This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Bunch of doc patches from Stas; plus regen.
[perl5.git] / pod / perlguts.pod
CommitLineData
a0d0e21e
LW
1=head1 NAME
2
954c1994 3perlguts - Introduction to the Perl API
a0d0e21e
LW
4
5=head1 DESCRIPTION
6
b3b6085d
PP
7This document attempts to describe how to use the Perl API, as well as
8containing some info on the basic workings of the Perl core. It is far
9from complete and probably contains many errors. Please refer any
10questions or comments to the author below.
a0d0e21e 11
0a753a76
PP
12=head1 Variables
13
5f05dabc 14=head2 Datatypes
a0d0e21e
LW
15
16Perl has three typedefs that handle Perl's three main data types:
17
18 SV Scalar Value
19 AV Array Value
20 HV Hash Value
21
d1b91892 22Each typedef has specific routines that manipulate the various data types.
a0d0e21e
LW
23
24=head2 What is an "IV"?
25
954c1994 26Perl uses a special typedef IV which is a simple signed integer type that is
5f05dabc 27guaranteed to be large enough to hold a pointer (as well as an integer).
954c1994 28Additionally, there is the UV, which is simply an unsigned IV.
a0d0e21e 29
d1b91892 30Perl also uses two special typedefs, I32 and I16, which will always be at
954c1994 31least 32-bits and 16-bits long, respectively. (Again, there are U32 and U16,
20dbd849
NC
32as well.) They will usually be exactly 32 and 16 bits long, but on Crays
33they will both be 64 bits.
a0d0e21e 34
54310121 35=head2 Working with SVs
a0d0e21e 36
20dbd849
NC
37An SV can be created and loaded with one command. There are five types of
38values that can be loaded: an integer value (IV), an unsigned integer
39value (UV), a double (NV), a string (PV), and another scalar (SV).
a0d0e21e 40
20dbd849 41The seven routines are:
a0d0e21e
LW
42
43 SV* newSViv(IV);
20dbd849 44 SV* newSVuv(UV);
a0d0e21e 45 SV* newSVnv(double);
08105a92
GS
46 SV* newSVpv(const char*, int);
47 SV* newSVpvn(const char*, int);
46fc3d4c 48 SV* newSVpvf(const char*, ...);
a0d0e21e
LW
49 SV* newSVsv(SV*);
50
20dbd849
NC
51If you require more complex initialisation you can create an empty SV with
52newSV(len). If C<len> is 0 an empty SV of type NULL is returned, else an
53SV of type PV is returned with len + 1 (for the NUL) bytes of storage
54allocated, accessible via SvPVX. In both cases the SV has value undef.
55
56 SV* newSV(0); /* no storage allocated */
57 SV* newSV(10); /* 10 (+1) bytes of uninitialised storage allocated */
58
59To change the value of an *already-existing* SV, there are eight routines:
a0d0e21e
LW
60
61 void sv_setiv(SV*, IV);
deb3007b 62 void sv_setuv(SV*, UV);
a0d0e21e 63 void sv_setnv(SV*, double);
08105a92
GS
64 void sv_setpv(SV*, const char*);
65 void sv_setpvn(SV*, const char*, int)
46fc3d4c 66 void sv_setpvf(SV*, const char*, ...);
5b3218b9 67 void sv_vsetpvfn(SV*, const char*, STRLEN, va_list *, SV **, I32, bool *);
a0d0e21e
LW
68 void sv_setsv(SV*, SV*);
69
70Notice that you can choose to specify the length of the string to be
9da1e3b5
MUN
71assigned by using C<sv_setpvn>, C<newSVpvn>, or C<newSVpv>, or you may
72allow Perl to calculate the length by using C<sv_setpv> or by specifying
730 as the second argument to C<newSVpv>. Be warned, though, that Perl will
74determine the string's length by using C<strlen>, which depends on the
9abd00ed
GS
75string terminating with a NUL character.
76
77The arguments of C<sv_setpvf> are processed like C<sprintf>, and the
78formatted output becomes the value.
79
328bf373 80C<sv_vsetpvfn> is an analogue of C<vsprintf>, but it allows you to specify
9abd00ed
GS
81either a pointer to a variable argument list or the address and length of
82an array of SVs. The last argument points to a boolean; on return, if that
83boolean is true, then locale-specific information has been used to format
c2611fb3 84the string, and the string's contents are therefore untrustworthy (see
9abd00ed
GS
85L<perlsec>). This pointer may be NULL if that information is not
86important. Note that this function requires you to specify the length of
87the format.
88
7c906a97 89STRLEN is an integer type (Size_t, usually defined as size_t in
00aadd71 90config.h) guaranteed to be large enough to represent the size of
7c906a97
AD
91any string that perl can handle.
92
9da1e3b5
MUN
93The C<sv_set*()> functions are not generic enough to operate on values
94that have "magic". See L<Magic Virtual Tables> later in this document.
a0d0e21e 95
a3cb178b
GS
96All SVs that contain strings should be terminated with a NUL character.
97If it is not NUL-terminated there is a risk of
5f05dabc
PP
98core dumps and corruptions from code which passes the string to C
99functions or system calls which expect a NUL-terminated string.
100Perl's own functions typically add a trailing NUL for this reason.
101Nevertheless, you should be very careful when you pass a string stored
102in an SV to a C function or system call.
103
a0d0e21e
LW
104To access the actual value that an SV points to, you can use the macros:
105
106 SvIV(SV*)
954c1994 107 SvUV(SV*)
a0d0e21e
LW
108 SvNV(SV*)
109 SvPV(SV*, STRLEN len)
1fa8b10d 110 SvPV_nolen(SV*)
a0d0e21e 111
954c1994 112which will automatically coerce the actual scalar type into an IV, UV, double,
a0d0e21e
LW
113or string.
114
115In the C<SvPV> macro, the length of the string returned is placed into the
1fa8b10d
JD
116variable C<len> (this is a macro, so you do I<not> use C<&len>). If you do
117not care what the length of the data is, use the C<SvPV_nolen> macro.
118Historically the C<SvPV> macro with the global variable C<PL_na> has been
119used in this case. But that can be quite inefficient because C<PL_na> must
120be accessed in thread-local storage in threaded Perl. In any case, remember
121that Perl allows arbitrary strings of data that may both contain NULs and
122might not be terminated by a NUL.
a0d0e21e 123
ce2f5d8f
KA
124Also remember that C doesn't allow you to safely say C<foo(SvPV(s, len),
125len);>. It might work with your compiler, but it won't work for everyone.
126Break this sort of statement up into separate assignments:
127
b2f5ed49 128 SV *s;
ce2f5d8f
KA
129 STRLEN len;
130 char * ptr;
b2f5ed49 131 ptr = SvPV(s, len);
ce2f5d8f
KA
132 foo(ptr, len);
133
07fa94a1 134If you want to know if the scalar value is TRUE, you can use:
a0d0e21e
LW
135
136 SvTRUE(SV*)
137
138Although Perl will automatically grow strings for you, if you need to force
139Perl to allocate more memory for your SV, you can use the macro
140
141 SvGROW(SV*, STRLEN newlen)
142
143which will determine if more memory needs to be allocated. If so, it will
144call the function C<sv_grow>. Note that C<SvGROW> can only increase, not
5f05dabc
PP
145decrease, the allocated memory of an SV and that it does not automatically
146add a byte for the a trailing NUL (perl's own string functions typically do
8ebc5c01 147C<SvGROW(sv, len + 1)>).
a0d0e21e
LW
148
149If you have an SV and want to know what kind of data Perl thinks is stored
150in it, you can use the following macros to check the type of SV you have.
151
152 SvIOK(SV*)
153 SvNOK(SV*)
154 SvPOK(SV*)
155
156You can get and set the current length of the string stored in an SV with
157the following macros:
158
159 SvCUR(SV*)
160 SvCUR_set(SV*, I32 val)
161
cb1a09d0
AD
162You can also get a pointer to the end of the string stored in the SV
163with the macro:
164
165 SvEND(SV*)
166
167But note that these last three macros are valid only if C<SvPOK()> is true.
a0d0e21e 168
d1b91892
AD
169If you want to append something to the end of string stored in an C<SV*>,
170you can use the following functions:
171
08105a92 172 void sv_catpv(SV*, const char*);
e65f3abd 173 void sv_catpvn(SV*, const char*, STRLEN);
46fc3d4c 174 void sv_catpvf(SV*, const char*, ...);
328bf373 175 void sv_vcatpvfn(SV*, const char*, STRLEN, va_list *, SV **, I32, bool);
d1b91892
AD
176 void sv_catsv(SV*, SV*);
177
178The first function calculates the length of the string to be appended by
179using C<strlen>. In the second, you specify the length of the string
46fc3d4c 180yourself. The third function processes its arguments like C<sprintf> and
9abd00ed
GS
181appends the formatted output. The fourth function works like C<vsprintf>.
182You can specify the address and length of an array of SVs instead of the
183va_list argument. The fifth function extends the string stored in the first
184SV with the string stored in the second SV. It also forces the second SV
185to be interpreted as a string.
186
187The C<sv_cat*()> functions are not generic enough to operate on values that
188have "magic". See L<Magic Virtual Tables> later in this document.
d1b91892 189
a0d0e21e
LW
190If you know the name of a scalar variable, you can get a pointer to its SV
191by using the following:
192
4929bf7b 193 SV* get_sv("package::varname", FALSE);
a0d0e21e
LW
194
195This returns NULL if the variable does not exist.
196
d1b91892 197If you want to know if this variable (or any other SV) is actually C<defined>,
a0d0e21e
LW
198you can call:
199
200 SvOK(SV*)
201
9cde0e7f 202The scalar C<undef> value is stored in an SV instance called C<PL_sv_undef>. Its
a0d0e21e
LW
203address can be used whenever an C<SV*> is needed.
204
9cde0e7f
GS
205There are also the two values C<PL_sv_yes> and C<PL_sv_no>, which contain Boolean
206TRUE and FALSE values, respectively. Like C<PL_sv_undef>, their addresses can
a0d0e21e
LW
207be used whenever an C<SV*> is needed.
208
9cde0e7f 209Do not be fooled into thinking that C<(SV *) 0> is the same as C<&PL_sv_undef>.
a0d0e21e
LW
210Take this code:
211
212 SV* sv = (SV*) 0;
213 if (I-am-to-return-a-real-value) {
214 sv = sv_2mortal(newSViv(42));
215 }
216 sv_setsv(ST(0), sv);
217
218This code tries to return a new SV (which contains the value 42) if it should
04343c6d 219return a real value, or undef otherwise. Instead it has returned a NULL
a0d0e21e 220pointer which, somewhere down the line, will cause a segmentation violation,
9cde0e7f 221bus error, or just weird results. Change the zero to C<&PL_sv_undef> in the first
5f05dabc 222line and all will be well.
a0d0e21e
LW
223
224To free an SV that you've created, call C<SvREFCNT_dec(SV*)>. Normally this
3fe9a6f1 225call is not necessary (see L<Reference Counts and Mortality>).
a0d0e21e 226
94dde4fb
SC
227=head2 Offsets
228
229Perl provides the function C<sv_chop> to efficiently remove characters
230from the beginning of a string; you give it an SV and a pointer to
da75cd15 231somewhere inside the PV, and it discards everything before the
94dde4fb
SC
232pointer. The efficiency comes by means of a little hack: instead of
233actually removing the characters, C<sv_chop> sets the flag C<OOK>
234(offset OK) to signal to other functions that the offset hack is in
235effect, and it puts the number of bytes chopped off into the IV field
236of the SV. It then moves the PV pointer (called C<SvPVX>) forward that
00aadd71 237many bytes, and adjusts C<SvCUR> and C<SvLEN>.
94dde4fb
SC
238
239Hence, at this point, the start of the buffer that we allocated lives
240at C<SvPVX(sv) - SvIV(sv)> in memory and the PV pointer is pointing
241into the middle of this allocated storage.
242
243This is best demonstrated by example:
244
245 % ./perl -Ilib -MDevel::Peek -le '$a="12345"; $a=~s/.//; Dump($a)'
246 SV = PVIV(0x8128450) at 0x81340f0
247 REFCNT = 1
248 FLAGS = (POK,OOK,pPOK)
249 IV = 1 (OFFSET)
250 PV = 0x8135781 ( "1" . ) "2345"\0
251 CUR = 4
252 LEN = 5
253
254Here the number of bytes chopped off (1) is put into IV, and
255C<Devel::Peek::Dump> helpfully reminds us that this is an offset. The
256portion of the string between the "real" and the "fake" beginnings is
257shown in parentheses, and the values of C<SvCUR> and C<SvLEN> reflect
258the fake beginning, not the real one.
259
fe854a6f 260Something similar to the offset hack is performed on AVs to enable
319cef53
SC
261efficient shifting and splicing off the beginning of the array; while
262C<AvARRAY> points to the first element in the array that is visible from
263Perl, C<AvALLOC> points to the real start of the C array. These are
264usually the same, but a C<shift> operation can be carried out by
265increasing C<AvARRAY> by one and decreasing C<AvFILL> and C<AvLEN>.
266Again, the location of the real start of the C array only comes into
267play when freeing the array. See C<av_shift> in F<av.c>.
268
d1b91892 269=head2 What's Really Stored in an SV?
a0d0e21e
LW
270
271Recall that the usual method of determining the type of scalar you have is
5f05dabc 272to use C<Sv*OK> macros. Because a scalar can be both a number and a string,
d1b91892 273usually these macros will always return TRUE and calling the C<Sv*V>
a0d0e21e
LW
274macros will do the appropriate conversion of string to integer/double or
275integer/double to string.
276
277If you I<really> need to know if you have an integer, double, or string
278pointer in an SV, you can use the following three macros instead:
279
280 SvIOKp(SV*)
281 SvNOKp(SV*)
282 SvPOKp(SV*)
283
284These will tell you if you truly have an integer, double, or string pointer
d1b91892 285stored in your SV. The "p" stands for private.
a0d0e21e 286
9e9796d6
JH
287The are various ways in which the private and public flags may differ.
288For example, a tied SV may have a valid underlying value in the IV slot
289(so SvIOKp is true), but the data should be accessed via the FETCH
290routine rather than directly, so SvIOK is false. Another is when
291numeric conversion has occured and precision has been lost: only the
292private flag is set on 'lossy' values. So when an NV is converted to an
293IV with loss, SvIOKp, SvNOKp and SvNOK will be set, while SvIOK wont be.
294
07fa94a1 295In general, though, it's best to use the C<Sv*V> macros.
a0d0e21e 296
54310121 297=head2 Working with AVs
a0d0e21e 298
07fa94a1
JO
299There are two ways to create and load an AV. The first method creates an
300empty AV:
a0d0e21e
LW
301
302 AV* newAV();
303
54310121 304The second method both creates the AV and initially populates it with SVs:
a0d0e21e
LW
305
306 AV* av_make(I32 num, SV **ptr);
307
5f05dabc 308The second argument points to an array containing C<num> C<SV*>'s. Once the
54310121 309AV has been created, the SVs can be destroyed, if so desired.
a0d0e21e 310
54310121 311Once the AV has been created, the following operations are possible on AVs:
a0d0e21e
LW
312
313 void av_push(AV*, SV*);
314 SV* av_pop(AV*);
315 SV* av_shift(AV*);
316 void av_unshift(AV*, I32 num);
317
318These should be familiar operations, with the exception of C<av_unshift>.
319This routine adds C<num> elements at the front of the array with the C<undef>
320value. You must then use C<av_store> (described below) to assign values
321to these new elements.
322
323Here are some other functions:
324
5f05dabc 325 I32 av_len(AV*);
a0d0e21e 326 SV** av_fetch(AV*, I32 key, I32 lval);
a0d0e21e 327 SV** av_store(AV*, I32 key, SV* val);
a0d0e21e 328
5f05dabc
PP
329The C<av_len> function returns the highest index value in array (just
330like $#array in Perl). If the array is empty, -1 is returned. The
331C<av_fetch> function returns the value at index C<key>, but if C<lval>
332is non-zero, then C<av_fetch> will store an undef value at that index.
04343c6d
GS
333The C<av_store> function stores the value C<val> at index C<key>, and does
334not increment the reference count of C<val>. Thus the caller is responsible
335for taking care of that, and if C<av_store> returns NULL, the caller will
336have to decrement the reference count to avoid a memory leak. Note that
337C<av_fetch> and C<av_store> both return C<SV**>'s, not C<SV*>'s as their
338return value.
d1b91892 339
a0d0e21e 340 void av_clear(AV*);
a0d0e21e 341 void av_undef(AV*);
cb1a09d0 342 void av_extend(AV*, I32 key);
5f05dabc
PP
343
344The C<av_clear> function deletes all the elements in the AV* array, but
345does not actually delete the array itself. The C<av_undef> function will
346delete all the elements in the array plus the array itself. The
adc882cf
GS
347C<av_extend> function extends the array so that it contains at least C<key+1>
348elements. If C<key+1> is less than the currently allocated length of the array,
349then nothing is done.
a0d0e21e
LW
350
351If you know the name of an array variable, you can get a pointer to its AV
352by using the following:
353
4929bf7b 354 AV* get_av("package::varname", FALSE);
a0d0e21e
LW
355
356This returns NULL if the variable does not exist.
357
04343c6d
GS
358See L<Understanding the Magic of Tied Hashes and Arrays> for more
359information on how to use the array access functions on tied arrays.
360
54310121 361=head2 Working with HVs
a0d0e21e
LW
362
363To create an HV, you use the following routine:
364
365 HV* newHV();
366
54310121 367Once the HV has been created, the following operations are possible on HVs:
a0d0e21e 368
08105a92
GS
369 SV** hv_store(HV*, const char* key, U32 klen, SV* val, U32 hash);
370 SV** hv_fetch(HV*, const char* key, U32 klen, I32 lval);
a0d0e21e 371
5f05dabc
PP
372The C<klen> parameter is the length of the key being passed in (Note that
373you cannot pass 0 in as a value of C<klen> to tell Perl to measure the
374length of the key). The C<val> argument contains the SV pointer to the
54310121 375scalar being stored, and C<hash> is the precomputed hash value (zero if
5f05dabc
PP
376you want C<hv_store> to calculate it for you). The C<lval> parameter
377indicates whether this fetch is actually a part of a store operation, in
378which case a new undefined value will be added to the HV with the supplied
379key and C<hv_fetch> will return as if the value had already existed.
a0d0e21e 380
5f05dabc
PP
381Remember that C<hv_store> and C<hv_fetch> return C<SV**>'s and not just
382C<SV*>. To access the scalar value, you must first dereference the return
383value. However, you should check to make sure that the return value is
384not NULL before dereferencing it.
a0d0e21e
LW
385
386These two functions check if a hash table entry exists, and deletes it.
387
08105a92
GS
388 bool hv_exists(HV*, const char* key, U32 klen);
389 SV* hv_delete(HV*, const char* key, U32 klen, I32 flags);
a0d0e21e 390
5f05dabc
PP
391If C<flags> does not include the C<G_DISCARD> flag then C<hv_delete> will
392create and return a mortal copy of the deleted value.
393
a0d0e21e
LW
394And more miscellaneous functions:
395
396 void hv_clear(HV*);
a0d0e21e 397 void hv_undef(HV*);
5f05dabc
PP
398
399Like their AV counterparts, C<hv_clear> deletes all the entries in the hash
400table but does not actually delete the hash table. The C<hv_undef> deletes
401both the entries and the hash table itself.
a0d0e21e 402
d1b91892
AD
403Perl keeps the actual data in linked list of structures with a typedef of HE.
404These contain the actual key and value pointers (plus extra administrative
405overhead). The key is a string pointer; the value is an C<SV*>. However,
406once you have an C<HE*>, to get the actual key and value, use the routines
407specified below.
408
a0d0e21e
LW
409 I32 hv_iterinit(HV*);
410 /* Prepares starting point to traverse hash table */
411 HE* hv_iternext(HV*);
412 /* Get the next entry, and return a pointer to a
413 structure that has both the key and value */
414 char* hv_iterkey(HE* entry, I32* retlen);
415 /* Get the key from an HE structure and also return
416 the length of the key string */
cb1a09d0 417 SV* hv_iterval(HV*, HE* entry);
d1be9408 418 /* Return an SV pointer to the value of the HE
a0d0e21e 419 structure */
cb1a09d0 420 SV* hv_iternextsv(HV*, char** key, I32* retlen);
d1b91892
AD
421 /* This convenience routine combines hv_iternext,
422 hv_iterkey, and hv_iterval. The key and retlen
423 arguments are return values for the key and its
424 length. The value is returned in the SV* argument */
a0d0e21e
LW
425
426If you know the name of a hash variable, you can get a pointer to its HV
427by using the following:
428
4929bf7b 429 HV* get_hv("package::varname", FALSE);
a0d0e21e
LW
430
431This returns NULL if the variable does not exist.
432
8ebc5c01 433The hash algorithm is defined in the C<PERL_HASH(hash, key, klen)> macro:
a0d0e21e 434
a0d0e21e 435 hash = 0;
ab192400
GS
436 while (klen--)
437 hash = (hash * 33) + *key++;
87275199 438 hash = hash + (hash >> 5); /* after 5.6 */
ab192400 439
87275199 440The last step was added in version 5.6 to improve distribution of
ab192400 441lower bits in the resulting hash value.
a0d0e21e 442
04343c6d
GS
443See L<Understanding the Magic of Tied Hashes and Arrays> for more
444information on how to use the hash access functions on tied hashes.
445
1e422769
PP
446=head2 Hash API Extensions
447
448Beginning with version 5.004, the following functions are also supported:
449
450 HE* hv_fetch_ent (HV* tb, SV* key, I32 lval, U32 hash);
451 HE* hv_store_ent (HV* tb, SV* key, SV* val, U32 hash);
c47ff5f1 452
1e422769
PP
453 bool hv_exists_ent (HV* tb, SV* key, U32 hash);
454 SV* hv_delete_ent (HV* tb, SV* key, I32 flags, U32 hash);
c47ff5f1 455
1e422769
PP
456 SV* hv_iterkeysv (HE* entry);
457
458Note that these functions take C<SV*> keys, which simplifies writing
459of extension code that deals with hash structures. These functions
460also allow passing of C<SV*> keys to C<tie> functions without forcing
461you to stringify the keys (unlike the previous set of functions).
462
463They also return and accept whole hash entries (C<HE*>), making their
464use more efficient (since the hash number for a particular string
4a4eefd0
GS
465doesn't have to be recomputed every time). See L<perlapi> for detailed
466descriptions.
1e422769
PP
467
468The following macros must always be used to access the contents of hash
469entries. Note that the arguments to these macros must be simple
470variables, since they may get evaluated more than once. See
4a4eefd0 471L<perlapi> for detailed descriptions of these macros.
1e422769
PP
472
473 HePV(HE* he, STRLEN len)
474 HeVAL(HE* he)
475 HeHASH(HE* he)
476 HeSVKEY(HE* he)
477 HeSVKEY_force(HE* he)
478 HeSVKEY_set(HE* he, SV* sv)
479
480These two lower level macros are defined, but must only be used when
481dealing with keys that are not C<SV*>s:
482
483 HeKEY(HE* he)
484 HeKLEN(HE* he)
485
04343c6d
GS
486Note that both C<hv_store> and C<hv_store_ent> do not increment the
487reference count of the stored C<val>, which is the caller's responsibility.
488If these functions return a NULL value, the caller will usually have to
489decrement the reference count of C<val> to avoid a memory leak.
1e422769 490
a0d0e21e
LW
491=head2 References
492
d1b91892
AD
493References are a special type of scalar that point to other data types
494(including references).
a0d0e21e 495
07fa94a1 496To create a reference, use either of the following functions:
a0d0e21e 497
5f05dabc
PP
498 SV* newRV_inc((SV*) thing);
499 SV* newRV_noinc((SV*) thing);
a0d0e21e 500
5f05dabc 501The C<thing> argument can be any of an C<SV*>, C<AV*>, or C<HV*>. The
07fa94a1
JO
502functions are identical except that C<newRV_inc> increments the reference
503count of the C<thing>, while C<newRV_noinc> does not. For historical
504reasons, C<newRV> is a synonym for C<newRV_inc>.
505
506Once you have a reference, you can use the following macro to dereference
507the reference:
a0d0e21e
LW
508
509 SvRV(SV*)
510
511then call the appropriate routines, casting the returned C<SV*> to either an
d1b91892 512C<AV*> or C<HV*>, if required.
a0d0e21e 513
d1b91892 514To determine if an SV is a reference, you can use the following macro:
a0d0e21e
LW
515
516 SvROK(SV*)
517
07fa94a1
JO
518To discover what type of value the reference refers to, use the following
519macro and then check the return value.
d1b91892
AD
520
521 SvTYPE(SvRV(SV*))
522
523The most useful types that will be returned are:
524
525 SVt_IV Scalar
526 SVt_NV Scalar
527 SVt_PV Scalar
5f05dabc 528 SVt_RV Scalar
d1b91892
AD
529 SVt_PVAV Array
530 SVt_PVHV Hash
531 SVt_PVCV Code
5f05dabc
PP
532 SVt_PVGV Glob (possible a file handle)
533 SVt_PVMG Blessed or Magical Scalar
534
535 See the sv.h header file for more details.
d1b91892 536
cb1a09d0
AD
537=head2 Blessed References and Class Objects
538
539References are also used to support object-oriented programming. In the
540OO lexicon, an object is simply a reference that has been blessed into a
541package (or class). Once blessed, the programmer may now use the reference
542to access the various methods in the class.
543
544A reference can be blessed into a package with the following function:
545
546 SV* sv_bless(SV* sv, HV* stash);
547
548The C<sv> argument must be a reference. The C<stash> argument specifies
3fe9a6f1 549which class the reference will belong to. See
2ae324a7 550L<Stashes and Globs> for information on converting class names into stashes.
cb1a09d0
AD
551
552/* Still under construction */
553
554Upgrades rv to reference if not already one. Creates new SV for rv to
8ebc5c01
PP
555point to. If C<classname> is non-null, the SV is blessed into the specified
556class. SV is returned.
cb1a09d0 557
08105a92 558 SV* newSVrv(SV* rv, const char* classname);
cb1a09d0 559
e1c57cef 560Copies integer, unsigned integer or double into an SV whose reference is C<rv>. SV is blessed
8ebc5c01 561if C<classname> is non-null.
cb1a09d0 562
08105a92 563 SV* sv_setref_iv(SV* rv, const char* classname, IV iv);
e1c57cef 564 SV* sv_setref_uv(SV* rv, const char* classname, UV uv);
08105a92 565 SV* sv_setref_nv(SV* rv, const char* classname, NV iv);
cb1a09d0 566
5f05dabc 567Copies the pointer value (I<the address, not the string!>) into an SV whose
8ebc5c01 568reference is rv. SV is blessed if C<classname> is non-null.
cb1a09d0 569
08105a92 570 SV* sv_setref_pv(SV* rv, const char* classname, PV iv);
cb1a09d0 571
8ebc5c01
PP
572Copies string into an SV whose reference is C<rv>. Set length to 0 to let
573Perl calculate the string length. SV is blessed if C<classname> is non-null.
cb1a09d0 574
e65f3abd 575 SV* sv_setref_pvn(SV* rv, const char* classname, PV iv, STRLEN length);
cb1a09d0 576
9abd00ed
GS
577Tests whether the SV is blessed into the specified class. It does not
578check inheritance relationships.
579
08105a92 580 int sv_isa(SV* sv, const char* name);
9abd00ed
GS
581
582Tests whether the SV is a reference to a blessed object.
583
584 int sv_isobject(SV* sv);
585
586Tests whether the SV is derived from the specified class. SV can be either
587a reference to a blessed object or a string containing a class name. This
588is the function implementing the C<UNIVERSAL::isa> functionality.
589
08105a92 590 bool sv_derived_from(SV* sv, const char* name);
9abd00ed 591
00aadd71 592To check if you've got an object derived from a specific class you have
9abd00ed
GS
593to write:
594
595 if (sv_isobject(sv) && sv_derived_from(sv, class)) { ... }
cb1a09d0 596
5f05dabc 597=head2 Creating New Variables
cb1a09d0 598
5f05dabc
PP
599To create a new Perl variable with an undef value which can be accessed from
600your Perl script, use the following routines, depending on the variable type.
cb1a09d0 601
4929bf7b
GS
602 SV* get_sv("package::varname", TRUE);
603 AV* get_av("package::varname", TRUE);
604 HV* get_hv("package::varname", TRUE);
cb1a09d0
AD
605
606Notice the use of TRUE as the second parameter. The new variable can now
607be set, using the routines appropriate to the data type.
608
5f05dabc
PP
609There are additional macros whose values may be bitwise OR'ed with the
610C<TRUE> argument to enable certain extra features. Those bits are:
cb1a09d0 611
9a68f1db
SB
612=over
613
614=item GV_ADDMULTI
615
616Marks the variable as multiply defined, thus preventing the:
617
618 Name <varname> used only once: possible typo
619
620warning.
621
9a68f1db
SB
622=item GV_ADDWARN
623
624Issues the warning:
625
626 Had to create <varname> unexpectedly
627
628if the variable did not exist before the function was called.
629
630=back
cb1a09d0 631
07fa94a1
JO
632If you do not specify a package name, the variable is created in the current
633package.
cb1a09d0 634
5f05dabc 635=head2 Reference Counts and Mortality
a0d0e21e 636
d1be9408 637Perl uses a reference count-driven garbage collection mechanism. SVs,
54310121 638AVs, or HVs (xV for short in the following) start their life with a
55497cff 639reference count of 1. If the reference count of an xV ever drops to 0,
07fa94a1 640then it will be destroyed and its memory made available for reuse.
55497cff
PP
641
642This normally doesn't happen at the Perl level unless a variable is
5f05dabc
PP
643undef'ed or the last variable holding a reference to it is changed or
644overwritten. At the internal level, however, reference counts can be
55497cff
PP
645manipulated with the following macros:
646
647 int SvREFCNT(SV* sv);
5f05dabc 648 SV* SvREFCNT_inc(SV* sv);
55497cff
PP
649 void SvREFCNT_dec(SV* sv);
650
651However, there is one other function which manipulates the reference
07fa94a1
JO
652count of its argument. The C<newRV_inc> function, you will recall,
653creates a reference to the specified argument. As a side effect,
654it increments the argument's reference count. If this is not what
655you want, use C<newRV_noinc> instead.
656
657For example, imagine you want to return a reference from an XSUB function.
658Inside the XSUB routine, you create an SV which initially has a reference
659count of one. Then you call C<newRV_inc>, passing it the just-created SV.
5f05dabc
PP
660This returns the reference as a new SV, but the reference count of the
661SV you passed to C<newRV_inc> has been incremented to two. Now you
07fa94a1
JO
662return the reference from the XSUB routine and forget about the SV.
663But Perl hasn't! Whenever the returned reference is destroyed, the
664reference count of the original SV is decreased to one and nothing happens.
665The SV will hang around without any way to access it until Perl itself
666terminates. This is a memory leak.
5f05dabc
PP
667
668The correct procedure, then, is to use C<newRV_noinc> instead of
faed5253
JO
669C<newRV_inc>. Then, if and when the last reference is destroyed,
670the reference count of the SV will go to zero and it will be destroyed,
07fa94a1 671stopping any memory leak.
55497cff 672
5f05dabc 673There are some convenience functions available that can help with the
54310121 674destruction of xVs. These functions introduce the concept of "mortality".
07fa94a1
JO
675An xV that is mortal has had its reference count marked to be decremented,
676but not actually decremented, until "a short time later". Generally the
677term "short time later" means a single Perl statement, such as a call to
54310121 678an XSUB function. The actual determinant for when mortal xVs have their
07fa94a1
JO
679reference count decremented depends on two macros, SAVETMPS and FREETMPS.
680See L<perlcall> and L<perlxs> for more details on these macros.
55497cff
PP
681
682"Mortalization" then is at its simplest a deferred C<SvREFCNT_dec>.
683However, if you mortalize a variable twice, the reference count will
684later be decremented twice.
685
00aadd71
NIS
686"Mortal" SVs are mainly used for SVs that are placed on perl's stack.
687For example an SV which is created just to pass a number to a called sub
688is made mortal to have it cleaned up automatically when stack is popped.
689Similarly results returned by XSUBs (which go in the stack) are often
690made mortal.
a0d0e21e
LW
691
692To create a mortal variable, use the functions:
693
694 SV* sv_newmortal()
695 SV* sv_2mortal(SV*)
696 SV* sv_mortalcopy(SV*)
697
00aadd71 698The first call creates a mortal SV (with no value), the second converts an existing
5f05dabc
PP
699SV to a mortal SV (and thus defers a call to C<SvREFCNT_dec>), and the
700third creates a mortal copy of an existing SV.
00aadd71 701Because C<sv_newmortal> gives the new SV no value,it must normally be given one
9a68f1db 702via C<sv_setpv>, C<sv_setiv>, etc. :
00aadd71
NIS
703
704 SV *tmp = sv_newmortal();
705 sv_setiv(tmp, an_integer);
706
707As that is multiple C statements it is quite common so see this idiom instead:
708
709 SV *tmp = sv_2mortal(newSViv(an_integer));
710
711
712You should be careful about creating mortal variables. Strange things
713can happen if you make the same value mortal within multiple contexts,
714or if you make a variable mortal multiple times. Thinking of "Mortalization"
715as deferred C<SvREFCNT_dec> should help to minimize such problems.
716For example if you are passing an SV which you I<know> has high enough REFCNT
717to survive its use on the stack you need not do any mortalization.
718If you are not sure then doing an C<SvREFCNT_inc> and C<sv_2mortal>, or
719making a C<sv_mortalcopy> is safer.
a0d0e21e 720
54310121 721The mortal routines are not just for SVs -- AVs and HVs can be
faed5253 722made mortal by passing their address (type-casted to C<SV*>) to the
07fa94a1 723C<sv_2mortal> or C<sv_mortalcopy> routines.
a0d0e21e 724
5f05dabc 725=head2 Stashes and Globs
a0d0e21e 726
aa689395
PP
727A "stash" is a hash that contains all of the different objects that
728are contained within a package. Each key of the stash is a symbol
729name (shared by all the different types of objects that have the same
730name), and each value in the hash table is a GV (Glob Value). This GV
731in turn contains references to the various objects of that name,
732including (but not limited to) the following:
cb1a09d0 733
a0d0e21e
LW
734 Scalar Value
735 Array Value
736 Hash Value
a3cb178b 737 I/O Handle
a0d0e21e
LW
738 Format
739 Subroutine
740
9cde0e7f 741There is a single stash called "PL_defstash" that holds the items that exist
5f05dabc
PP
742in the "main" package. To get at the items in other packages, append the
743string "::" to the package name. The items in the "Foo" package are in
9cde0e7f 744the stash "Foo::" in PL_defstash. The items in the "Bar::Baz" package are
5f05dabc 745in the stash "Baz::" in "Bar::"'s stash.
a0d0e21e 746
d1b91892 747To get the stash pointer for a particular package, use the function:
a0d0e21e 748
08105a92 749 HV* gv_stashpv(const char* name, I32 create)
a0d0e21e
LW
750 HV* gv_stashsv(SV*, I32 create)
751
752The first function takes a literal string, the second uses the string stored
d1b91892 753in the SV. Remember that a stash is just a hash table, so you get back an
cb1a09d0 754C<HV*>. The C<create> flag will create a new package if it is set.
a0d0e21e
LW
755
756The name that C<gv_stash*v> wants is the name of the package whose symbol table
757you want. The default package is called C<main>. If you have multiply nested
d1b91892
AD
758packages, pass their names to C<gv_stash*v>, separated by C<::> as in the Perl
759language itself.
a0d0e21e
LW
760
761Alternately, if you have an SV that is a blessed reference, you can find
762out the stash pointer by using:
763
764 HV* SvSTASH(SvRV(SV*));
765
766then use the following to get the package name itself:
767
768 char* HvNAME(HV* stash);
769
5f05dabc
PP
770If you need to bless or re-bless an object you can use the following
771function:
a0d0e21e
LW
772
773 SV* sv_bless(SV*, HV* stash)
774
775where the first argument, an C<SV*>, must be a reference, and the second
776argument is a stash. The returned C<SV*> can now be used in the same way
777as any other SV.
778
d1b91892
AD
779For more information on references and blessings, consult L<perlref>.
780
54310121 781=head2 Double-Typed SVs
0a753a76
PP
782
783Scalar variables normally contain only one type of value, an integer,
784double, pointer, or reference. Perl will automatically convert the
785actual scalar data from the stored type into the requested type.
786
787Some scalar variables contain more than one type of scalar data. For
788example, the variable C<$!> contains either the numeric value of C<errno>
789or its string equivalent from either C<strerror> or C<sys_errlist[]>.
790
791To force multiple data values into an SV, you must do two things: use the
792C<sv_set*v> routines to add the additional scalar type, then set a flag
793so that Perl will believe it contains more than one type of data. The
794four macros to set the flags are:
795
796 SvIOK_on
797 SvNOK_on
798 SvPOK_on
799 SvROK_on
800
801The particular macro you must use depends on which C<sv_set*v> routine
802you called first. This is because every C<sv_set*v> routine turns on
803only the bit for the particular type of data being set, and turns off
804all the rest.
805
806For example, to create a new Perl variable called "dberror" that contains
807both the numeric and descriptive string error values, you could use the
808following code:
809
810 extern int dberror;
811 extern char *dberror_list;
812
4929bf7b 813 SV* sv = get_sv("dberror", TRUE);
0a753a76
PP
814 sv_setiv(sv, (IV) dberror);
815 sv_setpv(sv, dberror_list[dberror]);
816 SvIOK_on(sv);
817
818If the order of C<sv_setiv> and C<sv_setpv> had been reversed, then the
819macro C<SvPOK_on> would need to be called instead of C<SvIOK_on>.
820
821=head2 Magic Variables
a0d0e21e 822
d1b91892
AD
823[This section still under construction. Ignore everything here. Post no
824bills. Everything not permitted is forbidden.]
825
d1b91892
AD
826Any SV may be magical, that is, it has special features that a normal
827SV does not have. These features are stored in the SV structure in a
5f05dabc 828linked list of C<struct magic>'s, typedef'ed to C<MAGIC>.
d1b91892
AD
829
830 struct magic {
831 MAGIC* mg_moremagic;
832 MGVTBL* mg_virtual;
833 U16 mg_private;
834 char mg_type;
835 U8 mg_flags;
836 SV* mg_obj;
837 char* mg_ptr;
838 I32 mg_len;
839 };
840
841Note this is current as of patchlevel 0, and could change at any time.
842
843=head2 Assigning Magic
844
845Perl adds magic to an SV using the sv_magic function:
846
08105a92 847 void sv_magic(SV* sv, SV* obj, int how, const char* name, I32 namlen);
d1b91892
AD
848
849The C<sv> argument is a pointer to the SV that is to acquire a new magical
850feature.
851
852If C<sv> is not already magical, Perl uses the C<SvUPGRADE> macro to
645c22ef
DM
853convert C<sv> to type C<SVt_PVMG>. Perl then continues by adding new magic
854to the beginning of the linked list of magical features. Any prior entry
855of the same type of magic is deleted. Note that this can be overridden,
856and multiple instances of the same type of magic can be associated with an
857SV.
d1b91892 858
54310121
PP
859The C<name> and C<namlen> arguments are used to associate a string with
860the magic, typically the name of a variable. C<namlen> is stored in the
9b5bb84f 861C<mg_len> field and if C<name> is non-null and C<namlen> E<gt>= 0 a malloc'd
d1b91892
AD
862copy of the name is stored in C<mg_ptr> field.
863
864The sv_magic function uses C<how> to determine which, if any, predefined
865"Magic Virtual Table" should be assigned to the C<mg_virtual> field.
cb1a09d0 866See the "Magic Virtual Table" section below. The C<how> argument is also
14befaf4
DM
867stored in the C<mg_type> field. The value of C<how> should be chosen
868from the set of macros C<PERL_MAGIC_foo> found perl.h. Note that before
645c22ef 869these macros were added, Perl internals used to directly use character
14befaf4 870literals, so you may occasionally come across old code or documentation
75d0f26d 871referring to 'U' magic rather than C<PERL_MAGIC_uvar> for example.
d1b91892
AD
872
873The C<obj> argument is stored in the C<mg_obj> field of the C<MAGIC>
874structure. If it is not the same as the C<sv> argument, the reference
875count of the C<obj> object is incremented. If it is the same, or if
645c22ef 876the C<how> argument is C<PERL_MAGIC_arylen>, or if it is a NULL pointer,
14befaf4 877then C<obj> is merely stored, without the reference count being incremented.
d1b91892 878
cb1a09d0
AD
879There is also a function to add magic to an C<HV>:
880
881 void hv_magic(HV *hv, GV *gv, int how);
882
883This simply calls C<sv_magic> and coerces the C<gv> argument into an C<SV>.
884
885To remove the magic from an SV, call the function sv_unmagic:
886
887 void sv_unmagic(SV *sv, int type);
888
889The C<type> argument should be equal to the C<how> value when the C<SV>
890was initially made magical.
891
d1b91892
AD
892=head2 Magic Virtual Tables
893
d1be9408 894The C<mg_virtual> field in the C<MAGIC> structure is a pointer to an
d1b91892
AD
895C<MGVTBL>, which is a structure of function pointers and stands for
896"Magic Virtual Table" to handle the various operations that might be
897applied to that variable.
898
899The C<MGVTBL> has five pointers to the following routine types:
900
901 int (*svt_get)(SV* sv, MAGIC* mg);
902 int (*svt_set)(SV* sv, MAGIC* mg);
903 U32 (*svt_len)(SV* sv, MAGIC* mg);
904 int (*svt_clear)(SV* sv, MAGIC* mg);
905 int (*svt_free)(SV* sv, MAGIC* mg);
906
907This MGVTBL structure is set at compile-time in C<perl.h> and there are
908currently 19 types (or 21 with overloading turned on). These different
909structures contain pointers to various routines that perform additional
910actions depending on which function is being called.
911
912 Function pointer Action taken
913 ---------------- ------------
8b0711c3 914 svt_get Do something before the value of the SV is retrieved.
d1b91892
AD
915 svt_set Do something after the SV is assigned a value.
916 svt_len Report on the SV's length.
917 svt_clear Clear something the SV represents.
918 svt_free Free any extra storage associated with the SV.
919
920For instance, the MGVTBL structure called C<vtbl_sv> (which corresponds
14befaf4 921to an C<mg_type> of C<PERL_MAGIC_sv>) contains:
d1b91892
AD
922
923 { magic_get, magic_set, magic_len, 0, 0 }
924
14befaf4
DM
925Thus, when an SV is determined to be magical and of type C<PERL_MAGIC_sv>,
926if a get operation is being performed, the routine C<magic_get> is
927called. All the various routines for the various magical types begin
928with C<magic_>. NOTE: the magic routines are not considered part of
929the Perl API, and may not be exported by the Perl library.
d1b91892
AD
930
931The current kinds of Magic Virtual Tables are:
932
14befaf4
DM
933 mg_type
934 (old-style char and macro) MGVTBL Type of magic
935 -------------------------- ------ ----------------------------
936 \0 PERL_MAGIC_sv vtbl_sv Special scalar variable
937 A PERL_MAGIC_overload vtbl_amagic %OVERLOAD hash
938 a PERL_MAGIC_overload_elem vtbl_amagicelem %OVERLOAD hash element
939 c PERL_MAGIC_overload_table (none) Holds overload table (AMT)
940 on stash
941 B PERL_MAGIC_bm vtbl_bm Boyer-Moore (fast string search)
942 D PERL_MAGIC_regdata vtbl_regdata Regex match position data
943 (@+ and @- vars)
944 d PERL_MAGIC_regdatum vtbl_regdatum Regex match position data
945 element
946 E PERL_MAGIC_env vtbl_env %ENV hash
947 e PERL_MAGIC_envelem vtbl_envelem %ENV hash element
948 f PERL_MAGIC_fm vtbl_fm Formline ('compiled' format)
949 g PERL_MAGIC_regex_global vtbl_mglob m//g target / study()ed string
950 I PERL_MAGIC_isa vtbl_isa @ISA array
951 i PERL_MAGIC_isaelem vtbl_isaelem @ISA array element
952 k PERL_MAGIC_nkeys vtbl_nkeys scalar(keys()) lvalue
953 L PERL_MAGIC_dbfile (none) Debugger %_<filename
954 l PERL_MAGIC_dbline vtbl_dbline Debugger %_<filename element
955 m PERL_MAGIC_mutex vtbl_mutex ???
645c22ef 956 o PERL_MAGIC_collxfrm vtbl_collxfrm Locale collate transformation
14befaf4
DM
957 P PERL_MAGIC_tied vtbl_pack Tied array or hash
958 p PERL_MAGIC_tiedelem vtbl_packelem Tied array or hash element
959 q PERL_MAGIC_tiedscalar vtbl_packelem Tied scalar or handle
960 r PERL_MAGIC_qr vtbl_qr precompiled qr// regex
961 S PERL_MAGIC_sig vtbl_sig %SIG hash
962 s PERL_MAGIC_sigelem vtbl_sigelem %SIG hash element
963 t PERL_MAGIC_taint vtbl_taint Taintedness
964 U PERL_MAGIC_uvar vtbl_uvar Available for use by extensions
965 v PERL_MAGIC_vec vtbl_vec vec() lvalue
92f0c265 966 V PERL_MAGIC_vstring (none) v-string scalars
836995da 967 w PERL_MAGIC_utf8 vtbl_utf8 UTF-8 length+offset cache
14befaf4
DM
968 x PERL_MAGIC_substr vtbl_substr substr() lvalue
969 y PERL_MAGIC_defelem vtbl_defelem Shadow "foreach" iterator
970 variable / smart parameter
971 vivification
972 * PERL_MAGIC_glob vtbl_glob GV (typeglob)
973 # PERL_MAGIC_arylen vtbl_arylen Array length ($#ary)
974 . PERL_MAGIC_pos vtbl_pos pos() lvalue
975 < PERL_MAGIC_backref vtbl_backref ???
976 ~ PERL_MAGIC_ext (none) Available for use by extensions
d1b91892 977
68dc0745 978When an uppercase and lowercase letter both exist in the table, then the
92f0c265
JP
979uppercase letter is typically used to represent some kind of composite type
980(a list or a hash), and the lowercase letter is used to represent an element
981of that composite type. Some internals code makes use of this case
982relationship. However, 'v' and 'V' (vec and v-string) are in no way related.
14befaf4
DM
983
984The C<PERL_MAGIC_ext> and C<PERL_MAGIC_uvar> magic types are defined
985specifically for use by extensions and will not be used by perl itself.
986Extensions can use C<PERL_MAGIC_ext> magic to 'attach' private information
987to variables (typically objects). This is especially useful because
988there is no way for normal perl code to corrupt this private information
989(unlike using extra elements of a hash object).
990
991Similarly, C<PERL_MAGIC_uvar> magic can be used much like tie() to call a
992C function any time a scalar's value is used or changed. The C<MAGIC>'s
bdbeb323
SM
993C<mg_ptr> field points to a C<ufuncs> structure:
994
995 struct ufuncs {
a9402793
AB
996 I32 (*uf_val)(pTHX_ IV, SV*);
997 I32 (*uf_set)(pTHX_ IV, SV*);
bdbeb323
SM
998 IV uf_index;
999 };
1000
1001When the SV is read from or written to, the C<uf_val> or C<uf_set>
14befaf4
DM
1002function will be called with C<uf_index> as the first arg and a pointer to
1003the SV as the second. A simple example of how to add C<PERL_MAGIC_uvar>
1526ead6
AB
1004magic is shown below. Note that the ufuncs structure is copied by
1005sv_magic, so you can safely allocate it on the stack.
1006
1007 void
1008 Umagic(sv)
1009 SV *sv;
1010 PREINIT:
1011 struct ufuncs uf;
1012 CODE:
1013 uf.uf_val = &my_get_fn;
1014 uf.uf_set = &my_set_fn;
1015 uf.uf_index = 0;
14befaf4 1016 sv_magic(sv, 0, PERL_MAGIC_uvar, (char*)&uf, sizeof(uf));
5f05dabc 1017
14befaf4
DM
1018Note that because multiple extensions may be using C<PERL_MAGIC_ext>
1019or C<PERL_MAGIC_uvar> magic, it is important for extensions to take
1020extra care to avoid conflict. Typically only using the magic on
1021objects blessed into the same class as the extension is sufficient.
1022For C<PERL_MAGIC_ext> magic, it may also be appropriate to add an I32
1023'signature' at the top of the private data area and check that.
5f05dabc 1024
ef50df4b
GS
1025Also note that the C<sv_set*()> and C<sv_cat*()> functions described
1026earlier do B<not> invoke 'set' magic on their targets. This must
1027be done by the user either by calling the C<SvSETMAGIC()> macro after
1028calling these functions, or by using one of the C<sv_set*_mg()> or
1029C<sv_cat*_mg()> functions. Similarly, generic C code must call the
1030C<SvGETMAGIC()> macro to invoke any 'get' magic if they use an SV
1031obtained from external sources in functions that don't handle magic.
4a4eefd0 1032See L<perlapi> for a description of these functions.
189b2af5
GS
1033For example, calls to the C<sv_cat*()> functions typically need to be
1034followed by C<SvSETMAGIC()>, but they don't need a prior C<SvGETMAGIC()>
1035since their implementation handles 'get' magic.
1036
d1b91892
AD
1037=head2 Finding Magic
1038
1039 MAGIC* mg_find(SV*, int type); /* Finds the magic pointer of that type */
1040
1041This routine returns a pointer to the C<MAGIC> structure stored in the SV.
1042If the SV does not have that magical feature, C<NULL> is returned. Also,
54310121 1043if the SV is not of type SVt_PVMG, Perl may core dump.
d1b91892 1044
08105a92 1045 int mg_copy(SV* sv, SV* nsv, const char* key, STRLEN klen);
d1b91892
AD
1046
1047This routine checks to see what types of magic C<sv> has. If the mg_type
68dc0745
PP
1048field is an uppercase letter, then the mg_obj is copied to C<nsv>, but
1049the mg_type field is changed to be the lowercase letter.
a0d0e21e 1050
04343c6d
GS
1051=head2 Understanding the Magic of Tied Hashes and Arrays
1052
14befaf4
DM
1053Tied hashes and arrays are magical beasts of the C<PERL_MAGIC_tied>
1054magic type.
9edb2b46
GS
1055
1056WARNING: As of the 5.004 release, proper usage of the array and hash
1057access functions requires understanding a few caveats. Some
1058of these caveats are actually considered bugs in the API, to be fixed
1059in later releases, and are bracketed with [MAYCHANGE] below. If
1060you find yourself actually applying such information in this section, be
1061aware that the behavior may change in the future, umm, without warning.
04343c6d 1062
1526ead6 1063The perl tie function associates a variable with an object that implements
9a68f1db 1064the various GET, SET, etc methods. To perform the equivalent of the perl
1526ead6
AB
1065tie function from an XSUB, you must mimic this behaviour. The code below
1066carries out the necessary steps - firstly it creates a new hash, and then
1067creates a second hash which it blesses into the class which will implement
1068the tie methods. Lastly it ties the two hashes together, and returns a
1069reference to the new tied hash. Note that the code below does NOT call the
1070TIEHASH method in the MyTie class -
1071see L<Calling Perl Routines from within C Programs> for details on how
1072to do this.
1073
1074 SV*
1075 mytie()
1076 PREINIT:
1077 HV *hash;
1078 HV *stash;
1079 SV *tie;
1080 CODE:
1081 hash = newHV();
1082 tie = newRV_noinc((SV*)newHV());
1083 stash = gv_stashpv("MyTie", TRUE);
1084 sv_bless(tie, stash);
899e16d0 1085 hv_magic(hash, (GV*)tie, PERL_MAGIC_tied);
1526ead6
AB
1086 RETVAL = newRV_noinc(hash);
1087 OUTPUT:
1088 RETVAL
1089
04343c6d
GS
1090The C<av_store> function, when given a tied array argument, merely
1091copies the magic of the array onto the value to be "stored", using
1092C<mg_copy>. It may also return NULL, indicating that the value did not
9edb2b46
GS
1093actually need to be stored in the array. [MAYCHANGE] After a call to
1094C<av_store> on a tied array, the caller will usually need to call
1095C<mg_set(val)> to actually invoke the perl level "STORE" method on the
1096TIEARRAY object. If C<av_store> did return NULL, a call to
1097C<SvREFCNT_dec(val)> will also be usually necessary to avoid a memory
1098leak. [/MAYCHANGE]
04343c6d
GS
1099
1100The previous paragraph is applicable verbatim to tied hash access using the
1101C<hv_store> and C<hv_store_ent> functions as well.
1102
1103C<av_fetch> and the corresponding hash functions C<hv_fetch> and
1104C<hv_fetch_ent> actually return an undefined mortal value whose magic
1105has been initialized using C<mg_copy>. Note the value so returned does not
9edb2b46
GS
1106need to be deallocated, as it is already mortal. [MAYCHANGE] But you will
1107need to call C<mg_get()> on the returned value in order to actually invoke
1108the perl level "FETCH" method on the underlying TIE object. Similarly,
04343c6d
GS
1109you may also call C<mg_set()> on the return value after possibly assigning
1110a suitable value to it using C<sv_setsv>, which will invoke the "STORE"
9edb2b46 1111method on the TIE object. [/MAYCHANGE]
04343c6d 1112
9edb2b46 1113[MAYCHANGE]
04343c6d
GS
1114In other words, the array or hash fetch/store functions don't really
1115fetch and store actual values in the case of tied arrays and hashes. They
1116merely call C<mg_copy> to attach magic to the values that were meant to be
1117"stored" or "fetched". Later calls to C<mg_get> and C<mg_set> actually
1118do the job of invoking the TIE methods on the underlying objects. Thus
9edb2b46 1119the magic mechanism currently implements a kind of lazy access to arrays
04343c6d
GS
1120and hashes.
1121
1122Currently (as of perl version 5.004), use of the hash and array access
1123functions requires the user to be aware of whether they are operating on
9edb2b46
GS
1124"normal" hashes and arrays, or on their tied variants. The API may be
1125changed to provide more transparent access to both tied and normal data
1126types in future versions.
1127[/MAYCHANGE]
04343c6d
GS
1128
1129You would do well to understand that the TIEARRAY and TIEHASH interfaces
1130are mere sugar to invoke some perl method calls while using the uniform hash
1131and array syntax. The use of this sugar imposes some overhead (typically
1132about two to four extra opcodes per FETCH/STORE operation, in addition to
1133the creation of all the mortal variables required to invoke the methods).
1134This overhead will be comparatively small if the TIE methods are themselves
1135substantial, but if they are only a few statements long, the overhead
1136will not be insignificant.
1137
d1c897a1
IZ
1138=head2 Localizing changes
1139
1140Perl has a very handy construction
1141
1142 {
1143 local $var = 2;
1144 ...
1145 }
1146
1147This construction is I<approximately> equivalent to
1148
1149 {
1150 my $oldvar = $var;
1151 $var = 2;
1152 ...
1153 $var = $oldvar;
1154 }
1155
1156The biggest difference is that the first construction would
1157reinstate the initial value of $var, irrespective of how control exits
9a68f1db 1158the block: C<goto>, C<return>, C<die>/C<eval>, etc. It is a little bit
d1c897a1
IZ
1159more efficient as well.
1160
1161There is a way to achieve a similar task from C via Perl API: create a
1162I<pseudo-block>, and arrange for some changes to be automatically
1163undone at the end of it, either explicit, or via a non-local exit (via
1164die()). A I<block>-like construct is created by a pair of
b687b08b
TC
1165C<ENTER>/C<LEAVE> macros (see L<perlcall/"Returning a Scalar">).
1166Such a construct may be created specially for some important localized
1167task, or an existing one (like boundaries of enclosing Perl
1168subroutine/block, or an existing pair for freeing TMPs) may be
1169used. (In the second case the overhead of additional localization must
1170be almost negligible.) Note that any XSUB is automatically enclosed in
1171an C<ENTER>/C<LEAVE> pair.
d1c897a1
IZ
1172
1173Inside such a I<pseudo-block> the following service is available:
1174
13a2d996 1175=over 4
d1c897a1
IZ
1176
1177=item C<SAVEINT(int i)>
1178
1179=item C<SAVEIV(IV i)>
1180
1181=item C<SAVEI32(I32 i)>
1182
1183=item C<SAVELONG(long i)>
1184
1185These macros arrange things to restore the value of integer variable
1186C<i> at the end of enclosing I<pseudo-block>.
1187
1188=item C<SAVESPTR(s)>
1189
1190=item C<SAVEPPTR(p)>
1191
1192These macros arrange things to restore the value of pointers C<s> and
1193C<p>. C<s> must be a pointer of a type which survives conversion to
1194C<SV*> and back, C<p> should be able to survive conversion to C<char*>
1195and back.
1196
1197=item C<SAVEFREESV(SV *sv)>
1198
1199The refcount of C<sv> would be decremented at the end of
26d9b02f
JH
1200I<pseudo-block>. This is similar to C<sv_2mortal> in that it is also a
1201mechanism for doing a delayed C<SvREFCNT_dec>. However, while C<sv_2mortal>
1202extends the lifetime of C<sv> until the beginning of the next statement,
1203C<SAVEFREESV> extends it until the end of the enclosing scope. These
1204lifetimes can be wildly different.
1205
1206Also compare C<SAVEMORTALIZESV>.
1207
1208=item C<SAVEMORTALIZESV(SV *sv)>
1209
1210Just like C<SAVEFREESV>, but mortalizes C<sv> at the end of the current
1211scope instead of decrementing its reference count. This usually has the
1212effect of keeping C<sv> alive until the statement that called the currently
1213live scope has finished executing.
d1c897a1
IZ
1214
1215=item C<SAVEFREEOP(OP *op)>
1216
1217The C<OP *> is op_free()ed at the end of I<pseudo-block>.
1218
1219=item C<SAVEFREEPV(p)>
1220
1221The chunk of memory which is pointed to by C<p> is Safefree()ed at the
1222end of I<pseudo-block>.
1223
1224=item C<SAVECLEARSV(SV *sv)>
1225
1226Clears a slot in the current scratchpad which corresponds to C<sv> at
1227the end of I<pseudo-block>.
1228
1229=item C<SAVEDELETE(HV *hv, char *key, I32 length)>
1230
1231The key C<key> of C<hv> is deleted at the end of I<pseudo-block>. The
1232string pointed to by C<key> is Safefree()ed. If one has a I<key> in
1233short-lived storage, the corresponding string may be reallocated like
1234this:
1235
9cde0e7f 1236 SAVEDELETE(PL_defstash, savepv(tmpbuf), strlen(tmpbuf));
d1c897a1 1237
c76ac1ee 1238=item C<SAVEDESTRUCTOR(DESTRUCTORFUNC_NOCONTEXT_t f, void *p)>
d1c897a1
IZ
1239
1240At the end of I<pseudo-block> the function C<f> is called with the
c76ac1ee
GS
1241only argument C<p>.
1242
1243=item C<SAVEDESTRUCTOR_X(DESTRUCTORFUNC_t f, void *p)>
1244
1245At the end of I<pseudo-block> the function C<f> is called with the
1246implicit context argument (if any), and C<p>.
d1c897a1
IZ
1247
1248=item C<SAVESTACK_POS()>
1249
1250The current offset on the Perl internal stack (cf. C<SP>) is restored
1251at the end of I<pseudo-block>.
1252
1253=back
1254
1255The following API list contains functions, thus one needs to
1256provide pointers to the modifiable data explicitly (either C pointers,
00aadd71 1257or Perlish C<GV *>s). Where the above macros take C<int>, a similar
d1c897a1
IZ
1258function takes C<int *>.
1259
13a2d996 1260=over 4
d1c897a1
IZ
1261
1262=item C<SV* save_scalar(GV *gv)>
1263
1264Equivalent to Perl code C<local $gv>.
1265
1266=item C<AV* save_ary(GV *gv)>
1267
1268=item C<HV* save_hash(GV *gv)>
1269
1270Similar to C<save_scalar>, but localize C<@gv> and C<%gv>.
1271
1272=item C<void save_item(SV *item)>
1273
1274Duplicates the current value of C<SV>, on the exit from the current
1275C<ENTER>/C<LEAVE> I<pseudo-block> will restore the value of C<SV>
1276using the stored value.
1277
1278=item C<void save_list(SV **sarg, I32 maxsarg)>
1279
1280A variant of C<save_item> which takes multiple arguments via an array
1281C<sarg> of C<SV*> of length C<maxsarg>.
1282
1283=item C<SV* save_svref(SV **sptr)>
1284
d1be9408 1285Similar to C<save_scalar>, but will reinstate an C<SV *>.
d1c897a1
IZ
1286
1287=item C<void save_aptr(AV **aptr)>
1288
1289=item C<void save_hptr(HV **hptr)>
1290
1291Similar to C<save_svref>, but localize C<AV *> and C<HV *>.
1292
1293=back
1294
1295The C<Alias> module implements localization of the basic types within the
1296I<caller's scope>. People who are interested in how to localize things in
1297the containing scope should take a look there too.
1298
0a753a76 1299=head1 Subroutines
a0d0e21e 1300
68dc0745 1301=head2 XSUBs and the Argument Stack
5f05dabc
PP
1302
1303The XSUB mechanism is a simple way for Perl programs to access C subroutines.
1304An XSUB routine will have a stack that contains the arguments from the Perl
1305program, and a way to map from the Perl data structures to a C equivalent.
1306
1307The stack arguments are accessible through the C<ST(n)> macro, which returns
1308the C<n>'th stack argument. Argument 0 is the first argument passed in the
1309Perl subroutine call. These arguments are C<SV*>, and can be used anywhere
1310an C<SV*> is used.
1311
1312Most of the time, output from the C routine can be handled through use of
1313the RETVAL and OUTPUT directives. However, there are some cases where the
1314argument stack is not already long enough to handle all the return values.
1315An example is the POSIX tzname() call, which takes no arguments, but returns
1316two, the local time zone's standard and summer time abbreviations.
1317
1318To handle this situation, the PPCODE directive is used and the stack is
1319extended using the macro:
1320
924508f0 1321 EXTEND(SP, num);
5f05dabc 1322
924508f0
GS
1323where C<SP> is the macro that represents the local copy of the stack pointer,
1324and C<num> is the number of elements the stack should be extended by.
5f05dabc 1325
00aadd71 1326Now that there is room on the stack, values can be pushed on it using C<PUSHs>
484ce0c5 1327macro. The values pushed will often need to be "mortal" (See L</Reference Counts and Mortality>).
5f05dabc 1328
00aadd71
NIS
1329 PUSHs(sv_2mortal(newSViv(an_integer)))
1330 PUSHs(sv_2mortal(newSVpv("Some String",0)))
1331 PUSHs(sv_2mortal(newSVnv(3.141592)))
5f05dabc
PP
1332
1333And now the Perl program calling C<tzname>, the two values will be assigned
1334as in:
1335
1336 ($standard_abbrev, $summer_abbrev) = POSIX::tzname;
1337
1338An alternate (and possibly simpler) method to pushing values on the stack is
00aadd71 1339to use the macro:
5f05dabc 1340
5f05dabc
PP
1341 XPUSHs(SV*)
1342
00aadd71 1343This macro automatically adjust the stack for you, if needed. Thus, you
5f05dabc 1344do not need to call C<EXTEND> to extend the stack.
00aadd71
NIS
1345
1346Despite their suggestions in earlier versions of this document the macros
1347C<PUSHi>, C<PUSHn> and C<PUSHp> are I<not> suited to XSUBs which return
1348multiple results, see L</Putting a C value on Perl stack>.
5f05dabc
PP
1349
1350For more information, consult L<perlxs> and L<perlxstut>.
1351
1352=head2 Calling Perl Routines from within C Programs
a0d0e21e
LW
1353
1354There are four routines that can be used to call a Perl subroutine from
1355within a C program. These four are:
1356
954c1994
GS
1357 I32 call_sv(SV*, I32);
1358 I32 call_pv(const char*, I32);
1359 I32 call_method(const char*, I32);
1360 I32 call_argv(const char*, I32, register char**);
a0d0e21e 1361
954c1994 1362The routine most often used is C<call_sv>. The C<SV*> argument
d1b91892
AD
1363contains either the name of the Perl subroutine to be called, or a
1364reference to the subroutine. The second argument consists of flags
1365that control the context in which the subroutine is called, whether
1366or not the subroutine is being passed arguments, how errors should be
1367trapped, and how to treat return values.
a0d0e21e
LW
1368
1369All four routines return the number of arguments that the subroutine returned
1370on the Perl stack.
1371
9a68f1db 1372These routines used to be called C<perl_call_sv>, etc., before Perl v5.6.0,
954c1994
GS
1373but those names are now deprecated; macros of the same name are provided for
1374compatibility.
1375
1376When using any of these routines (except C<call_argv>), the programmer
d1b91892
AD
1377must manipulate the Perl stack. These include the following macros and
1378functions:
a0d0e21e
LW
1379
1380 dSP
924508f0 1381 SP
a0d0e21e
LW
1382 PUSHMARK()
1383 PUTBACK
1384 SPAGAIN
1385 ENTER
1386 SAVETMPS
1387 FREETMPS
1388 LEAVE
1389 XPUSH*()
cb1a09d0 1390 POP*()
a0d0e21e 1391
5f05dabc
PP
1392For a detailed description of calling conventions from C to Perl,
1393consult L<perlcall>.
a0d0e21e 1394
5f05dabc 1395=head2 Memory Allocation
a0d0e21e 1396
86058a2d
GS
1397All memory meant to be used with the Perl API functions should be manipulated
1398using the macros described in this section. The macros provide the necessary
1399transparency between differences in the actual malloc implementation that is
1400used within perl.
1401
1402It is suggested that you enable the version of malloc that is distributed
5f05dabc 1403with Perl. It keeps pools of various sizes of unallocated memory in
07fa94a1
JO
1404order to satisfy allocation requests more quickly. However, on some
1405platforms, it may cause spurious malloc or free errors.
d1b91892
AD
1406
1407 New(x, pointer, number, type);
1408 Newc(x, pointer, number, type, cast);
1409 Newz(x, pointer, number, type);
1410
07fa94a1 1411These three macros are used to initially allocate memory.
5f05dabc
PP
1412
1413The first argument C<x> was a "magic cookie" that was used to keep track
1414of who called the macro, to help when debugging memory problems. However,
07fa94a1
JO
1415the current code makes no use of this feature (most Perl developers now
1416use run-time memory checkers), so this argument can be any number.
5f05dabc
PP
1417
1418The second argument C<pointer> should be the name of a variable that will
1419point to the newly allocated memory.
d1b91892 1420
d1b91892
AD
1421The third and fourth arguments C<number> and C<type> specify how many of
1422the specified type of data structure should be allocated. The argument
1423C<type> is passed to C<sizeof>. The final argument to C<Newc>, C<cast>,
1424should be used if the C<pointer> argument is different from the C<type>
1425argument.
1426
1427Unlike the C<New> and C<Newc> macros, the C<Newz> macro calls C<memzero>
1428to zero out all the newly allocated memory.
1429
1430 Renew(pointer, number, type);
1431 Renewc(pointer, number, type, cast);
1432 Safefree(pointer)
1433
1434These three macros are used to change a memory buffer size or to free a
1435piece of memory no longer needed. The arguments to C<Renew> and C<Renewc>
1436match those of C<New> and C<Newc> with the exception of not needing the
1437"magic cookie" argument.
1438
1439 Move(source, dest, number, type);
1440 Copy(source, dest, number, type);
1441 Zero(dest, number, type);
1442
1443These three macros are used to move, copy, or zero out previously allocated
1444memory. The C<source> and C<dest> arguments point to the source and
1445destination starting points. Perl will move, copy, or zero out C<number>
1446instances of the size of the C<type> data structure (using the C<sizeof>
1447function).
a0d0e21e 1448
5f05dabc 1449=head2 PerlIO
ce3d39e2 1450
5f05dabc
PP
1451The most recent development releases of Perl has been experimenting with
1452removing Perl's dependency on the "normal" standard I/O suite and allowing
1453other stdio implementations to be used. This involves creating a new
1454abstraction layer that then calls whichever implementation of stdio Perl
68dc0745 1455was compiled with. All XSUBs should now use the functions in the PerlIO
5f05dabc
PP
1456abstraction layer and not make any assumptions about what kind of stdio
1457is being used.
1458
1459For a complete description of the PerlIO abstraction, consult L<perlapio>.
1460
8ebc5c01 1461=head2 Putting a C value on Perl stack
ce3d39e2
IZ
1462
1463A lot of opcodes (this is an elementary operation in the internal perl
1464stack machine) put an SV* on the stack. However, as an optimization
1465the corresponding SV is (usually) not recreated each time. The opcodes
1466reuse specially assigned SVs (I<target>s) which are (as a corollary)
1467not constantly freed/created.
1468
0a753a76 1469Each of the targets is created only once (but see
ce3d39e2
IZ
1470L<Scratchpads and recursion> below), and when an opcode needs to put
1471an integer, a double, or a string on stack, it just sets the
1472corresponding parts of its I<target> and puts the I<target> on stack.
1473
1474The macro to put this target on stack is C<PUSHTARG>, and it is
1475directly used in some opcodes, as well as indirectly in zillions of
1476others, which use it via C<(X)PUSH[pni]>.
1477
1bd1c0d5
SC
1478Because the target is reused, you must be careful when pushing multiple
1479values on the stack. The following code will not do what you think:
1480
1481 XPUSHi(10);
1482 XPUSHi(20);
1483
1484This translates as "set C<TARG> to 10, push a pointer to C<TARG> onto
1485the stack; set C<TARG> to 20, push a pointer to C<TARG> onto the stack".
1486At the end of the operation, the stack does not contain the values 10
1487and 20, but actually contains two pointers to C<TARG>, which we have set
1488to 20. If you need to push multiple different values, use C<XPUSHs>,
1489which bypasses C<TARG>.
1490
1491On a related note, if you do use C<(X)PUSH[npi]>, then you're going to
1492need a C<dTARG> in your variable declarations so that the C<*PUSH*>
00aadd71 1493macros can make use of the local variable C<TARG>.
1bd1c0d5 1494
8ebc5c01 1495=head2 Scratchpads
ce3d39e2 1496
54310121 1497The question remains on when the SVs which are I<target>s for opcodes
5f05dabc
PP
1498are created. The answer is that they are created when the current unit --
1499a subroutine or a file (for opcodes for statements outside of
1500subroutines) -- is compiled. During this time a special anonymous Perl
ce3d39e2
IZ
1501array is created, which is called a scratchpad for the current
1502unit.
1503
54310121 1504A scratchpad keeps SVs which are lexicals for the current unit and are
ce3d39e2
IZ
1505targets for opcodes. One can deduce that an SV lives on a scratchpad
1506by looking on its flags: lexicals have C<SVs_PADMY> set, and
1507I<target>s have C<SVs_PADTMP> set.
1508
54310121
PP
1509The correspondence between OPs and I<target>s is not 1-to-1. Different
1510OPs in the compile tree of the unit can use the same target, if this
ce3d39e2
IZ
1511would not conflict with the expected life of the temporary.
1512
2ae324a7 1513=head2 Scratchpads and recursion
ce3d39e2
IZ
1514
1515In fact it is not 100% true that a compiled unit contains a pointer to
1516the scratchpad AV. In fact it contains a pointer to an AV of
1517(initially) one element, and this element is the scratchpad AV. Why do
1518we need an extra level of indirection?
1519
9a68f1db 1520The answer is B<recursion>, and maybe B<threads>. Both
ce3d39e2
IZ
1521these can create several execution pointers going into the same
1522subroutine. For the subroutine-child not write over the temporaries
1523for the subroutine-parent (lifespan of which covers the call to the
1524child), the parent and the child should have different
1525scratchpads. (I<And> the lexicals should be separate anyway!)
1526
5f05dabc
PP
1527So each subroutine is born with an array of scratchpads (of length 1).
1528On each entry to the subroutine it is checked that the current
ce3d39e2
IZ
1529depth of the recursion is not more than the length of this array, and
1530if it is, new scratchpad is created and pushed into the array.
1531
1532The I<target>s on this scratchpad are C<undef>s, but they are already
1533marked with correct flags.
1534
0a753a76
PP
1535=head1 Compiled code
1536
1537=head2 Code tree
1538
1539Here we describe the internal form your code is converted to by
1540Perl. Start with a simple example:
1541
1542 $a = $b + $c;
1543
1544This is converted to a tree similar to this one:
1545
1546 assign-to
1547 / \
1548 + $a
1549 / \
1550 $b $c
1551
7b8d334a 1552(but slightly more complicated). This tree reflects the way Perl
0a753a76
PP
1553parsed your code, but has nothing to do with the execution order.
1554There is an additional "thread" going through the nodes of the tree
1555which shows the order of execution of the nodes. In our simplified
1556example above it looks like:
1557
1558 $b ---> $c ---> + ---> $a ---> assign-to
1559
1560But with the actual compile tree for C<$a = $b + $c> it is different:
1561some nodes I<optimized away>. As a corollary, though the actual tree
1562contains more nodes than our simplified example, the execution order
1563is the same as in our example.
1564
1565=head2 Examining the tree
1566
1567If you have your perl compiled for debugging (usually done with C<-D
1568optimize=-g> on C<Configure> command line), you may examine the
1569compiled tree by specifying C<-Dx> on the Perl command line. The
1570output takes several lines per node, and for C<$b+$c> it looks like
1571this:
1572
1573 5 TYPE = add ===> 6
1574 TARG = 1
1575 FLAGS = (SCALAR,KIDS)
1576 {
1577 TYPE = null ===> (4)
1578 (was rv2sv)
1579 FLAGS = (SCALAR,KIDS)
1580 {
1581 3 TYPE = gvsv ===> 4
1582 FLAGS = (SCALAR)
1583 GV = main::b
1584 }
1585 }
1586 {
1587 TYPE = null ===> (5)
1588 (was rv2sv)
1589 FLAGS = (SCALAR,KIDS)
1590 {
1591 4 TYPE = gvsv ===> 5
1592 FLAGS = (SCALAR)
1593 GV = main::c
1594 }
1595 }
1596
1597This tree has 5 nodes (one per C<TYPE> specifier), only 3 of them are
1598not optimized away (one per number in the left column). The immediate
1599children of the given node correspond to C<{}> pairs on the same level
1600of indentation, thus this listing corresponds to the tree:
1601
1602 add
1603 / \
1604 null null
1605 | |
1606 gvsv gvsv
1607
1608The execution order is indicated by C<===E<gt>> marks, thus it is C<3
16094 5 6> (node C<6> is not included into above listing), i.e.,
1610C<gvsv gvsv add whatever>.
1611
9afa14e3
SC
1612Each of these nodes represents an op, a fundamental operation inside the
1613Perl core. The code which implements each operation can be found in the
1614F<pp*.c> files; the function which implements the op with type C<gvsv>
1615is C<pp_gvsv>, and so on. As the tree above shows, different ops have
1616different numbers of children: C<add> is a binary operator, as one would
1617expect, and so has two children. To accommodate the various different
1618numbers of children, there are various types of op data structure, and
1619they link together in different ways.
1620
1621The simplest type of op structure is C<OP>: this has no children. Unary
1622operators, C<UNOP>s, have one child, and this is pointed to by the
1623C<op_first> field. Binary operators (C<BINOP>s) have not only an
1624C<op_first> field but also an C<op_last> field. The most complex type of
1625op is a C<LISTOP>, which has any number of children. In this case, the
1626first child is pointed to by C<op_first> and the last child by
1627C<op_last>. The children in between can be found by iteratively
1628following the C<op_sibling> pointer from the first child to the last.
1629
1630There are also two other op types: a C<PMOP> holds a regular expression,
1631and has no children, and a C<LOOP> may or may not have children. If the
1632C<op_children> field is non-zero, it behaves like a C<LISTOP>. To
1633complicate matters, if a C<UNOP> is actually a C<null> op after
1634optimization (see L</Compile pass 2: context propagation>) it will still
1635have children in accordance with its former type.
1636
0a753a76
PP
1637=head2 Compile pass 1: check routines
1638
8870b5c7
GS
1639The tree is created by the compiler while I<yacc> code feeds it
1640the constructions it recognizes. Since I<yacc> works bottom-up, so does
0a753a76
PP
1641the first pass of perl compilation.
1642
1643What makes this pass interesting for perl developers is that some
1644optimization may be performed on this pass. This is optimization by
8870b5c7 1645so-called "check routines". The correspondence between node names
0a753a76
PP
1646and corresponding check routines is described in F<opcode.pl> (do not
1647forget to run C<make regen_headers> if you modify this file).
1648
1649A check routine is called when the node is fully constructed except
7b8d334a 1650for the execution-order thread. Since at this time there are no
0a753a76
PP
1651back-links to the currently constructed node, one can do most any
1652operation to the top-level node, including freeing it and/or creating
1653new nodes above/below it.
1654
1655The check routine returns the node which should be inserted into the
1656tree (if the top-level node was not modified, check routine returns
1657its argument).
1658
1659By convention, check routines have names C<ck_*>. They are usually
1660called from C<new*OP> subroutines (or C<convert>) (which in turn are
1661called from F<perly.y>).
1662
1663=head2 Compile pass 1a: constant folding
1664
1665Immediately after the check routine is called the returned node is
1666checked for being compile-time executable. If it is (the value is
1667judged to be constant) it is immediately executed, and a I<constant>
1668node with the "return value" of the corresponding subtree is
1669substituted instead. The subtree is deleted.
1670
1671If constant folding was not performed, the execution-order thread is
1672created.
1673
1674=head2 Compile pass 2: context propagation
1675
1676When a context for a part of compile tree is known, it is propagated
a3cb178b 1677down through the tree. At this time the context can have 5 values
0a753a76
PP
1678(instead of 2 for runtime context): void, boolean, scalar, list, and
1679lvalue. In contrast with the pass 1 this pass is processed from top
1680to bottom: a node's context determines the context for its children.
1681
1682Additional context-dependent optimizations are performed at this time.
1683Since at this moment the compile tree contains back-references (via
1684"thread" pointers), nodes cannot be free()d now. To allow
1685optimized-away nodes at this stage, such nodes are null()ified instead
1686of free()ing (i.e. their type is changed to OP_NULL).
1687
1688=head2 Compile pass 3: peephole optimization
1689
1690After the compile tree for a subroutine (or for an C<eval> or a file)
1691is created, an additional pass over the code is performed. This pass
1692is neither top-down or bottom-up, but in the execution order (with
7b8d334a 1693additional complications for conditionals). These optimizations are
0a753a76
PP
1694done in the subroutine peep(). Optimizations performed at this stage
1695are subject to the same restrictions as in the pass 2.
1696
1ba7f851
PJ
1697=head2 Pluggable runops
1698
1699The compile tree is executed in a runops function. There are two runops
1700functions in F<run.c>. C<Perl_runops_debug> is used with DEBUGGING and
1701C<Perl_runops_standard> is used otherwise. For fine control over the
1702execution of the compile tree it is possible to provide your own runops
1703function.
1704
1705It's probably best to copy one of the existing runops functions and
1706change it to suit your needs. Then, in the BOOT section of your XS
1707file, add the line:
1708
1709 PL_runops = my_runops;
1710
1711This function should be as efficient as possible to keep your programs
1712running as fast as possible.
1713
9afa14e3
SC
1714=head1 Examining internal data structures with the C<dump> functions
1715
1716To aid debugging, the source file F<dump.c> contains a number of
1717functions which produce formatted output of internal data structures.
1718
1719The most commonly used of these functions is C<Perl_sv_dump>; it's used
1720for dumping SVs, AVs, HVs, and CVs. The C<Devel::Peek> module calls
1721C<sv_dump> to produce debugging output from Perl-space, so users of that
00aadd71 1722module should already be familiar with its format.
9afa14e3
SC
1723
1724C<Perl_op_dump> can be used to dump an C<OP> structure or any of its
210b36aa 1725derivatives, and produces output similar to C<perl -Dx>; in fact,
9afa14e3
SC
1726C<Perl_dump_eval> will dump the main root of the code being evaluated,
1727exactly like C<-Dx>.
1728
1729Other useful functions are C<Perl_dump_sub>, which turns a C<GV> into an
1730op tree, C<Perl_dump_packsubs> which calls C<Perl_dump_sub> on all the
1731subroutines in a package like so: (Thankfully, these are all xsubs, so
1732there is no op tree)
1733
1734 (gdb) print Perl_dump_packsubs(PL_defstash)
1735
1736 SUB attributes::bootstrap = (xsub 0x811fedc 0)
1737
1738 SUB UNIVERSAL::can = (xsub 0x811f50c 0)
1739
1740 SUB UNIVERSAL::isa = (xsub 0x811f304 0)
1741
1742 SUB UNIVERSAL::VERSION = (xsub 0x811f7ac 0)
1743
1744 SUB DynaLoader::boot_DynaLoader = (xsub 0x805b188 0)
1745
1746and C<Perl_dump_all>, which dumps all the subroutines in the stash and
1747the op tree of the main root.
1748
954c1994 1749=head1 How multiple interpreters and concurrency are supported
ee072b34 1750
ee072b34
GS
1751=head2 Background and PERL_IMPLICIT_CONTEXT
1752
1753The Perl interpreter can be regarded as a closed box: it has an API
1754for feeding it code or otherwise making it do things, but it also has
1755functions for its own use. This smells a lot like an object, and
1756there are ways for you to build Perl so that you can have multiple
acfe0abc
GS
1757interpreters, with one interpreter represented either as a C structure,
1758or inside a thread-specific structure. These structures contain all
1759the context, the state of that interpreter.
1760
9a68f1db 1761Two macros control the major Perl build flavors: MULTIPLICITY and
acfe0abc
GS
1762USE_5005THREADS. The MULTIPLICITY build has a C structure
1763that packages all the interpreter state, and there is a similar thread-specific
1764data structure under USE_5005THREADS. In both cases,
54aff467
GS
1765PERL_IMPLICIT_CONTEXT is also normally defined, and enables the
1766support for passing in a "hidden" first argument that represents all three
651a3225 1767data structures.
54aff467
GS
1768
1769All this obviously requires a way for the Perl internal functions to be
acfe0abc 1770either subroutines taking some kind of structure as the first
ee072b34 1771argument, or subroutines taking nothing as the first argument. To
acfe0abc 1772enable these two very different ways of building the interpreter,
ee072b34
GS
1773the Perl source (as it does in so many other situations) makes heavy
1774use of macros and subroutine naming conventions.
1775
54aff467 1776First problem: deciding which functions will be public API functions and
00aadd71 1777which will be private. All functions whose names begin C<S_> are private
954c1994
GS
1778(think "S" for "secret" or "static"). All other functions begin with
1779"Perl_", but just because a function begins with "Perl_" does not mean it is
00aadd71
NIS
1780part of the API. (See L</Internal Functions>.) The easiest way to be B<sure> a
1781function is part of the API is to find its entry in L<perlapi>.
1782If it exists in L<perlapi>, it's part of the API. If it doesn't, and you
1783think it should be (i.e., you need it for your extension), send mail via
a422fd2d 1784L<perlbug> explaining why you think it should be.
ee072b34
GS
1785
1786Second problem: there must be a syntax so that the same subroutine
1787declarations and calls can pass a structure as their first argument,
1788or pass nothing. To solve this, the subroutines are named and
1789declared in a particular way. Here's a typical start of a static
1790function used within the Perl guts:
1791
1792 STATIC void
1793 S_incline(pTHX_ char *s)
1794
acfe0abc
GS
1795STATIC becomes "static" in C, and may be #define'd to nothing in some
1796configurations in future.
ee072b34 1797
651a3225
GS
1798A public function (i.e. part of the internal API, but not necessarily
1799sanctioned for use in extensions) begins like this:
ee072b34
GS
1800
1801 void
2307c6d0 1802 Perl_sv_setiv(pTHX_ SV* dsv, IV num)
ee072b34
GS
1803
1804C<pTHX_> is one of a number of macros (in perl.h) that hide the
1805details of the interpreter's context. THX stands for "thread", "this",
1806or "thingy", as the case may be. (And no, George Lucas is not involved. :-)
1807The first character could be 'p' for a B<p>rototype, 'a' for B<a>rgument,
a7486cbb
JH
1808or 'd' for B<d>eclaration, so we have C<pTHX>, C<aTHX> and C<dTHX>, and
1809their variants.
ee072b34 1810
a7486cbb
JH
1811When Perl is built without options that set PERL_IMPLICIT_CONTEXT, there is no
1812first argument containing the interpreter's context. The trailing underscore
ee072b34
GS
1813in the pTHX_ macro indicates that the macro expansion needs a comma
1814after the context argument because other arguments follow it. If
1815PERL_IMPLICIT_CONTEXT is not defined, pTHX_ will be ignored, and the
54aff467
GS
1816subroutine is not prototyped to take the extra argument. The form of the
1817macro without the trailing underscore is used when there are no additional
ee072b34
GS
1818explicit arguments.
1819
54aff467 1820When a core function calls another, it must pass the context. This
2307c6d0 1821is normally hidden via macros. Consider C<sv_setiv>. It expands into
ee072b34
GS
1822something like this:
1823
2307c6d0
SB
1824 #ifdef PERL_IMPLICIT_CONTEXT
1825 #define sv_setiv(a,b) Perl_sv_setiv(aTHX_ a, b)
ee072b34 1826 /* can't do this for vararg functions, see below */
2307c6d0
SB
1827 #else
1828 #define sv_setiv Perl_sv_setiv
1829 #endif
ee072b34
GS
1830
1831This works well, and means that XS authors can gleefully write:
1832
2307c6d0 1833 sv_setiv(foo, bar);
ee072b34
GS
1834
1835and still have it work under all the modes Perl could have been
1836compiled with.
1837
ee072b34
GS
1838This doesn't work so cleanly for varargs functions, though, as macros
1839imply that the number of arguments is known in advance. Instead we
1840either need to spell them out fully, passing C<aTHX_> as the first
1841argument (the Perl core tends to do this with functions like
1842Perl_warner), or use a context-free version.
1843
1844The context-free version of Perl_warner is called
1845Perl_warner_nocontext, and does not take the extra argument. Instead
1846it does dTHX; to get the context from thread-local storage. We
1847C<#define warner Perl_warner_nocontext> so that extensions get source
1848compatibility at the expense of performance. (Passing an arg is
1849cheaper than grabbing it from thread-local storage.)
1850
acfe0abc 1851You can ignore [pad]THXx when browsing the Perl headers/sources.
ee072b34
GS
1852Those are strictly for use within the core. Extensions and embedders
1853need only be aware of [pad]THX.
1854
a7486cbb
JH
1855=head2 So what happened to dTHR?
1856
1857C<dTHR> was introduced in perl 5.005 to support the older thread model.
1858The older thread model now uses the C<THX> mechanism to pass context
1859pointers around, so C<dTHR> is not useful any more. Perl 5.6.0 and
1860later still have it for backward source compatibility, but it is defined
1861to be a no-op.
1862
ee072b34
GS
1863=head2 How do I use all this in extensions?
1864
1865When Perl is built with PERL_IMPLICIT_CONTEXT, extensions that call
1866any functions in the Perl API will need to pass the initial context
1867argument somehow. The kicker is that you will need to write it in
1868such a way that the extension still compiles when Perl hasn't been
1869built with PERL_IMPLICIT_CONTEXT enabled.
1870
1871There are three ways to do this. First, the easy but inefficient way,
1872which is also the default, in order to maintain source compatibility
1873with extensions: whenever XSUB.h is #included, it redefines the aTHX
1874and aTHX_ macros to call a function that will return the context.
1875Thus, something like:
1876
2307c6d0 1877 sv_setiv(sv, num);
ee072b34 1878
4375e838 1879in your extension will translate to this when PERL_IMPLICIT_CONTEXT is
54aff467 1880in effect:
ee072b34 1881
2307c6d0 1882 Perl_sv_setiv(Perl_get_context(), sv, num);
ee072b34 1883
54aff467 1884or to this otherwise:
ee072b34 1885
2307c6d0 1886 Perl_sv_setiv(sv, num);
ee072b34
GS
1887
1888You have to do nothing new in your extension to get this; since
2fa86c13 1889the Perl library provides Perl_get_context(), it will all just
ee072b34
GS
1890work.
1891
1892The second, more efficient way is to use the following template for
1893your Foo.xs:
1894
c52f9dcd
JH
1895 #define PERL_NO_GET_CONTEXT /* we want efficiency */
1896 #include "EXTERN.h"
1897 #include "perl.h"
1898 #include "XSUB.h"
ee072b34
GS
1899
1900 static my_private_function(int arg1, int arg2);
1901
c52f9dcd
JH
1902 static SV *
1903 my_private_function(int arg1, int arg2)
1904 {
1905 dTHX; /* fetch context */
1906 ... call many Perl API functions ...
1907 }
ee072b34
GS
1908
1909 [... etc ...]
1910
c52f9dcd 1911 MODULE = Foo PACKAGE = Foo
ee072b34 1912
c52f9dcd 1913 /* typical XSUB */
ee072b34 1914
c52f9dcd
JH
1915 void
1916 my_xsub(arg)
1917 int arg
1918 CODE:
1919 my_private_function(arg, 10);
ee072b34
GS
1920
1921Note that the only two changes from the normal way of writing an
1922extension is the addition of a C<#define PERL_NO_GET_CONTEXT> before
1923including the Perl headers, followed by a C<dTHX;> declaration at
1924the start of every function that will call the Perl API. (You'll
1925know which functions need this, because the C compiler will complain
1926that there's an undeclared identifier in those functions.) No changes
1927are needed for the XSUBs themselves, because the XS() macro is
1928correctly defined to pass in the implicit context if needed.
1929
1930The third, even more efficient way is to ape how it is done within
1931the Perl guts:
1932
1933
c52f9dcd
JH
1934 #define PERL_NO_GET_CONTEXT /* we want efficiency */
1935 #include "EXTERN.h"
1936 #include "perl.h"
1937 #include "XSUB.h"
ee072b34
GS
1938
1939 /* pTHX_ only needed for functions that call Perl API */
1940 static my_private_function(pTHX_ int arg1, int arg2);
1941
c52f9dcd
JH
1942 static SV *
1943 my_private_function(pTHX_ int arg1, int arg2)
1944 {
1945 /* dTHX; not needed here, because THX is an argument */
1946 ... call Perl API functions ...
1947 }
ee072b34
GS
1948
1949 [... etc ...]
1950
c52f9dcd 1951 MODULE = Foo PACKAGE = Foo
ee072b34 1952
c52f9dcd 1953 /* typical XSUB */
ee072b34 1954
c52f9dcd
JH
1955 void
1956 my_xsub(arg)
1957 int arg
1958 CODE:
1959 my_private_function(aTHX_ arg, 10);
ee072b34
GS
1960
1961This implementation never has to fetch the context using a function
1962call, since it is always passed as an extra argument. Depending on
1963your needs for simplicity or efficiency, you may mix the previous
1964two approaches freely.
1965
651a3225
GS
1966Never add a comma after C<pTHX> yourself--always use the form of the
1967macro with the underscore for functions that take explicit arguments,
1968or the form without the argument for functions with no explicit arguments.
ee072b34 1969
a7486cbb
JH
1970=head2 Should I do anything special if I call perl from multiple threads?
1971
1972If you create interpreters in one thread and then proceed to call them in
1973another, you need to make sure perl's own Thread Local Storage (TLS) slot is
1974initialized correctly in each of those threads.
1975
1976The C<perl_alloc> and C<perl_clone> API functions will automatically set
1977the TLS slot to the interpreter they created, so that there is no need to do
1978anything special if the interpreter is always accessed in the same thread that
1979created it, and that thread did not create or call any other interpreters
1980afterwards. If that is not the case, you have to set the TLS slot of the
1981thread before calling any functions in the Perl API on that particular
1982interpreter. This is done by calling the C<PERL_SET_CONTEXT> macro in that
1983thread as the first thing you do:
1984
1985 /* do this before doing anything else with some_perl */
1986 PERL_SET_CONTEXT(some_perl);
1987
1988 ... other Perl API calls on some_perl go here ...
1989
ee072b34
GS
1990=head2 Future Plans and PERL_IMPLICIT_SYS
1991
1992Just as PERL_IMPLICIT_CONTEXT provides a way to bundle up everything
1993that the interpreter knows about itself and pass it around, so too are
1994there plans to allow the interpreter to bundle up everything it knows
1995about the environment it's running on. This is enabled with the
acfe0abc 1996PERL_IMPLICIT_SYS macro. Currently it only works with USE_ITHREADS
4d1ff10f 1997and USE_5005THREADS on Windows (see inside iperlsys.h).
ee072b34
GS
1998
1999This allows the ability to provide an extra pointer (called the "host"
2000environment) for all the system calls. This makes it possible for
2001all the system stuff to maintain their own state, broken down into
2002seven C structures. These are thin wrappers around the usual system
2003calls (see win32/perllib.c) for the default perl executable, but for a
2004more ambitious host (like the one that would do fork() emulation) all
2005the extra work needed to pretend that different interpreters are
2006actually different "processes", would be done here.
2007
2008The Perl engine/interpreter and the host are orthogonal entities.
2009There could be one or more interpreters in a process, and one or
2010more "hosts", with free association between them.
2011
a422fd2d
SC
2012=head1 Internal Functions
2013
2014All of Perl's internal functions which will be exposed to the outside
2015world are be prefixed by C<Perl_> so that they will not conflict with XS
2016functions or functions used in a program in which Perl is embedded.
2017Similarly, all global variables begin with C<PL_>. (By convention,
2018static functions start with C<S_>)
2019
2020Inside the Perl core, you can get at the functions either with or
2021without the C<Perl_> prefix, thanks to a bunch of defines that live in
2022F<embed.h>. This header file is generated automatically from
2023F<embed.pl>. F<embed.pl> also creates the prototyping header files for
2024the internal functions, generates the documentation and a lot of other
2025bits and pieces. It's important that when you add a new function to the
2026core or change an existing one, you change the data in the table at the
2027end of F<embed.pl> as well. Here's a sample entry from that table:
2028
2029 Apd |SV** |av_fetch |AV* ar|I32 key|I32 lval
2030
2031The second column is the return type, the third column the name. Columns
2032after that are the arguments. The first column is a set of flags:
2033
2034=over 3
2035
2036=item A
2037
2038This function is a part of the public API.
2039
2040=item p
2041
2042This function has a C<Perl_> prefix; ie, it is defined as C<Perl_av_fetch>
2043
2044=item d
2045
2046This function has documentation using the C<apidoc> feature which we'll
2047look at in a second.
2048
2049=back
2050
2051Other available flags are:
2052
2053=over 3
2054
2055=item s
2056
a7486cbb
JH
2057This is a static function and is defined as C<S_whatever>, and usually
2058called within the sources as C<whatever(...)>.
a422fd2d
SC
2059
2060=item n
2061
2062This does not use C<aTHX_> and C<pTHX> to pass interpreter context. (See
2063L<perlguts/Background and PERL_IMPLICIT_CONTEXT>.)
2064
2065=item r
2066
2067This function never returns; C<croak>, C<exit> and friends.
2068
2069=item f
2070
2071This function takes a variable number of arguments, C<printf> style.
2072The argument list should end with C<...>, like this:
2073
2074 Afprd |void |croak |const char* pat|...
2075
a7486cbb 2076=item M
a422fd2d 2077
00aadd71 2078This function is part of the experimental development API, and may change
a422fd2d
SC
2079or disappear without notice.
2080
2081=item o
2082
2083This function should not have a compatibility macro to define, say,
2084C<Perl_parse> to C<parse>. It must be called as C<Perl_parse>.
2085
2086=item j
2087
2088This function is not a member of C<CPerlObj>. If you don't know
2089what this means, don't use it.
2090
2091=item x
2092
2093This function isn't exported out of the Perl core.
2094
2095=back
2096
2097If you edit F<embed.pl>, you will need to run C<make regen_headers> to
2098force a rebuild of F<embed.h> and other auto-generated files.
2099
6b4667fc 2100=head2 Formatted Printing of IVs, UVs, and NVs
9dd9db0b 2101
6b4667fc
A
2102If you are printing IVs, UVs, or NVS instead of the stdio(3) style
2103formatting codes like C<%d>, C<%ld>, C<%f>, you should use the
2104following macros for portability
9dd9db0b 2105
c52f9dcd
JH
2106 IVdf IV in decimal
2107 UVuf UV in decimal
2108 UVof UV in octal
2109 UVxf UV in hexadecimal
2110 NVef NV %e-like
2111 NVff NV %f-like
2112 NVgf NV %g-like
9dd9db0b 2113
6b4667fc
A
2114These will take care of 64-bit integers and long doubles.
2115For example:
2116
c52f9dcd 2117 printf("IV is %"IVdf"\n", iv);
6b4667fc
A
2118
2119The IVdf will expand to whatever is the correct format for the IVs.
9dd9db0b 2120
8908e76d
JH
2121If you are printing addresses of pointers, use UVxf combined
2122with PTR2UV(), do not use %lx or %p.
2123
2124=head2 Pointer-To-Integer and Integer-To-Pointer
2125
2126Because pointer size does not necessarily equal integer size,
2127use the follow macros to do it right.
2128
c52f9dcd
JH
2129 PTR2UV(pointer)
2130 PTR2IV(pointer)
2131 PTR2NV(pointer)
2132 INT2PTR(pointertotype, integer)
8908e76d
JH
2133
2134For example:
2135
c52f9dcd
JH
2136 IV iv = ...;
2137 SV *sv = INT2PTR(SV*, iv);
8908e76d
JH
2138
2139and
2140
c52f9dcd
JH
2141 AV *av = ...;
2142 UV uv = PTR2UV(av);
8908e76d 2143
a422fd2d
SC
2144=head2 Source Documentation
2145
2146There's an effort going on to document the internal functions and
2147automatically produce reference manuals from them - L<perlapi> is one
2148such manual which details all the functions which are available to XS
2149writers. L<perlintern> is the autogenerated manual for the functions
2150which are not part of the API and are supposedly for internal use only.
2151
2152Source documentation is created by putting POD comments into the C
2153source, like this:
2154
2155 /*
2156 =for apidoc sv_setiv
2157
2158 Copies an integer into the given SV. Does not handle 'set' magic. See
2159 C<sv_setiv_mg>.
2160
2161 =cut
2162 */
2163
2164Please try and supply some documentation if you add functions to the
2165Perl core.
2166
2167=head1 Unicode Support
2168
2169Perl 5.6.0 introduced Unicode support. It's important for porters and XS
2170writers to understand this support and make sure that the code they
2171write does not corrupt Unicode data.
2172
2173=head2 What B<is> Unicode, anyway?
2174
2175In the olden, less enlightened times, we all used to use ASCII. Most of
2176us did, anyway. The big problem with ASCII is that it's American. Well,
2177no, that's not actually the problem; the problem is that it's not
2178particularly useful for people who don't use the Roman alphabet. What
2179used to happen was that particular languages would stick their own
2180alphabet in the upper range of the sequence, between 128 and 255. Of
2181course, we then ended up with plenty of variants that weren't quite
2182ASCII, and the whole point of it being a standard was lost.
2183
2184Worse still, if you've got a language like Chinese or
2185Japanese that has hundreds or thousands of characters, then you really
2186can't fit them into a mere 256, so they had to forget about ASCII
2187altogether, and build their own systems using pairs of numbers to refer
2188to one character.
2189
2190To fix this, some people formed Unicode, Inc. and
2191produced a new character set containing all the characters you can
2192possibly think of and more. There are several ways of representing these
2193characters, and the one Perl uses is called UTF8. UTF8 uses
2194a variable number of bytes to represent a character, instead of just
b3b6085d 2195one. You can learn more about Unicode at http://www.unicode.org/
a422fd2d
SC
2196
2197=head2 How can I recognise a UTF8 string?
2198
2199You can't. This is because UTF8 data is stored in bytes just like
2200non-UTF8 data. The Unicode character 200, (C<0xC8> for you hex types)
2201capital E with a grave accent, is represented by the two bytes
2202C<v196.172>. Unfortunately, the non-Unicode string C<chr(196).chr(172)>
2203has that byte sequence as well. So you can't tell just by looking - this
2204is what makes Unicode input an interesting problem.
2205
2206The API function C<is_utf8_string> can help; it'll tell you if a string
2207contains only valid UTF8 characters. However, it can't do the work for
2208you. On a character-by-character basis, C<is_utf8_char> will tell you
2209whether the current character in a string is valid UTF8.
2210
2211=head2 How does UTF8 represent Unicode characters?
2212
2213As mentioned above, UTF8 uses a variable number of bytes to store a
2214character. Characters with values 1...128 are stored in one byte, just
2215like good ol' ASCII. Character 129 is stored as C<v194.129>; this
a31a806a 2216continues up to character 191, which is C<v194.191>. Now we've run out of
a422fd2d
SC
2217bits (191 is binary C<10111111>) so we move on; 192 is C<v195.128>. And
2218so it goes on, moving to three bytes at character 2048.
2219
2220Assuming you know you're dealing with a UTF8 string, you can find out
2221how long the first character in it is with the C<UTF8SKIP> macro:
2222
2223 char *utf = "\305\233\340\240\201";
2224 I32 len;
2225
2226 len = UTF8SKIP(utf); /* len is 2 here */
2227 utf += len;
2228 len = UTF8SKIP(utf); /* len is 3 here */
2229
2230Another way to skip over characters in a UTF8 string is to use
2231C<utf8_hop>, which takes a string and a number of characters to skip
2232over. You're on your own about bounds checking, though, so don't use it
2233lightly.
2234
3a2263fe
RGS
2235All bytes in a multi-byte UTF8 character will have the high bit set,
2236so you can test if you need to do something special with this
2237character like this (the UTF8_IS_INVARIANT() is a macro that tests
2238whether the byte can be encoded as a single byte even in UTF-8):
a422fd2d 2239
3a2263fe
RGS
2240 U8 *utf;
2241 UV uv; /* Note: a UV, not a U8, not a char */
a422fd2d 2242
3a2263fe 2243 if (!UTF8_IS_INVARIANT(*utf))
a422fd2d
SC
2244 /* Must treat this as UTF8 */
2245 uv = utf8_to_uv(utf);
2246 else
2247 /* OK to treat this character as a byte */
2248 uv = *utf;
2249
2250You can also see in that example that we use C<utf8_to_uv> to get the
2251value of the character; the inverse function C<uv_to_utf8> is available
2252for putting a UV into UTF8:
2253
3a2263fe 2254 if (!UTF8_IS_INVARIANT(uv))
a422fd2d
SC
2255 /* Must treat this as UTF8 */
2256 utf8 = uv_to_utf8(utf8, uv);
2257 else
2258 /* OK to treat this character as a byte */
2259 *utf8++ = uv;
2260
2261You B<must> convert characters to UVs using the above functions if
2262you're ever in a situation where you have to match UTF8 and non-UTF8
2263characters. You may not skip over UTF8 characters in this case. If you
2264do this, you'll lose the ability to match hi-bit non-UTF8 characters;
2265for instance, if your UTF8 string contains C<v196.172>, and you skip
2266that character, you can never match a C<chr(200)> in a non-UTF8 string.
2267So don't do that!
2268
2269=head2 How does Perl store UTF8 strings?
2270
2271Currently, Perl deals with Unicode strings and non-Unicode strings
2272slightly differently. If a string has been identified as being UTF-8
2273encoded, Perl will set a flag in the SV, C<SVf_UTF8>. You can check and
2274manipulate this flag with the following macros:
2275
2276 SvUTF8(sv)
2277 SvUTF8_on(sv)
2278 SvUTF8_off(sv)
2279
2280This flag has an important effect on Perl's treatment of the string: if
2281Unicode data is not properly distinguished, regular expressions,
2282C<length>, C<substr> and other string handling operations will have
2283undesirable results.
2284
2285The problem comes when you have, for instance, a string that isn't
2286flagged is UTF8, and contains a byte sequence that could be UTF8 -
2287especially when combining non-UTF8 and UTF8 strings.
2288
2289Never forget that the C<SVf_UTF8> flag is separate to the PV value; you
2290need be sure you don't accidentally knock it off while you're
2291manipulating SVs. More specifically, you cannot expect to do this:
2292
2293 SV *sv;
2294 SV *nsv;
2295 STRLEN len;
2296 char *p;
2297
2298 p = SvPV(sv, len);
2299 frobnicate(p);
2300 nsv = newSVpvn(p, len);
2301
2302The C<char*> string does not tell you the whole story, and you can't
2303copy or reconstruct an SV just by copying the string value. Check if the
2304old SV has the UTF8 flag set, and act accordingly:
2305
2306 p = SvPV(sv, len);
2307 frobnicate(p);
2308 nsv = newSVpvn(p, len);
2309 if (SvUTF8(sv))
2310 SvUTF8_on(nsv);
2311
2312In fact, your C<frobnicate> function should be made aware of whether or
2313not it's dealing with UTF8 data, so that it can handle the string
2314appropriately.
2315
3a2263fe
RGS
2316Since just passing an SV to an XS function and copying the data of
2317the SV is not enough to copy the UTF8 flags, even less right is just
2318passing a C<char *> to an XS function.
2319
a422fd2d
SC
2320=head2 How do I convert a string to UTF8?
2321
2322If you're mixing UTF8 and non-UTF8 strings, you might find it necessary
2323to upgrade one of the strings to UTF8. If you've got an SV, the easiest
2324way to do this is:
2325
2326 sv_utf8_upgrade(sv);
2327
2328However, you must not do this, for example:
2329
2330 if (!SvUTF8(left))
2331 sv_utf8_upgrade(left);
2332
2333If you do this in a binary operator, you will actually change one of the
b1866b2d 2334strings that came into the operator, and, while it shouldn't be noticeable
a422fd2d
SC
2335by the end user, it can cause problems.
2336
2337Instead, C<bytes_to_utf8> will give you a UTF8-encoded B<copy> of its
2338string argument. This is useful for having the data available for
b1866b2d 2339comparisons and so on, without harming the original SV. There's also
a422fd2d
SC
2340C<utf8_to_bytes> to go the other way, but naturally, this will fail if
2341the string contains any characters above 255 that can't be represented
2342in a single byte.
2343
2344=head2 Is there anything else I need to know?
2345
2346Not really. Just remember these things:
2347
2348=over 3
2349
2350=item *
2351
2352There's no way to tell if a string is UTF8 or not. You can tell if an SV
2353is UTF8 by looking at is C<SvUTF8> flag. Don't forget to set the flag if
2354something should be UTF8. Treat the flag as part of the PV, even though
2355it's not - if you pass on the PV to somewhere, pass on the flag too.
2356
2357=item *
2358
2359If a string is UTF8, B<always> use C<utf8_to_uv> to get at the value,
3a2263fe 2360unless C<UTF8_IS_INVARIANT(*s)> in which case you can use C<*s>.
a422fd2d
SC
2361
2362=item *
2363
3a2263fe
RGS
2364When writing a character C<uv> to a UTF8 string, B<always> use
2365C<uv_to_utf8>, unless C<UTF8_IS_INVARIANT(uv))> in which case
2366you can use C<*s = uv>.
a422fd2d
SC
2367
2368=item *
2369
2370Mixing UTF8 and non-UTF8 strings is tricky. Use C<bytes_to_utf8> to get
2371a new string which is UTF8 encoded. There are tricks you can use to
2372delay deciding whether you need to use a UTF8 string until you get to a
2373high character - C<HALF_UPGRADE> is one of those.
2374
2375=back
2376
53e06cf0
SC
2377=head1 Custom Operators
2378
9a68f1db 2379Custom operator support is a new experimental feature that allows you to
53e06cf0
SC
2380define your own ops. This is primarily to allow the building of
2381interpreters for other languages in the Perl core, but it also allows
2382optimizations through the creation of "macro-ops" (ops which perform the
2383functions of multiple ops which are usually executed together, such as
b7cb320d 2384C<gvsv, gvsv, add>.)
53e06cf0 2385
b455bf3f 2386This feature is implemented as a new op type, C<OP_CUSTOM>. The Perl
53e06cf0
SC
2387core does not "know" anything special about this op type, and so it will
2388not be involved in any optimizations. This also means that you can
2389define your custom ops to be any op structure - unary, binary, list and
2390so on - you like.
2391
2392It's important to know what custom operators won't do for you. They
2393won't let you add new syntax to Perl, directly. They won't even let you
2394add new keywords, directly. In fact, they won't change the way Perl
2395compiles a program at all. You have to do those changes yourself, after
2396Perl has compiled the program. You do this either by manipulating the op
2397tree using a C<CHECK> block and the C<B::Generate> module, or by adding
2398a custom peephole optimizer with the C<optimize> module.
2399
2400When you do this, you replace ordinary Perl ops with custom ops by
2401creating ops with the type C<OP_CUSTOM> and the C<pp_addr> of your own
2402PP function. This should be defined in XS code, and should look like
2403the PP ops in C<pp_*.c>. You are responsible for ensuring that your op
2404takes the appropriate number of values from the stack, and you are
2405responsible for adding stack marks if necessary.
2406
2407You should also "register" your op with the Perl interpreter so that it
2408can produce sensible error and warning messages. Since it is possible to
2409have multiple custom ops within the one "logical" op type C<OP_CUSTOM>,
2410Perl uses the value of C<< o->op_ppaddr >> as a key into the
2411C<PL_custom_op_descs> and C<PL_custom_op_names> hashes. This means you
2412need to enter a name and description for your op at the appropriate
2413place in the C<PL_custom_op_names> and C<PL_custom_op_descs> hashes.
2414
2415Forthcoming versions of C<B::Generate> (version 1.0 and above) should
2416directly support the creation of custom ops by name; C<Opcodes::Custom>
2417will provide functions which make it trivial to "register" custom ops to
2418the Perl interpreter.
2419
954c1994 2420=head1 AUTHORS
e89caa19 2421
954c1994 2422Until May 1997, this document was maintained by Jeff Okamoto
9b5bb84f
SB
2423E<lt>okamoto@corp.hp.comE<gt>. It is now maintained as part of Perl
2424itself by the Perl 5 Porters E<lt>perl5-porters@perl.orgE<gt>.
cb1a09d0 2425
954c1994
GS
2426With lots of help and suggestions from Dean Roehrich, Malcolm Beattie,
2427Andreas Koenig, Paul Hudson, Ilya Zakharevich, Paul Marquess, Neil
2428Bowers, Matthew Green, Tim Bunce, Spider Boardman, Ulrich Pfeifer,
2429Stephen McCamant, and Gurusamy Sarathy.
cb1a09d0 2430
9b5bb84f 2431API Listing originally by Dean Roehrich E<lt>roehrich@cray.comE<gt>.
cb1a09d0 2432
954c1994
GS
2433Modifications to autogenerate the API listing (L<perlapi>) by Benjamin
2434Stuhl.
cb1a09d0 2435
954c1994 2436=head1 SEE ALSO
cb1a09d0 2437
954c1994 2438perlapi(1), perlintern(1), perlxs(1), perlembed(1)