This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
better documentation of reference counts
[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 7This document attempts to describe how to use the Perl API, as well as
10e2eb10
FC
8to provide some info on the basic workings of the Perl core. It is far
9from complete and probably contains many errors. Please refer any
b3b6085d 10questions or comments to the author below.
a0d0e21e 11
0a753a76 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
10e2eb10 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).
61984ee1
KW
40("PV" stands for "Pointer Value". You might think that it is misnamed
41because it is described as pointing only to strings. However, it is
3ee1a09c 42possible to have it point to other things. For example, it could point
d6605d24 43to an array of UVs. But,
61984ee1
KW
44using it for non-strings requires care, as the underlying assumption of
45much of the internals is that PVs are just for strings. Often, for
6602b933 46example, a trailing C<NUL> is tacked on automatically. The non-string use
61984ee1 47is documented only in this paragraph.)
a0d0e21e 48
20dbd849 49The seven routines are:
a0d0e21e
LW
50
51 SV* newSViv(IV);
20dbd849 52 SV* newSVuv(UV);
a0d0e21e 53 SV* newSVnv(double);
06f6df17
RGS
54 SV* newSVpv(const char*, STRLEN);
55 SV* newSVpvn(const char*, STRLEN);
46fc3d4c 56 SV* newSVpvf(const char*, ...);
a0d0e21e
LW
57 SV* newSVsv(SV*);
58
e613617c 59C<STRLEN> is an integer type (C<Size_t>, usually defined as C<size_t> in
06f6df17
RGS
60F<config.h>) guaranteed to be large enough to represent the size of
61any string that perl can handle.
62
3bf17896 63In the unlikely case of a SV requiring more complex initialization, you
06f6df17
RGS
64can create an empty SV with newSV(len). If C<len> is 0 an empty SV of
65type NULL is returned, else an SV of type PV is returned with len + 1 (for
6602b933 66the C<NUL>) bytes of storage allocated, accessible via SvPVX. In both cases
da8c5729 67the SV has the undef value.
20dbd849 68
06f6df17 69 SV *sv = newSV(0); /* no storage allocated */
a9b0660e
KW
70 SV *sv = newSV(10); /* 10 (+1) bytes of uninitialised storage
71 * allocated */
20dbd849 72
06f6df17 73To change the value of an I<already-existing> SV, there are eight routines:
a0d0e21e
LW
74
75 void sv_setiv(SV*, IV);
deb3007b 76 void sv_setuv(SV*, UV);
a0d0e21e 77 void sv_setnv(SV*, double);
08105a92 78 void sv_setpv(SV*, const char*);
06f6df17 79 void sv_setpvn(SV*, const char*, STRLEN)
46fc3d4c 80 void sv_setpvf(SV*, const char*, ...);
a9b0660e 81 void sv_vsetpvfn(SV*, const char*, STRLEN, va_list *,
03a22d83 82 SV **, Size_t, bool *);
a0d0e21e
LW
83 void sv_setsv(SV*, SV*);
84
85Notice that you can choose to specify the length of the string to be
9da1e3b5
MUN
86assigned by using C<sv_setpvn>, C<newSVpvn>, or C<newSVpv>, or you may
87allow Perl to calculate the length by using C<sv_setpv> or by specifying
880 as the second argument to C<newSVpv>. Be warned, though, that Perl will
89determine the string's length by using C<strlen>, which depends on the
6602b933 90string terminating with a C<NUL> character, and not otherwise containing
a9b0660e 91NULs.
9abd00ed
GS
92
93The arguments of C<sv_setpvf> are processed like C<sprintf>, and the
94formatted output becomes the value.
95
328bf373 96C<sv_vsetpvfn> is an analogue of C<vsprintf>, but it allows you to specify
9abd00ed
GS
97either a pointer to a variable argument list or the address and length of
98an array of SVs. The last argument points to a boolean; on return, if that
99boolean is true, then locale-specific information has been used to format
c2611fb3 100the string, and the string's contents are therefore untrustworthy (see
9abd00ed
GS
101L<perlsec>). This pointer may be NULL if that information is not
102important. Note that this function requires you to specify the length of
103the format.
104
9da1e3b5 105The C<sv_set*()> functions are not generic enough to operate on values
5a0de581 106that have "magic". See L</Magic Virtual Tables> later in this document.
a0d0e21e 107
6602b933
KW
108All SVs that contain strings should be terminated with a C<NUL> character.
109If it is not C<NUL>-terminated there is a risk of
5f05dabc 110core dumps and corruptions from code which passes the string to C
6602b933
KW
111functions or system calls which expect a C<NUL>-terminated string.
112Perl's own functions typically add a trailing C<NUL> for this reason.
5f05dabc 113Nevertheless, you should be very careful when you pass a string stored
114in an SV to a C function or system call.
115
a0d0e21e
LW
116To access the actual value that an SV points to, you can use the macros:
117
118 SvIV(SV*)
954c1994 119 SvUV(SV*)
a0d0e21e
LW
120 SvNV(SV*)
121 SvPV(SV*, STRLEN len)
1fa8b10d 122 SvPV_nolen(SV*)
a0d0e21e 123
954c1994 124which will automatically coerce the actual scalar type into an IV, UV, double,
a0d0e21e
LW
125or string.
126
127In the C<SvPV> macro, the length of the string returned is placed into the
1fa8b10d
JD
128variable C<len> (this is a macro, so you do I<not> use C<&len>). If you do
129not care what the length of the data is, use the C<SvPV_nolen> macro.
130Historically the C<SvPV> macro with the global variable C<PL_na> has been
131used in this case. But that can be quite inefficient because C<PL_na> must
132be accessed in thread-local storage in threaded Perl. In any case, remember
133that Perl allows arbitrary strings of data that may both contain NULs and
6602b933 134might not be terminated by a C<NUL>.
a0d0e21e 135
ce2f5d8f 136Also remember that C doesn't allow you to safely say C<foo(SvPV(s, len),
10e2eb10
FC
137len);>. It might work with your
138compiler, but it won't work for everyone.
ce2f5d8f
KA
139Break this sort of statement up into separate assignments:
140
1aa6ea50
JC
141 SV *s;
142 STRLEN len;
61955433 143 char *ptr;
1aa6ea50
JC
144 ptr = SvPV(s, len);
145 foo(ptr, len);
ce2f5d8f 146
07fa94a1 147If you want to know if the scalar value is TRUE, you can use:
a0d0e21e
LW
148
149 SvTRUE(SV*)
150
151Although Perl will automatically grow strings for you, if you need to force
152Perl to allocate more memory for your SV, you can use the macro
153
154 SvGROW(SV*, STRLEN newlen)
155
156which will determine if more memory needs to be allocated. If so, it will
157call the function C<sv_grow>. Note that C<SvGROW> can only increase, not
5f05dabc 158decrease, the allocated memory of an SV and that it does not automatically
6602b933 159add space for the trailing C<NUL> byte (perl's own string functions typically do
8ebc5c01 160C<SvGROW(sv, len + 1)>).
a0d0e21e 161
21134f66
TC
162If you want to write to an existing SV's buffer and set its value to a
163string, use SvPV_force() or one of its variants to force the SV to be
164a PV. This will remove any of various types of non-stringness from
165the SV while preserving the content of the SV in the PV. This can be
166used, for example, to append data from an API function to a buffer
167without extra copying:
168
169 (void)SvPVbyte_force(sv, len);
170 s = SvGROW(sv, len + needlen + 1);
171 /* something that modifies up to needlen bytes at s+len, but
172 modifies newlen bytes
173 eg. newlen = read(fd, s + len, needlen);
174 ignoring errors for these examples
175 */
176 s[len + newlen] = '\0';
177 SvCUR_set(sv, len + newlen);
178 SvUTF8_off(sv);
179 SvSETMAGIC(sv);
180
181If you already have the data in memory or if you want to keep your
182code simple, you can use one of the sv_cat*() variants, such as
183sv_catpvn(). If you want to insert anywhere in the string you can use
184sv_insert() or sv_insert_flags().
185
186If you don't need the existing content of the SV, you can avoid some
187copying with:
188
5b1fede8 189 SvPVCLEAR(sv);
21134f66
TC
190 s = SvGROW(sv, needlen + 1);
191 /* something that modifies up to needlen bytes at s, but modifies
192 newlen bytes
193 eg. newlen = read(fd, s. needlen);
194 */
195 s[newlen] = '\0';
196 SvCUR_set(sv, newlen);
197 SvPOK_only(sv); /* also clears SVf_UTF8 */
198 SvSETMAGIC(sv);
199
200Again, if you already have the data in memory or want to avoid the
201complexity of the above, you can use sv_setpvn().
202
203If you have a buffer allocated with Newx() and want to set that as the
204SV's value, you can use sv_usepvn_flags(). That has some requirements
205if you want to avoid perl re-allocating the buffer to fit the trailing
206NUL:
207
208 Newx(buf, somesize+1, char);
209 /* ... fill in buf ... */
210 buf[somesize] = '\0';
211 sv_usepvn_flags(sv, buf, somesize, SV_SMAGIC | SV_HAS_TRAILING_NUL);
212 /* buf now belongs to perl, don't release it */
213
a0d0e21e
LW
214If you have an SV and want to know what kind of data Perl thinks is stored
215in it, you can use the following macros to check the type of SV you have.
216
217 SvIOK(SV*)
218 SvNOK(SV*)
219 SvPOK(SV*)
220
221You can get and set the current length of the string stored in an SV with
222the following macros:
223
224 SvCUR(SV*)
225 SvCUR_set(SV*, I32 val)
226
cb1a09d0
AD
227You can also get a pointer to the end of the string stored in the SV
228with the macro:
229
230 SvEND(SV*)
231
232But note that these last three macros are valid only if C<SvPOK()> is true.
a0d0e21e 233
d1b91892
AD
234If you want to append something to the end of string stored in an C<SV*>,
235you can use the following functions:
236
08105a92 237 void sv_catpv(SV*, const char*);
e65f3abd 238 void sv_catpvn(SV*, const char*, STRLEN);
46fc3d4c 239 void sv_catpvf(SV*, const char*, ...);
a9b0660e
KW
240 void sv_vcatpvfn(SV*, const char*, STRLEN, va_list *, SV **,
241 I32, bool);
d1b91892
AD
242 void sv_catsv(SV*, SV*);
243
244The first function calculates the length of the string to be appended by
245using C<strlen>. In the second, you specify the length of the string
46fc3d4c 246yourself. The third function processes its arguments like C<sprintf> and
9abd00ed
GS
247appends the formatted output. The fourth function works like C<vsprintf>.
248You can specify the address and length of an array of SVs instead of the
10e2eb10
FC
249va_list argument. The fifth function
250extends the string stored in the first
9abd00ed
GS
251SV with the string stored in the second SV. It also forces the second SV
252to be interpreted as a string.
253
254The C<sv_cat*()> functions are not generic enough to operate on values that
5a0de581 255have "magic". See L</Magic Virtual Tables> later in this document.
d1b91892 256
a0d0e21e
LW
257If you know the name of a scalar variable, you can get a pointer to its SV
258by using the following:
259
64ace3f8 260 SV* get_sv("package::varname", 0);
a0d0e21e
LW
261
262This returns NULL if the variable does not exist.
263
d1b91892 264If you want to know if this variable (or any other SV) is actually C<defined>,
a0d0e21e
LW
265you can call:
266
267 SvOK(SV*)
268
06f6df17 269The scalar C<undef> value is stored in an SV instance called C<PL_sv_undef>.
9adebda4 270
10e2eb10
FC
271Its address can be used whenever an C<SV*> is needed. Make sure that
272you don't try to compare a random sv with C<&PL_sv_undef>. For example
9adebda4
SB
273when interfacing Perl code, it'll work correctly for:
274
275 foo(undef);
276
277But won't work when called as:
278
279 $x = undef;
280 foo($x);
281
282So to repeat always use SvOK() to check whether an sv is defined.
283
284Also you have to be careful when using C<&PL_sv_undef> as a value in
5a0de581 285AVs or HVs (see L</AVs, HVs and undefined values>).
a0d0e21e 286
06f6df17
RGS
287There are also the two values C<PL_sv_yes> and C<PL_sv_no>, which contain
288boolean TRUE and FALSE values, respectively. Like C<PL_sv_undef>, their
289addresses can be used whenever an C<SV*> is needed.
a0d0e21e 290
9cde0e7f 291Do not be fooled into thinking that C<(SV *) 0> is the same as C<&PL_sv_undef>.
a0d0e21e
LW
292Take this code:
293
294 SV* sv = (SV*) 0;
295 if (I-am-to-return-a-real-value) {
296 sv = sv_2mortal(newSViv(42));
297 }
298 sv_setsv(ST(0), sv);
299
300This code tries to return a new SV (which contains the value 42) if it should
04343c6d 301return a real value, or undef otherwise. Instead it has returned a NULL
a0d0e21e 302pointer which, somewhere down the line, will cause a segmentation violation,
06f6df17
RGS
303bus error, or just weird results. Change the zero to C<&PL_sv_undef> in the
304first line and all will be well.
a0d0e21e
LW
305
306To free an SV that you've created, call C<SvREFCNT_dec(SV*)>. Normally this
5a0de581 307call is not necessary (see L</Reference Counts and Mortality>).
a0d0e21e 308
94dde4fb
SC
309=head2 Offsets
310
311Perl provides the function C<sv_chop> to efficiently remove characters
312from the beginning of a string; you give it an SV and a pointer to
da75cd15 313somewhere inside the PV, and it discards everything before the
10e2eb10 314pointer. The efficiency comes by means of a little hack: instead of
94dde4fb
SC
315actually removing the characters, C<sv_chop> sets the flag C<OOK>
316(offset OK) to signal to other functions that the offset hack is in
883bb8c0
KW
317effect, and it moves the PV pointer (called C<SvPVX>) forward
318by the number of bytes chopped off, and adjusts C<SvCUR> and C<SvLEN>
319accordingly. (A portion of the space between the old and new PV
320pointers is used to store the count of chopped bytes.)
94dde4fb
SC
321
322Hence, at this point, the start of the buffer that we allocated lives
323at C<SvPVX(sv) - SvIV(sv)> in memory and the PV pointer is pointing
324into the middle of this allocated storage.
325
f942a0df
FC
326This is best demonstrated by example. Normally copy-on-write will prevent
327the substitution from operator from using this hack, but if you can craft a
328string for which copy-on-write is not possible, you can see it in play. In
329the current implementation, the final byte of a string buffer is used as a
330copy-on-write reference count. If the buffer is not big enough, then
331copy-on-write is skipped. First have a look at an empty string:
332
333 % ./perl -Ilib -MDevel::Peek -le '$a=""; $a .= ""; Dump $a'
334 SV = PV(0x7ffb7c008a70) at 0x7ffb7c030390
335 REFCNT = 1
336 FLAGS = (POK,pPOK)
337 PV = 0x7ffb7bc05b50 ""\0
338 CUR = 0
339 LEN = 10
340
341Notice here the LEN is 10. (It may differ on your platform.) Extend the
342length of the string to one less than 10, and do a substitution:
94dde4fb 343
e46aa1dd
KW
344 % ./perl -Ilib -MDevel::Peek -le '$a=""; $a.="123456789"; $a=~s/.//; \
345 Dump($a)'
346 SV = PV(0x7ffa04008a70) at 0x7ffa04030390
347 REFCNT = 1
348 FLAGS = (POK,OOK,pPOK)
349 OFFSET = 1
350 PV = 0x7ffa03c05b61 ( "\1" . ) "23456789"\0
351 CUR = 8
352 LEN = 9
94dde4fb 353
f942a0df 354Here the number of bytes chopped off (1) is shown next as the OFFSET. The
94dde4fb
SC
355portion of the string between the "real" and the "fake" beginnings is
356shown in parentheses, and the values of C<SvCUR> and C<SvLEN> reflect
f942a0df
FC
357the fake beginning, not the real one. (The first character of the string
358buffer happens to have changed to "\1" here, not "1", because the current
359implementation stores the offset count in the string buffer. This is
360subject to change.)
94dde4fb 361
fe854a6f 362Something similar to the offset hack is performed on AVs to enable
319cef53
SC
363efficient shifting and splicing off the beginning of the array; while
364C<AvARRAY> points to the first element in the array that is visible from
10e2eb10 365Perl, C<AvALLOC> points to the real start of the C array. These are
319cef53 366usually the same, but a C<shift> operation can be carried out by
6de131f0 367increasing C<AvARRAY> by one and decreasing C<AvFILL> and C<AvMAX>.
319cef53 368Again, the location of the real start of the C array only comes into
10e2eb10 369play when freeing the array. See C<av_shift> in F<av.c>.
319cef53 370
d1b91892 371=head2 What's Really Stored in an SV?
a0d0e21e
LW
372
373Recall that the usual method of determining the type of scalar you have is
5f05dabc 374to use C<Sv*OK> macros. Because a scalar can be both a number and a string,
d1b91892 375usually these macros will always return TRUE and calling the C<Sv*V>
a0d0e21e
LW
376macros will do the appropriate conversion of string to integer/double or
377integer/double to string.
378
379If you I<really> need to know if you have an integer, double, or string
380pointer in an SV, you can use the following three macros instead:
381
382 SvIOKp(SV*)
383 SvNOKp(SV*)
384 SvPOKp(SV*)
385
386These will tell you if you truly have an integer, double, or string pointer
d1b91892 387stored in your SV. The "p" stands for private.
a0d0e21e 388
da8c5729 389There are various ways in which the private and public flags may differ.
9090718a
FC
390For example, in perl 5.16 and earlier a tied SV may have a valid
391underlying value in the IV slot (so SvIOKp is true), but the data
392should be accessed via the FETCH routine rather than directly,
393so SvIOK is false. (In perl 5.18 onwards, tied scalars use
394the flags the same way as untied scalars.) Another is when
d7f8936a 395numeric conversion has occurred and precision has been lost: only the
10e2eb10 396private flag is set on 'lossy' values. So when an NV is converted to an
9e9796d6
JH
397IV with loss, SvIOKp, SvNOKp and SvNOK will be set, while SvIOK wont be.
398
07fa94a1 399In general, though, it's best to use the C<Sv*V> macros.
a0d0e21e 400
54310121 401=head2 Working with AVs
a0d0e21e 402
07fa94a1
JO
403There are two ways to create and load an AV. The first method creates an
404empty AV:
a0d0e21e
LW
405
406 AV* newAV();
407
54310121 408The second method both creates the AV and initially populates it with SVs:
a0d0e21e 409
c70927a6 410 AV* av_make(SSize_t num, SV **ptr);
a0d0e21e 411
5f05dabc 412The second argument points to an array containing C<num> C<SV*>'s. Once the
54310121 413AV has been created, the SVs can be destroyed, if so desired.
a0d0e21e 414
da8c5729 415Once the AV has been created, the following operations are possible on it:
a0d0e21e
LW
416
417 void av_push(AV*, SV*);
418 SV* av_pop(AV*);
419 SV* av_shift(AV*);
c70927a6 420 void av_unshift(AV*, SSize_t num);
a0d0e21e
LW
421
422These should be familiar operations, with the exception of C<av_unshift>.
423This routine adds C<num> elements at the front of the array with the C<undef>
424value. You must then use C<av_store> (described below) to assign values
425to these new elements.
426
427Here are some other functions:
428
c70927a6
FC
429 SSize_t av_top_index(AV*);
430 SV** av_fetch(AV*, SSize_t key, I32 lval);
431 SV** av_store(AV*, SSize_t key, SV* val);
a0d0e21e 432
dab460cd 433The C<av_top_index> function returns the highest index value in an array (just
5f05dabc 434like $#array in Perl). If the array is empty, -1 is returned. The
435C<av_fetch> function returns the value at index C<key>, but if C<lval>
436is non-zero, then C<av_fetch> will store an undef value at that index.
04343c6d
GS
437The C<av_store> function stores the value C<val> at index C<key>, and does
438not increment the reference count of C<val>. Thus the caller is responsible
439for taking care of that, and if C<av_store> returns NULL, the caller will
440have to decrement the reference count to avoid a memory leak. Note that
441C<av_fetch> and C<av_store> both return C<SV**>'s, not C<SV*>'s as their
442return value.
d1b91892 443
da8c5729
MH
444A few more:
445
a0d0e21e 446 void av_clear(AV*);
a0d0e21e 447 void av_undef(AV*);
c70927a6 448 void av_extend(AV*, SSize_t key);
5f05dabc 449
450The C<av_clear> function deletes all the elements in the AV* array, but
451does not actually delete the array itself. The C<av_undef> function will
452delete all the elements in the array plus the array itself. The
adc882cf
GS
453C<av_extend> function extends the array so that it contains at least C<key+1>
454elements. If C<key+1> is less than the currently allocated length of the array,
455then nothing is done.
a0d0e21e
LW
456
457If you know the name of an array variable, you can get a pointer to its AV
458by using the following:
459
cbfd0a87 460 AV* get_av("package::varname", 0);
a0d0e21e
LW
461
462This returns NULL if the variable does not exist.
463
5a0de581 464See L</Understanding the Magic of Tied Hashes and Arrays> for more
04343c6d
GS
465information on how to use the array access functions on tied arrays.
466
54310121 467=head2 Working with HVs
a0d0e21e
LW
468
469To create an HV, you use the following routine:
470
471 HV* newHV();
472
da8c5729 473Once the HV has been created, the following operations are possible on it:
a0d0e21e 474
08105a92
GS
475 SV** hv_store(HV*, const char* key, U32 klen, SV* val, U32 hash);
476 SV** hv_fetch(HV*, const char* key, U32 klen, I32 lval);
a0d0e21e 477
5f05dabc 478The C<klen> parameter is the length of the key being passed in (Note that
479you cannot pass 0 in as a value of C<klen> to tell Perl to measure the
480length of the key). The C<val> argument contains the SV pointer to the
54310121 481scalar being stored, and C<hash> is the precomputed hash value (zero if
5f05dabc 482you want C<hv_store> to calculate it for you). The C<lval> parameter
483indicates whether this fetch is actually a part of a store operation, in
484which case a new undefined value will be added to the HV with the supplied
485key and C<hv_fetch> will return as if the value had already existed.
a0d0e21e 486
5f05dabc 487Remember that C<hv_store> and C<hv_fetch> return C<SV**>'s and not just
488C<SV*>. To access the scalar value, you must first dereference the return
489value. However, you should check to make sure that the return value is
490not NULL before dereferencing it.
a0d0e21e 491
da8c5729
MH
492The first of these two functions checks if a hash table entry exists, and the
493second deletes it.
a0d0e21e 494
08105a92
GS
495 bool hv_exists(HV*, const char* key, U32 klen);
496 SV* hv_delete(HV*, const char* key, U32 klen, I32 flags);
a0d0e21e 497
5f05dabc 498If C<flags> does not include the C<G_DISCARD> flag then C<hv_delete> will
499create and return a mortal copy of the deleted value.
500
a0d0e21e
LW
501And more miscellaneous functions:
502
503 void hv_clear(HV*);
a0d0e21e 504 void hv_undef(HV*);
5f05dabc 505
506Like their AV counterparts, C<hv_clear> deletes all the entries in the hash
507table but does not actually delete the hash table. The C<hv_undef> deletes
508both the entries and the hash table itself.
a0d0e21e 509
a9b0660e 510Perl keeps the actual data in a linked list of structures with a typedef of HE.
d1b91892
AD
511These contain the actual key and value pointers (plus extra administrative
512overhead). The key is a string pointer; the value is an C<SV*>. However,
513once you have an C<HE*>, to get the actual key and value, use the routines
514specified below.
515
a0d0e21e
LW
516 I32 hv_iterinit(HV*);
517 /* Prepares starting point to traverse hash table */
518 HE* hv_iternext(HV*);
519 /* Get the next entry, and return a pointer to a
520 structure that has both the key and value */
521 char* hv_iterkey(HE* entry, I32* retlen);
522 /* Get the key from an HE structure and also return
523 the length of the key string */
cb1a09d0 524 SV* hv_iterval(HV*, HE* entry);
d1be9408 525 /* Return an SV pointer to the value of the HE
a0d0e21e 526 structure */
cb1a09d0 527 SV* hv_iternextsv(HV*, char** key, I32* retlen);
d1b91892
AD
528 /* This convenience routine combines hv_iternext,
529 hv_iterkey, and hv_iterval. The key and retlen
530 arguments are return values for the key and its
531 length. The value is returned in the SV* argument */
a0d0e21e
LW
532
533If you know the name of a hash variable, you can get a pointer to its HV
534by using the following:
535
6673a63c 536 HV* get_hv("package::varname", 0);
a0d0e21e
LW
537
538This returns NULL if the variable does not exist.
539
a43e7901 540The hash algorithm is defined in the C<PERL_HASH> macro:
a0d0e21e 541
a43e7901 542 PERL_HASH(hash, key, klen)
ab192400 543
a43e7901
YO
544The exact implementation of this macro varies by architecture and version
545of perl, and the return value may change per invocation, so the value
546is only valid for the duration of a single perl process.
a0d0e21e 547
5a0de581 548See L</Understanding the Magic of Tied Hashes and Arrays> for more
04343c6d
GS
549information on how to use the hash access functions on tied hashes.
550
1e422769 551=head2 Hash API Extensions
552
553Beginning with version 5.004, the following functions are also supported:
554
555 HE* hv_fetch_ent (HV* tb, SV* key, I32 lval, U32 hash);
556 HE* hv_store_ent (HV* tb, SV* key, SV* val, U32 hash);
c47ff5f1 557
1e422769 558 bool hv_exists_ent (HV* tb, SV* key, U32 hash);
559 SV* hv_delete_ent (HV* tb, SV* key, I32 flags, U32 hash);
c47ff5f1 560
1e422769 561 SV* hv_iterkeysv (HE* entry);
562
563Note that these functions take C<SV*> keys, which simplifies writing
564of extension code that deals with hash structures. These functions
565also allow passing of C<SV*> keys to C<tie> functions without forcing
566you to stringify the keys (unlike the previous set of functions).
567
568They also return and accept whole hash entries (C<HE*>), making their
569use more efficient (since the hash number for a particular string
4a4eefd0
GS
570doesn't have to be recomputed every time). See L<perlapi> for detailed
571descriptions.
1e422769 572
573The following macros must always be used to access the contents of hash
574entries. Note that the arguments to these macros must be simple
575variables, since they may get evaluated more than once. See
4a4eefd0 576L<perlapi> for detailed descriptions of these macros.
1e422769 577
578 HePV(HE* he, STRLEN len)
579 HeVAL(HE* he)
580 HeHASH(HE* he)
581 HeSVKEY(HE* he)
582 HeSVKEY_force(HE* he)
583 HeSVKEY_set(HE* he, SV* sv)
584
585These two lower level macros are defined, but must only be used when
586dealing with keys that are not C<SV*>s:
587
588 HeKEY(HE* he)
589 HeKLEN(HE* he)
590
04343c6d
GS
591Note that both C<hv_store> and C<hv_store_ent> do not increment the
592reference count of the stored C<val>, which is the caller's responsibility.
593If these functions return a NULL value, the caller will usually have to
594decrement the reference count of C<val> to avoid a memory leak.
1e422769 595
a9381218
MHM
596=head2 AVs, HVs and undefined values
597
10e2eb10
FC
598Sometimes you have to store undefined values in AVs or HVs. Although
599this may be a rare case, it can be tricky. That's because you're
a9381218
MHM
600used to using C<&PL_sv_undef> if you need an undefined SV.
601
602For example, intuition tells you that this XS code:
603
604 AV *av = newAV();
605 av_store( av, 0, &PL_sv_undef );
606
607is equivalent to this Perl code:
608
609 my @av;
610 $av[0] = undef;
611
f3c4ec28 612Unfortunately, this isn't true. In perl 5.18 and earlier, AVs use C<&PL_sv_undef> as a marker
a9381218
MHM
613for indicating that an array element has not yet been initialized.
614Thus, C<exists $av[0]> would be true for the above Perl code, but
f3c4ec28
FC
615false for the array generated by the XS code. In perl 5.20, storing
616&PL_sv_undef will create a read-only element, because the scalar
617&PL_sv_undef itself is stored, not a copy.
a9381218 618
f3c4ec28 619Similar problems can occur when storing C<&PL_sv_undef> in HVs:
a9381218
MHM
620
621 hv_store( hv, "key", 3, &PL_sv_undef, 0 );
622
623This will indeed make the value C<undef>, but if you try to modify
624the value of C<key>, you'll get the following error:
625
626 Modification of non-creatable hash value attempted
627
628In perl 5.8.0, C<&PL_sv_undef> was also used to mark placeholders
10e2eb10 629in restricted hashes. This caused such hash entries not to appear
a9381218
MHM
630when iterating over the hash or when checking for the keys
631with the C<hv_exists> function.
632
8abccac8 633You can run into similar problems when you store C<&PL_sv_yes> or
10e2eb10 634C<&PL_sv_no> into AVs or HVs. Trying to modify such elements
a9381218
MHM
635will give you the following error:
636
637 Modification of a read-only value attempted
638
639To make a long story short, you can use the special variables
8abccac8 640C<&PL_sv_undef>, C<&PL_sv_yes> and C<&PL_sv_no> with AVs and
a9381218
MHM
641HVs, but you have to make sure you know what you're doing.
642
643Generally, if you want to store an undefined value in an AV
644or HV, you should not use C<&PL_sv_undef>, but rather create a
645new undefined value using the C<newSV> function, for example:
646
647 av_store( av, 42, newSV(0) );
648 hv_store( hv, "foo", 3, newSV(0), 0 );
649
a0d0e21e
LW
650=head2 References
651
d1b91892 652References are a special type of scalar that point to other data types
a9b0660e 653(including other references).
a0d0e21e 654
07fa94a1 655To create a reference, use either of the following functions:
a0d0e21e 656
5f05dabc 657 SV* newRV_inc((SV*) thing);
658 SV* newRV_noinc((SV*) thing);
a0d0e21e 659
5f05dabc 660The C<thing> argument can be any of an C<SV*>, C<AV*>, or C<HV*>. The
07fa94a1
JO
661functions are identical except that C<newRV_inc> increments the reference
662count of the C<thing>, while C<newRV_noinc> does not. For historical
663reasons, C<newRV> is a synonym for C<newRV_inc>.
664
665Once you have a reference, you can use the following macro to dereference
666the reference:
a0d0e21e
LW
667
668 SvRV(SV*)
669
670then call the appropriate routines, casting the returned C<SV*> to either an
d1b91892 671C<AV*> or C<HV*>, if required.
a0d0e21e 672
d1b91892 673To determine if an SV is a reference, you can use the following macro:
a0d0e21e
LW
674
675 SvROK(SV*)
676
07fa94a1
JO
677To discover what type of value the reference refers to, use the following
678macro and then check the return value.
d1b91892
AD
679
680 SvTYPE(SvRV(SV*))
681
682The most useful types that will be returned are:
683
a5e62da0
FC
684 < SVt_PVAV Scalar
685 SVt_PVAV Array
686 SVt_PVHV Hash
687 SVt_PVCV Code
688 SVt_PVGV Glob (possibly a file handle)
689
690See L<perlapi/svtype> for more details.
d1b91892 691
cb1a09d0
AD
692=head2 Blessed References and Class Objects
693
06f6df17 694References are also used to support object-oriented programming. In perl's
cb1a09d0
AD
695OO lexicon, an object is simply a reference that has been blessed into a
696package (or class). Once blessed, the programmer may now use the reference
697to access the various methods in the class.
698
699A reference can be blessed into a package with the following function:
700
701 SV* sv_bless(SV* sv, HV* stash);
702
06f6df17
RGS
703The C<sv> argument must be a reference value. The C<stash> argument
704specifies which class the reference will belong to. See
5a0de581 705L</Stashes and Globs> for information on converting class names into stashes.
cb1a09d0
AD
706
707/* Still under construction */
708
ddd2cc91
DM
709The following function upgrades rv to reference if not already one.
710Creates a new SV for rv to point to. If C<classname> is non-null, the SV
711is blessed into the specified class. SV is returned.
cb1a09d0 712
08105a92 713 SV* newSVrv(SV* rv, const char* classname);
cb1a09d0 714
ddd2cc91
DM
715The following three functions copy integer, unsigned integer or double
716into an SV whose reference is C<rv>. SV is blessed if C<classname> is
717non-null.
cb1a09d0 718
08105a92 719 SV* sv_setref_iv(SV* rv, const char* classname, IV iv);
e1c57cef 720 SV* sv_setref_uv(SV* rv, const char* classname, UV uv);
08105a92 721 SV* sv_setref_nv(SV* rv, const char* classname, NV iv);
cb1a09d0 722
ddd2cc91
DM
723The following function copies the pointer value (I<the address, not the
724string!>) into an SV whose reference is rv. SV is blessed if C<classname>
725is non-null.
cb1a09d0 726
ddd2cc91 727 SV* sv_setref_pv(SV* rv, const char* classname, void* pv);
cb1a09d0 728
a9b0660e 729The following function copies a string into an SV whose reference is C<rv>.
ddd2cc91
DM
730Set length to 0 to let Perl calculate the string length. SV is blessed if
731C<classname> is non-null.
cb1a09d0 732
a9b0660e
KW
733 SV* sv_setref_pvn(SV* rv, const char* classname, char* pv,
734 STRLEN length);
cb1a09d0 735
ddd2cc91
DM
736The following function tests whether the SV is blessed into the specified
737class. It does not check inheritance relationships.
9abd00ed 738
08105a92 739 int sv_isa(SV* sv, const char* name);
9abd00ed 740
ddd2cc91 741The following function tests whether the SV is a reference to a blessed object.
9abd00ed
GS
742
743 int sv_isobject(SV* sv);
744
ddd2cc91 745The following function tests whether the SV is derived from the specified
10e2eb10
FC
746class. SV can be either a reference to a blessed object or a string
747containing a class name. This is the function implementing the
ddd2cc91 748C<UNIVERSAL::isa> functionality.
9abd00ed 749
08105a92 750 bool sv_derived_from(SV* sv, const char* name);
9abd00ed 751
00aadd71 752To check if you've got an object derived from a specific class you have
9abd00ed
GS
753to write:
754
755 if (sv_isobject(sv) && sv_derived_from(sv, class)) { ... }
cb1a09d0 756
5f05dabc 757=head2 Creating New Variables
cb1a09d0 758
5f05dabc 759To create a new Perl variable with an undef value which can be accessed from
760your Perl script, use the following routines, depending on the variable type.
cb1a09d0 761
64ace3f8 762 SV* get_sv("package::varname", GV_ADD);
cbfd0a87 763 AV* get_av("package::varname", GV_ADD);
6673a63c 764 HV* get_hv("package::varname", GV_ADD);
cb1a09d0 765
058a5f6c 766Notice the use of GV_ADD as the second parameter. The new variable can now
cb1a09d0
AD
767be set, using the routines appropriate to the data type.
768
5f05dabc 769There are additional macros whose values may be bitwise OR'ed with the
058a5f6c 770C<GV_ADD> argument to enable certain extra features. Those bits are:
cb1a09d0 771
9a68f1db
SB
772=over
773
774=item GV_ADDMULTI
775
776Marks the variable as multiply defined, thus preventing the:
777
778 Name <varname> used only once: possible typo
779
780warning.
781
9a68f1db
SB
782=item GV_ADDWARN
783
784Issues the warning:
785
786 Had to create <varname> unexpectedly
787
788if the variable did not exist before the function was called.
789
790=back
cb1a09d0 791
07fa94a1
JO
792If you do not specify a package name, the variable is created in the current
793package.
cb1a09d0 794
5f05dabc 795=head2 Reference Counts and Mortality
a0d0e21e 796
10e2eb10 797Perl uses a reference count-driven garbage collection mechanism. SVs,
54310121 798AVs, or HVs (xV for short in the following) start their life with a
55497cff 799reference count of 1. If the reference count of an xV ever drops to 0,
07fa94a1 800then it will be destroyed and its memory made available for reuse.
3d2ba989
Z
801At the most basic internal level, reference counts can be manipulated
802with the following macros:
55497cff 803
804 int SvREFCNT(SV* sv);
5f05dabc 805 SV* SvREFCNT_inc(SV* sv);
55497cff 806 void SvREFCNT_dec(SV* sv);
807
3d2ba989
Z
808(There are also suffixed versions of the increment and decrement macros,
809for situations where the full generality of these basic macros can be
810exchanged for some performance.)
811
812However, the way a programmer should think about references is not so
813much in terms of the bare reference count, but in terms of I<ownership>
814of references. A reference to an xV can be owned by any of a variety
815of entities: another xV, the Perl interpreter, an XS data structure,
816a piece of running code, or a dynamic scope. An xV generally does not
817know what entities own the references to it; it only knows how many
818references there are, which is the reference count.
819
820To correctly maintain reference counts, it is essential to keep track
821of what references the XS code is manipulating. The programmer should
822always know where a reference has come from and who owns it, and be
823aware of any creation or destruction of references, and any transfers
824of ownership. Because ownership isn't represented explicitly in the xV
825data structures, only the reference count need be actually maintained
826by the code, and that means that this understanding of ownership is not
827actually evident in the code. For example, transferring ownership of a
828reference from one owner to another doesn't change the reference count
829at all, so may be achieved with no actual code. (The transferring code
830doesn't touch the referenced object, but does need to ensure that the
831former owner knows that it no longer owns the reference, and that the
832new owner knows that it now does.)
833
834An xV that is visible at the Perl level should not become unreferenced
835and thus be destroyed. Normally, an object will only become unreferenced
836when it is no longer visible, often by the same means that makes it
837invisible. For example, a Perl reference value (RV) owns a reference to
838its referent, so if the RV is overwritten that reference gets destroyed,
839and the no-longer-reachable referent may be destroyed as a result.
840
841Many functions have some kind of reference manipulation as
842part of their purpose. Sometimes this is documented in terms
843of ownership of references, and sometimes it is (less helpfully)
844documented in terms of changes to reference counts. For example, the
845L<newRV_inc()|perlapi/newRV_inc> function is documented to create a new RV
846(with reference count 1) and increment the reference count of the referent
847that was supplied by the caller. This is best understood as creating
848a new reference to the referent, which is owned by the created RV,
849and returning to the caller ownership of the sole reference to the RV.
850The L<newRV_noinc()|perlapi/newRV_noinc> function instead does not
851increment the reference count of the referent, but the RV nevertheless
852ends up owning a reference to the referent. It is therefore implied
853that the caller of C<newRV_noinc()> is relinquishing a reference to the
854referent, making this conceptually a more complicated operation even
855though it does less to the data structures.
856
857For example, imagine you want to return a reference from an XSUB
858function. Inside the XSUB routine, you create an SV which initially
859has just a single reference, owned by the XSUB routine. This reference
860needs to be disposed of before the routine is complete, otherwise it
861will leak, preventing the SV from ever being destroyed. So to create
862an RV referencing the SV, it is most convenient to pass the SV to
863C<newRV_noinc()>, which consumes that reference. Now the XSUB routine
864no longer owns a reference to the SV, but does own a reference to the RV,
865which in turn owns a reference to the SV. The ownership of the reference
866to the RV is then transferred by the process of returning the RV from
867the XSUB.
55497cff 868
5f05dabc 869There are some convenience functions available that can help with the
54310121 870destruction of xVs. These functions introduce the concept of "mortality".
3d2ba989
Z
871Much documentation speaks of an xV itself being mortal, but this is
872misleading. It is really I<a reference to> an xV that is mortal, and it
873is possible for there to be more than one mortal reference to a single xV.
874For a reference to be mortal means that it is owned by the temps stack,
875one of perl's many internal stacks, which will destroy that reference
876"a short time later". Usually the "short time later" is the end of
877the current Perl statement. However, it gets more complicated around
878dynamic scopes: there can be multiple sets of mortal references hanging
879around at the same time, with different death dates. Internally, the
880actual determinant for when mortal xV references are destroyed depends
881on two macros, SAVETMPS and FREETMPS. See L<perlcall> and L<perlxs>
882for more details on these macros.
883
884Mortal references are mainly used for xVs that are placed on perl's
885main stack. The stack is problematic for reference tracking, because it
886contains a lot of xV references, but doesn't own those references: they
887are not counted. Currently, there are many bugs resulting from xVs being
888destroyed while referenced by the stack, because the stack's uncounted
889references aren't enough to keep the xVs alive. So when putting an
890(uncounted) reference on the stack, it is vitally important to ensure that
891there will be a counted reference to the same xV that will last at least
892as long as the uncounted reference. But it's also important that that
893counted reference be cleaned up at an appropriate time, and not unduly
894prolong the xV's life. For there to be a mortal reference is often the
895best way to satisfy this requirement, especially if the xV was created
896especially to be put on the stack and would otherwise be unreferenced.
897
898To create a mortal reference, use the functions:
a0d0e21e
LW
899
900 SV* sv_newmortal()
a0d0e21e 901 SV* sv_mortalcopy(SV*)
3d2ba989 902 SV* sv_2mortal(SV*)
a0d0e21e 903
3d2ba989
Z
904C<sv_newmortal()> creates an SV (with the undefined value) whose sole
905reference is mortal. C<sv_mortalcopy()> creates an xV whose value is a
906copy of a supplied xV and whose sole reference is mortal. C<sv_2mortal()>
907mortalises an existing xV reference: it transfers ownership of a reference
908from the caller to the temps stack. Because C<sv_newmortal> gives the new
909SV no value, it must normally be given one via C<sv_setpv>, C<sv_setiv>,
910etc. :
00aadd71
NIS
911
912 SV *tmp = sv_newmortal();
913 sv_setiv(tmp, an_integer);
914
915As that is multiple C statements it is quite common so see this idiom instead:
916
917 SV *tmp = sv_2mortal(newSViv(an_integer));
918
ac036724 919The mortal routines are not just for SVs; AVs and HVs can be
faed5253 920made mortal by passing their address (type-casted to C<SV*>) to the
07fa94a1 921C<sv_2mortal> or C<sv_mortalcopy> routines.
a0d0e21e 922
5f05dabc 923=head2 Stashes and Globs
a0d0e21e 924
06f6df17
RGS
925A B<stash> is a hash that contains all variables that are defined
926within a package. Each key of the stash is a symbol
aa689395 927name (shared by all the different types of objects that have the same
928name), and each value in the hash table is a GV (Glob Value). This GV
929in turn contains references to the various objects of that name,
930including (but not limited to) the following:
cb1a09d0 931
a0d0e21e
LW
932 Scalar Value
933 Array Value
934 Hash Value
a3cb178b 935 I/O Handle
a0d0e21e
LW
936 Format
937 Subroutine
938
06f6df17
RGS
939There is a single stash called C<PL_defstash> that holds the items that exist
940in the C<main> package. To get at the items in other packages, append the
941string "::" to the package name. The items in the C<Foo> package are in
942the stash C<Foo::> in PL_defstash. The items in the C<Bar::Baz> package are
943in the stash C<Baz::> in C<Bar::>'s stash.
a0d0e21e 944
d1b91892 945To get the stash pointer for a particular package, use the function:
a0d0e21e 946
da51bb9b
NC
947 HV* gv_stashpv(const char* name, I32 flags)
948 HV* gv_stashsv(SV*, I32 flags)
a0d0e21e
LW
949
950The first function takes a literal string, the second uses the string stored
d1b91892 951in the SV. Remember that a stash is just a hash table, so you get back an
da51bb9b 952C<HV*>. The C<flags> flag will create a new package if it is set to GV_ADD.
a0d0e21e
LW
953
954The name that C<gv_stash*v> wants is the name of the package whose symbol table
955you want. The default package is called C<main>. If you have multiply nested
d1b91892
AD
956packages, pass their names to C<gv_stash*v>, separated by C<::> as in the Perl
957language itself.
a0d0e21e
LW
958
959Alternately, if you have an SV that is a blessed reference, you can find
960out the stash pointer by using:
961
962 HV* SvSTASH(SvRV(SV*));
963
964then use the following to get the package name itself:
965
966 char* HvNAME(HV* stash);
967
5f05dabc 968If you need to bless or re-bless an object you can use the following
969function:
a0d0e21e
LW
970
971 SV* sv_bless(SV*, HV* stash)
972
973where the first argument, an C<SV*>, must be a reference, and the second
974argument is a stash. The returned C<SV*> can now be used in the same way
975as any other SV.
976
d1b91892
AD
977For more information on references and blessings, consult L<perlref>.
978
54310121 979=head2 Double-Typed SVs
0a753a76 980
981Scalar variables normally contain only one type of value, an integer,
982double, pointer, or reference. Perl will automatically convert the
983actual scalar data from the stored type into the requested type.
984
985Some scalar variables contain more than one type of scalar data. For
986example, the variable C<$!> contains either the numeric value of C<errno>
987or its string equivalent from either C<strerror> or C<sys_errlist[]>.
988
989To force multiple data values into an SV, you must do two things: use the
990C<sv_set*v> routines to add the additional scalar type, then set a flag
991so that Perl will believe it contains more than one type of data. The
992four macros to set the flags are:
993
994 SvIOK_on
995 SvNOK_on
996 SvPOK_on
997 SvROK_on
998
999The particular macro you must use depends on which C<sv_set*v> routine
1000you called first. This is because every C<sv_set*v> routine turns on
1001only the bit for the particular type of data being set, and turns off
1002all the rest.
1003
1004For example, to create a new Perl variable called "dberror" that contains
1005both the numeric and descriptive string error values, you could use the
1006following code:
1007
1008 extern int dberror;
1009 extern char *dberror_list;
1010
64ace3f8 1011 SV* sv = get_sv("dberror", GV_ADD);
0a753a76 1012 sv_setiv(sv, (IV) dberror);
1013 sv_setpv(sv, dberror_list[dberror]);
1014 SvIOK_on(sv);
1015
1016If the order of C<sv_setiv> and C<sv_setpv> had been reversed, then the
1017macro C<SvPOK_on> would need to be called instead of C<SvIOK_on>.
1018
4f4531b8
FC
1019=head2 Read-Only Values
1020
1021In Perl 5.16 and earlier, copy-on-write (see the next section) shared a
1022flag bit with read-only scalars. So the only way to test whether
1023C<sv_setsv>, etc., will raise a "Modification of a read-only value" error
1024in those versions is:
1025
1026 SvREADONLY(sv) && !SvIsCOW(sv)
1027
1028Under Perl 5.18 and later, SvREADONLY only applies to read-only variables,
1029and, under 5.20, copy-on-write scalars can also be read-only, so the above
1030check is incorrect. You just want:
1031
1032 SvREADONLY(sv)
1033
1034If you need to do this check often, define your own macro like this:
1035
1036 #if PERL_VERSION >= 18
1037 # define SvTRULYREADONLY(sv) SvREADONLY(sv)
1038 #else
1039 # define SvTRULYREADONLY(sv) (SvREADONLY(sv) && !SvIsCOW(sv))
1040 #endif
1041
1042=head2 Copy on Write
1043
1044Perl implements a copy-on-write (COW) mechanism for scalars, in which
1045string copies are not immediately made when requested, but are deferred
1046until made necessary by one or the other scalar changing. This is mostly
1047transparent, but one must take care not to modify string buffers that are
1048shared by multiple SVs.
1049
1050You can test whether an SV is using copy-on-write with C<SvIsCOW(sv)>.
1051
1052You can force an SV to make its own copy of its string buffer by calling C<sv_force_normal(sv)> or SvPV_force_nolen(sv).
1053
1054If you want to make the SV drop its string buffer, use
1055C<sv_force_normal_flags(sv, SV_COW_DROP_PV)> or simply
1056C<sv_setsv(sv, NULL)>.
1057
1058All of these functions will croak on read-only scalars (see the previous
1059section for more on those).
1060
1061To test that your code is behaving correctly and not modifying COW buffers,
1062on systems that support L<mmap(2)> (i.e., Unix) you can configure perl with
1063C<-Accflags=-DPERL_DEBUG_READONLY_COW> and it will turn buffer violations
1064into crashes. You will find it to be marvellously slow, so you may want to
1065skip perl's own tests.
1066
0a753a76 1067=head2 Magic Variables
a0d0e21e 1068
d1b91892
AD
1069[This section still under construction. Ignore everything here. Post no
1070bills. Everything not permitted is forbidden.]
1071
d1b91892
AD
1072Any SV may be magical, that is, it has special features that a normal
1073SV does not have. These features are stored in the SV structure in a
5f05dabc 1074linked list of C<struct magic>'s, typedef'ed to C<MAGIC>.
d1b91892
AD
1075
1076 struct magic {
1077 MAGIC* mg_moremagic;
1078 MGVTBL* mg_virtual;
1079 U16 mg_private;
1080 char mg_type;
1081 U8 mg_flags;
b205eb13 1082 I32 mg_len;
d1b91892
AD
1083 SV* mg_obj;
1084 char* mg_ptr;
d1b91892
AD
1085 };
1086
1087Note this is current as of patchlevel 0, and could change at any time.
1088
1089=head2 Assigning Magic
1090
1091Perl adds magic to an SV using the sv_magic function:
1092
a9b0660e 1093 void sv_magic(SV* sv, SV* obj, int how, const char* name, I32 namlen);
d1b91892
AD
1094
1095The C<sv> argument is a pointer to the SV that is to acquire a new magical
1096feature.
1097
1098If C<sv> is not already magical, Perl uses the C<SvUPGRADE> macro to
10e2eb10
FC
1099convert C<sv> to type C<SVt_PVMG>.
1100Perl then continues by adding new magic
645c22ef
DM
1101to the beginning of the linked list of magical features. Any prior entry
1102of the same type of magic is deleted. Note that this can be overridden,
1103and multiple instances of the same type of magic can be associated with an
1104SV.
d1b91892 1105
54310121 1106The C<name> and C<namlen> arguments are used to associate a string with
10e2eb10 1107the magic, typically the name of a variable. C<namlen> is stored in the
2d8d5d5a
SH
1108C<mg_len> field and if C<name> is non-null then either a C<savepvn> copy of
1109C<name> or C<name> itself is stored in the C<mg_ptr> field, depending on
1110whether C<namlen> is greater than zero or equal to zero respectively. As a
1111special case, if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
1112to contain an C<SV*> and is stored as-is with its REFCNT incremented.
d1b91892
AD
1113
1114The sv_magic function uses C<how> to determine which, if any, predefined
1115"Magic Virtual Table" should be assigned to the C<mg_virtual> field.
5a0de581 1116See the L</Magic Virtual Tables> section below. The C<how> argument is also
10e2eb10
FC
1117stored in the C<mg_type> field. The value of
1118C<how> should be chosen from the set of macros
1119C<PERL_MAGIC_foo> found in F<perl.h>. Note that before
645c22ef 1120these macros were added, Perl internals used to directly use character
14befaf4 1121literals, so you may occasionally come across old code or documentation
75d0f26d 1122referring to 'U' magic rather than C<PERL_MAGIC_uvar> for example.
d1b91892
AD
1123
1124The C<obj> argument is stored in the C<mg_obj> field of the C<MAGIC>
1125structure. If it is not the same as the C<sv> argument, the reference
1126count of the C<obj> object is incremented. If it is the same, or if
27deb0cf
YO
1127the C<how> argument is C<PERL_MAGIC_arylen>, C<PERL_MAGIC_regdatum>,
1128C<PERL_MAGIC_regdata>, or if it is a NULL pointer, then C<obj> is merely
1129stored, without the reference count being incremented.
d1b91892 1130
2d8d5d5a
SH
1131See also C<sv_magicext> in L<perlapi> for a more flexible way to add magic
1132to an SV.
1133
cb1a09d0
AD
1134There is also a function to add magic to an C<HV>:
1135
1136 void hv_magic(HV *hv, GV *gv, int how);
1137
1138This simply calls C<sv_magic> and coerces the C<gv> argument into an C<SV>.
1139
1140To remove the magic from an SV, call the function sv_unmagic:
1141
70a53b35 1142 int sv_unmagic(SV *sv, int type);
cb1a09d0
AD
1143
1144The C<type> argument should be equal to the C<how> value when the C<SV>
1145was initially made magical.
1146
f6ee7b17 1147However, note that C<sv_unmagic> removes all magic of a certain C<type> from the
10e2eb10
FC
1148C<SV>. If you want to remove only certain
1149magic of a C<type> based on the magic
f6ee7b17
FR
1150virtual table, use C<sv_unmagicext> instead:
1151
1152 int sv_unmagicext(SV *sv, int type, MGVTBL *vtbl);
1153
d1b91892
AD
1154=head2 Magic Virtual Tables
1155
d1be9408 1156The C<mg_virtual> field in the C<MAGIC> structure is a pointer to an
d1b91892
AD
1157C<MGVTBL>, which is a structure of function pointers and stands for
1158"Magic Virtual Table" to handle the various operations that might be
1159applied to that variable.
1160
301cb7e8
DM
1161The C<MGVTBL> has five (or sometimes eight) pointers to the following
1162routine types:
d1b91892 1163
e97ca230
DM
1164 int (*svt_get) (pTHX_ SV* sv, MAGIC* mg);
1165 int (*svt_set) (pTHX_ SV* sv, MAGIC* mg);
1166 U32 (*svt_len) (pTHX_ SV* sv, MAGIC* mg);
1167 int (*svt_clear)(pTHX_ SV* sv, MAGIC* mg);
1168 int (*svt_free) (pTHX_ SV* sv, MAGIC* mg);
d1b91892 1169
e97ca230 1170 int (*svt_copy) (pTHX_ SV *sv, MAGIC* mg, SV *nsv,
a9b0660e 1171 const char *name, I32 namlen);
e97ca230
DM
1172 int (*svt_dup) (pTHX_ MAGIC *mg, CLONE_PARAMS *param);
1173 int (*svt_local)(pTHX_ SV *nsv, MAGIC *mg);
301cb7e8
DM
1174
1175
06f6df17 1176This MGVTBL structure is set at compile-time in F<perl.h> and there are
b7a0f54c
S
1177currently 32 types. These different structures contain pointers to various
1178routines that perform additional actions depending on which function is
1179being called.
d1b91892 1180
a9b0660e
KW
1181 Function pointer Action taken
1182 ---------------- ------------
1183 svt_get Do something before the value of the SV is
1184 retrieved.
1185 svt_set Do something after the SV is assigned a value.
1186 svt_len Report on the SV's length.
1187 svt_clear Clear something the SV represents.
1188 svt_free Free any extra storage associated with the SV.
d1b91892 1189
a9b0660e
KW
1190 svt_copy copy tied variable magic to a tied element
1191 svt_dup duplicate a magic structure during thread cloning
1192 svt_local copy magic to local value during 'local'
301cb7e8 1193
d1b91892 1194For instance, the MGVTBL structure called C<vtbl_sv> (which corresponds
14befaf4 1195to an C<mg_type> of C<PERL_MAGIC_sv>) contains:
d1b91892
AD
1196
1197 { magic_get, magic_set, magic_len, 0, 0 }
1198
14befaf4
DM
1199Thus, when an SV is determined to be magical and of type C<PERL_MAGIC_sv>,
1200if a get operation is being performed, the routine C<magic_get> is
1201called. All the various routines for the various magical types begin
1202with C<magic_>. NOTE: the magic routines are not considered part of
1203the Perl API, and may not be exported by the Perl library.
d1b91892 1204
301cb7e8
DM
1205The last three slots are a recent addition, and for source code
1206compatibility they are only checked for if one of the three flags
10e2eb10
FC
1207MGf_COPY, MGf_DUP or MGf_LOCAL is set in mg_flags.
1208This means that most code can continue declaring
1209a vtable as a 5-element value. These three are
301cb7e8
DM
1210currently used exclusively by the threading code, and are highly subject
1211to change.
1212
d1b91892
AD
1213The current kinds of Magic Virtual Tables are:
1214
f1f5ddd7
FC
1215=for comment
1216This table is generated by regen/mg_vtable.pl. Any changes made here
1217will be lost.
1218
1219=for mg_vtable.pl begin
1220
a9b0660e 1221 mg_type
bd6e6c12
FC
1222 (old-style char and macro) MGVTBL Type of magic
1223 -------------------------- ------ -------------
1224 \0 PERL_MAGIC_sv vtbl_sv Special scalar variable
1225 # PERL_MAGIC_arylen vtbl_arylen Array length ($#ary)
e5e1ee61 1226 % PERL_MAGIC_rhash (none) Extra data for restricted
bd6e6c12 1227 hashes
a6d69523
TC
1228 * PERL_MAGIC_debugvar vtbl_debugvar $DB::single, signal, trace
1229 vars
bd6e6c12 1230 . PERL_MAGIC_pos vtbl_pos pos() lvalue
e5e1ee61 1231 : PERL_MAGIC_symtab (none) Extra data for symbol
bd6e6c12 1232 tables
e5e1ee61
FC
1233 < PERL_MAGIC_backref vtbl_backref For weak ref data
1234 @ PERL_MAGIC_arylen_p (none) To move arylen out of XPVAV
bd6e6c12
FC
1235 B PERL_MAGIC_bm vtbl_regexp Boyer-Moore
1236 (fast string search)
1237 c PERL_MAGIC_overload_table vtbl_ovrld Holds overload table
1238 (AMT) on stash
1239 D PERL_MAGIC_regdata vtbl_regdata Regex match position data
1240 (@+ and @- vars)
1241 d PERL_MAGIC_regdatum vtbl_regdatum Regex match position data
1242 element
1243 E PERL_MAGIC_env vtbl_env %ENV hash
1244 e PERL_MAGIC_envelem vtbl_envelem %ENV hash element
eccba044 1245 f PERL_MAGIC_fm vtbl_regexp Formline
bd6e6c12 1246 ('compiled' format)
bd6e6c12
FC
1247 g PERL_MAGIC_regex_global vtbl_mglob m//g target
1248 H PERL_MAGIC_hints vtbl_hints %^H hash
1249 h PERL_MAGIC_hintselem vtbl_hintselem %^H hash element
1250 I PERL_MAGIC_isa vtbl_isa @ISA array
1251 i PERL_MAGIC_isaelem vtbl_isaelem @ISA array element
1252 k PERL_MAGIC_nkeys vtbl_nkeys scalar(keys()) lvalue
1253 L PERL_MAGIC_dbfile (none) Debugger %_<filename
1254 l PERL_MAGIC_dbline vtbl_dbline Debugger %_<filename
1255 element
1256 N PERL_MAGIC_shared (none) Shared between threads
1257 n PERL_MAGIC_shared_scalar (none) Shared between threads
1258 o PERL_MAGIC_collxfrm vtbl_collxfrm Locale transformation
1259 P PERL_MAGIC_tied vtbl_pack Tied array or hash
1260 p PERL_MAGIC_tiedelem vtbl_packelem Tied array or hash element
1261 q PERL_MAGIC_tiedscalar vtbl_packelem Tied scalar or handle
e5e1ee61 1262 r PERL_MAGIC_qr vtbl_regexp Precompiled qr// regex
bd6e6c12
FC
1263 S PERL_MAGIC_sig (none) %SIG hash
1264 s PERL_MAGIC_sigelem vtbl_sigelem %SIG hash element
1265 t PERL_MAGIC_taint vtbl_taint Taintedness
1266 U PERL_MAGIC_uvar vtbl_uvar Available for use by
1267 extensions
1268 u PERL_MAGIC_uvar_elem (none) Reserved for use by
1269 extensions
4499db73 1270 V PERL_MAGIC_vstring (none) SV was vstring literal
bd6e6c12
FC
1271 v PERL_MAGIC_vec vtbl_vec vec() lvalue
1272 w PERL_MAGIC_utf8 vtbl_utf8 Cached UTF-8 information
1273 x PERL_MAGIC_substr vtbl_substr substr() lvalue
1274 y PERL_MAGIC_defelem vtbl_defelem Shadow "foreach" iterator
1275 variable / smart parameter
1276 vivification
baabe3fb
FC
1277 \ PERL_MAGIC_lvref vtbl_lvref Lvalue reference
1278 constructor
e5e1ee61 1279 ] PERL_MAGIC_checkcall vtbl_checkcall Inlining/mutation of call
bd6e6c12
FC
1280 to this CV
1281 ~ PERL_MAGIC_ext (none) Available for use by
1282 extensions
0cbee0a4 1283
f1f5ddd7 1284=for mg_vtable.pl end
d1b91892 1285
68dc0745 1286When an uppercase and lowercase letter both exist in the table, then the
92f0c265
JP
1287uppercase letter is typically used to represent some kind of composite type
1288(a list or a hash), and the lowercase letter is used to represent an element
10e2eb10 1289of that composite type. Some internals code makes use of this case
92f0c265 1290relationship. However, 'v' and 'V' (vec and v-string) are in no way related.
14befaf4
DM
1291
1292The C<PERL_MAGIC_ext> and C<PERL_MAGIC_uvar> magic types are defined
1293specifically for use by extensions and will not be used by perl itself.
1294Extensions can use C<PERL_MAGIC_ext> magic to 'attach' private information
1295to variables (typically objects). This is especially useful because
1296there is no way for normal perl code to corrupt this private information
1297(unlike using extra elements of a hash object).
1298
1299Similarly, C<PERL_MAGIC_uvar> magic can be used much like tie() to call a
1300C function any time a scalar's value is used or changed. The C<MAGIC>'s
bdbeb323
SM
1301C<mg_ptr> field points to a C<ufuncs> structure:
1302
1303 struct ufuncs {
a9402793
AB
1304 I32 (*uf_val)(pTHX_ IV, SV*);
1305 I32 (*uf_set)(pTHX_ IV, SV*);
bdbeb323
SM
1306 IV uf_index;
1307 };
1308
1309When the SV is read from or written to, the C<uf_val> or C<uf_set>
14befaf4
DM
1310function will be called with C<uf_index> as the first arg and a pointer to
1311the SV as the second. A simple example of how to add C<PERL_MAGIC_uvar>
1526ead6
AB
1312magic is shown below. Note that the ufuncs structure is copied by
1313sv_magic, so you can safely allocate it on the stack.
1314
1315 void
1316 Umagic(sv)
1317 SV *sv;
1318 PREINIT:
1319 struct ufuncs uf;
1320 CODE:
1321 uf.uf_val = &my_get_fn;
1322 uf.uf_set = &my_set_fn;
1323 uf.uf_index = 0;
14befaf4 1324 sv_magic(sv, 0, PERL_MAGIC_uvar, (char*)&uf, sizeof(uf));
5f05dabc 1325
1e73acc8
AS
1326Attaching C<PERL_MAGIC_uvar> to arrays is permissible but has no effect.
1327
1328For hashes there is a specialized hook that gives control over hash
1329keys (but not values). This hook calls C<PERL_MAGIC_uvar> 'get' magic
1330if the "set" function in the C<ufuncs> structure is NULL. The hook
1331is activated whenever the hash is accessed with a key specified as
1332an C<SV> through the functions C<hv_store_ent>, C<hv_fetch_ent>,
1333C<hv_delete_ent>, and C<hv_exists_ent>. Accessing the key as a string
1334through the functions without the C<..._ent> suffix circumvents the
4509d391 1335hook. See L<Hash::Util::FieldHash/GUTS> for a detailed description.
1e73acc8 1336
14befaf4
DM
1337Note that because multiple extensions may be using C<PERL_MAGIC_ext>
1338or C<PERL_MAGIC_uvar> magic, it is important for extensions to take
1339extra care to avoid conflict. Typically only using the magic on
1340objects blessed into the same class as the extension is sufficient.
2f07f21a
FR
1341For C<PERL_MAGIC_ext> magic, it is usually a good idea to define an
1342C<MGVTBL>, even if all its fields will be C<0>, so that individual
1343C<MAGIC> pointers can be identified as a particular kind of magic
10e2eb10 1344using their magic virtual table. C<mg_findext> provides an easy way
f6ee7b17 1345to do that:
2f07f21a
FR
1346
1347 STATIC MGVTBL my_vtbl = { 0, 0, 0, 0, 0, 0, 0, 0 };
1348
1349 MAGIC *mg;
f6ee7b17
FR
1350 if ((mg = mg_findext(sv, PERL_MAGIC_ext, &my_vtbl))) {
1351 /* this is really ours, not another module's PERL_MAGIC_ext */
1352 my_priv_data_t *priv = (my_priv_data_t *)mg->mg_ptr;
1353 ...
2f07f21a 1354 }
5f05dabc 1355
ef50df4b
GS
1356Also note that the C<sv_set*()> and C<sv_cat*()> functions described
1357earlier do B<not> invoke 'set' magic on their targets. This must
1358be done by the user either by calling the C<SvSETMAGIC()> macro after
1359calling these functions, or by using one of the C<sv_set*_mg()> or
1360C<sv_cat*_mg()> functions. Similarly, generic C code must call the
1361C<SvGETMAGIC()> macro to invoke any 'get' magic if they use an SV
1362obtained from external sources in functions that don't handle magic.
4a4eefd0 1363See L<perlapi> for a description of these functions.
189b2af5
GS
1364For example, calls to the C<sv_cat*()> functions typically need to be
1365followed by C<SvSETMAGIC()>, but they don't need a prior C<SvGETMAGIC()>
1366since their implementation handles 'get' magic.
1367
d1b91892
AD
1368=head2 Finding Magic
1369
a9b0660e
KW
1370 MAGIC *mg_find(SV *sv, int type); /* Finds the magic pointer of that
1371 * type */
f6ee7b17
FR
1372
1373This routine returns a pointer to a C<MAGIC> structure stored in the SV.
10e2eb10
FC
1374If the SV does not have that magical
1375feature, C<NULL> is returned. If the
f6ee7b17 1376SV has multiple instances of that magical feature, the first one will be
10e2eb10
FC
1377returned. C<mg_findext> can be used
1378to find a C<MAGIC> structure of an SV
da8c5729 1379based on both its magic type and its magic virtual table:
f6ee7b17
FR
1380
1381 MAGIC *mg_findext(SV *sv, int type, MGVTBL *vtbl);
d1b91892 1382
f6ee7b17
FR
1383Also, if the SV passed to C<mg_find> or C<mg_findext> is not of type
1384SVt_PVMG, Perl may core dump.
d1b91892 1385
08105a92 1386 int mg_copy(SV* sv, SV* nsv, const char* key, STRLEN klen);
d1b91892
AD
1387
1388This routine checks to see what types of magic C<sv> has. If the mg_type
68dc0745 1389field is an uppercase letter, then the mg_obj is copied to C<nsv>, but
1390the mg_type field is changed to be the lowercase letter.
a0d0e21e 1391
04343c6d
GS
1392=head2 Understanding the Magic of Tied Hashes and Arrays
1393
14befaf4
DM
1394Tied hashes and arrays are magical beasts of the C<PERL_MAGIC_tied>
1395magic type.
9edb2b46
GS
1396
1397WARNING: As of the 5.004 release, proper usage of the array and hash
1398access functions requires understanding a few caveats. Some
1399of these caveats are actually considered bugs in the API, to be fixed
10e2eb10 1400in later releases, and are bracketed with [MAYCHANGE] below. If
9edb2b46
GS
1401you find yourself actually applying such information in this section, be
1402aware that the behavior may change in the future, umm, without warning.
04343c6d 1403
1526ead6 1404The perl tie function associates a variable with an object that implements
9a68f1db 1405the various GET, SET, etc methods. To perform the equivalent of the perl
1526ead6 1406tie function from an XSUB, you must mimic this behaviour. The code below
61ad4b94 1407carries out the necessary steps -- firstly it creates a new hash, and then
1526ead6 1408creates a second hash which it blesses into the class which will implement
10e2eb10 1409the tie methods. Lastly it ties the two hashes together, and returns a
1526ead6
AB
1410reference to the new tied hash. Note that the code below does NOT call the
1411TIEHASH method in the MyTie class -
5a0de581 1412see L</Calling Perl Routines from within C Programs> for details on how
1526ead6
AB
1413to do this.
1414
1415 SV*
1416 mytie()
1417 PREINIT:
1418 HV *hash;
1419 HV *stash;
1420 SV *tie;
1421 CODE:
1422 hash = newHV();
1423 tie = newRV_noinc((SV*)newHV());
da51bb9b 1424 stash = gv_stashpv("MyTie", GV_ADD);
1526ead6 1425 sv_bless(tie, stash);
899e16d0 1426 hv_magic(hash, (GV*)tie, PERL_MAGIC_tied);
1526ead6
AB
1427 RETVAL = newRV_noinc(hash);
1428 OUTPUT:
1429 RETVAL
1430
04343c6d
GS
1431The C<av_store> function, when given a tied array argument, merely
1432copies the magic of the array onto the value to be "stored", using
1433C<mg_copy>. It may also return NULL, indicating that the value did not
9edb2b46
GS
1434actually need to be stored in the array. [MAYCHANGE] After a call to
1435C<av_store> on a tied array, the caller will usually need to call
1436C<mg_set(val)> to actually invoke the perl level "STORE" method on the
1437TIEARRAY object. If C<av_store> did return NULL, a call to
1438C<SvREFCNT_dec(val)> will also be usually necessary to avoid a memory
1439leak. [/MAYCHANGE]
04343c6d
GS
1440
1441The previous paragraph is applicable verbatim to tied hash access using the
1442C<hv_store> and C<hv_store_ent> functions as well.
1443
1444C<av_fetch> and the corresponding hash functions C<hv_fetch> and
1445C<hv_fetch_ent> actually return an undefined mortal value whose magic
1446has been initialized using C<mg_copy>. Note the value so returned does not
9edb2b46
GS
1447need to be deallocated, as it is already mortal. [MAYCHANGE] But you will
1448need to call C<mg_get()> on the returned value in order to actually invoke
1449the perl level "FETCH" method on the underlying TIE object. Similarly,
04343c6d
GS
1450you may also call C<mg_set()> on the return value after possibly assigning
1451a suitable value to it using C<sv_setsv>, which will invoke the "STORE"
9edb2b46 1452method on the TIE object. [/MAYCHANGE]
04343c6d 1453
9edb2b46 1454[MAYCHANGE]
04343c6d
GS
1455In other words, the array or hash fetch/store functions don't really
1456fetch and store actual values in the case of tied arrays and hashes. They
1457merely call C<mg_copy> to attach magic to the values that were meant to be
1458"stored" or "fetched". Later calls to C<mg_get> and C<mg_set> actually
1459do the job of invoking the TIE methods on the underlying objects. Thus
9edb2b46 1460the magic mechanism currently implements a kind of lazy access to arrays
04343c6d
GS
1461and hashes.
1462
1463Currently (as of perl version 5.004), use of the hash and array access
1464functions requires the user to be aware of whether they are operating on
9edb2b46
GS
1465"normal" hashes and arrays, or on their tied variants. The API may be
1466changed to provide more transparent access to both tied and normal data
1467types in future versions.
1468[/MAYCHANGE]
04343c6d
GS
1469
1470You would do well to understand that the TIEARRAY and TIEHASH interfaces
1471are mere sugar to invoke some perl method calls while using the uniform hash
1472and array syntax. The use of this sugar imposes some overhead (typically
1473about two to four extra opcodes per FETCH/STORE operation, in addition to
1474the creation of all the mortal variables required to invoke the methods).
1475This overhead will be comparatively small if the TIE methods are themselves
1476substantial, but if they are only a few statements long, the overhead
1477will not be insignificant.
1478
d1c897a1
IZ
1479=head2 Localizing changes
1480
1481Perl has a very handy construction
1482
1483 {
1484 local $var = 2;
1485 ...
1486 }
1487
1488This construction is I<approximately> equivalent to
1489
1490 {
1491 my $oldvar = $var;
1492 $var = 2;
1493 ...
1494 $var = $oldvar;
1495 }
1496
1497The biggest difference is that the first construction would
1498reinstate the initial value of $var, irrespective of how control exits
10e2eb10 1499the block: C<goto>, C<return>, C<die>/C<eval>, etc. It is a little bit
d1c897a1
IZ
1500more efficient as well.
1501
1502There is a way to achieve a similar task from C via Perl API: create a
1503I<pseudo-block>, and arrange for some changes to be automatically
1504undone at the end of it, either explicit, or via a non-local exit (via
10e2eb10 1505die()). A I<block>-like construct is created by a pair of
b687b08b
TC
1506C<ENTER>/C<LEAVE> macros (see L<perlcall/"Returning a Scalar">).
1507Such a construct may be created specially for some important localized
1508task, or an existing one (like boundaries of enclosing Perl
1509subroutine/block, or an existing pair for freeing TMPs) may be
10e2eb10
FC
1510used. (In the second case the overhead of additional localization must
1511be almost negligible.) Note that any XSUB is automatically enclosed in
b687b08b 1512an C<ENTER>/C<LEAVE> pair.
d1c897a1
IZ
1513
1514Inside such a I<pseudo-block> the following service is available:
1515
13a2d996 1516=over 4
d1c897a1
IZ
1517
1518=item C<SAVEINT(int i)>
1519
1520=item C<SAVEIV(IV i)>
1521
1522=item C<SAVEI32(I32 i)>
1523
1524=item C<SAVELONG(long i)>
1525
1526These macros arrange things to restore the value of integer variable
1527C<i> at the end of enclosing I<pseudo-block>.
1528
1529=item C<SAVESPTR(s)>
1530
1531=item C<SAVEPPTR(p)>
1532
1533These macros arrange things to restore the value of pointers C<s> and
10e2eb10 1534C<p>. C<s> must be a pointer of a type which survives conversion to
d1c897a1
IZ
1535C<SV*> and back, C<p> should be able to survive conversion to C<char*>
1536and back.
1537
1538=item C<SAVEFREESV(SV *sv)>
1539
06f1e0b6 1540The refcount of C<sv> will be decremented at the end of
26d9b02f
JH
1541I<pseudo-block>. This is similar to C<sv_2mortal> in that it is also a
1542mechanism for doing a delayed C<SvREFCNT_dec>. However, while C<sv_2mortal>
1543extends the lifetime of C<sv> until the beginning of the next statement,
1544C<SAVEFREESV> extends it until the end of the enclosing scope. These
1545lifetimes can be wildly different.
1546
1547Also compare C<SAVEMORTALIZESV>.
1548
1549=item C<SAVEMORTALIZESV(SV *sv)>
1550
1551Just like C<SAVEFREESV>, but mortalizes C<sv> at the end of the current
1552scope instead of decrementing its reference count. This usually has the
1553effect of keeping C<sv> alive until the statement that called the currently
1554live scope has finished executing.
d1c897a1
IZ
1555
1556=item C<SAVEFREEOP(OP *op)>
1557
1558The C<OP *> is op_free()ed at the end of I<pseudo-block>.
1559
1560=item C<SAVEFREEPV(p)>
1561
1562The chunk of memory which is pointed to by C<p> is Safefree()ed at the
1563end of I<pseudo-block>.
1564
1565=item C<SAVECLEARSV(SV *sv)>
1566
1567Clears a slot in the current scratchpad which corresponds to C<sv> at
1568the end of I<pseudo-block>.
1569
1570=item C<SAVEDELETE(HV *hv, char *key, I32 length)>
1571
10e2eb10 1572The key C<key> of C<hv> is deleted at the end of I<pseudo-block>. The
d1c897a1
IZ
1573string pointed to by C<key> is Safefree()ed. If one has a I<key> in
1574short-lived storage, the corresponding string may be reallocated like
1575this:
1576
9cde0e7f 1577 SAVEDELETE(PL_defstash, savepv(tmpbuf), strlen(tmpbuf));
d1c897a1 1578
c76ac1ee 1579=item C<SAVEDESTRUCTOR(DESTRUCTORFUNC_NOCONTEXT_t f, void *p)>
d1c897a1
IZ
1580
1581At the end of I<pseudo-block> the function C<f> is called with the
c76ac1ee
GS
1582only argument C<p>.
1583
1584=item C<SAVEDESTRUCTOR_X(DESTRUCTORFUNC_t f, void *p)>
1585
1586At the end of I<pseudo-block> the function C<f> is called with the
1587implicit context argument (if any), and C<p>.
d1c897a1
IZ
1588
1589=item C<SAVESTACK_POS()>
1590
1591The current offset on the Perl internal stack (cf. C<SP>) is restored
1592at the end of I<pseudo-block>.
1593
1594=back
1595
1596The following API list contains functions, thus one needs to
1597provide pointers to the modifiable data explicitly (either C pointers,
00aadd71 1598or Perlish C<GV *>s). Where the above macros take C<int>, a similar
d1c897a1
IZ
1599function takes C<int *>.
1600
13a2d996 1601=over 4
d1c897a1
IZ
1602
1603=item C<SV* save_scalar(GV *gv)>
1604
1605Equivalent to Perl code C<local $gv>.
1606
1607=item C<AV* save_ary(GV *gv)>
1608
1609=item C<HV* save_hash(GV *gv)>
1610
1611Similar to C<save_scalar>, but localize C<@gv> and C<%gv>.
1612
1613=item C<void save_item(SV *item)>
1614
1615Duplicates the current value of C<SV>, on the exit from the current
1616C<ENTER>/C<LEAVE> I<pseudo-block> will restore the value of C<SV>
10e2eb10 1617using the stored value. It doesn't handle magic. Use C<save_scalar> if
038fcae3 1618magic is affected.
d1c897a1
IZ
1619
1620=item C<void save_list(SV **sarg, I32 maxsarg)>
1621
1622A variant of C<save_item> which takes multiple arguments via an array
1623C<sarg> of C<SV*> of length C<maxsarg>.
1624
1625=item C<SV* save_svref(SV **sptr)>
1626
d1be9408 1627Similar to C<save_scalar>, but will reinstate an C<SV *>.
d1c897a1
IZ
1628
1629=item C<void save_aptr(AV **aptr)>
1630
1631=item C<void save_hptr(HV **hptr)>
1632
1633Similar to C<save_svref>, but localize C<AV *> and C<HV *>.
1634
1635=back
1636
1637The C<Alias> module implements localization of the basic types within the
1638I<caller's scope>. People who are interested in how to localize things in
1639the containing scope should take a look there too.
1640
0a753a76 1641=head1 Subroutines
a0d0e21e 1642
68dc0745 1643=head2 XSUBs and the Argument Stack
5f05dabc 1644
1645The XSUB mechanism is a simple way for Perl programs to access C subroutines.
1646An XSUB routine will have a stack that contains the arguments from the Perl
1647program, and a way to map from the Perl data structures to a C equivalent.
1648
1649The stack arguments are accessible through the C<ST(n)> macro, which returns
1650the C<n>'th stack argument. Argument 0 is the first argument passed in the
1651Perl subroutine call. These arguments are C<SV*>, and can be used anywhere
1652an C<SV*> is used.
1653
1654Most of the time, output from the C routine can be handled through use of
1655the RETVAL and OUTPUT directives. However, there are some cases where the
1656argument stack is not already long enough to handle all the return values.
1657An example is the POSIX tzname() call, which takes no arguments, but returns
1658two, the local time zone's standard and summer time abbreviations.
1659
1660To handle this situation, the PPCODE directive is used and the stack is
1661extended using the macro:
1662
924508f0 1663 EXTEND(SP, num);
5f05dabc 1664
924508f0
GS
1665where C<SP> is the macro that represents the local copy of the stack pointer,
1666and C<num> is the number of elements the stack should be extended by.
5f05dabc 1667
00aadd71 1668Now that there is room on the stack, values can be pushed on it using C<PUSHs>
10e2eb10 1669macro. The pushed values will often need to be "mortal" (See
d82b684c 1670L</Reference Counts and Mortality>):
5f05dabc 1671
00aadd71 1672 PUSHs(sv_2mortal(newSViv(an_integer)))
d82b684c
SH
1673 PUSHs(sv_2mortal(newSVuv(an_unsigned_integer)))
1674 PUSHs(sv_2mortal(newSVnv(a_double)))
00aadd71 1675 PUSHs(sv_2mortal(newSVpv("Some String",0)))
a9b0660e
KW
1676 /* Although the last example is better written as the more
1677 * efficient: */
a3179684 1678 PUSHs(newSVpvs_flags("Some String", SVs_TEMP))
5f05dabc 1679
1680And now the Perl program calling C<tzname>, the two values will be assigned
1681as in:
1682
1683 ($standard_abbrev, $summer_abbrev) = POSIX::tzname;
1684
1685An alternate (and possibly simpler) method to pushing values on the stack is
00aadd71 1686to use the macro:
5f05dabc 1687
5f05dabc 1688 XPUSHs(SV*)
1689
da8c5729 1690This macro automatically adjusts the stack for you, if needed. Thus, you
5f05dabc 1691do not need to call C<EXTEND> to extend the stack.
00aadd71
NIS
1692
1693Despite their suggestions in earlier versions of this document the macros
d82b684c
SH
1694C<(X)PUSH[iunp]> are I<not> suited to XSUBs which return multiple results.
1695For that, either stick to the C<(X)PUSHs> macros shown above, or use the new
1696C<m(X)PUSH[iunp]> macros instead; see L</Putting a C value on Perl stack>.
5f05dabc 1697
1698For more information, consult L<perlxs> and L<perlxstut>.
1699
5b36e945
FC
1700=head2 Autoloading with XSUBs
1701
1702If an AUTOLOAD routine is an XSUB, as with Perl subroutines, Perl puts the
1703fully-qualified name of the autoloaded subroutine in the $AUTOLOAD variable
1704of the XSUB's package.
1705
1706But it also puts the same information in certain fields of the XSUB itself:
1707
1708 HV *stash = CvSTASH(cv);
1709 const char *subname = SvPVX(cv);
1710 STRLEN name_length = SvCUR(cv); /* in bytes */
1711 U32 is_utf8 = SvUTF8(cv);
f703fc96 1712
5b36e945 1713C<SvPVX(cv)> contains just the sub name itself, not including the package.
d8893903
FC
1714For an AUTOLOAD routine in UNIVERSAL or one of its superclasses,
1715C<CvSTASH(cv)> returns NULL during a method call on a nonexistent package.
5b36e945
FC
1716
1717B<Note>: Setting $AUTOLOAD stopped working in 5.6.1, which did not support
1718XS AUTOLOAD subs at all. Perl 5.8.0 introduced the use of fields in the
1719XSUB itself. Perl 5.16.0 restored the setting of $AUTOLOAD. If you need
1720to support 5.8-5.14, use the XSUB's fields.
1721
5f05dabc 1722=head2 Calling Perl Routines from within C Programs
a0d0e21e
LW
1723
1724There are four routines that can be used to call a Perl subroutine from
1725within a C program. These four are:
1726
954c1994
GS
1727 I32 call_sv(SV*, I32);
1728 I32 call_pv(const char*, I32);
1729 I32 call_method(const char*, I32);
5aaab254 1730 I32 call_argv(const char*, I32, char**);
a0d0e21e 1731
954c1994 1732The routine most often used is C<call_sv>. The C<SV*> argument
d1b91892
AD
1733contains either the name of the Perl subroutine to be called, or a
1734reference to the subroutine. The second argument consists of flags
1735that control the context in which the subroutine is called, whether
1736or not the subroutine is being passed arguments, how errors should be
1737trapped, and how to treat return values.
a0d0e21e
LW
1738
1739All four routines return the number of arguments that the subroutine returned
1740on the Perl stack.
1741
9a68f1db 1742These routines used to be called C<perl_call_sv>, etc., before Perl v5.6.0,
954c1994
GS
1743but those names are now deprecated; macros of the same name are provided for
1744compatibility.
1745
1746When using any of these routines (except C<call_argv>), the programmer
d1b91892
AD
1747must manipulate the Perl stack. These include the following macros and
1748functions:
a0d0e21e
LW
1749
1750 dSP
924508f0 1751 SP
a0d0e21e
LW
1752 PUSHMARK()
1753 PUTBACK
1754 SPAGAIN
1755 ENTER
1756 SAVETMPS
1757 FREETMPS
1758 LEAVE
1759 XPUSH*()
cb1a09d0 1760 POP*()
a0d0e21e 1761
5f05dabc 1762For a detailed description of calling conventions from C to Perl,
1763consult L<perlcall>.
a0d0e21e 1764
8ebc5c01 1765=head2 Putting a C value on Perl stack
ce3d39e2
IZ
1766
1767A lot of opcodes (this is an elementary operation in the internal perl
10e2eb10
FC
1768stack machine) put an SV* on the stack. However, as an optimization
1769the corresponding SV is (usually) not recreated each time. The opcodes
ce3d39e2
IZ
1770reuse specially assigned SVs (I<target>s) which are (as a corollary)
1771not constantly freed/created.
1772
0a753a76 1773Each of the targets is created only once (but see
5a0de581 1774L</Scratchpads and recursion> below), and when an opcode needs to put
ce3d39e2
IZ
1775an integer, a double, or a string on stack, it just sets the
1776corresponding parts of its I<target> and puts the I<target> on stack.
1777
1778The macro to put this target on stack is C<PUSHTARG>, and it is
1779directly used in some opcodes, as well as indirectly in zillions of
d82b684c 1780others, which use it via C<(X)PUSH[iunp]>.
ce3d39e2 1781
1bd1c0d5 1782Because the target is reused, you must be careful when pushing multiple
10e2eb10 1783values on the stack. The following code will not do what you think:
1bd1c0d5
SC
1784
1785 XPUSHi(10);
1786 XPUSHi(20);
1787
1788This translates as "set C<TARG> to 10, push a pointer to C<TARG> onto
1789the stack; set C<TARG> to 20, push a pointer to C<TARG> onto the stack".
1790At the end of the operation, the stack does not contain the values 10
1791and 20, but actually contains two pointers to C<TARG>, which we have set
d82b684c 1792to 20.
1bd1c0d5 1793
d82b684c
SH
1794If you need to push multiple different values then you should either use
1795the C<(X)PUSHs> macros, or else use the new C<m(X)PUSH[iunp]> macros,
1796none of which make use of C<TARG>. The C<(X)PUSHs> macros simply push an
1797SV* on the stack, which, as noted under L</XSUBs and the Argument Stack>,
1798will often need to be "mortal". The new C<m(X)PUSH[iunp]> macros make
1799this a little easier to achieve by creating a new mortal for you (via
1800C<(X)PUSHmortal>), pushing that onto the stack (extending it if necessary
1801in the case of the C<mXPUSH[iunp]> macros), and then setting its value.
1802Thus, instead of writing this to "fix" the example above:
1803
1804 XPUSHs(sv_2mortal(newSViv(10)))
1805 XPUSHs(sv_2mortal(newSViv(20)))
1806
1807you can simply write:
1808
1809 mXPUSHi(10)
1810 mXPUSHi(20)
1811
1812On a related note, if you do use C<(X)PUSH[iunp]>, then you're going to
1bd1c0d5 1813need a C<dTARG> in your variable declarations so that the C<*PUSH*>
d82b684c
SH
1814macros can make use of the local variable C<TARG>. See also C<dTARGET>
1815and C<dXSTARG>.
1bd1c0d5 1816
8ebc5c01 1817=head2 Scratchpads
ce3d39e2 1818
54310121 1819The question remains on when the SVs which are I<target>s for opcodes
10e2eb10 1820are created. The answer is that they are created when the current
ac036724 1821unit--a subroutine or a file (for opcodes for statements outside of
10e2eb10 1822subroutines)--is compiled. During this time a special anonymous Perl
ac036724 1823array is created, which is called a scratchpad for the current unit.
ce3d39e2 1824
54310121 1825A scratchpad keeps SVs which are lexicals for the current unit and are
d777b41a
FC
1826targets for opcodes. A previous version of this document
1827stated that one can deduce that an SV lives on a scratchpad
ce3d39e2 1828by looking on its flags: lexicals have C<SVs_PADMY> set, and
eee3e302 1829I<target>s have C<SVs_PADTMP> set. But this has never been fully true.
d777b41a
FC
1830C<SVs_PADMY> could be set on a variable that no longer resides in any pad.
1831While I<target>s do have C<SVs_PADTMP> set, it can also be set on variables
eed77337
FC
1832that have never resided in a pad, but nonetheless act like I<target>s. As
1833of perl 5.21.5, the C<SVs_PADMY> flag is no longer used and is defined as
18340. C<SvPADMY()> now returns true for anything without C<SVs_PADTMP>.
ce3d39e2 1835
10e2eb10 1836The correspondence between OPs and I<target>s is not 1-to-1. Different
54310121 1837OPs in the compile tree of the unit can use the same target, if this
ce3d39e2
IZ
1838would not conflict with the expected life of the temporary.
1839
2ae324a7 1840=head2 Scratchpads and recursion
ce3d39e2
IZ
1841
1842In fact it is not 100% true that a compiled unit contains a pointer to
10e2eb10
FC
1843the scratchpad AV. In fact it contains a pointer to an AV of
1844(initially) one element, and this element is the scratchpad AV. Why do
ce3d39e2
IZ
1845we need an extra level of indirection?
1846
10e2eb10 1847The answer is B<recursion>, and maybe B<threads>. Both
ce3d39e2 1848these can create several execution pointers going into the same
10e2eb10 1849subroutine. For the subroutine-child not write over the temporaries
ce3d39e2
IZ
1850for the subroutine-parent (lifespan of which covers the call to the
1851child), the parent and the child should have different
10e2eb10 1852scratchpads. (I<And> the lexicals should be separate anyway!)
ce3d39e2 1853
5f05dabc 1854So each subroutine is born with an array of scratchpads (of length 1).
1855On each entry to the subroutine it is checked that the current
ce3d39e2
IZ
1856depth of the recursion is not more than the length of this array, and
1857if it is, new scratchpad is created and pushed into the array.
1858
1859The I<target>s on this scratchpad are C<undef>s, but they are already
1860marked with correct flags.
1861
22d36020
FC
1862=head1 Memory Allocation
1863
1864=head2 Allocation
1865
1866All memory meant to be used with the Perl API functions should be manipulated
1867using the macros described in this section. The macros provide the necessary
1868transparency between differences in the actual malloc implementation that is
1869used within perl.
1870
1871It is suggested that you enable the version of malloc that is distributed
1872with Perl. It keeps pools of various sizes of unallocated memory in
1873order to satisfy allocation requests more quickly. However, on some
1874platforms, it may cause spurious malloc or free errors.
1875
1876The following three macros are used to initially allocate memory :
1877
1878 Newx(pointer, number, type);
1879 Newxc(pointer, number, type, cast);
1880 Newxz(pointer, number, type);
1881
1882The first argument C<pointer> should be the name of a variable that will
1883point to the newly allocated memory.
1884
1885The second and third arguments C<number> and C<type> specify how many of
1886the specified type of data structure should be allocated. The argument
1887C<type> is passed to C<sizeof>. The final argument to C<Newxc>, C<cast>,
1888should be used if the C<pointer> argument is different from the C<type>
1889argument.
1890
1891Unlike the C<Newx> and C<Newxc> macros, the C<Newxz> macro calls C<memzero>
1892to zero out all the newly allocated memory.
1893
1894=head2 Reallocation
1895
1896 Renew(pointer, number, type);
1897 Renewc(pointer, number, type, cast);
1898 Safefree(pointer)
1899
1900These three macros are used to change a memory buffer size or to free a
1901piece of memory no longer needed. The arguments to C<Renew> and C<Renewc>
1902match those of C<New> and C<Newc> with the exception of not needing the
1903"magic cookie" argument.
1904
1905=head2 Moving
1906
1907 Move(source, dest, number, type);
1908 Copy(source, dest, number, type);
1909 Zero(dest, number, type);
1910
1911These three macros are used to move, copy, or zero out previously allocated
1912memory. The C<source> and C<dest> arguments point to the source and
1913destination starting points. Perl will move, copy, or zero out C<number>
1914instances of the size of the C<type> data structure (using the C<sizeof>
1915function).
1916
1917=head1 PerlIO
1918
1919The most recent development releases of Perl have been experimenting with
1920removing Perl's dependency on the "normal" standard I/O suite and allowing
1921other stdio implementations to be used. This involves creating a new
1922abstraction layer that then calls whichever implementation of stdio Perl
1923was compiled with. All XSUBs should now use the functions in the PerlIO
1924abstraction layer and not make any assumptions about what kind of stdio
1925is being used.
1926
1927For a complete description of the PerlIO abstraction, consult L<perlapio>.
1928
0a753a76 1929=head1 Compiled code
1930
1931=head2 Code tree
1932
1933Here we describe the internal form your code is converted to by
10e2eb10 1934Perl. Start with a simple example:
0a753a76 1935
1936 $a = $b + $c;
1937
1938This is converted to a tree similar to this one:
1939
1940 assign-to
1941 / \
1942 + $a
1943 / \
1944 $b $c
1945
7b8d334a 1946(but slightly more complicated). This tree reflects the way Perl
0a753a76 1947parsed your code, but has nothing to do with the execution order.
1948There is an additional "thread" going through the nodes of the tree
1949which shows the order of execution of the nodes. In our simplified
1950example above it looks like:
1951
1952 $b ---> $c ---> + ---> $a ---> assign-to
1953
1954But with the actual compile tree for C<$a = $b + $c> it is different:
1955some nodes I<optimized away>. As a corollary, though the actual tree
1956contains more nodes than our simplified example, the execution order
1957is the same as in our example.
1958
1959=head2 Examining the tree
1960
06f6df17
RGS
1961If you have your perl compiled for debugging (usually done with
1962C<-DDEBUGGING> on the C<Configure> command line), you may examine the
0a753a76 1963compiled tree by specifying C<-Dx> on the Perl command line. The
1964output takes several lines per node, and for C<$b+$c> it looks like
1965this:
1966
1967 5 TYPE = add ===> 6
1968 TARG = 1
1969 FLAGS = (SCALAR,KIDS)
1970 {
1971 TYPE = null ===> (4)
1972 (was rv2sv)
1973 FLAGS = (SCALAR,KIDS)
1974 {
1975 3 TYPE = gvsv ===> 4
1976 FLAGS = (SCALAR)
1977 GV = main::b
1978 }
1979 }
1980 {
1981 TYPE = null ===> (5)
1982 (was rv2sv)
1983 FLAGS = (SCALAR,KIDS)
1984 {
1985 4 TYPE = gvsv ===> 5
1986 FLAGS = (SCALAR)
1987 GV = main::c
1988 }
1989 }
1990
1991This tree has 5 nodes (one per C<TYPE> specifier), only 3 of them are
1992not optimized away (one per number in the left column). The immediate
1993children of the given node correspond to C<{}> pairs on the same level
1994of indentation, thus this listing corresponds to the tree:
1995
1996 add
1997 / \
1998 null null
1999 | |
2000 gvsv gvsv
2001
2002The execution order is indicated by C<===E<gt>> marks, thus it is C<3
20034 5 6> (node C<6> is not included into above listing), i.e.,
2004C<gvsv gvsv add whatever>.
2005
9afa14e3 2006Each of these nodes represents an op, a fundamental operation inside the
10e2eb10 2007Perl core. The code which implements each operation can be found in the
9afa14e3 2008F<pp*.c> files; the function which implements the op with type C<gvsv>
10e2eb10 2009is C<pp_gvsv>, and so on. As the tree above shows, different ops have
9afa14e3 2010different numbers of children: C<add> is a binary operator, as one would
10e2eb10 2011expect, and so has two children. To accommodate the various different
9afa14e3
SC
2012numbers of children, there are various types of op data structure, and
2013they link together in different ways.
2014
10e2eb10 2015The simplest type of op structure is C<OP>: this has no children. Unary
9afa14e3 2016operators, C<UNOP>s, have one child, and this is pointed to by the
10e2eb10
FC
2017C<op_first> field. Binary operators (C<BINOP>s) have not only an
2018C<op_first> field but also an C<op_last> field. The most complex type of
2019op is a C<LISTOP>, which has any number of children. In this case, the
9afa14e3 2020first child is pointed to by C<op_first> and the last child by
10e2eb10 2021C<op_last>. The children in between can be found by iteratively
86cd3a13 2022following the C<OpSIBLING> pointer from the first child to the last (but
29e61fd9 2023see below).
9afa14e3 2024
29e61fd9 2025There are also some other op types: a C<PMOP> holds a regular expression,
10e2eb10
FC
2026and has no children, and a C<LOOP> may or may not have children. If the
2027C<op_children> field is non-zero, it behaves like a C<LISTOP>. To
9afa14e3
SC
2028complicate matters, if a C<UNOP> is actually a C<null> op after
2029optimization (see L</Compile pass 2: context propagation>) it will still
2030have children in accordance with its former type.
2031
29e61fd9
DM
2032Finally, there is a C<LOGOP>, or logic op. Like a C<LISTOP>, this has one
2033or more children, but it doesn't have an C<op_last> field: so you have to
86cd3a13 2034follow C<op_first> and then the C<OpSIBLING> chain itself to find the
29e61fd9
DM
2035last child. Instead it has an C<op_other> field, which is comparable to
2036the C<op_next> field described below, and represents an alternate
2037execution path. Operators like C<and>, C<or> and C<?> are C<LOGOP>s. Note
2038that in general, C<op_other> may not point to any of the direct children
2039of the C<LOGOP>.
2040
2041Starting in version 5.21.2, perls built with the experimental
2042define C<-DPERL_OP_PARENT> add an extra boolean flag for each op,
87b5a8b9 2043C<op_moresib>. When not set, this indicates that this is the last op in an
86cd3a13
DM
2044C<OpSIBLING> chain. This frees up the C<op_sibling> field on the last
2045sibling to point back to the parent op. Under this build, that field is
2046also renamed C<op_sibparent> to reflect its joint role. The macro
2047C<OpSIBLING(o)> wraps this special behaviour, and always returns NULL on
2048the last sibling. With this build the C<op_parent(o)> function can be
2049used to find the parent of any op. Thus for forward compatibility, you
2050should always use the C<OpSIBLING(o)> macro rather than accessing
2051C<op_sibling> directly.
29e61fd9 2052
06f6df17
RGS
2053Another way to examine the tree is to use a compiler back-end module, such
2054as L<B::Concise>.
2055
0a753a76 2056=head2 Compile pass 1: check routines
2057
8870b5c7 2058The tree is created by the compiler while I<yacc> code feeds it
10e2eb10 2059the constructions it recognizes. Since I<yacc> works bottom-up, so does
0a753a76 2060the first pass of perl compilation.
2061
2062What makes this pass interesting for perl developers is that some
2063optimization may be performed on this pass. This is optimization by
8870b5c7 2064so-called "check routines". The correspondence between node names
0a753a76 2065and corresponding check routines is described in F<opcode.pl> (do not
2066forget to run C<make regen_headers> if you modify this file).
2067
2068A check routine is called when the node is fully constructed except
7b8d334a 2069for the execution-order thread. Since at this time there are no
0a753a76 2070back-links to the currently constructed node, one can do most any
2071operation to the top-level node, including freeing it and/or creating
2072new nodes above/below it.
2073
2074The check routine returns the node which should be inserted into the
2075tree (if the top-level node was not modified, check routine returns
2076its argument).
2077
10e2eb10 2078By convention, check routines have names C<ck_*>. They are usually
0a753a76 2079called from C<new*OP> subroutines (or C<convert>) (which in turn are
2080called from F<perly.y>).
2081
2082=head2 Compile pass 1a: constant folding
2083
2084Immediately after the check routine is called the returned node is
2085checked for being compile-time executable. If it is (the value is
2086judged to be constant) it is immediately executed, and a I<constant>
2087node with the "return value" of the corresponding subtree is
2088substituted instead. The subtree is deleted.
2089
2090If constant folding was not performed, the execution-order thread is
2091created.
2092
2093=head2 Compile pass 2: context propagation
2094
2095When a context for a part of compile tree is known, it is propagated
a3cb178b 2096down through the tree. At this time the context can have 5 values
0a753a76 2097(instead of 2 for runtime context): void, boolean, scalar, list, and
2098lvalue. In contrast with the pass 1 this pass is processed from top
2099to bottom: a node's context determines the context for its children.
2100
2101Additional context-dependent optimizations are performed at this time.
2102Since at this moment the compile tree contains back-references (via
2103"thread" pointers), nodes cannot be free()d now. To allow
2104optimized-away nodes at this stage, such nodes are null()ified instead
2105of free()ing (i.e. their type is changed to OP_NULL).
2106
2107=head2 Compile pass 3: peephole optimization
2108
2109After the compile tree for a subroutine (or for an C<eval> or a file)
10e2eb10 2110is created, an additional pass over the code is performed. This pass
0a753a76 2111is neither top-down or bottom-up, but in the execution order (with
9ea12537
Z
2112additional complications for conditionals). Optimizations performed
2113at this stage are subject to the same restrictions as in the pass 2.
2114
2115Peephole optimizations are done by calling the function pointed to
2116by the global variable C<PL_peepp>. By default, C<PL_peepp> just
2117calls the function pointed to by the global variable C<PL_rpeepp>.
2118By default, that performs some basic op fixups and optimisations along
2119the execution-order op chain, and recursively calls C<PL_rpeepp> for
2120each side chain of ops (resulting from conditionals). Extensions may
2121provide additional optimisations or fixups, hooking into either the
2122per-subroutine or recursive stage, like this:
2123
2124 static peep_t prev_peepp;
2125 static void my_peep(pTHX_ OP *o)
2126 {
2127 /* custom per-subroutine optimisation goes here */
f0358462 2128 prev_peepp(aTHX_ o);
9ea12537
Z
2129 /* custom per-subroutine optimisation may also go here */
2130 }
2131 BOOT:
2132 prev_peepp = PL_peepp;
2133 PL_peepp = my_peep;
2134
2135 static peep_t prev_rpeepp;
2136 static void my_rpeep(pTHX_ OP *o)
2137 {
2138 OP *orig_o = o;
2139 for(; o; o = o->op_next) {
2140 /* custom per-op optimisation goes here */
2141 }
f0358462 2142 prev_rpeepp(aTHX_ orig_o);
9ea12537
Z
2143 }
2144 BOOT:
2145 prev_rpeepp = PL_rpeepp;
2146 PL_rpeepp = my_rpeep;
0a753a76 2147
1ba7f851
PJ
2148=head2 Pluggable runops
2149
2150The compile tree is executed in a runops function. There are two runops
1388f78e
RGS
2151functions, in F<run.c> and in F<dump.c>. C<Perl_runops_debug> is used
2152with DEBUGGING and C<Perl_runops_standard> is used otherwise. For fine
2153control over the execution of the compile tree it is possible to provide
2154your own runops function.
1ba7f851
PJ
2155
2156It's probably best to copy one of the existing runops functions and
2157change it to suit your needs. Then, in the BOOT section of your XS
2158file, add the line:
2159
2160 PL_runops = my_runops;
2161
2162This function should be as efficient as possible to keep your programs
2163running as fast as possible.
2164
fd85fad2
BM
2165=head2 Compile-time scope hooks
2166
2167As of perl 5.14 it is possible to hook into the compile-time lexical
10e2eb10 2168scope mechanism using C<Perl_blockhook_register>. This is used like
fd85fad2
BM
2169this:
2170
2171 STATIC void my_start_hook(pTHX_ int full);
2172 STATIC BHK my_hooks;
2173
2174 BOOT:
a88d97bf 2175 BhkENTRY_set(&my_hooks, bhk_start, my_start_hook);
fd85fad2
BM
2176 Perl_blockhook_register(aTHX_ &my_hooks);
2177
2178This will arrange to have C<my_start_hook> called at the start of
10e2eb10 2179compiling every lexical scope. The available hooks are:
fd85fad2
BM
2180
2181=over 4
2182
a88d97bf 2183=item C<void bhk_start(pTHX_ int full)>
fd85fad2 2184
10e2eb10 2185This is called just after starting a new lexical scope. Note that Perl
fd85fad2
BM
2186code like
2187
2188 if ($x) { ... }
2189
2190creates two scopes: the first starts at the C<(> and has C<full == 1>,
10e2eb10 2191the second starts at the C<{> and has C<full == 0>. Both end at the
f185f654 2192C<}>, so calls to C<start> and C<pre>/C<post_end> will match. Anything
fd85fad2
BM
2193pushed onto the save stack by this hook will be popped just before the
2194scope ends (between the C<pre_> and C<post_end> hooks, in fact).
2195
a88d97bf 2196=item C<void bhk_pre_end(pTHX_ OP **o)>
fd85fad2
BM
2197
2198This is called at the end of a lexical scope, just before unwinding the
10e2eb10 2199stack. I<o> is the root of the optree representing the scope; it is a
fd85fad2
BM
2200double pointer so you can replace the OP if you need to.
2201
a88d97bf 2202=item C<void bhk_post_end(pTHX_ OP **o)>
fd85fad2
BM
2203
2204This is called at the end of a lexical scope, just after unwinding the
10e2eb10 2205stack. I<o> is as above. Note that it is possible for calls to C<pre_>
fd85fad2
BM
2206and C<post_end> to nest, if there is something on the save stack that
2207calls string eval.
2208
a88d97bf 2209=item C<void bhk_eval(pTHX_ OP *const o)>
fd85fad2
BM
2210
2211This is called just before starting to compile an C<eval STRING>, C<do
10e2eb10 2212FILE>, C<require> or C<use>, after the eval has been set up. I<o> is the
fd85fad2
BM
2213OP that requested the eval, and will normally be an C<OP_ENTEREVAL>,
2214C<OP_DOFILE> or C<OP_REQUIRE>.
2215
2216=back
2217
2218Once you have your hook functions, you need a C<BHK> structure to put
10e2eb10
FC
2219them in. It's best to allocate it statically, since there is no way to
2220free it once it's registered. The function pointers should be inserted
fd85fad2 2221into this structure using the C<BhkENTRY_set> macro, which will also set
10e2eb10 2222flags indicating which entries are valid. If you do need to allocate
fd85fad2
BM
2223your C<BHK> dynamically for some reason, be sure to zero it before you
2224start.
2225
2226Once registered, there is no mechanism to switch these hooks off, so if
10e2eb10 2227that is necessary you will need to do this yourself. An entry in C<%^H>
a3e07c87
BM
2228is probably the best way, so the effect is lexically scoped; however it
2229is also possible to use the C<BhkDISABLE> and C<BhkENABLE> macros to
10e2eb10 2230temporarily switch entries on and off. You should also be aware that
a3e07c87 2231generally speaking at least one scope will have opened before your
f185f654 2232extension is loaded, so you will see some C<pre>/C<post_end> pairs that
a3e07c87 2233didn't have a matching C<start>.
fd85fad2 2234
9afa14e3
SC
2235=head1 Examining internal data structures with the C<dump> functions
2236
2237To aid debugging, the source file F<dump.c> contains a number of
2238functions which produce formatted output of internal data structures.
2239
2240The most commonly used of these functions is C<Perl_sv_dump>; it's used
10e2eb10 2241for dumping SVs, AVs, HVs, and CVs. The C<Devel::Peek> module calls
9afa14e3 2242C<sv_dump> to produce debugging output from Perl-space, so users of that
00aadd71 2243module should already be familiar with its format.
9afa14e3
SC
2244
2245C<Perl_op_dump> can be used to dump an C<OP> structure or any of its
210b36aa 2246derivatives, and produces output similar to C<perl -Dx>; in fact,
9afa14e3
SC
2247C<Perl_dump_eval> will dump the main root of the code being evaluated,
2248exactly like C<-Dx>.
2249
2250Other useful functions are C<Perl_dump_sub>, which turns a C<GV> into an
2251op tree, C<Perl_dump_packsubs> which calls C<Perl_dump_sub> on all the
2252subroutines in a package like so: (Thankfully, these are all xsubs, so
2253there is no op tree)
2254
2255 (gdb) print Perl_dump_packsubs(PL_defstash)
2256
2257 SUB attributes::bootstrap = (xsub 0x811fedc 0)
2258
2259 SUB UNIVERSAL::can = (xsub 0x811f50c 0)
2260
2261 SUB UNIVERSAL::isa = (xsub 0x811f304 0)
2262
2263 SUB UNIVERSAL::VERSION = (xsub 0x811f7ac 0)
2264
2265 SUB DynaLoader::boot_DynaLoader = (xsub 0x805b188 0)
2266
2267and C<Perl_dump_all>, which dumps all the subroutines in the stash and
2268the op tree of the main root.
2269
954c1994 2270=head1 How multiple interpreters and concurrency are supported
ee072b34 2271
ee072b34
GS
2272=head2 Background and PERL_IMPLICIT_CONTEXT
2273
2274The Perl interpreter can be regarded as a closed box: it has an API
2275for feeding it code or otherwise making it do things, but it also has
2276functions for its own use. This smells a lot like an object, and
2277there are ways for you to build Perl so that you can have multiple
acfe0abc
GS
2278interpreters, with one interpreter represented either as a C structure,
2279or inside a thread-specific structure. These structures contain all
2280the context, the state of that interpreter.
2281
10e2eb10 2282One macro controls the major Perl build flavor: MULTIPLICITY. The
7b52221d 2283MULTIPLICITY build has a C structure that packages all the interpreter
10e2eb10 2284state. With multiplicity-enabled perls, PERL_IMPLICIT_CONTEXT is also
7b52221d 2285normally defined, and enables the support for passing in a "hidden" first
10e2eb10 2286argument that represents all three data structures. MULTIPLICITY makes
1a64a5e6 2287multi-threaded perls possible (with the ithreads threading model, related
7b52221d 2288to the macro USE_ITHREADS.)
54aff467 2289
27da23d5
JH
2290Two other "encapsulation" macros are the PERL_GLOBAL_STRUCT and
2291PERL_GLOBAL_STRUCT_PRIVATE (the latter turns on the former, and the
2292former turns on MULTIPLICITY.) The PERL_GLOBAL_STRUCT causes all the
2293internal variables of Perl to be wrapped inside a single global struct,
2294struct perl_vars, accessible as (globals) &PL_Vars or PL_VarsPtr or
2295the function Perl_GetVars(). The PERL_GLOBAL_STRUCT_PRIVATE goes
2296one step further, there is still a single struct (allocated in main()
2297either from heap or from stack) but there are no global data symbols
3bf17896 2298pointing to it. In either case the global struct should be initialized
27da23d5
JH
2299as the very first thing in main() using Perl_init_global_struct() and
2300correspondingly tear it down after perl_free() using Perl_free_global_struct(),
2301please see F<miniperlmain.c> for usage details. You may also need
2302to use C<dVAR> in your coding to "declare the global variables"
2303when you are using them. dTHX does this for you automatically.
2304
9aa97215
JH
2305To see whether you have non-const data you can use a BSD (or GNU)
2306compatible C<nm>:
bc028b6b
JH
2307
2308 nm libperl.a | grep -v ' [TURtr] '
2309
9aa97215
JH
2310If this displays any C<D> or C<d> symbols (or possibly C<C> or C<c>),
2311you have non-const data. The symbols the C<grep> removed are as follows:
2312C<Tt> are I<text>, or code, the C<Rr> are I<read-only> (const) data,
2313and the C<U> is <undefined>, external symbols referred to.
2314
2315The test F<t/porting/libperl.t> does this kind of symbol sanity
2316checking on C<libperl.a>.
bc028b6b 2317
27da23d5
JH
2318For backward compatibility reasons defining just PERL_GLOBAL_STRUCT
2319doesn't actually hide all symbols inside a big global struct: some
2320PerlIO_xxx vtables are left visible. The PERL_GLOBAL_STRUCT_PRIVATE
2321then hides everything (see how the PERLIO_FUNCS_DECL is used).
2322
54aff467 2323All this obviously requires a way for the Perl internal functions to be
acfe0abc 2324either subroutines taking some kind of structure as the first
ee072b34 2325argument, or subroutines taking nothing as the first argument. To
acfe0abc 2326enable these two very different ways of building the interpreter,
ee072b34
GS
2327the Perl source (as it does in so many other situations) makes heavy
2328use of macros and subroutine naming conventions.
2329
54aff467 2330First problem: deciding which functions will be public API functions and
00aadd71 2331which will be private. All functions whose names begin C<S_> are private
954c1994
GS
2332(think "S" for "secret" or "static"). All other functions begin with
2333"Perl_", but just because a function begins with "Perl_" does not mean it is
10e2eb10
FC
2334part of the API. (See L</Internal
2335Functions>.) The easiest way to be B<sure> a
00aadd71
NIS
2336function is part of the API is to find its entry in L<perlapi>.
2337If it exists in L<perlapi>, it's part of the API. If it doesn't, and you
2338think it should be (i.e., you need it for your extension), send mail via
a422fd2d 2339L<perlbug> explaining why you think it should be.
ee072b34
GS
2340
2341Second problem: there must be a syntax so that the same subroutine
2342declarations and calls can pass a structure as their first argument,
2343or pass nothing. To solve this, the subroutines are named and
2344declared in a particular way. Here's a typical start of a static
2345function used within the Perl guts:
2346
2347 STATIC void
2348 S_incline(pTHX_ char *s)
2349
acfe0abc 2350STATIC becomes "static" in C, and may be #define'd to nothing in some
da8c5729 2351configurations in the future.
ee072b34 2352
651a3225
GS
2353A public function (i.e. part of the internal API, but not necessarily
2354sanctioned for use in extensions) begins like this:
ee072b34
GS
2355
2356 void
2307c6d0 2357 Perl_sv_setiv(pTHX_ SV* dsv, IV num)
ee072b34 2358
0147cd53 2359C<pTHX_> is one of a number of macros (in F<perl.h>) that hide the
ee072b34
GS
2360details of the interpreter's context. THX stands for "thread", "this",
2361or "thingy", as the case may be. (And no, George Lucas is not involved. :-)
2362The first character could be 'p' for a B<p>rototype, 'a' for B<a>rgument,
a7486cbb
JH
2363or 'd' for B<d>eclaration, so we have C<pTHX>, C<aTHX> and C<dTHX>, and
2364their variants.
ee072b34 2365
a7486cbb
JH
2366When Perl is built without options that set PERL_IMPLICIT_CONTEXT, there is no
2367first argument containing the interpreter's context. The trailing underscore
ee072b34
GS
2368in the pTHX_ macro indicates that the macro expansion needs a comma
2369after the context argument because other arguments follow it. If
2370PERL_IMPLICIT_CONTEXT is not defined, pTHX_ will be ignored, and the
54aff467
GS
2371subroutine is not prototyped to take the extra argument. The form of the
2372macro without the trailing underscore is used when there are no additional
ee072b34
GS
2373explicit arguments.
2374
54aff467 2375When a core function calls another, it must pass the context. This
2307c6d0 2376is normally hidden via macros. Consider C<sv_setiv>. It expands into
ee072b34
GS
2377something like this:
2378
2307c6d0
SB
2379 #ifdef PERL_IMPLICIT_CONTEXT
2380 #define sv_setiv(a,b) Perl_sv_setiv(aTHX_ a, b)
ee072b34 2381 /* can't do this for vararg functions, see below */
2307c6d0
SB
2382 #else
2383 #define sv_setiv Perl_sv_setiv
2384 #endif
ee072b34
GS
2385
2386This works well, and means that XS authors can gleefully write:
2387
2307c6d0 2388 sv_setiv(foo, bar);
ee072b34
GS
2389
2390and still have it work under all the modes Perl could have been
2391compiled with.
2392
ee072b34
GS
2393This doesn't work so cleanly for varargs functions, though, as macros
2394imply that the number of arguments is known in advance. Instead we
2395either need to spell them out fully, passing C<aTHX_> as the first
2396argument (the Perl core tends to do this with functions like
2397Perl_warner), or use a context-free version.
2398
2399The context-free version of Perl_warner is called
2400Perl_warner_nocontext, and does not take the extra argument. Instead
2401it does dTHX; to get the context from thread-local storage. We
2402C<#define warner Perl_warner_nocontext> so that extensions get source
2403compatibility at the expense of performance. (Passing an arg is
2404cheaper than grabbing it from thread-local storage.)
2405
acfe0abc 2406You can ignore [pad]THXx when browsing the Perl headers/sources.
ee072b34
GS
2407Those are strictly for use within the core. Extensions and embedders
2408need only be aware of [pad]THX.
2409
a7486cbb
JH
2410=head2 So what happened to dTHR?
2411
2412C<dTHR> was introduced in perl 5.005 to support the older thread model.
2413The older thread model now uses the C<THX> mechanism to pass context
2414pointers around, so C<dTHR> is not useful any more. Perl 5.6.0 and
2415later still have it for backward source compatibility, but it is defined
2416to be a no-op.
2417
ee072b34
GS
2418=head2 How do I use all this in extensions?
2419
2420When Perl is built with PERL_IMPLICIT_CONTEXT, extensions that call
2421any functions in the Perl API will need to pass the initial context
2422argument somehow. The kicker is that you will need to write it in
2423such a way that the extension still compiles when Perl hasn't been
2424built with PERL_IMPLICIT_CONTEXT enabled.
2425
2426There are three ways to do this. First, the easy but inefficient way,
2427which is also the default, in order to maintain source compatibility
0147cd53 2428with extensions: whenever F<XSUB.h> is #included, it redefines the aTHX
ee072b34
GS
2429and aTHX_ macros to call a function that will return the context.
2430Thus, something like:
2431
2307c6d0 2432 sv_setiv(sv, num);
ee072b34 2433
4375e838 2434in your extension will translate to this when PERL_IMPLICIT_CONTEXT is
54aff467 2435in effect:
ee072b34 2436
2307c6d0 2437 Perl_sv_setiv(Perl_get_context(), sv, num);
ee072b34 2438
54aff467 2439or to this otherwise:
ee072b34 2440
2307c6d0 2441 Perl_sv_setiv(sv, num);
ee072b34 2442
da8c5729 2443You don't have to do anything new in your extension to get this; since
2fa86c13 2444the Perl library provides Perl_get_context(), it will all just
ee072b34
GS
2445work.
2446
2447The second, more efficient way is to use the following template for
2448your Foo.xs:
2449
c52f9dcd
JH
2450 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2451 #include "EXTERN.h"
2452 #include "perl.h"
2453 #include "XSUB.h"
ee072b34 2454
fd061412 2455 STATIC void my_private_function(int arg1, int arg2);
ee072b34 2456
fd061412 2457 STATIC void
c52f9dcd
JH
2458 my_private_function(int arg1, int arg2)
2459 {
2460 dTHX; /* fetch context */
2461 ... call many Perl API functions ...
2462 }
ee072b34
GS
2463
2464 [... etc ...]
2465
c52f9dcd 2466 MODULE = Foo PACKAGE = Foo
ee072b34 2467
c52f9dcd 2468 /* typical XSUB */
ee072b34 2469
c52f9dcd
JH
2470 void
2471 my_xsub(arg)
2472 int arg
2473 CODE:
2474 my_private_function(arg, 10);
ee072b34
GS
2475
2476Note that the only two changes from the normal way of writing an
2477extension is the addition of a C<#define PERL_NO_GET_CONTEXT> before
2478including the Perl headers, followed by a C<dTHX;> declaration at
2479the start of every function that will call the Perl API. (You'll
2480know which functions need this, because the C compiler will complain
2481that there's an undeclared identifier in those functions.) No changes
2482are needed for the XSUBs themselves, because the XS() macro is
2483correctly defined to pass in the implicit context if needed.
2484
2485The third, even more efficient way is to ape how it is done within
2486the Perl guts:
2487
2488
c52f9dcd
JH
2489 #define PERL_NO_GET_CONTEXT /* we want efficiency */
2490 #include "EXTERN.h"
2491 #include "perl.h"
2492 #include "XSUB.h"
ee072b34
GS
2493
2494 /* pTHX_ only needed for functions that call Perl API */
fd061412 2495 STATIC void my_private_function(pTHX_ int arg1, int arg2);
ee072b34 2496
fd061412 2497 STATIC void
c52f9dcd
JH
2498 my_private_function(pTHX_ int arg1, int arg2)
2499 {
2500 /* dTHX; not needed here, because THX is an argument */
2501 ... call Perl API functions ...
2502 }
ee072b34
GS
2503
2504 [... etc ...]
2505
c52f9dcd 2506 MODULE = Foo PACKAGE = Foo
ee072b34 2507
c52f9dcd 2508 /* typical XSUB */
ee072b34 2509
c52f9dcd
JH
2510 void
2511 my_xsub(arg)
2512 int arg
2513 CODE:
2514 my_private_function(aTHX_ arg, 10);
ee072b34
GS
2515
2516This implementation never has to fetch the context using a function
2517call, since it is always passed as an extra argument. Depending on
2518your needs for simplicity or efficiency, you may mix the previous
2519two approaches freely.
2520
651a3225
GS
2521Never add a comma after C<pTHX> yourself--always use the form of the
2522macro with the underscore for functions that take explicit arguments,
2523or the form without the argument for functions with no explicit arguments.
ee072b34 2524
27da23d5
JH
2525If one is compiling Perl with the C<-DPERL_GLOBAL_STRUCT> the C<dVAR>
2526definition is needed if the Perl global variables (see F<perlvars.h>
2527or F<globvar.sym>) are accessed in the function and C<dTHX> is not
2528used (the C<dTHX> includes the C<dVAR> if necessary). One notices
2529the need for C<dVAR> only with the said compile-time define, because
2530otherwise the Perl global variables are visible as-is.
2531
a7486cbb
JH
2532=head2 Should I do anything special if I call perl from multiple threads?
2533
2534If you create interpreters in one thread and then proceed to call them in
2535another, you need to make sure perl's own Thread Local Storage (TLS) slot is
2536initialized correctly in each of those threads.
2537
2538The C<perl_alloc> and C<perl_clone> API functions will automatically set
2539the TLS slot to the interpreter they created, so that there is no need to do
2540anything special if the interpreter is always accessed in the same thread that
2541created it, and that thread did not create or call any other interpreters
2542afterwards. If that is not the case, you have to set the TLS slot of the
2543thread before calling any functions in the Perl API on that particular
2544interpreter. This is done by calling the C<PERL_SET_CONTEXT> macro in that
2545thread as the first thing you do:
2546
2547 /* do this before doing anything else with some_perl */
2548 PERL_SET_CONTEXT(some_perl);
2549
2550 ... other Perl API calls on some_perl go here ...
2551
ee072b34
GS
2552=head2 Future Plans and PERL_IMPLICIT_SYS
2553
2554Just as PERL_IMPLICIT_CONTEXT provides a way to bundle up everything
2555that the interpreter knows about itself and pass it around, so too are
2556there plans to allow the interpreter to bundle up everything it knows
2557about the environment it's running on. This is enabled with the
7b52221d
RGS
2558PERL_IMPLICIT_SYS macro. Currently it only works with USE_ITHREADS on
2559Windows.
ee072b34
GS
2560
2561This allows the ability to provide an extra pointer (called the "host"
2562environment) for all the system calls. This makes it possible for
2563all the system stuff to maintain their own state, broken down into
2564seven C structures. These are thin wrappers around the usual system
0147cd53 2565calls (see F<win32/perllib.c>) for the default perl executable, but for a
ee072b34
GS
2566more ambitious host (like the one that would do fork() emulation) all
2567the extra work needed to pretend that different interpreters are
2568actually different "processes", would be done here.
2569
2570The Perl engine/interpreter and the host are orthogonal entities.
2571There could be one or more interpreters in a process, and one or
2572more "hosts", with free association between them.
2573
a422fd2d
SC
2574=head1 Internal Functions
2575
2576All of Perl's internal functions which will be exposed to the outside
06f6df17 2577world are prefixed by C<Perl_> so that they will not conflict with XS
a422fd2d 2578functions or functions used in a program in which Perl is embedded.
10e2eb10 2579Similarly, all global variables begin with C<PL_>. (By convention,
06f6df17 2580static functions start with C<S_>.)
a422fd2d 2581
0972ecdf
DM
2582Inside the Perl core (C<PERL_CORE> defined), you can get at the functions
2583either with or without the C<Perl_> prefix, thanks to a bunch of defines
10e2eb10 2584that live in F<embed.h>. Note that extension code should I<not> set
0972ecdf
DM
2585C<PERL_CORE>; this exposes the full perl internals, and is likely to cause
2586breakage of the XS in each new perl release.
2587
2588The file F<embed.h> is generated automatically from
10e2eb10 2589F<embed.pl> and F<embed.fnc>. F<embed.pl> also creates the prototyping
dc9b1d22 2590header files for the internal functions, generates the documentation
10e2eb10 2591and a lot of other bits and pieces. It's important that when you add
dc9b1d22 2592a new function to the core or change an existing one, you change the
10e2eb10 2593data in the table in F<embed.fnc> as well. Here's a sample entry from
dc9b1d22 2594that table:
a422fd2d
SC
2595
2596 Apd |SV** |av_fetch |AV* ar|I32 key|I32 lval
2597
10e2eb10
FC
2598The second column is the return type, the third column the name. Columns
2599after that are the arguments. The first column is a set of flags:
a422fd2d
SC
2600
2601=over 3
2602
2603=item A
2604
10e2eb10
FC
2605This function is a part of the public
2606API. All such functions should also
1aa6ea50 2607have 'd', very few do not.
a422fd2d
SC
2608
2609=item p
2610
1aa6ea50
JC
2611This function has a C<Perl_> prefix; i.e. it is defined as
2612C<Perl_av_fetch>.
a422fd2d
SC
2613
2614=item d
2615
2616This function has documentation using the C<apidoc> feature which we'll
1aa6ea50 2617look at in a second. Some functions have 'd' but not 'A'; docs are good.
a422fd2d
SC
2618
2619=back
2620
2621Other available flags are:
2622
2623=over 3
2624
2625=item s
2626
1aa6ea50
JC
2627This is a static function and is defined as C<STATIC S_whatever>, and
2628usually called within the sources as C<whatever(...)>.
a422fd2d
SC
2629
2630=item n
2631
da8c5729 2632This does not need an interpreter context, so the definition has no
1aa6ea50 2633C<pTHX>, and it follows that callers don't use C<aTHX>. (See
d3a43cd8 2634L</Background and PERL_IMPLICIT_CONTEXT>.)
a422fd2d
SC
2635
2636=item r
2637
2638This function never returns; C<croak>, C<exit> and friends.
2639
2640=item f
2641
2642This function takes a variable number of arguments, C<printf> style.
2643The argument list should end with C<...>, like this:
2644
2645 Afprd |void |croak |const char* pat|...
2646
a7486cbb 2647=item M
a422fd2d 2648
00aadd71 2649This function is part of the experimental development API, and may change
a422fd2d
SC
2650or disappear without notice.
2651
2652=item o
2653
2654This function should not have a compatibility macro to define, say,
10e2eb10 2655C<Perl_parse> to C<parse>. It must be called as C<Perl_parse>.
a422fd2d 2656
a422fd2d
SC
2657=item x
2658
2659This function isn't exported out of the Perl core.
2660
dc9b1d22
MHM
2661=item m
2662
2663This is implemented as a macro.
2664
2665=item X
2666
2667This function is explicitly exported.
2668
2669=item E
2670
2671This function is visible to extensions included in the Perl core.
2672
2673=item b
2674
2675Binary backward compatibility; this function is a macro but also has
2676a C<Perl_> implementation (which is exported).
2677
1aa6ea50
JC
2678=item others
2679
2680See the comments at the top of C<embed.fnc> for others.
2681
a422fd2d
SC
2682=back
2683
dc9b1d22
MHM
2684If you edit F<embed.pl> or F<embed.fnc>, you will need to run
2685C<make regen_headers> to force a rebuild of F<embed.h> and other
2686auto-generated files.
a422fd2d 2687
6b4667fc 2688=head2 Formatted Printing of IVs, UVs, and NVs
9dd9db0b 2689
6b4667fc
A
2690If you are printing IVs, UVs, or NVS instead of the stdio(3) style
2691formatting codes like C<%d>, C<%ld>, C<%f>, you should use the
2692following macros for portability
9dd9db0b 2693
c52f9dcd
JH
2694 IVdf IV in decimal
2695 UVuf UV in decimal
2696 UVof UV in octal
2697 UVxf UV in hexadecimal
2698 NVef NV %e-like
2699 NVff NV %f-like
2700 NVgf NV %g-like
9dd9db0b 2701
6b4667fc
A
2702These will take care of 64-bit integers and long doubles.
2703For example:
2704
c52f9dcd 2705 printf("IV is %"IVdf"\n", iv);
6b4667fc
A
2706
2707The IVdf will expand to whatever is the correct format for the IVs.
9dd9db0b 2708
aacf4ea2
JH
2709Note that there are different "long doubles": Perl will use
2710whatever the compiler has.
2711
8908e76d
JH
2712If you are printing addresses of pointers, use UVxf combined
2713with PTR2UV(), do not use %lx or %p.
2714
e613617c 2715=head2 Formatted Printing of C<Size_t> and C<SSize_t>
5862f74e
KW
2716
2717The most general way to do this is to cast them to a UV or IV, and
2718print as in the
2719L<previous section|/Formatted Printing of IVs, UVs, and NVs>.
2720
2721But if you're using C<PerlIO_printf()>, it's less typing and visual
2722clutter to use the C<"%z"> length modifier (for I<siZe>):
2723
2724 PerlIO_printf("STRLEN is %zu\n", len);
2725
2726This modifier is not portable, so its use should be restricted to
2727C<PerlIO_printf()>.
2728
8908e76d
JH
2729=head2 Pointer-To-Integer and Integer-To-Pointer
2730
2731Because pointer size does not necessarily equal integer size,
2732use the follow macros to do it right.
2733
c52f9dcd
JH
2734 PTR2UV(pointer)
2735 PTR2IV(pointer)
2736 PTR2NV(pointer)
2737 INT2PTR(pointertotype, integer)
8908e76d
JH
2738
2739For example:
2740
c52f9dcd
JH
2741 IV iv = ...;
2742 SV *sv = INT2PTR(SV*, iv);
8908e76d
JH
2743
2744and
2745
c52f9dcd
JH
2746 AV *av = ...;
2747 UV uv = PTR2UV(av);
8908e76d 2748
0ca3a874
MHM
2749=head2 Exception Handling
2750
9b5c3821 2751There are a couple of macros to do very basic exception handling in XS
10e2eb10 2752modules. You have to define C<NO_XSLOCKS> before including F<XSUB.h> to
9b5c3821
MHM
2753be able to use these macros:
2754
2755 #define NO_XSLOCKS
2756 #include "XSUB.h"
2757
2758You can use these macros if you call code that may croak, but you need
10e2eb10 2759to do some cleanup before giving control back to Perl. For example:
0ca3a874 2760
d7f8936a 2761 dXCPT; /* set up necessary variables */
0ca3a874
MHM
2762
2763 XCPT_TRY_START {
2764 code_that_may_croak();
2765 } XCPT_TRY_END
2766
2767 XCPT_CATCH
2768 {
2769 /* do cleanup here */
2770 XCPT_RETHROW;
2771 }
2772
2773Note that you always have to rethrow an exception that has been
10e2eb10
FC
2774caught. Using these macros, it is not possible to just catch the
2775exception and ignore it. If you have to ignore the exception, you
0ca3a874
MHM
2776have to use the C<call_*> function.
2777
2778The advantage of using the above macros is that you don't have
2779to setup an extra function for C<call_*>, and that using these
2780macros is faster than using C<call_*>.
2781
a422fd2d
SC
2782=head2 Source Documentation
2783
2784There's an effort going on to document the internal functions and
61ad4b94 2785automatically produce reference manuals from them -- L<perlapi> is one
a422fd2d 2786such manual which details all the functions which are available to XS
10e2eb10 2787writers. L<perlintern> is the autogenerated manual for the functions
a422fd2d
SC
2788which are not part of the API and are supposedly for internal use only.
2789
2790Source documentation is created by putting POD comments into the C
2791source, like this:
2792
2793 /*
2794 =for apidoc sv_setiv
2795
2796 Copies an integer into the given SV. Does not handle 'set' magic. See
a95b3d6a 2797 L<perlapi/sv_setiv_mg>.
a422fd2d
SC
2798
2799 =cut
2800 */
2801
2802Please try and supply some documentation if you add functions to the
2803Perl core.
2804
0d098d33
MHM
2805=head2 Backwards compatibility
2806
10e2eb10
FC
2807The Perl API changes over time. New functions are
2808added or the interfaces of existing functions are
2809changed. The C<Devel::PPPort> module tries to
0d098d33
MHM
2810provide compatibility code for some of these changes, so XS writers don't
2811have to code it themselves when supporting multiple versions of Perl.
2812
2813C<Devel::PPPort> generates a C header file F<ppport.h> that can also
10e2eb10 2814be run as a Perl script. To generate F<ppport.h>, run:
0d098d33
MHM
2815
2816 perl -MDevel::PPPort -eDevel::PPPort::WriteFile
2817
2818Besides checking existing XS code, the script can also be used to retrieve
2819compatibility information for various API calls using the C<--api-info>
10e2eb10 2820command line switch. For example:
0d098d33
MHM
2821
2822 % perl ppport.h --api-info=sv_magicext
2823
2824For details, see C<perldoc ppport.h>.
2825
a422fd2d
SC
2826=head1 Unicode Support
2827
10e2eb10 2828Perl 5.6.0 introduced Unicode support. It's important for porters and XS
a422fd2d
SC
2829writers to understand this support and make sure that the code they
2830write does not corrupt Unicode data.
2831
2832=head2 What B<is> Unicode, anyway?
2833
10e2eb10
FC
2834In the olden, less enlightened times, we all used to use ASCII. Most of
2835us did, anyway. The big problem with ASCII is that it's American. Well,
a422fd2d 2836no, that's not actually the problem; the problem is that it's not
10e2eb10 2837particularly useful for people who don't use the Roman alphabet. What
a422fd2d 2838used to happen was that particular languages would stick their own
10e2eb10 2839alphabet in the upper range of the sequence, between 128 and 255. Of
a422fd2d
SC
2840course, we then ended up with plenty of variants that weren't quite
2841ASCII, and the whole point of it being a standard was lost.
2842
2843Worse still, if you've got a language like Chinese or
2844Japanese that has hundreds or thousands of characters, then you really
2845can't fit them into a mere 256, so they had to forget about ASCII
2846altogether, and build their own systems using pairs of numbers to refer
2847to one character.
2848
2849To fix this, some people formed Unicode, Inc. and
2850produced a new character set containing all the characters you can
10e2eb10
FC
2851possibly think of and more. There are several ways of representing these
2852characters, and the one Perl uses is called UTF-8. UTF-8 uses
2853a variable number of bytes to represent a character. You can learn more
2575c402 2854about Unicode and Perl's Unicode model in L<perlunicode>.
a422fd2d 2855
3ad86f0e
KW
2856(On EBCDIC platforms, Perl uses instead UTF-EBCDIC, which is a form of
2857UTF-8 adapted for EBCDIC platforms. Below, we just talk about UTF-8.
2858UTF-EBCDIC is like UTF-8, but the details are different. The macros
2859hide the differences from you, just remember that the particular numbers
2860and bit patterns presented below will differ in UTF-EBCDIC.)
2861
1e54db1a 2862=head2 How can I recognise a UTF-8 string?
a422fd2d 2863
10e2eb10
FC
2864You can't. This is because UTF-8 data is stored in bytes just like
2865non-UTF-8 data. The Unicode character 200, (C<0xC8> for you hex types)
a422fd2d 2866capital E with a grave accent, is represented by the two bytes
10e2eb10 2867C<v196.172>. Unfortunately, the non-Unicode string C<chr(196).chr(172)>
61ad4b94 2868has that byte sequence as well. So you can't tell just by looking -- this
a422fd2d
SC
2869is what makes Unicode input an interesting problem.
2870
2575c402
JW
2871In general, you either have to know what you're dealing with, or you
2872have to guess. The API function C<is_utf8_string> can help; it'll tell
61ad4b94
KW
2873you if a string contains only valid UTF-8 characters, and the chances
2874of a non-UTF-8 string looking like valid UTF-8 become very small very
2875quickly with increasing string length. On a character-by-character
2876basis, C<isUTF8_CHAR>
2575c402 2877will tell you whether the current character in a string is valid UTF-8.
a422fd2d 2878
1e54db1a 2879=head2 How does UTF-8 represent Unicode characters?
a422fd2d 2880
1e54db1a 2881As mentioned above, UTF-8 uses a variable number of bytes to store a
10e2eb10
FC
2882character. Characters with values 0...127 are stored in one
2883byte, just like good ol' ASCII. Character 128 is stored as
2884C<v194.128>; this continues up to character 191, which is
2885C<v194.191>. Now we've run out of bits (191 is binary
61ad4b94 2886C<10111111>) so we move on; character 192 is C<v195.128>. And
a422fd2d 2887so it goes on, moving to three bytes at character 2048.
6e31cdd1 2888L<perlunicode/Unicode Encodings> has pictures of how this works.
a422fd2d 2889
1e54db1a 2890Assuming you know you're dealing with a UTF-8 string, you can find out
a422fd2d
SC
2891how long the first character in it is with the C<UTF8SKIP> macro:
2892
2893 char *utf = "\305\233\340\240\201";
2894 I32 len;
2895
2896 len = UTF8SKIP(utf); /* len is 2 here */
2897 utf += len;
2898 len = UTF8SKIP(utf); /* len is 3 here */
2899
1e54db1a 2900Another way to skip over characters in a UTF-8 string is to use
a422fd2d 2901C<utf8_hop>, which takes a string and a number of characters to skip
10e2eb10 2902over. You're on your own about bounds checking, though, so don't use it
a422fd2d
SC
2903lightly.
2904
1e54db1a 2905All bytes in a multi-byte UTF-8 character will have the high bit set,
3a2263fe 2906so you can test if you need to do something special with this
61ad4b94 2907character like this (the C<UTF8_IS_INVARIANT()> is a macro that tests
9f98c7fe 2908whether the byte is encoded as a single byte even in UTF-8):
a422fd2d 2909
3a2263fe 2910 U8 *utf;
4b88fb76 2911 U8 *utf_end; /* 1 beyond buffer pointed to by utf */
3a2263fe 2912 UV uv; /* Note: a UV, not a U8, not a char */
95701e00 2913 STRLEN len; /* length of character in bytes */
a422fd2d 2914
3a2263fe 2915 if (!UTF8_IS_INVARIANT(*utf))
1e54db1a 2916 /* Must treat this as UTF-8 */
4b88fb76 2917 uv = utf8_to_uvchr_buf(utf, utf_end, &len);
a422fd2d
SC
2918 else
2919 /* OK to treat this character as a byte */
2920 uv = *utf;
2921
4b88fb76 2922You can also see in that example that we use C<utf8_to_uvchr_buf> to get the
95701e00 2923value of the character; the inverse function C<uvchr_to_utf8> is available
1e54db1a 2924for putting a UV into UTF-8:
a422fd2d 2925
61ad4b94 2926 if (!UVCHR_IS_INVARIANT(uv))
a422fd2d 2927 /* Must treat this as UTF8 */
95701e00 2928 utf8 = uvchr_to_utf8(utf8, uv);
a422fd2d
SC
2929 else
2930 /* OK to treat this character as a byte */
2931 *utf8++ = uv;
2932
2933You B<must> convert characters to UVs using the above functions if
1e54db1a 2934you're ever in a situation where you have to match UTF-8 and non-UTF-8
10e2eb10 2935characters. You may not skip over UTF-8 characters in this case. If you
1e54db1a
JH
2936do this, you'll lose the ability to match hi-bit non-UTF-8 characters;
2937for instance, if your UTF-8 string contains C<v196.172>, and you skip
2938that character, you can never match a C<chr(200)> in a non-UTF-8 string.
a422fd2d
SC
2939So don't do that!
2940
61ad4b94
KW
2941(Note that we don't have to test for invariant characters in the
2942examples above. The functions work on any well-formed UTF-8 input.
2943It's just that its faster to avoid the function overhead when it's not
2944needed.)
2945
1e54db1a 2946=head2 How does Perl store UTF-8 strings?
a422fd2d 2947
61ad4b94 2948Currently, Perl deals with UTF-8 strings and non-UTF-8 strings
10e2eb10
FC
2949slightly differently. A flag in the SV, C<SVf_UTF8>, indicates that the
2950string is internally encoded as UTF-8. Without it, the byte value is the
61ad4b94
KW
2951codepoint number and vice versa. This flag is only meaningful if the SV
2952is C<SvPOK> or immediately after stringification via C<SvPV> or a
2953similar macro. You can check and manipulate this flag with the
2575c402 2954following macros:
a422fd2d
SC
2955
2956 SvUTF8(sv)
2957 SvUTF8_on(sv)
2958 SvUTF8_off(sv)
2959
2960This flag has an important effect on Perl's treatment of the string: if
61ad4b94 2961UTF-8 data is not properly distinguished, regular expressions,
a422fd2d 2962C<length>, C<substr> and other string handling operations will have
61ad4b94 2963undesirable (wrong) results.
a422fd2d
SC
2964
2965The problem comes when you have, for instance, a string that isn't
61ad4b94 2966flagged as UTF-8, and contains a byte sequence that could be UTF-8 --
1e54db1a 2967especially when combining non-UTF-8 and UTF-8 strings.
a422fd2d 2968
61ad4b94
KW
2969Never forget that the C<SVf_UTF8> flag is separate from the PV value; you
2970need to be sure you don't accidentally knock it off while you're
10e2eb10 2971manipulating SVs. More specifically, you cannot expect to do this:
a422fd2d
SC
2972
2973 SV *sv;
2974 SV *nsv;
2975 STRLEN len;
2976 char *p;
2977
2978 p = SvPV(sv, len);
2979 frobnicate(p);
2980 nsv = newSVpvn(p, len);
2981
2982The C<char*> string does not tell you the whole story, and you can't
10e2eb10 2983copy or reconstruct an SV just by copying the string value. Check if the
c31cc9fc
FC
2984old SV has the UTF8 flag set (I<after> the C<SvPV> call), and act
2985accordingly:
a422fd2d
SC
2986
2987 p = SvPV(sv, len);
6db25795
KW
2988 is_utf8 = SvUTF8(sv);
2989 frobnicate(p, is_utf8);
a422fd2d 2990 nsv = newSVpvn(p, len);
6db25795 2991 if (is_utf8)
a422fd2d
SC
2992 SvUTF8_on(nsv);
2993
6db25795
KW
2994In the above, your C<frobnicate> function has been changed to be made
2995aware of whether or not it's dealing with UTF-8 data, so that it can
2996handle the string appropriately.
a422fd2d 2997
3a2263fe 2998Since just passing an SV to an XS function and copying the data of
2575c402 2999the SV is not enough to copy the UTF8 flags, even less right is just
61ad4b94 3000passing a S<C<char *>> to an XS function.
3a2263fe 3001
dc83bf8e 3002For full generality, use the L<C<DO_UTF8>|perlapi/DO_UTF8> macro to see if the
6db25795
KW
3003string in an SV is to be I<treated> as UTF-8. This takes into account
3004if the call to the XS function is being made from within the scope of
3005L<S<C<use bytes>>|bytes>. If so, the underlying bytes that comprise the
3006UTF-8 string are to be exposed, rather than the character they
3007represent. But this pragma should only really be used for debugging and
3008perhaps low-level testing at the byte level. Hence most XS code need
3009not concern itself with this, but various areas of the perl core do need
3010to support it.
3011
3012And this isn't the whole story. Starting in Perl v5.12, strings that
3013aren't encoded in UTF-8 may also be treated as Unicode under various
6e31cdd1 3014conditions (see L<perlunicode/ASCII Rules versus Unicode Rules>).
6db25795
KW
3015This is only really a problem for characters whose ordinals are between
3016128 and 255, and their behavior varies under ASCII versus Unicode rules
3017in ways that your code cares about (see L<perlunicode/The "Unicode Bug">).
3018There is no published API for dealing with this, as it is subject to
3019change, but you can look at the code for C<pp_lc> in F<pp.c> for an
3020example as to how it's currently done.
3021
1e54db1a 3022=head2 How do I convert a string to UTF-8?
a422fd2d 3023
2575c402 3024If you're mixing UTF-8 and non-UTF-8 strings, it is necessary to upgrade
61ad4b94 3025the non-UTF-8 strings to UTF-8. If you've got an SV, the easiest way to do
2575c402 3026this is:
a422fd2d
SC
3027
3028 sv_utf8_upgrade(sv);
3029
3030However, you must not do this, for example:
3031
3032 if (!SvUTF8(left))
3033 sv_utf8_upgrade(left);
3034
3035If you do this in a binary operator, you will actually change one of the
b1866b2d 3036strings that came into the operator, and, while it shouldn't be noticeable
2575c402 3037by the end user, it can cause problems in deficient code.
a422fd2d 3038
1e54db1a 3039Instead, C<bytes_to_utf8> will give you a UTF-8-encoded B<copy> of its
10e2eb10
FC
3040string argument. This is useful for having the data available for
3041comparisons and so on, without harming the original SV. There's also
a422fd2d
SC
3042C<utf8_to_bytes> to go the other way, but naturally, this will fail if
3043the string contains any characters above 255 that can't be represented
3044in a single byte.
3045
6db25795
KW
3046=head2 How do I compare strings?
3047
3048L<perlapi/sv_cmp> and L<perlapi/sv_cmp_flags> do a lexigraphic
3049comparison of two SV's, and handle UTF-8ness properly. Note, however,
3050that Unicode specifies a much fancier mechanism for collation, available
3051via the L<Unicode::Collate> module.
3052
3053To just compare two strings for equality/non-equality, you can just use
3054L<C<memEQ()>|perlapi/memEQ> and L<C<memNE()>|perlapi/memEQ> as usual,
3055except the strings must be both UTF-8 or not UTF-8 encoded.
3056
3057To compare two strings case-insensitively, use
3058L<C<foldEQ_utf8()>|perlapi/foldEQ_utf8> (the strings don't have to have
3059the same UTF-8ness).
3060
a422fd2d
SC
3061=head2 Is there anything else I need to know?
3062
10e2eb10 3063Not really. Just remember these things:
a422fd2d
SC
3064
3065=over 3
3066
3067=item *
3068
6db25795
KW
3069There's no way to tell if a S<C<char *>> or S<C<U8 *>> string is UTF-8
3070or not. But you can tell if an SV is to be treated as UTF-8 by calling
3071C<DO_UTF8> on it, after stringifying it with C<SvPV> or a similar
3072macro. And, you can tell if SV is actually UTF-8 (even if it is not to
3073be treated as such) by looking at its C<SvUTF8> flag (again after
3074stringifying it). Don't forget to set the flag if something should be
3075UTF-8.
3076Treat the flag as part of the PV, even though it's not -- if you pass on
3077the PV to somewhere, pass on the flag too.
a422fd2d
SC
3078
3079=item *
3080
4b88fb76 3081If a string is UTF-8, B<always> use C<utf8_to_uvchr_buf> to get at the value,
3a2263fe 3082unless C<UTF8_IS_INVARIANT(*s)> in which case you can use C<*s>.
a422fd2d
SC
3083
3084=item *
3085
61ad4b94
KW
3086When writing a character UV to a UTF-8 string, B<always> use
3087C<uvchr_to_utf8>, unless C<UVCHR_IS_INVARIANT(uv))> in which case
3a2263fe 3088you can use C<*s = uv>.
a422fd2d
SC
3089
3090=item *
3091
10e2eb10
FC
3092Mixing UTF-8 and non-UTF-8 strings is
3093tricky. Use C<bytes_to_utf8> to get
2bbc8d55 3094a new string which is UTF-8 encoded, and then combine them.
a422fd2d
SC
3095
3096=back
3097
53e06cf0
SC
3098=head1 Custom Operators
3099
2a0fd0f1 3100Custom operator support is an experimental feature that allows you to
10e2eb10 3101define your own ops. This is primarily to allow the building of
53e06cf0
SC
3102interpreters for other languages in the Perl core, but it also allows
3103optimizations through the creation of "macro-ops" (ops which perform the
3104functions of multiple ops which are usually executed together, such as
1aa6ea50 3105C<gvsv, gvsv, add>.)
53e06cf0 3106
10e2eb10 3107This feature is implemented as a new op type, C<OP_CUSTOM>. The Perl
53e06cf0 3108core does not "know" anything special about this op type, and so it will
10e2eb10 3109not be involved in any optimizations. This also means that you can
61ad4b94
KW
3110define your custom ops to be any op structure -- unary, binary, list and
3111so on -- you like.
53e06cf0 3112
10e2eb10
FC
3113It's important to know what custom operators won't do for you. They
3114won't let you add new syntax to Perl, directly. They won't even let you
3115add new keywords, directly. In fact, they won't change the way Perl
3116compiles a program at all. You have to do those changes yourself, after
3117Perl has compiled the program. You do this either by manipulating the op
53e06cf0
SC
3118tree using a C<CHECK> block and the C<B::Generate> module, or by adding
3119a custom peephole optimizer with the C<optimize> module.
3120
3121When you do this, you replace ordinary Perl ops with custom ops by
407f86e1 3122creating ops with the type C<OP_CUSTOM> and the C<op_ppaddr> of your own
10e2eb10
FC
3123PP function. This should be defined in XS code, and should look like
3124the PP ops in C<pp_*.c>. You are responsible for ensuring that your op
53e06cf0
SC
3125takes the appropriate number of values from the stack, and you are
3126responsible for adding stack marks if necessary.
3127
3128You should also "register" your op with the Perl interpreter so that it
10e2eb10 3129can produce sensible error and warning messages. Since it is possible to
53e06cf0 3130have multiple custom ops within the one "logical" op type C<OP_CUSTOM>,
9733086d 3131Perl uses the value of C<< o->op_ppaddr >> to determine which custom op
10e2eb10 3132it is dealing with. You should create an C<XOP> structure for each
9733086d
BM
3133ppaddr you use, set the properties of the custom op with
3134C<XopENTRY_set>, and register the structure against the ppaddr using
10e2eb10 3135C<Perl_custom_op_register>. A trivial example might look like:
9733086d
BM
3136
3137 static XOP my_xop;
3138 static OP *my_pp(pTHX);
3139
3140 BOOT:
3141 XopENTRY_set(&my_xop, xop_name, "myxop");
3142 XopENTRY_set(&my_xop, xop_desc, "Useless custom op");
3143 Perl_custom_op_register(aTHX_ my_pp, &my_xop);
3144
3145The available fields in the structure are:
3146
3147=over 4
3148
3149=item xop_name
3150
10e2eb10 3151A short name for your op. This will be included in some error messages,
9733086d
BM
3152and will also be returned as C<< $op->name >> by the L<B|B> module, so
3153it will appear in the output of module like L<B::Concise|B::Concise>.
3154
3155=item xop_desc
3156
3157A short description of the function of the op.
3158
3159=item xop_class
3160
10e2eb10 3161Which of the various C<*OP> structures this op uses. This should be one of
9733086d
BM
3162the C<OA_*> constants from F<op.h>, namely
3163
3164=over 4
3165
3166=item OA_BASEOP
3167
3168=item OA_UNOP
3169
3170=item OA_BINOP
3171
3172=item OA_LOGOP
3173
3174=item OA_LISTOP
3175
3176=item OA_PMOP
3177
3178=item OA_SVOP
3179
3180=item OA_PADOP
3181
3182=item OA_PVOP_OR_SVOP
3183
10e2eb10 3184This should be interpreted as 'C<PVOP>' only. The C<_OR_SVOP> is because
9733086d
BM
3185the only core C<PVOP>, C<OP_TRANS>, can sometimes be a C<SVOP> instead.
3186
3187=item OA_LOOP
3188
3189=item OA_COP
3190
3191=back
3192
3193The other C<OA_*> constants should not be used.
3194
3195=item xop_peep
3196
3197This member is of type C<Perl_cpeep_t>, which expands to C<void
10e2eb10 3198(*Perl_cpeep_t)(aTHX_ OP *o, OP *oldop)>. If it is set, this function
9733086d 3199will be called from C<Perl_rpeep> when ops of this type are encountered
10e2eb10 3200by the peephole optimizer. I<o> is the OP that needs optimizing;
9733086d
BM
3201I<oldop> is the previous OP optimized, whose C<op_next> points to I<o>.
3202
3203=back
53e06cf0 3204
e7d4c058 3205C<B::Generate> directly supports the creation of custom ops by name.
53e06cf0 3206
bf5e9371
DM
3207
3208=head1 Dynamic Scope and the Context Stack
3209
3210B<Note:> this section describes a non-public internal API that is subject
3211to change without notice.
3212
3213=head2 Introduction to the context stack
3214
3215In Perl, dynamic scoping refers to the runtime nesting of things like
3216subroutine calls, evals etc, as well as the entering and exiting of block
3217scopes. For example, the restoring of a C<local>ised variable is
3218determined by the dynamic scope.
3219
3220Perl tracks the dynamic scope by a data structure called the context
3221stack, which is an array of C<PERL_CONTEXT> structures, and which is
3222itself a big union for all the types of context. Whenever a new scope is
3223entered (such as a block, a C<for> loop, or a subroutine call), a new
3224context entry is pushed onto the stack. Similarly when leaving a block or
3225returning from a subroutine call etc. a context is popped. Since the
3226context stack represents the current dynamic scope, it can be searched.
3227For example, C<next LABEL> searches back through the stack looking for a
3228loop context that matches the label; C<return> pops contexts until it
3229finds a sub or eval context or similar; C<caller> examines sub contexts on
3230the stack.
3231
3232Each context entry is labelled with a context type, C<cx_type>. Typical
3233context types are C<CXt_SUB>, C<CXt_EVAL> etc., as well as C<CXt_BLOCK>
3234and C<CXt_NULL> which represent a basic scope (as pushed by C<pp_enter>)
3235and a sort block. The type determines which part of the context union are
3236valid.
3237
3238The main division in the context struct is between a substitution scope
3239(C<CXt_SUBST>) and block scopes, which are everything else. The former is
d7c7f8cb 3240just used while executing C<s///e>, and won't be discussed further
bf5e9371
DM
3241here.
3242
3243All the block scope types share a common base, which corresponds to
3244C<CXt_BLOCK>. This stores the old values of various scope-related
3245variables like C<PL_curpm>, as well as information about the current
3246scope, such as C<gimme>. On scope exit, the old variables are restored.
3247
3248Particular block scope types store extra per-type information. For
3249example, C<CXt_SUB> stores the currently executing CV, while the various
3250for loop types might hold the original loop variable SV. On scope exit,
3251the per-type data is processed; for example the CV has its reference count
3252decremented, and the original loop variable is restored.
3253
3254The macro C<cxstack> returns the base of the current context stack, while
3255C<cxstack_ix> is the index of the current frame within that stack.
3256
3257In fact, the context stack is actually part of a stack-of-stacks system;
3258whenever something unusual is done such as calling a C<DESTROY> or tie
3259handler, a new stack is pushed, then popped at the end.
3260
3261Note that the API described here changed considerably in perl 5.24; prior
3262to that, big macros like C<PUSHBLOCK> and C<POPSUB> were used; in 5.24
3263they were replaced by the inline static functions described below. In
3264addition, the ordering and detail of how these macros/function work
3265changed in many ways, often subtly. In particular they didn't handle
3266saving the savestack and temps stack positions, and required additional
3267C<ENTER>, C<SAVETMPS> and C<LEAVE> compared to the new functions. The
3268old-style macros will not be described further.
3269
3270
3271=head2 Pushing contexts
3272
3273For pushing a new context, the two basic functions are
3274C<cx = cx_pushblock()>, which pushes a new basic context block and returns
3275its address, and a family of similar functions with names like
3276C<cx_pushsub(cx)> which populate the additional type-dependent fields in
3277the C<cx> struct. Note that C<CXt_NULL> and C<CXt_BLOCK> don't have their
3278own push functions, as they don't store any data beyond that pushed by
3279C<cx_pushblock>.
3280
3281The fields of the context struct and the arguments to the C<cx_*>
3282functions are subject to change between perl releases, representing
3283whatever is convenient or efficient for that release.
3284
3285A typical context stack pushing can be found in C<pp_entersub>; the
3286following shows a simplified and stripped-down example of a non-XS call,
3287along with comments showing roughly what each function does.
3288
61f554bd
KW
3289 dMARK;
3290 U8 gimme = GIMME_V;
3291 bool hasargs = cBOOL(PL_op->op_flags & OPf_STACKED);
3292 OP *retop = PL_op->op_next;
3293 I32 old_ss_ix = PL_savestack_ix;
3294 CV *cv = ....;
3295
3296 /* ... make mortal copies of stack args which are PADTMPs here ... */
3297
3298 /* ... do any additional savestack pushes here ... */
3299
3300 /* Now push a new context entry of type 'CXt_SUB'; initially just
3301 * doing the actions common to all block types: */
3302
3303 cx = cx_pushblock(CXt_SUB, gimme, MARK, old_ss_ix);
3304
3305 /* this does (approximately):
3306 CXINC; /* cxstack_ix++ (grow if necessary) */
3307 cx = CX_CUR(); /* and get the address of new frame */
3308 cx->cx_type = CXt_SUB;
3309 cx->blk_gimme = gimme;
3310 cx->blk_oldsp = MARK - PL_stack_base;
3311 cx->blk_oldsaveix = old_ss_ix;
3312 cx->blk_oldcop = PL_curcop;
3313 cx->blk_oldmarksp = PL_markstack_ptr - PL_markstack;
3314 cx->blk_oldscopesp = PL_scopestack_ix;
3315 cx->blk_oldpm = PL_curpm;
3316 cx->blk_old_tmpsfloor = PL_tmps_floor;
3317
3318 PL_tmps_floor = PL_tmps_ix;
3319 */
bf5e9371
DM
3320
3321
61f554bd
KW
3322 /* then update the new context frame with subroutine-specific info,
3323 * such as the CV about to be executed: */
bf5e9371 3324
61f554bd 3325 cx_pushsub(cx, cv, retop, hasargs);
bf5e9371 3326
61f554bd
KW
3327 /* this does (approximately):
3328 cx->blk_sub.cv = cv;
3329 cx->blk_sub.olddepth = CvDEPTH(cv);
3330 cx->blk_sub.prevcomppad = PL_comppad;
3331 cx->cx_type |= (hasargs) ? CXp_HASARGS : 0;
3332 cx->blk_sub.retop = retop;
3333 SvREFCNT_inc_simple_void_NN(cv);
3334 */
bf5e9371
DM
3335
3336Note that C<cx_pushblock()> sets two new floors: for the args stack (to
3337C<MARK>) and the temps stack (to C<PL_tmps_ix>). While executing at this
3338scope level, every C<nextstate> (amongst others) will reset the args and
3339tmps stack levels to these floors. Note that since C<cx_pushblock> uses
3340the current value of C<PL_tmps_ix> rather than it being passed as an arg,
3341this dictates at what point C<cx_pushblock> should be called. In
3342particular, any new mortals which should be freed only on scope exit
3343(rather than at the next C<nextstate>) should be created first.
3344
3345Most callers of C<cx_pushblock> simply set the new args stack floor to the
3346top of the previous stack frame, but for C<CXt_LOOP_LIST> it stores the
3347items being iterated over on the stack, and so sets C<blk_oldsp> to the
3348top of these items instead. Note that, contrary to its name, C<blk_oldsp>
3349doesn't always represent the value to restore C<PL_stack_sp> to on scope
3350exit.
3351
3352Note the early capture of C<PL_savestack_ix> to C<old_ss_ix>, which is
3353later passed as an arg to C<cx_pushblock>. In the case of C<pp_entersub>,
3354this is because, although most values needing saving are stored in fields
3355of the context struct, an extra value needs saving only when the debugger
3356is running, and it doesn't make sense to bloat the struct for this rare
3357case. So instead it is saved on the savestack. Since this value gets
3358calculated and saved before the context is pushed, it is necessary to pass
3359the old value of C<PL_savestack_ix> to C<cx_pushblock>, to ensure that the
3360saved value gets freed during scope exit. For most users of
3361C<cx_pushblock>, where nothing needs pushing on the save stack,
3362C<PL_savestack_ix> is just passed directly as an arg to C<cx_pushblock>.
3363
3364Note that where possible, values should be saved in the context struct
3365rather than on the save stack; it's much faster that way.
3366
3367Normally C<cx_pushblock> should be immediately followed by the appropriate
3368C<cx_pushfoo>, with nothing between them; this is because if code
3369in-between could die (e.g. a warning upgraded to fatal), then the context
3370stack unwinding code in C<dounwind> would see (in the example above) a
3371C<CXt_SUB> context frame, but without all the subroutine-specific fields
3372set, and crashes would soon ensue.
3373
3374Where the two must be separate, initially set the type to C<CXt_NULL> or
3375C<CXt_BLOCK>, and later change it to C<CXt_foo> when doing the
3376C<cx_pushfoo>. This is exactly what C<pp_enteriter> does, once it's
3377determined which type of loop it's pushing.
3378
3379=head2 Popping contexts
3380
3381Contexts are popped using C<cx_popsub()> etc. and C<cx_popblock()>. Note
3382however, that unlike C<cx_pushblock>, neither of these functions actually
3383decrement the current context stack index; this is done separately using
3384C<CX_POP()>.
3385
3386There are two main ways that contexts are popped. During normal execution
3387as scopes are exited, functions like C<pp_leave>, C<pp_leaveloop> and
3388C<pp_leavesub> process and pop just one context using C<cx_popfoo> and
3389C<cx_popblock>. On the other hand, things like C<pp_return> and C<next>
3390may have to pop back several scopes until a sub or loop context is found,
3391and exceptions (such as C<die>) need to pop back contexts until an eval
3392context is found. Both of these are accomplished by C<dounwind()>, which
3393is capable of processing and popping all contexts above the target one.
3394
3395Here is a typical example of context popping, as found in C<pp_leavesub>
3396(simplified slightly):
3397
61f554bd
KW
3398 U8 gimme;
3399 PERL_CONTEXT *cx;
3400 SV **oldsp;
3401 OP *retop;
bf5e9371 3402
61f554bd 3403 cx = CX_CUR();
bf5e9371 3404
61f554bd
KW
3405 gimme = cx->blk_gimme;
3406 oldsp = PL_stack_base + cx->blk_oldsp; /* last arg of previous frame */
bf5e9371 3407
61f554bd
KW
3408 if (gimme == G_VOID)
3409 PL_stack_sp = oldsp;
3410 else
3411 leave_adjust_stacks(oldsp, oldsp, gimme, 0);
bf5e9371 3412
61f554bd
KW
3413 CX_LEAVE_SCOPE(cx);
3414 cx_popsub(cx);
3415 cx_popblock(cx);
3416 retop = cx->blk_sub.retop;
3417 CX_POP(cx);
bf5e9371 3418
61f554bd 3419 return retop;
bf5e9371
DM
3420
3421The steps above are in a very specific order, designed to be the reverse
3422order of when the context was pushed. The first thing to do is to copy
3423and/or protect any any return arguments and free any temps in the current
3424scope. Scope exits like an rvalue sub normally return a mortal copy of
3425their return args (as opposed to lvalue subs). It is important to make
3426this copy before the save stack is popped or variables are restored, or
3427bad things like the following can happen:
3428
3429 sub f { my $x =...; $x } # $x freed before we get to copy it
3430 sub f { /(...)/; $1 } # PL_curpm restored before $1 copied
3431
3432Although we wish to free any temps at the same time, we have to be careful
3433not to free any temps which are keeping return args alive; nor to free the
3434temps we have just created while mortal copying return args. Fortunately,
3435C<leave_adjust_stacks()> is capable of making mortal copies of return args,
3436shifting args down the stack, and only processing those entries on the
3437temps stack that are safe to do so.
3438
3439In void context no args are returned, so it's more efficient to skip
3440calling C<leave_adjust_stacks()>. Also in void context, a C<nextstate> op
3441is likely to be imminently called which will do a C<FREETMPS>, so there's
3442no need to do that either.
3443
3444The next step is to pop savestack entries: C<CX_LEAVE_SCOPE(cx)> is just
3445defined as C<<LEAVE_SCOPE(cx->blk_oldsaveix)>>. Note that during the
3446popping, it's possible for perl to call destructors, call C<STORE> to undo
3447localisations of tied vars, and so on. Any of these can die or call
3448C<exit()>. In this case, C<dounwind()> will be called, and the current
3449context stack frame will be re-processed. Thus it is vital that all steps
3450in popping a context are done in such a way to support reentrancy. The
3451other alternative, of decrementing C<cxstack_ix> I<before> processing the
3452frame, would lead to leaks and the like if something died halfway through,
3453or overwriting of the current frame.
3454
3455C<CX_LEAVE_SCOPE> itself is safely re-entrant: if only half the savestack
3456items have been popped before dying and getting trapped by eval, then the
3457C<CX_LEAVE_SCOPE>s in C<dounwind> or C<pp_leaveeval> will continue where
3458the first one left off.
3459
3460The next step is the type-specific context processing; in this case
3461C<cx_popsub>. In part, this looks like:
3462
3463 cv = cx->blk_sub.cv;
3464 CvDEPTH(cv) = cx->blk_sub.olddepth;
3465 cx->blk_sub.cv = NULL;
3466 SvREFCNT_dec(cv);
3467
3468where its processing the just-executed CV. Note that before it decrements
3469the CV's reference count, it nulls the C<blk_sub.cv>. This means that if
3470it re-enters, the CV won't be freed twice. It also means that you can't
3471rely on such type-specific fields having useful values after the return
3472from C<cx_popfoo>.
3473
3474Next, C<cx_popblock> restores all the various interpreter vars to their
3475previous values or previous high water marks; it expands to:
3476
3477 PL_markstack_ptr = PL_markstack + cx->blk_oldmarksp;
3478 PL_scopestack_ix = cx->blk_oldscopesp;
3479 PL_curpm = cx->blk_oldpm;
3480 PL_curcop = cx->blk_oldcop;
3481 PL_tmps_floor = cx->blk_old_tmpsfloor;
3482
3483Note that it I<doesn't> restore C<PL_stack_sp>; as mentioned earlier,
3484which value to restore it to depends on the context type (specifically
3485C<for (list) {}>), and what args (if any) it returns; and that will
3486already have been sorted out earlier by C<leave_adjust_stacks()>.
3487
3488Finally, the context stack pointer is actually decremented by C<CX_POP(cx)>.
3489After this point, it's possible that that the current context frame could
3490be overwritten by other contexts being pushed. Although things like ties
3491and C<DESTROY> are supposed to work within a new context stack, it's best
3492not to assume this. Indeed on debugging builds, C<CX_POP(cx)> deliberately
3493sets C<cx> to null to detect code that is still relying on the field
3494values in that context frame. Note in the C<pp_leavesub()> example above,
3495we grab C<blk_sub.retop> I<before> calling C<CX_POP>.
3496
3497=head2 Redoing contexts
3498
3499Finally, there is C<cx_topblock(cx)>, which acts like a super-C<nextstate>
3500as regards to resetting various vars to their base values. It is used in
3501places like C<pp_next>, C<pp_redo> and C<pp_goto> where rather than
3502exiting a scope, we want to re-initialise the scope. As well as resetting