CC_PUNCT, p, e)
2093#define isSPACE_LC_utf8_safe(p, e) \
2094 _generic_LC_non_swash_utf8_safe(isSPACE_LC, is_XPERLSPACE_high, p, e)
2095#define isUPPER_LC_utf8_safe(p, e) \
2096 _generic_LC_swash_utf8_safe(isUPPER_LC, _CC_UPPER, p, e)
2097#define isWORDCHAR_LC_utf8_safe(p, e) \
2098 _generic_LC_swash_utf8_safe(isWORDCHAR_LC, _CC_WORDCHAR, p, e)
2099#define isXDIGIT_LC_utf8_safe(p, e) \
2100 _generic_LC_non_swash_utf8_safe(isXDIGIT_LC, is_XDIGIT_high, p, e)
aaa51d5e 2101
fbc19f27
KW
2102/* Macros for backwards compatibility and for completeness when the ASCII and
2103 * Latin1 values are identical */
b7d90381
KW
2104#define isALPHAU(c) isALPHA_L1(c)
2105#define isDIGIT_L1(c) isDIGIT_A(c)
2106#define isOCTAL(c) isOCTAL_A(c)
2107#define isOCTAL_L1(c) isOCTAL_A(c)
2108#define isXDIGIT_L1(c) isXDIGIT_A(c)
2109#define isALNUM(c) isWORDCHAR(c)
a377c856 2110#define isALNUM_A(c) isALNUM(c)
b7d90381
KW
2111#define isALNUMU(c) isWORDCHAR_L1(c)
2112#define isALNUM_LC(c) isWORDCHAR_LC(c)
2113#define isALNUM_uni(c) isWORDCHAR_uni(c)
2e28f0b9 2114#define isALNUM_LC_uvchr(c) isWORDCHAR_LC_uvchr(c)
b7d90381
KW
2115#define isALNUM_utf8(p) isWORDCHAR_utf8(p)
2116#define isALNUM_LC_utf8(p) isWORDCHAR_LC_utf8(p)
2117#define isALNUMC_A(c) isALPHANUMERIC_A(c) /* Mnemonic: "C's alnum" */
2118#define isALNUMC_L1(c) isALPHANUMERIC_L1(c)
2119#define isALNUMC(c) isALPHANUMERIC(c)
2120#define isALNUMC_LC(c) isALPHANUMERIC_LC(c)
2121#define isALNUMC_uni(c) isALPHANUMERIC_uni(c)
15861f94 2122#define isALNUMC_LC_uvchr(c) isALPHANUMERIC_LC_uvchr(c)
b7d90381 2123#define isALNUMC_utf8(p) isALPHANUMERIC_utf8(p)
15861f94 2124#define isALNUMC_LC_utf8(p) isALPHANUMERIC_LC_utf8(p)
fbc19f27 2125
2bd1cbf6
KW
2126/* On EBCDIC platforms, CTRL-@ is 0, CTRL-A is 1, etc, just like on ASCII,
2127 * except that they don't necessarily mean the same characters, e.g. CTRL-D is
2128 * 4 on both systems, but that is EOT on ASCII; ST on EBCDIC.
2129 * '?' is special-cased on EBCDIC to APC, which is the control there that is
2130 * the outlier from the block that contains the other controls, just like
2131 * toCTRL('?') on ASCII yields DEL, the control that is the outlier from the C0
2132 * block. If it weren't special cased, it would yield a non-control.
88794300
KW
2133 * The conversion works both ways, so toCTRL('D') is 4, and toCTRL(4) is D,
2134 * etc. */
2bd1cbf6 2135#ifndef EBCDIC
75763b3a 2136# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) toUPPER(((U8)(c))) ^ 64)
2bd1cbf6 2137#else
75763b3a
KW
2138# define toCTRL(c) (__ASSERT_(FITS_IN_8_BITS(c)) \
2139 ((isPRINT_A(c)) \
2140 ? (UNLIKELY((c) == '?') \
2141 ? QUESTION_MARK_CTRL \
2142 : (NATIVE_TO_LATIN1(toUPPER((U8) (c))) ^ 64)) \
2143 : (UNLIKELY((c) == QUESTION_MARK_CTRL) \
2144 ? '?' \
2145 : (LATIN1_TO_NATIVE(((U8) (c)) ^ 64)))))
2bd1cbf6 2146#endif
bbce6d69 2147
dea28490
JJ
2148/* Line numbers are unsigned, 32 bits. */
2149typedef U32 line_t;
e5dcd934 2150#define NOLINE ((line_t) 4294967295UL) /* = FFFFFFFF */
378cc40b 2151
91152fc1
DG
2152/* Helpful alias for version prescan */
2153#define is_LAX_VERSION(a,b) \
2154 (a != Perl_prescan_version(aTHX_ a, FALSE, b, NULL, NULL, NULL, NULL))
2155
2156#define is_STRICT_VERSION(a,b) \
2157 (a != Perl_prescan_version(aTHX_ a, TRUE, b, NULL, NULL, NULL, NULL))
2158
2159#define BADVERSION(a,b,c) \
2160 if (b) { \
2161 *b = c; \
2162 } \
2163 return a;
8c52afec 2164
cb27eebd
KW
2165/* Converts a character known to represent a hexadecimal digit (0-9, A-F, or
2166 * a-f) to its numeric value. READ_XDIGIT's argument is a string pointer,
2167 * which is advanced. The input is validated only by an assert() in DEBUGGING
2168 * builds. In both ASCII and EBCDIC the last 4 bits of the digits are 0-9; and
2169 * the last 4 bits of A-F and a-f are 1-6, so adding 9 yields 10-15 */
2170#define XDIGIT_VALUE(c) (__ASSERT_(isXDIGIT(c)) (0xf & (isDIGIT(c) \
2171 ? (c) \
2172 : ((c) + 9))))
902008b8
KW
2173#define READ_XDIGIT(s) (__ASSERT_(isXDIGIT(*s)) (0xf & (isDIGIT(*(s)) \
2174 ? (*(s)++) \
2175 : (*(s)++ + 9))))
95a59cab 2176
cb27eebd
KW
2177/* Converts a character known to represent an octal digit (0-7) to its numeric
2178 * value. The input is validated only by an assert() in DEBUGGING builds. In
2179 * both ASCII and EBCDIC the last 3 bits of the octal digits range from 0-7. */
2180#define OCTAL_VALUE(c) (__ASSERT_(isOCTAL(c)) (7 & (c)))
2181
305b8651
KW
2182/* Efficiently returns a boolean as to if two native characters are equivalent
2183 * case-insenstively. At least one of the characters must be one of [A-Za-z];
2184 * the ALPHA in the name is to remind you of that. This is asserted() in
2185 * DEBUGGING builds. Because [A-Za-z] are invariant under UTF-8, this macro
2186 * works (on valid input) for both non- and UTF-8-encoded bytes.
2187 *
2188 * When one of the inputs is a compile-time constant and gets folded by the
2189 * compiler, this reduces to an AND and a TEST. On both EBCDIC and ASCII
2190 * machines, 'A' and 'a' differ by a single bit; the same with the upper and
2191 * lower case of all other ASCII-range alphabetics. On ASCII platforms, they
96ca48da
KW
2192 * are 32 apart; on EBCDIC, they are 64. At compile time, this uses an
2193 * exclusive 'or' to find that bit and then inverts it to form a mask, with
2194 * just a single 0, in the bit position where the upper- and lowercase differ.
2195 * */
305b8651
KW
2196#define isALPHA_FOLD_EQ(c1, c2) \
2197 (__ASSERT_(isALPHA_A(c1) || isALPHA_A(c2)) \
2198 ((c1) & ~('A' ^ 'a')) == ((c2) & ~('A' ^ 'a')))
2199#define isALPHA_FOLD_NE(c1, c2) (! isALPHA_FOLD_EQ((c1), (c2)))
2200
8e84507e 2201/*
ccfc67b7
JH
2202=head1 Memory Management
2203
a02a5408 2204=for apidoc Am|void|Newx|void* ptr|int nitems|type
954c1994
GS
2205The XSUB-writer's interface to the C C<malloc> function.
2206
596f7718 2207Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2208
c5008215
JC
2209In 5.9.3, Newx() and friends replace the older New() API, and drops
2210the first parameter, I<x>, a debug aid which allowed callers to identify
37b8b4c9 2211themselves. This aid has been superseded by a new build option,
d10b4965 2212PERL_MEM_LOG (see L<perlhacktips/PERL_MEM_LOG>). The older API is still
c5008215
JC
2213there for use in XS modules supporting older perls.
2214
a02a5408 2215=for apidoc Am|void|Newxc|void* ptr|int nitems|type|cast
954c1994 2216The XSUB-writer's interface to the C C<malloc> function, with
fbe13c60 2217cast. See also C<L</Newx>>.
954c1994 2218
596f7718 2219Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2220
a02a5408 2221=for apidoc Am|void|Newxz|void* ptr|int nitems|type
954c1994 2222The XSUB-writer's interface to the C C<malloc> function. The allocated
fbe13c60 2223memory is zeroed with C<memzero>. See also C<L</Newx>>.
a02a5408 2224
596f7718 2225Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2226
954c1994
GS
2227=for apidoc Am|void|Renew|void* ptr|int nitems|type
2228The XSUB-writer's interface to the C C<realloc> function.
2229
596f7718 2230Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2231
954c1994
GS
2232=for apidoc Am|void|Renewc|void* ptr|int nitems|type|cast
2233The XSUB-writer's interface to the C C<realloc> function, with
2234cast.
2235
596f7718 2236Memory obtained by this should B<ONLY> be freed with L</"Safefree">.
0d7b2759 2237
49b8b560 2238=for apidoc Am|void|Safefree|void* ptr
954c1994
GS
2239The XSUB-writer's interface to the C C<free> function.
2240
596f7718 2241This should B<ONLY> be used on memory obtained using L</"Newx"> and friends.
0d7b2759 2242
954c1994
GS
2243=for apidoc Am|void|Move|void* src|void* dest|int nitems|type
2244The XSUB-writer's interface to the C C<memmove> function. The C<src> is the
926bb54c 2245source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2246C<type> is the type. Can do overlapping moves. See also C<L</Copy>>.
954c1994 2247
e90e2364 2248=for apidoc Am|void *|MoveD|void* src|void* dest|int nitems|type
796b6530 2249Like C<Move> but returns C<dest>. Useful
72d33970 2250for encouraging compilers to tail-call
e90e2364
NC
2251optimise.
2252
954c1994
GS
2253=for apidoc Am|void|Copy|void* src|void* dest|int nitems|type
2254The XSUB-writer's interface to the C C<memcpy> function. The C<src> is the
926bb54c 2255source, C<dest> is the destination, C<nitems> is the number of items, and
fbe13c60 2256C<type> is the type. May fail on overlapping copies. See also C<L</Move>>.
954c1994 2257
e90e2364
NC
2258=for apidoc Am|void *|CopyD|void* src|void* dest|int nitems|type
2259
796b6530 2260Like C<Copy> but returns C<dest>. Useful
72d33970 2261for encouraging compilers to tail-call
e90e2364
NC
2262optimise.
2263
954c1994
GS
2264=for apidoc Am|void|Zero|void* dest|int nitems|type
2265
2266The XSUB-writer's interface to the C C<memzero> function. The C<dest> is the
2267destination, C<nitems> is the number of items, and C<type> is the type.
2268
e90e2364
NC
2269=for apidoc Am|void *|ZeroD|void* dest|int nitems|type
2270
72d33970
FC
2271Like C<Zero> but returns dest. Useful
2272for encouraging compilers to tail-call
e90e2364
NC
2273optimise.
2274
da5d8dbb 2275=for apidoc Am|void|StructCopy|type *src|type *dest|type
4375e838 2276This is an architecture-independent macro to copy one structure to another.
954c1994 2277
7e337ee0
JH
2278=for apidoc Am|void|PoisonWith|void* dest|int nitems|type|U8 byte
2279
2280Fill up memory with a byte pattern (a byte repeated over and over
2281again) that hopefully catches attempts to access uninitialized memory.
2282
2283=for apidoc Am|void|PoisonNew|void* dest|int nitems|type
2284
2285PoisonWith(0xAB) for catching access to allocated but uninitialized memory.
2286
1c12ffb4 2287=for apidoc Am|void|PoisonFree|void* dest|int nitems|type
7e337ee0
JH
2288
2289PoisonWith(0xEF) for catching access to freed memory.
2290
9965345d
JH
2291=for apidoc Am|void|Poison|void* dest|int nitems|type
2292
7e337ee0 2293PoisonWith(0xEF) for catching access to freed memory.
9965345d
JH
2294
2295=cut */
954c1994 2296
561b68a9
SH
2297/* Maintained for backwards-compatibility only. Use newSV() instead. */
2298#ifndef PERL_CORE
ff06c60c 2299#define NEWSV(x,len) newSV(len)
561b68a9 2300#endif
ff06c60c 2301
b7112dce 2302#define MEM_SIZE_MAX ((MEM_SIZE)-1)
19a94d75 2303
a500027b 2304#define _PERL_STRLEN_ROUNDUP_UNCHECKED(n) (((n) - 1 + PERL_STRLEN_ROUNDUP_QUANTUM) & ~((MEM_SIZE)PERL_STRLEN_ROUNDUP_QUANTUM - 1))
e6bdf523 2305
27d5b266 2306#ifdef PERL_MALLOC_WRAP
e6bdf523
DM
2307
2308/* This expression will be constant-folded at compile time. It checks
2309 * whether or not the type of the count n is so small (e.g. U8 or U16, or
2310 * U32 on 64-bit systems) that there's no way a wrap-around could occur.
2311 * As well as avoiding the need for a run-time check in some cases, it's
2312 * designed to avoid compiler warnings like:
2313 * comparison is always false due to limited range of data type
73e8ff00
DM
2314 * It's mathematically equivalent to
2315 * max(n) * sizeof(t) > MEM_SIZE_MAX
e6bdf523
DM
2316 */
2317
2318# define _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) \
445198b9
LM
2319 ( sizeof(MEM_SIZE) < sizeof(n) \
2320 || sizeof(t) > ((MEM_SIZE)1 << 8*(sizeof(MEM_SIZE) - sizeof(n))))
e6bdf523 2321
88f9f128 2322/* This is written in a slightly odd way to avoid various spurious
d98e5cde
DM
2323 * compiler warnings. We *want* to write the expression as
2324 * _MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) && (n > C)
2325 * (for some compile-time constant C), but even when the LHS
2326 * constant-folds to false at compile-time, g++ insists on emitting
2327 * warnings about the RHS (e.g. "comparison is always false"), so instead
2328 * we write it as
e6bdf523 2329 *
d98e5cde 2330 * (cond ? n : X) > C
88f9f128 2331 *
d98e5cde
DM
2332 * where X is a constant with X > C always false. Choosing a value for X
2333 * is tricky. If 0, some compilers will complain about 0 > C always being
2334 * false; if 1, Coverity complains when n happens to be the constant value
2335 * '1', that cond ? 1 : 1 has the same value on both branches; so use C
2336 * for X and hope that nothing else whines.
e6bdf523
DM
2337 */
2338
2339# define _MEM_WRAP_WILL_WRAP(n,t) \
88f9f128
DM
2340 ((_MEM_WRAP_NEEDS_RUNTIME_CHECK(n,t) ? (MEM_SIZE)(n) : \
2341 MEM_SIZE_MAX/sizeof(t)) > MEM_SIZE_MAX/sizeof(t))
e6bdf523
DM
2342
2343# define MEM_WRAP_CHECK(n,t) \
2344 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2345 && (croak_memory_wrap(),0))
2346
2347# define MEM_WRAP_CHECK_1(n,t,a) \
2348 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2349 && (Perl_croak_nocontext("%s",(a)),0))
2350
814eedc8
DD
2351/* "a" arg must be a string literal */
2352# define MEM_WRAP_CHECK_s(n,t,a) \
2353 (void)(UNLIKELY(_MEM_WRAP_WILL_WRAP(n,t)) \
2354 && (Perl_croak_nocontext("" a ""),0))
2355
8b44ba4c 2356#define MEM_WRAP_CHECK_(n,t) MEM_WRAP_CHECK(n,t),
27d5b266 2357
a500027b 2358#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
2359#else
2360
410319be
NC
2361#define MEM_WRAP_CHECK(n,t)
2362#define MEM_WRAP_CHECK_1(n,t,a)
814eedc8 2363#define MEM_WRAP_CHECK_s(n,t,a)
8b44ba4c
NC
2364#define MEM_WRAP_CHECK_(n,t)
2365
a500027b 2366#define PERL_STRLEN_ROUNDUP(n) _PERL_STRLEN_ROUNDUP_UNCHECKED(n)
27d5b266 2367
1936d2a7 2368#endif
8b44ba4c 2369
fe4f188c 2370#ifdef PERL_MEM_LOG
46c6c7e2 2371/*
9f653bb5 2372 * If PERL_MEM_LOG is defined, all Newx()s, Renew()s, and Safefree()s
46c6c7e2
JH
2373 * go through functions, which are handy for debugging breakpoints, but
2374 * which more importantly get the immediate calling environment (file and
e352bcff
JH
2375 * line number, and C function name if available) passed in. This info can
2376 * then be used for logging the calls, for which one gets a sample
73d1d973 2377 * implementation unless -DPERL_MEM_LOG_NOIMPL is also defined.
3609ea0d 2378 *
46c6c7e2 2379 * Known problems:
94e892a6 2380 * - not all memory allocs get logged, only those
46c6c7e2 2381 * that go through Newx() and derivatives (while all
94e892a6 2382 * Safefrees do get logged)
46c6c7e2
JH
2383 * - __FILE__ and __LINE__ do not work everywhere
2384 * - __func__ or __FUNCTION__ even less so
2385 * - I think more goes on after the perlio frees but
2386 * the thing is that STDERR gets closed (as do all
2387 * the file descriptors)
2388 * - no deeper calling stack than the caller of the Newx()
2389 * or the kind, but do I look like a C reflection/introspection
2390 * utility to you?
2391 * - the function prototypes for the logging functions
2392 * probably should maybe be somewhere else than handy.h
2393 * - one could consider inlining (macrofying) the logging
2394 * for speed, but I am too lazy
2395 * - one could imagine recording the allocations in a hash,
2396 * (keyed by the allocation address?), and maintain that
2397 * through reallocs and frees, but how to do that without
2398 * any News() happening...?
73d1d973 2399 * - lots of -Ddefines to get useful/controllable output
b953482e 2400 * - lots of ENV reads
46c6c7e2
JH
2401 */
2402
0b0ab801 2403# ifdef PERL_CORE
73d1d973 2404# ifndef PERL_MEM_LOG_NOIMPL
0b0ab801
MHM
2405enum mem_log_type {
2406 MLT_ALLOC,
2407 MLT_REALLOC,
d7a2c63c
MHM
2408 MLT_FREE,
2409 MLT_NEW_SV,
2410 MLT_DEL_SV
0b0ab801
MHM
2411};
2412# endif
12754f92 2413# if defined(PERL_IN_SV_C) /* those are only used in sv.c */
d7a2c63c
MHM
2414void Perl_mem_log_new_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
2415void Perl_mem_log_del_sv(const SV *sv, const char *filename, const int linenumber, const char *funcname);
12754f92 2416# endif
0b0ab801
MHM
2417# endif
2418
fe4f188c
JH
2419#endif
2420
2421#ifdef PERL_MEM_LOG
d1401ee9
MHM
2422#define MEM_LOG_ALLOC(n,t,a) Perl_mem_log_alloc(n,sizeof(t),STRINGIFY(t),a,__FILE__,__LINE__,FUNCTION__)
2423#define MEM_LOG_REALLOC(n,t,v,a) Perl_mem_log_realloc(n,sizeof(t),STRINGIFY(t),v,a,__FILE__,__LINE__,FUNCTION__)
46c6c7e2 2424#define MEM_LOG_FREE(a) Perl_mem_log_free(a,__FILE__,__LINE__,FUNCTION__)
fe4f188c
JH
2425#endif
2426
2427#ifndef MEM_LOG_ALLOC
2428#define MEM_LOG_ALLOC(n,t,a) (a)
2429#endif
2430#ifndef MEM_LOG_REALLOC
2431#define MEM_LOG_REALLOC(n,t,v,a) (a)
2432#endif
2433#ifndef MEM_LOG_FREE
2434#define MEM_LOG_FREE(a) (a)
2435#endif
2436
d1401ee9
MHM
2437#define Newx(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2438#define Newxc(v,n,t,c) (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_ALLOC(n,t,safemalloc((MEM_SIZE)((n)*sizeof(t))))))
2439#define Newxz(v,n,t) (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_ALLOC(n,t,safecalloc((n),sizeof(t)))))
a6f6820f
NC
2440
2441#ifndef PERL_CORE
a02a5408
JC
2442/* pre 5.9.x compatibility */
2443#define New(x,v,n,t) Newx(v,n,t)
2444#define Newc(x,v,n,t,c) Newxc(v,n,t,c)
4541904d 2445#define Newz(x,v,n,t) Newxz(v,n,t)
a6f6820f 2446#endif
a02a5408 2447
ff68c719 2448#define Renew(v,n,t) \
d1401ee9 2449 (v = (MEM_WRAP_CHECK_(n,t) (t*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
ff68c719 2450#define Renewc(v,n,t,c) \
d1401ee9 2451 (v = (MEM_WRAP_CHECK_(n,t) (c*)MEM_LOG_REALLOC(n,t,v,saferealloc((Malloc_t)(v),(MEM_SIZE)((n)*sizeof(t))))))
94010e71
NC
2452
2453#ifdef PERL_POISON
2454#define Safefree(d) \
06c0cc96 2455 ((d) ? (void)(safefree(MEM_LOG_FREE((Malloc_t)(d))), Poison(&(d), 1, Malloc_t)) : (void) 0)
94010e71 2456#else
fe4f188c 2457#define Safefree(d) safefree(MEM_LOG_FREE((Malloc_t)(d)))
94010e71 2458#endif
55497cff 2459
dbb57106
YO
2460/* assert that a valid ptr has been supplied - use this instead of assert(ptr) *
2461 * as it handles cases like constant string arguments without throwing warnings *
2462 * the cast is required, as is the inequality check, to avoid warnings */
45908e4d 2463#define perl_assert_ptr(p) assert( ((void*)(p)) != 0 )
55497cff 2464
45908e4d
YO
2465
2466#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)))
2467#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)))
2468#define Zero(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), (void)memzero((char*)(d), (n) * sizeof(t)))
2469
bdd1531d 2470/* Like above, but returns a pointer to 'd' */
45908e4d
YO
2471#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)))
2472#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 2473#define ZeroD(d,n,t) (MEM_WRAP_CHECK_(n,t) perl_assert_ptr(d), memzero((char*)(d), (n) * sizeof(t)))
e90e2364 2474
7e337ee0
JH
2475#define PoisonWith(d,n,t,b) (MEM_WRAP_CHECK_(n,t) (void)memset((char*)(d), (U8)(b), (n) * sizeof(t)))
2476#define PoisonNew(d,n,t) PoisonWith(d,n,t,0xAB)
2477#define PoisonFree(d,n,t) PoisonWith(d,n,t,0xEF)
2478#define Poison(d,n,t) PoisonFree(d,n,t)
27d5b266 2479
caa674f3
DD
2480#ifdef PERL_POISON
2481# define PERL_POISON_EXPR(x) x
2482#else
2483# define PERL_POISON_EXPR(x)
2484#endif
2485
ff68c719 2486#define StructCopy(s,d,t) (*((t*)(d)) = *((t*)(s)))
2cc61e15 2487
c3caa5c3
JH
2488/* C_ARRAY_LENGTH is the number of elements in the C array (so you
2489 * want your zero-based indices to be less than but not equal to).
2490 *
2491 * C_ARRAY_END is one past the last: half-open/half-closed range,
2492 * not last-inclusive range. */
622913ab 2493#define C_ARRAY_LENGTH(a) (sizeof(a)/sizeof((a)[0]))
c3caa5c3 2494#define C_ARRAY_END(a) ((a) + C_ARRAY_LENGTH(a))
622913ab 2495
2cc61e15
DD
2496#ifdef NEED_VA_COPY
2497# ifdef va_copy
2498# define Perl_va_copy(s, d) va_copy(d, s)
07798b17
AC
2499# elif defined(__va_copy)
2500# define Perl_va_copy(s, d) __va_copy(d, s)
2cc61e15 2501# else
07798b17 2502# define Perl_va_copy(s, d) Copy(s, d, 1, va_list)
2cc61e15
DD
2503# endif
2504#endif
2505
472d47bc
SB
2506/* convenience debug macros */
2507#ifdef USE_ITHREADS
2508#define pTHX_FORMAT "Perl interpreter: 0x%p"
2509#define pTHX__FORMAT ", Perl interpreter: 0x%p"
f54cb97a
AL
2510#define pTHX_VALUE_ (void *)my_perl,
2511#define pTHX_VALUE (void *)my_perl
2512#define pTHX__VALUE_ ,(void *)my_perl,
2513#define pTHX__VALUE ,(void *)my_perl
472d47bc 2514#else
3609ea0d 2515#define pTHX_FORMAT
472d47bc 2516#define pTHX__FORMAT
3609ea0d 2517#define pTHX_VALUE_
472d47bc 2518#define pTHX_VALUE
3609ea0d 2519#define pTHX__VALUE_
472d47bc
SB
2520#define pTHX__VALUE
2521#endif /* USE_ITHREADS */
3609ea0d 2522
2acdbac1
NC
2523/* Perl_deprecate was not part of the public API, and did not have a deprecate()
2524 shortcut macro defined without -DPERL_CORE. Neither codesearch.google.com nor
2525 CPAN::Unpack show any users outside the core. */
2526#ifdef PERL_CORE
dc6e8de0
A
2527# define deprecate(s) Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2528 "Use of " s " is deprecated")
c9680906
A
2529# define deprecate_disappears_in(when,message) \
2530 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2531 message ", and will disappear in Perl " when)
ac641426
A
2532# define deprecate_fatal_in(when,message) \
2533 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
2534 message ". Its use will be fatal in Perl " when)
2acdbac1
NC
2535#endif
2536
dfff4baf
BF
2537/* Internal macros to deal with gids and uids */
2538#ifdef PERL_CORE
2539
2540# if Uid_t_size > IVSIZE
2541# define sv_setuid(sv, uid) sv_setnv((sv), (NV)(uid))
2542# define SvUID(sv) SvNV(sv)
07798b17
AC
2543# elif Uid_t_sign <= 0
2544# define sv_setuid(sv, uid) sv_setiv((sv), (IV)(uid))
2545# define SvUID(sv) SvIV(sv)
dfff4baf 2546# else
07798b17
AC
2547# define sv_setuid(sv, uid) sv_setuv((sv), (UV)(uid))
2548# define SvUID(sv) SvUV(sv)
dfff4baf
BF
2549# endif /* Uid_t_size */
2550
2551# if Gid_t_size > IVSIZE
2552# define sv_setgid(sv, gid) sv_setnv((sv), (NV)(gid))
2553# define SvGID(sv) SvNV(sv)
07798b17
AC
2554# elif Gid_t_sign <= 0
2555# define sv_setgid(sv, gid) sv_setiv((sv), (IV)(gid))
2556# define SvGID(sv) SvIV(sv)
dfff4baf 2557# else
07798b17
AC
2558# define sv_setgid(sv, gid) sv_setuv((sv), (UV)(gid))
2559# define SvGID(sv) SvUV(sv)
dfff4baf
BF
2560# endif /* Gid_t_size */
2561
2562#endif
2563
6a5bc5ac 2564#endif /* PERL_HANDY_H_ */
9d745869 2565
e9a8c099 2566/*
14d04a33 2567 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 2568 */