This is a live mirror of the Perl 5 development currently hosted at
5.6.0 & 5.7.0 VMS TZ fix for VMS6.2 and earlier
[perl5.git] / pod / perlhack.pod
1=head1 NAME
3perlhack - How to hack at the Perl internals
7This document attempts to explain how Perl development takes place,
8and ends with some suggestions for people wanting to become bona fide
11The perl5-porters mailing list is where the Perl standard distribution
12is maintained and developed. The list can get anywhere from 10 to 150
13messages a day, depending on the heatedness of the debate. Most days
14there are two or three patches, extensions, features, or bugs being
15discussed at a time.
17A searchable archive of the list is at:
21The list is also archived under the usenet group name
22C<perl.porters-gw> at:
26List subscribers (the porters themselves) come in several flavours.
27Some are quiet curious lurkers, who rarely pitch in and instead watch
28the ongoing development to ensure they're forewarned of new changes or
29features in Perl. Some are representatives of vendors, who are there
30to make sure that Perl continues to compile and work on their
31platforms. Some patch any reported bug that they know how to fix,
32some are actively patching their pet area (threads, Win32, the regexp
33engine), while others seem to do nothing but complain. In other
34words, it's your usual mix of technical people.
36Over this group of porters presides Larry Wall. He has the final word
37in what does and does not change in the Perl language. Various
38releases of Perl are shepherded by a ``pumpking'', a porter
39responsible for gathering patches, deciding on a patch-by-patch
40feature-by-feature basis what will and will not go into the release.
41For instance, Gurusamy Sarathy is the pumpking for the 5.6 release of
44In addition, various people are pumpkings for different things. For
45instance, Andy Dougherty and Jarkko Hietaniemi share the I<Configure>
46pumpkin, and Tom Christiansen is the documentation pumpking.
48Larry sees Perl development along the lines of the US government:
49there's the Legislature (the porters), the Executive branch (the
50pumpkings), and the Supreme Court (Larry). The legislature can
51discuss and submit patches to the executive branch all they like, but
52the executive branch is free to veto them. Rarely, the Supreme Court
53will side with the executive branch over the legislature, or the
54legislature over the executive branch. Mostly, however, the
55legislature and the executive branch are supposed to get along and
56work out their differences without impeachment or court cases.
58You might sometimes see reference to Rule 1 and Rule 2. Larry's power
59as Supreme Court is expressed in The Rules:
61=over 4
63=item 1
65Larry is always by definition right about how Perl should behave.
66This means he has final veto power on the core functionality.
68=item 2
70Larry is allowed to change his mind about any matter at a later date,
71regardless of whether he previously invoked Rule 1.
75Got that? Larry is always right, even when he was wrong. It's rare
76to see either Rule exercised, but they are often alluded to.
78New features and extensions to the language are contentious, because
79the criteria used by the pumpkings, Larry, and other porters to decide
80which features should be implemented and incorporated are not codified
81in a few small design goals as with some other languages. Instead,
82the heuristics are flexible and often difficult to fathom. Here is
83one person's list, roughly in decreasing order of importance, of
84heuristics that new features have to be weighed against:
86=over 4
88=item Does concept match the general goals of Perl?
90These haven't been written anywhere in stone, but one approximation
93 1. Keep it fast, simple, and useful.
94 2. Keep features/concepts as orthogonal as possible.
95 3. No arbitrary limits (platforms, data sizes, cultures).
96 4. Keep it open and exciting to use/patch/advocate Perl everywhere.
97 5. Either assimilate new technologies, or build bridges to them.
99=item Where is the implementation?
101All the talk in the world is useless without an implementation. In
102almost every case, the person or people who argue for a new feature
103will be expected to be the ones who implement it. Porters capable
104of coding new features have their own agendas, and are not available
105to implement your (possibly good) idea.
107=item Backwards compatibility
109It's a cardinal sin to break existing Perl programs. New warnings are
110contentious--some say that a program that emits warnings is not
111broken, while others say it is. Adding keywords has the potential to
112break programs, changing the meaning of existing token sequences or
113functions might break programs.
115=item Could it be a module instead?
117Perl 5 has extension mechanisms, modules and XS, specifically to avoid
118the need to keep changing the Perl interpreter. You can write modules
119that export functions, you can give those functions prototypes so they
120can be called like built-in functions, you can even write XS code to
121mess with the runtime data structures of the Perl interpreter if you
122want to implement really complicated things. If it can be done in a
123module instead of in the core, it's highly unlikely to be added.
125=item Is the feature generic enough?
127Is this something that only the submitter wants added to the language,
128or would it be broadly useful? Sometimes, instead of adding a feature
129with a tight focus, the porters might decide to wait until someone
130implements the more generalized feature. For instance, instead of
131implementing a ``delayed evaluation'' feature, the porters are waiting
132for a macro system that would permit delayed evaluation and much more.
134=item Does it potentially introduce new bugs?
136Radical rewrites of large chunks of the Perl interpreter have the
137potential to introduce new bugs. The smaller and more localized the
138change, the better.
140=item Does it preclude other desirable features?
142A patch is likely to be rejected if it closes off future avenues of
143development. For instance, a patch that placed a true and final
144interpretation on prototypes is likely to be rejected because there
145are still options for the future of prototypes that haven't been
148=item Is the implementation robust?
150Good patches (tight code, complete, correct) stand more chance of
151going in. Sloppy or incorrect patches might be placed on the back
152burner until the pumpking has time to fix, or might be discarded
153altogether without further notice.
155=item Is the implementation generic enough to be portable?
157The worst patches make use of a system-specific features. It's highly
158unlikely that nonportable additions to the Perl language will be
161=item Is there enough documentation?
163Patches without documentation are probably ill-thought out or
164incomplete. Nothing can be added without documentation, so submitting
165a patch for the appropriate manpages as well as the source code is
166always a good idea. If appropriate, patches should add to the test
167suite as well.
169=item Is there another way to do it?
171Larry said ``Although the Perl Slogan is I<There's More Than One Way
172to Do It>, I hesitate to make 10 ways to do something''. This is a
173tricky heuristic to navigate, though--one man's essential addition is
174another man's pointless cruft.
176=item Does it create too much work?
178Work for the pumpking, work for Perl programmers, work for module
179authors, ... Perl is supposed to be easy.
181=item Patches speak louder than words
183Working code is always preferred to pie-in-the-sky ideas. A patch to
184add a feature stands a much higher chance of making it to the language
185than does a random feature request, no matter how fervently argued the
186request might be. This ties into ``Will it be useful?'', as the fact
187that someone took the time to make the patch demonstrates a strong
188desire for the feature.
192If you're on the list, you might hear the word ``core'' bandied
193around. It refers to the standard distribution. ``Hacking on the
194core'' means you're changing the C source code to the Perl
195interpreter. ``A core module'' is one that ships with Perl.
197The source code to the Perl interpreter, in its different versions, is
198kept in a repository managed by a revision control system (which is
199currently the Perforce program, see The
200pumpkings and a few others have access to the repository to check in
201changes. Periodically the pumpking for the development version of Perl
202will release a new version, so the rest of the porters can see what's
203changed. The current state of the main trunk of repository, and patches
204that describe the individual changes that have happened since the last
205public release are available at this location:
209Selective parts are also visible via the rsync protocol. To get all
210the individual changes to the mainline since the last development
211release, use the following command:
6c7df0a8 213 rsync -avz rsync:// perl-diffs
215Use this to get the latest source tree in full:
6c7df0a8 217 rsync -avz rsync:// perl-current
219Needless to say, the source code in perl-current is usually in a perpetual
220state of evolution. You should expect it to be very buggy. Do B<not> use
221it for any purpose other than testing and development.
223Always submit patches to I<>. This lets other
224porters review your patch, which catches a surprising number of errors
225in patches. Either use the diff program (available in source code
226form from I<>), or use Johan Vromans'
227I<makepatch> (available from I<CPAN/authors/id/JV/>). Unified diffs
228are preferred, but context diffs are accepted. Do not send RCS-style
229diffs or diffs without context lines. More information is given in
230the I<Porting/patching.pod> file in the Perl source distribution.
231Please patch against the latest B<development> version (e.g., if
232you're fixing a bug in the 5.005 track, patch against the latest
2335.005_5x version). Only patches that survive the heat of the
234development branch get applied to maintenance versions.
236Your patch should update the documentation and test suite.
238To report a bug in Perl, use the program I<perlbug> which comes with
239Perl (if you can't get Perl to work, send mail to the address
240I<> or I<>). Reporting bugs through
241I<perlbug> feeds into the automated bug-tracking system, access to
242which is provided through the web at I<>. It
243often pays to check the archives of the perl5-porters mailing list to
244see whether the bug you're reporting has been reported before, and if
245so whether it was considered a bug. See above for the location of
246the searchable archives.
248The CPAN testers (I<>) are a group of
249volunteers who test CPAN modules on a variety of platforms. Perl Labs
250(I<>) automatically tests Perl source releases on
251platforms and gives feedback to the CPAN testers mailing list. Both
252efforts welcome volunteers.
e8cd7eae 253
254It's a good idea to read and lurk for a while before chipping in.
255That way you'll get to see the dynamic of the conversations, learn the
256personalities of the players, and hopefully be better prepared to make
257a useful contribution when do you speak up.
259If after all this you still think you want to join the perl5-porters
260mailing list, send mail to I<>. To
261unsubscribe, send mail to I<>.
e8cd7eae 262
263To hack on the Perl guts, you'll need to read the following things:
265=over 3
267=item L<perlguts>
269This is of paramount importance, since it's the documentation of what
270goes where in the Perl source. Read it over a couple of times and it
271might start to make sense - don't worry if it doesn't yet, because the
272best way to study it is to read it in conjunction with poking at Perl
273source, and we'll do that later on.
275You might also want to look at Gisle Aas's illustrated perlguts -
276there's no guarantee that this will be absolutely up-to-date with the
277latest documentation in the Perl core, but the fundamentals will be
278right. (
280=item L<perlxstut> and L<perlxs>
282A working knowledge of XSUB programming is incredibly useful for core
283hacking; XSUBs use techniques drawn from the PP code, the portion of the
284guts that actually executes a Perl program. It's a lot gentler to learn
285those techniques from simple examples and explanation than from the core
288=item L<perlapi>
290The documentation for the Perl API explains what some of the internal
291functions do, as well as the many macros used in the source.
293=item F<Porting/pumpkin.pod>
295This is a collection of words of wisdom for a Perl porter; some of it is
296only useful to the pumpkin holder, but most of it applies to anyone
297wanting to go about Perl development.
299=item The perl5-porters FAQ
301This is posted to perl5-porters at the beginning on every month, and
302should be available from; alternatively,
303you can get the FAQ emailed to you by sending mail to
304C<>. It contains hints on reading
305perl5-porters, information on how perl5-porters works and how Perl
306development in general works.
310=head2 Finding Your Way Around
312Perl maintenance can be split into a number of areas, and certain people
313(pumpkins) will have responsibility for each area. These areas sometimes
314correspond to files or directories in the source kit. Among the areas are:
316=over 3
318=item Core modules
320Modules shipped as part of the Perl core live in the F<lib/> and F<ext/>
321subdirectories: F<lib/> is for the pure-Perl modules, and F<ext/>
322contains the core XS modules.
324=item Documentation
326Documentation maintenance includes looking after everything in the
327F<pod/> directory, (as well as contributing new documentation) and
328the documentation to the modules in core.
330=item Configure
332The configure process is the way we make Perl portable across the
333myriad of operating systems it supports. Responsibility for the
334configure, build and installation process, as well as the overall
335portability of the core code rests with the configure pumpkin - others
336help out with individual operating systems.
338The files involved are the operating system directories, (F<win32/>,
339F<os2/>, F<vms/> and so on) the shell scripts which generate F<config.h>
340and F<Makefile>, as well as the metaconfig files which generate
341F<Configure>. (metaconfig isn't included in the core distribution.)
343=item Interpreter
345And of course, there's the core of the Perl interpreter itself. Let's
346have a look at that in a little more detail.
350Before we leave looking at the layout, though, don't forget that
351F<MANIFEST> contains not only the file names in the Perl distribution,
352but short descriptions of what's in them, too. For an overview of the
353important files, try this:
355 perl -lne 'print if /^[^\/]+\.[ch]\s+/' MANIFEST
357=head2 Elements of the interpreter
359The work of the interpreter has two main stages: compiling the code
360into the internal representation, or bytecode, and then executing it.
361L<perlguts/Compiled code> explains exactly how the compilation stage
364Here is a short breakdown of perl's operation:
366=over 3
368=item Startup
370The action begins in F<perlmain.c>. (or F<miniperlmain.c> for miniperl)
371This is very high-level code, enough to fit on a single screen, and it
372resembles the code found in L<perlembed>; most of the real action takes
373place in F<perl.c>
375First, F<perlmain.c> allocates some memory and constructs a Perl
378 1 PERL_SYS_INIT3(&argc,&argv,&env);
379 2
380 3 if (!PL_do_undump) {
381 4 my_perl = perl_alloc();
382 5 if (!my_perl)
383 6 exit(1);
384 7 perl_construct(my_perl);
385 8 PL_perl_destruct_level = 0;
386 9 }
388Line 1 is a macro, and its definition is dependent on your operating
389system. Line 3 references C<PL_do_undump>, a global variable - all
390global variables in Perl start with C<PL_>. This tells you whether the
391current running program was created with the C<-u> flag to perl and then
392F<undump>, which means it's going to be false in any sane context.
394Line 4 calls a function in F<perl.c> to allocate memory for a Perl
395interpreter. It's quite a simple function, and the guts of it looks like
398 my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
400Here you see an example of Perl's system abstraction, which we'll see
401later: C<PerlMem_malloc> is either your system's C<malloc>, or Perl's
402own C<malloc> as defined in F<malloc.c> if you selected that option at
403configure time.
405Next, in line 7, we construct the interpreter; this sets up all the
406special variables that Perl needs, the stacks, and so on.
408Now we pass Perl the command line options, and tell it to go:
410 exitstatus = perl_parse(my_perl, xs_init, argc, argv, (char **)NULL);
411 if (!exitstatus) {
412 exitstatus = perl_run(my_perl);
413 }
416C<perl_parse> is actually a wrapper around C<S_parse_body>, as defined
417in F<perl.c>, which processes the command line options, sets up any
418statically linked XS modules, opens the program and calls C<yyparse> to
419parse it.
421=item Parsing
423The aim of this stage is to take the Perl source, and turn it into an op
424tree. We'll see what one of those looks like later. Strictly speaking,
425there's three things going on here.
427C<yyparse>, the parser, lives in F<perly.c>, although you're better off
428reading the original YACC input in F<perly.y>. (Yes, Virginia, there
429B<is> a YACC grammar for Perl!) The job of the parser is to take your
430code and `understand' it, splitting it into sentences, deciding which
431operands go with which operators and so on.
433The parser is nobly assisted by the lexer, which chunks up your input
434into tokens, and decides what type of thing each token is: a variable
435name, an operator, a bareword, a subroutine, a core function, and so on.
436The main point of entry to the lexer is C<yylex>, and that and its
437associated routines can be found in F<toke.c>. Perl isn't much like
438other computer languages; it's highly context sensitive at times, it can
439be tricky to work out what sort of token something is, or where a token
440ends. As such, there's a lot of interplay between the tokeniser and the
441parser, which can get pretty frightening if you're not used to it.
443As the parser understands a Perl program, it builds up a tree of
444operations for the interpreter to perform during execution. The routines
445which construct and link together the various operations are to be found
446in F<op.c>, and will be examined later.
448=item Optimization
450Now the parsing stage is complete, and the finished tree represents
451the operations that the Perl interpreter needs to perform to execute our
452program. Next, Perl does a dry run over the tree looking for
453optimisations: constant expressions such as C<3 + 4> will be computed
454now, and the optimizer will also see if any multiple operations can be
455replaced with a single one. For instance, to fetch the variable C<$foo>,
456instead of grabbing the glob C<*foo> and looking at the scalar
457component, the optimizer fiddles the op tree to use a function which
458directly looks up the scalar in question. The main optimizer is C<peep>
459in F<op.c>, and many ops have their own optimizing functions.
461=item Running
463Now we're finally ready to go: we have compiled Perl byte code, and all
464that's left to do is run it. The actual execution is done by the
465C<runops_standard> function in F<run.c>; more specifically, it's done by
466these three innocent looking lines:
468 while ((PL_op = CALL_FPTR(PL_op->op_ppaddr)(aTHX))) {
470 }
472You may be more comfortable with the Perl version of that:
474 PERL_ASYNC_CHECK() while $Perl::op = &{$Perl::op->{function}};
476Well, maybe not. Anyway, each op contains a function pointer, which
477stipulates the function which will actually carry out the operation.
478This function will return the next op in the sequence - this allows for
479things like C<if> which choose the next op dynamically at run time.
480The C<PERL_ASYNC_CHECK> makes sure that things like signals interrupt
481execution if required.
483The actual functions called are known as PP code, and they're spread
484between four files: F<pp_hot.c> contains the `hot' code, which is most
485often used and highly optimized, F<pp_sys.c> contains all the
486system-specific functions, F<pp_ctl.c> contains the functions which
487implement control structures (C<if>, C<while> and the like) and F<pp.c>
488contains everything else. These are, if you like, the C code for Perl's
489built-in functions and operators.
493=head2 Internal Variable Types
495You should by now have had a look at L<perlguts>, which tells you about
496Perl's internal variable types: SVs, HVs, AVs and the rest. If not, do
497that now.
499These variables are used not only to represent Perl-space variables, but
500also any constants in the code, as well as some structures completely
501internal to Perl. The symbol table, for instance, is an ordinary Perl
502hash. Your code is represented by an SV as it's read into the parser;
503any program files you call are opened via ordinary Perl filehandles, and
504so on.
506The core L<Devel::Peek|Devel::Peek> module lets us examine SVs from a
507Perl program. Let's see, for instance, how Perl treats the constant
510 % perl -MDevel::Peek -e 'Dump("hello")'
511 1 SV = PV(0xa041450) at 0xa04ecbc
512 2 REFCNT = 1
514 4 PV = 0xa0484e0 "hello"\0
515 5 CUR = 5
516 6 LEN = 6
518Reading C<Devel::Peek> output takes a bit of practise, so let's go
519through it line by line.
521Line 1 tells us we're looking at an SV which lives at C<0xa04ecbc> in
522memory. SVs themselves are very simple structures, but they contain a
523pointer to a more complex structure. In this case, it's a PV, a
524structure which holds a string value, at location C<0xa041450>. Line 2
525is the reference count; there are no other references to this data, so
526it's 1.
528Line 3 are the flags for this SV - it's OK to use it as a PV, it's a
529read-only SV (because it's a constant) and the data is a PV internally.
530Next we've got the contents of the string, starting at location
533Line 5 gives us the current length of the string - note that this does
534B<not> include the null terminator. Line 6 is not the length of the
535string, but the length of the currently allocated buffer; as the string
536grows, Perl automatically extends the available storage via a routine
537called C<SvGROW>.
539You can get at any of these quantities from C very easily; just add
540C<Sv> to the name of the field shown in the snippet, and you've got a
541macro which will return the value: C<SvCUR(sv)> returns the current
542length of the string, C<SvREFCOUNT(sv)> returns the reference count,
543C<SvPV(sv, len)> returns the string itself with its length, and so on.
544More macros to manipulate these properties can be found in L<perlguts>.
546Let's take an example of manipulating a PV, from C<sv_catpvn>, in F<sv.c>
548 1 void
549 2 Perl_sv_catpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
550 3 {
551 4 STRLEN tlen;
552 5 char *junk;
554 6 junk = SvPV_force(sv, tlen);
555 7 SvGROW(sv, tlen + len + 1);
556 8 if (ptr == junk)
557 9 ptr = SvPVX(sv);
558 10 Move(ptr,SvPVX(sv)+tlen,len,char);
559 11 SvCUR(sv) += len;
560 12 *SvEND(sv) = '\0';
561 13 (void)SvPOK_only_UTF8(sv); /* validate pointer */
562 14 SvTAINT(sv);
563 15 }
565This is a function which adds a string, C<ptr>, of length C<len> onto
566the end of the PV stored in C<sv>. The first thing we do in line 6 is
567make sure that the SV B<has> a valid PV, by calling the C<SvPV_force>
568macro to force a PV. As a side effect, C<tlen> gets set to the current
569value of the PV, and the PV itself is returned to C<junk>.
b1866b2d 571In line 7, we make sure that the SV will have enough room to accommodate
572the old string, the new string and the null terminator. If C<LEN> isn't
573big enough, C<SvGROW> will reallocate space for us.
575Now, if C<junk> is the same as the string we're trying to add, we can
576grab the string directly from the SV; C<SvPVX> is the address of the PV
577in the SV.
579Line 10 does the actual catenation: the C<Move> macro moves a chunk of
580memory around: we move the string C<ptr> to the end of the PV - that's
581the start of the PV plus its current length. We're moving C<len> bytes
582of type C<char>. After doing so, we need to tell Perl we've extended the
583string, by altering C<CUR> to reflect the new length. C<SvEND> is a
584macro which gives us the end of the string, so that needs to be a
587Line 13 manipulates the flags; since we've changed the PV, any IV or NV
588values will no longer be valid: if we have C<$a=10; $a.="6";> we don't
589want to use the old IV of 10. C<SvPOK_only_utf8> is a special UTF8-aware
590version of C<SvPOK_only>, a macro which turns off the IOK and NOK flags
591and turns on POK. The final C<SvTAINT> is a macro which launders tainted
592data if taint mode is turned on.
594AVs and HVs are more complicated, but SVs are by far the most common
595variable type being thrown around. Having seen something of how we
596manipulate these, let's go on and look at how the op tree is
599=head2 Op Trees
601First, what is the op tree, anyway? The op tree is the parsed
602representation of your program, as we saw in our section on parsing, and
603it's the sequence of operations that Perl goes through to execute your
604program, as we saw in L</Running>.
606An op is a fundamental operation that Perl can perform: all the built-in
607functions and operators are ops, and there are a series of ops which
608deal with concepts the interpreter needs internally - entering and
609leaving a block, ending a statement, fetching a variable, and so on.
611The op tree is connected in two ways: you can imagine that there are two
612"routes" through it, two orders in which you can traverse the tree.
613First, parse order reflects how the parser understood the code, and
614secondly, execution order tells perl what order to perform the
615operations in.
617The easiest way to examine the op tree is to stop Perl after it has
618finished parsing, and get it to dump out the tree. This is exactly what
619the compiler backends L<B::Terse|B::Terse> and L<B::Debug|B::Debug> do.
621Let's have a look at how Perl sees C<$a = $b + $c>:
623 % perl -MO=Terse -e '$a=$b+$c'
624 1 LISTOP (0x8179888) leave
625 2 OP (0x81798b0) enter
626 3 COP (0x8179850) nextstate
627 4 BINOP (0x8179828) sassign
628 5 BINOP (0x8179800) add [1]
629 6 UNOP (0x81796e0) null [15]
630 7 SVOP (0x80fafe0) gvsv GV (0x80fa4cc) *b
631 8 UNOP (0x81797e0) null [15]
632 9 SVOP (0x8179700) gvsv GV (0x80efeb0) *c
633 10 UNOP (0x816b4f0) null [15]
634 11 SVOP (0x816dcf0) gvsv GV (0x80fa460) *a
636Let's start in the middle, at line 4. This is a BINOP, a binary
637operator, which is at location C<0x8179828>. The specific operator in
638question is C<sassign> - scalar assignment - and you can find the code
639which implements it in the function C<pp_sassign> in F<pp_hot.c>. As a
640binary operator, it has two children: the add operator, providing the
641result of C<$b+$c>, is uppermost on line 5, and the left hand side is on
642line 10.
644Line 10 is the null op: this does exactly nothing. What is that doing
645there? If you see the null op, it's a sign that something has been
646optimized away after parsing. As we mentioned in L</Optimization>,
647the optimization stage sometimes converts two operations into one, for
648example when fetching a scalar variable. When this happens, instead of
649rewriting the op tree and cleaning up the dangling pointers, it's easier
650just to replace the redundant operation with the null op. Originally,
651the tree would have looked like this:
653 10 SVOP (0x816b4f0) rv2sv [15]
654 11 SVOP (0x816dcf0) gv GV (0x80fa460) *a
656That is, fetch the C<a> entry from the main symbol table, and then look
657at the scalar component of it: C<gvsv> (C<pp_gvsv> into F<pp_hot.c>)
658happens to do both these things.
660The right hand side, starting at line 5 is similar to what we've just
661seen: we have the C<add> op (C<pp_add> also in F<pp_hot.c>) add together
662two C<gvsv>s.
664Now, what's this about?
666 1 LISTOP (0x8179888) leave
667 2 OP (0x81798b0) enter
668 3 COP (0x8179850) nextstate
670C<enter> and C<leave> are scoping ops, and their job is to perform any
671housekeeping every time you enter and leave a block: lexical variables
672are tidied up, unreferenced variables are destroyed, and so on. Every
673program will have those first three lines: C<leave> is a list, and its
674children are all the statements in the block. Statements are delimited
675by C<nextstate>, so a block is a collection of C<nextstate> ops, with
676the ops to be performed for each statement being the children of
677C<nextstate>. C<enter> is a single op which functions as a marker.
679That's how Perl parsed the program, from top to bottom:
681 Program
682 |
683 Statement
684 |
685 =
686 / \
687 / \
688 $a +
689 / \
690 $b $c
692However, it's impossible to B<perform> the operations in this order:
693you have to find the values of C<$b> and C<$c> before you add them
694together, for instance. So, the other thread that runs through the op
695tree is the execution order: each op has a field C<op_next> which points
696to the next op to be run, so following these pointers tells us how perl
697executes the code. We can traverse the tree in this order using
698the C<exec> option to C<B::Terse>:
700 % perl -MO=Terse,exec -e '$a=$b+$c'
701 1 OP (0x8179928) enter
702 2 COP (0x81798c8) nextstate
703 3 SVOP (0x81796c8) gvsv GV (0x80fa4d4) *b
704 4 SVOP (0x8179798) gvsv GV (0x80efeb0) *c
705 5 BINOP (0x8179878) add [1]
706 6 SVOP (0x816dd38) gvsv GV (0x80fa468) *a
707 7 BINOP (0x81798a0) sassign
708 8 LISTOP (0x8179900) leave
710This probably makes more sense for a human: enter a block, start a
711statement. Get the values of C<$b> and C<$c>, and add them together.
712Find C<$a>, and assign one to the other. Then leave.
714The way Perl builds up these op trees in the parsing process can be
715unravelled by examining F<perly.y>, the YACC grammar. Let's take the
716piece we need to construct the tree for C<$a = $b + $c>
718 1 term : term ASSIGNOP term
719 2 { $$ = newASSIGNOP(OPf_STACKED, $1, $2, $3); }
720 3 | term ADDOP term
721 4 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
723If you're not used to reading BNF grammars, this is how it works: You're
724fed certain things by the tokeniser, which generally end up in upper
725case. Here, C<ADDOP>, is provided when the tokeniser sees C<+> in your
726code. C<ASSIGNOP> is provided when C<=> is used for assigning. These are
727`terminal symbols', because you can't get any simpler than them.
729The grammar, lines one and three of the snippet above, tells you how to
730build up more complex forms. These complex forms, `non-terminal symbols'
731are generally placed in lower case. C<term> here is a non-terminal
732symbol, representing a single expression.
734The grammar gives you the following rule: you can make the thing on the
735left of the colon if you see all the things on the right in sequence.
736This is called a "reduction", and the aim of parsing is to completely
737reduce the input. There are several different ways you can perform a
738reduction, separated by vertical bars: so, C<term> followed by C<=>
739followed by C<term> makes a C<term>, and C<term> followed by C<+>
740followed by C<term> can also make a C<term>.
742So, if you see two terms with an C<=> or C<+>, between them, you can
743turn them into a single expression. When you do this, you execute the
744code in the block on the next line: if you see C<=>, you'll do the code
745in line 2. If you see C<+>, you'll do the code in line 4. It's this code
746which contributes to the op tree.
748 | term ADDOP term
749 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
751What this does is creates a new binary op, and feeds it a number of
752variables. The variables refer to the tokens: C<$1> is the first token in
753the input, C<$2> the second, and so on - think regular expression
754backreferences. C<$$> is the op returned from this reduction. So, we
755call C<newBINOP> to create a new binary operator. The first parameter to
756C<newBINOP>, a function in F<op.c>, is the op type. It's an addition
757operator, so we want the type to be C<ADDOP>. We could specify this
758directly, but it's right there as the second token in the input, so we
759use C<$2>. The second parameter is the op's flags: 0 means `nothing
760special'. Then the things to add: the left and right hand side of our
761expression, in scalar context.
763=head2 Stacks
765When perl executes something like C<addop>, how does it pass on its
766results to the next op? The answer is, through the use of stacks. Perl
767has a number of stacks to store things it's currently working on, and
768we'll look at the three most important ones here.
770=over 3
772=item Argument stack
774Arguments are passed to PP code and returned from PP code using the
775argument stack, C<ST>. The typical way to handle arguments is to pop
776them off the stack, deal with them how you wish, and then push the result
777back onto the stack. This is how, for instance, the cosine operator
780 NV value;
781 value = POPn;
782 value = Perl_cos(value);
783 XPUSHn(value);
785We'll see a more tricky example of this when we consider Perl's macros
786below. C<POPn> gives you the NV (floating point value) of the top SV on
787the stack: the C<$x> in C<cos($x)>. Then we compute the cosine, and push
788the result back as an NV. The C<X> in C<XPUSHn> means that the stack
789should be extended if necessary - it can't be necessary here, because we
790know there's room for one more item on the stack, since we've just
791removed one! The C<XPUSH*> macros at least guarantee safety.
793Alternatively, you can fiddle with the stack directly: C<SP> gives you
794the first element in your portion of the stack, and C<TOP*> gives you
795the top SV/IV/NV/etc. on the stack. So, for instance, to do unary
796negation of an integer:
798 SETi(-TOPi);
800Just set the integer value of the top stack entry to its negation.
802Argument stack manipulation in the core is exactly the same as it is in
803XSUBs - see L<perlxstut>, L<perlxs> and L<perlguts> for a longer
804description of the macros used in stack manipulation.
806=item Mark stack
808I say `your portion of the stack' above because PP code doesn't
809necessarily get the whole stack to itself: if your function calls
810another function, you'll only want to expose the arguments aimed for the
811called function, and not (necessarily) let it get at your own data. The
812way we do this is to have a `virtual' bottom-of-stack, exposed to each
813function. The mark stack keeps bookmarks to locations in the argument
814stack usable by each function. For instance, when dealing with a tied
815variable, (internally, something with `P' magic) Perl has to call
816methods for accesses to the tied variables. However, we need to separate
817the arguments exposed to the method to the argument exposed to the
818original function - the store or fetch or whatever it may be. Here's how
819the tied C<push> is implemented; see C<av_push> in F<av.c>:
822 2 EXTEND(SP,2);
823 3 PUSHs(SvTIED_obj((SV*)av, mg));
824 4 PUSHs(val);
825 5 PUTBACK;
826 6 ENTER;
827 7 call_method("PUSH", G_SCALAR|G_DISCARD);
828 8 LEAVE;
831The lines which concern the mark stack are the first, fifth and last
832lines: they save away, restore and remove the current position of the
833argument stack.
835Let's examine the whole implementation, for practice:
839Push the current state of the stack pointer onto the mark stack. This is
840so that when we've finished adding items to the argument stack, Perl
841knows how many things we've added recently.
843 2 EXTEND(SP,2);
844 3 PUSHs(SvTIED_obj((SV*)av, mg));
845 4 PUSHs(val);
847We're going to add two more items onto the argument stack: when you have
848a tied array, the C<PUSH> subroutine receives the object and the value
849to be pushed, and that's exactly what we have here - the tied object,
850retrieved with C<SvTIED_obj>, and the value, the SV C<val>.
852 5 PUTBACK;
854Next we tell Perl to make the change to the global stack pointer: C<dSP>
855only gave us a local copy, not a reference to the global.
857 6 ENTER;
858 7 call_method("PUSH", G_SCALAR|G_DISCARD);
859 8 LEAVE;
861C<ENTER> and C<LEAVE> localise a block of code - they make sure that all
862variables are tidied up, everything that has been localised gets
863its previous value returned, and so on. Think of them as the C<{> and
864C<}> of a Perl block.
866To actually do the magic method call, we have to call a subroutine in
867Perl space: C<call_method> takes care of that, and it's described in
868L<perlcall>. We call the C<PUSH> method in scalar context, and we're
869going to discard its return value.
873Finally, we remove the value we placed on the mark stack, since we
874don't need it any more.
876=item Save stack
878C doesn't have a concept of local scope, so perl provides one. We've
879seen that C<ENTER> and C<LEAVE> are used as scoping braces; the save
880stack implements the C equivalent of, for example:
882 {
883 local $foo = 42;
884 ...
885 }
887See L<perlguts/Localising Changes> for how to use the save stack.
891=head2 Millions of Macros
893One thing you'll notice about the Perl source is that it's full of
894macros. Some have called the pervasive use of macros the hardest thing
895to understand, others find it adds to clarity. Let's take an example,
896the code which implements the addition operator:
898 1 PP(pp_add)
899 2 {
900 3 djSP; dATARGET; tryAMAGICbin(add,opASSIGN);
901 4 {
902 5 dPOPTOPnnrl_ul;
903 6 SETn( left + right );
904 7 RETURN;
905 8 }
906 9 }
908Every line here (apart from the braces, of course) contains a macro. The
909first line sets up the function declaration as Perl expects for PP code;
910line 3 sets up variable declarations for the argument stack and the
911target, the return value of the operation. Finally, it tries to see if
912the addition operation is overloaded; if so, the appropriate subroutine
913is called.
915Line 5 is another variable declaration - all variable declarations start
916with C<d> - which pops from the top of the argument stack two NVs (hence
917C<nn>) and puts them into the variables C<right> and C<left>, hence the
918C<rl>. These are the two operands to the addition operator. Next, we
919call C<SETn> to set the NV of the return value to the result of adding
920the two values. This done, we return - the C<RETURN> macro makes sure
921that our return value is properly handled, and we pass the next operator
922to run back to the main run loop.
924Most of these macros are explained in L<perlapi>, and some of the more
925important ones are explained in L<perlxs> as well. Pay special attention
926to L<perlguts/Background and PERL_IMPLICIT_CONTEXT> for information on
927the C<[pad]THX_?> macros.
930=head2 Poking at Perl
932To really poke around with Perl, you'll probably want to build Perl for
933debugging, like this:
935 ./Configure -d -D optimize=-g
936 make
938C<-g> is a flag to the C compiler to have it produce debugging
939information which will allow us to step through a running program.
940F<Configure> will also turn on the C<DEBUGGING> compilation symbol which
941enables all the internal debugging code in Perl. There are a whole bunch
942of things you can debug with this: L<perlrun> lists them all, and the
943best way to find out about them is to play about with them. The most
944useful options are probably
946 l Context (loop) stack processing
947 t Trace execution
948 o Method and overloading resolution
949 c String/numeric conversions
951Some of the functionality of the debugging code can be achieved using XS
954 -Dr => use re 'debug'
955 -Dx => use O 'Debug'
957=head2 Using a source-level debugger
959If the debugging output of C<-D> doesn't help you, it's time to step
960through perl's execution with a source-level debugger.
962=over 3
964=item *
966We'll use C<gdb> for our examples here; the principles will apply to any
967debugger, but check the manual of the one you're using.
971To fire up the debugger, type
973 gdb ./perl
975You'll want to do that in your Perl source tree so the debugger can read
976the source code. You should see the copyright message, followed by the
979 (gdb)
981C<help> will get you into the documentation, but here are the most
982useful commands:
984=over 3
986=item run [args]
988Run the program with the given arguments.
990=item break function_name
992=item break source.c:xxx
994Tells the debugger that we'll want to pause execution when we reach
995either the named function (but see L</Function names>!) or the given
996line in the named source file.
998=item step
1000Steps through the program a line at a time.
1002=item next
1004Steps through the program a line at a time, without descending into
1007=item continue
1009Run until the next breakpoint.
1011=item finish
1013Run until the end of the current function, then stop again.
1017Just pressing Enter will do the most recent operation again - it's a
1018blessing when stepping through miles of source code.
1020=item print
1022Execute the given C code and print its results. B<WARNING>: Perl makes
1023heavy use of macros, and F<gdb> is not aware of macros. You'll have to
1024substitute them yourself. So, for instance, you can't say
1026 print SvPV_nolen(sv)
1028but you have to say
1030 print Perl_sv_2pv_nolen(sv)
1032You may find it helpful to have a "macro dictionary", which you can
1033produce by saying C<cpp -dM perl.c | sort>. Even then, F<cpp> won't
1034recursively apply the macros for you.
1038=head2 Dumping Perl Data Structures
1040One way to get around this macro hell is to use the dumping functions in
1041F<dump.c>; these work a little like an internal
1042L<Devel::Peek|Devel::Peek>, but they also cover OPs and other structures
1043that you can't get at from Perl. Let's take an example. We'll use the
1044C<$a = $b + $c> we used before, but give it a bit of context:
1045C<$b = "6XXXX"; $c = 2.3;>. Where's a good place to stop and poke around?
1047What about C<pp_add>, the function we examined earlier to implement the
1048C<+> operator:
1050 (gdb) break Perl_pp_add
1051 Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.
1053Notice we use C<Perl_pp_add> and not C<pp_add> - see L<perlguts/Function Names>.
1054With the breakpoint in place, we can run our program:
1056 (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'
1058Lots of junk will go past as gdb reads in the relevant source files and
1059libraries, and then:
1061 Breakpoint 1, Perl_pp_add () at pp_hot.c:309
1062 309 djSP; dATARGET; tryAMAGICbin(add,opASSIGN);
1063 (gdb) step
1064 311 dPOPTOPnnrl_ul;
1065 (gdb)
1067We looked at this bit of code before, and we said that C<dPOPTOPnnrl_ul>
1068arranges for two C<NV>s to be placed into C<left> and C<right> - let's
1069slightly expand it:
1071 #define dPOPTOPnnrl_ul NV right = POPn; \
1072 SV *leftsv = TOPs; \
1073 NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0
1075C<POPn> takes the SV from the top of the stack and obtains its NV either
1076directly (if C<SvNOK> is set) or by calling the C<sv_2nv> function.
1077C<TOPs> takes the next SV from the top of the stack - yes, C<POPn> uses
1078C<TOPs> - but doesn't remove it. We then use C<SvNV> to get the NV from
1079C<leftsv> in the same way as before - yes, C<POPn> uses C<SvNV>.
1081Since we don't have an NV for C<$b>, we'll have to use C<sv_2nv> to
1082convert it. If we step again, we'll find ourselves there:
1084 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
1085 1669 if (!sv)
1086 (gdb)
1088We can now use C<Perl_sv_dump> to investigate the SV:
1090 SV = PV(0xa057cc0) at 0xa0675d0
1091 REFCNT = 1
1092 FLAGS = (POK,pPOK)
1093 PV = 0xa06a510 "6XXXX"\0
1094 CUR = 5
1095 LEN = 6
1096 $1 = void
1098We know we're going to get C<6> from this, so let's finish the
1101 (gdb) finish
1102 Run till exit from #0 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
1103 0x462669 in Perl_pp_add () at pp_hot.c:311
1104 311 dPOPTOPnnrl_ul;
1106We can also dump out this op: the current op is always stored in
1107C<PL_op>, and we can dump it with C<Perl_op_dump>. This'll give us
1108similar output to L<B::Debug|B::Debug>.
1110 {
1111 13 TYPE = add ===> 14
1112 TARG = 1
1114 {
1115 TYPE = null ===> (12)
1116 (was rv2sv)
1118 {
1119 11 TYPE = gvsv ===> 12
1121 GV = main::b
1122 }
1123 }
1125< finish this later >
1127=head2 Patching
1129All right, we've now had a look at how to navigate the Perl sources and
1130some things you'll need to know when fiddling with them. Let's now get
1131on and create a simple patch. Here's something Larry suggested: if a
1132C<U> is the first active format during a C<pack>, (for example,
1133C<pack "U3C8", @stuff>) then the resulting string should be treated as
1134UTF8 encoded.
1136How do we prepare to fix this up? First we locate the code in question -
1137the C<pack> happens at runtime, so it's going to be in one of the F<pp>
1138files. Sure enough, C<pp_pack> is in F<pp.c>. Since we're going to be
1139altering this file, let's copy it to F<pp.c~>.
1141Now let's look over C<pp_pack>: we take a pattern into C<pat>, and then
1142loop over the pattern, taking each format character in turn into
1143C<datum_type>. Then for each possible format character, we swallow up
1144the other arguments in the pattern (a field width, an asterisk, and so
1145on) and convert the next chunk input into the specified format, adding
1146it onto the output SV C<cat>.
1148How do we know if the C<U> is the first format in the C<pat>? Well, if
1149we have a pointer to the start of C<pat> then, if we see a C<U> we can
1150test whether we're still at the start of the string. So, here's where
1151C<pat> is set up:
1153 STRLEN fromlen;
1154 register char *pat = SvPVx(*++MARK, fromlen);
1155 register char *patend = pat + fromlen;
1156 register I32 len;
1157 I32 datumtype;
1158 SV *fromstr;
1160We'll have another string pointer in there:
1162 STRLEN fromlen;
1163 register char *pat = SvPVx(*++MARK, fromlen);
1164 register char *patend = pat + fromlen;
1165 + char *patcopy;
1166 register I32 len;
1167 I32 datumtype;
1168 SV *fromstr;
1170And just before we start the loop, we'll set C<patcopy> to be the start
1171of C<pat>:
1173 items = SP - MARK;
1174 MARK++;
1175 sv_setpvn(cat, "", 0);
1176 + patcopy = pat;
1177 while (pat < patend) {
1179Now if we see a C<U> which was at the start of the string, we turn on
1180the UTF8 flag for the output SV, C<cat>:
1182 + if (datumtype == 'U' && pat==patcopy+1)
1183 + SvUTF8_on(cat);
1184 if (datumtype == '#') {
1185 while (pat < patend && *pat != '\n')
1186 pat++;
1188Remember that it has to be C<patcopy+1> because the first character of
1189the string is the C<U> which has been swallowed into C<datumtype!>
1191Oops, we forgot one thing: what if there are spaces at the start of the
1192pattern? C<pack(" U*", @stuff)> will have C<U> as the first active
1193character, even though it's not the first thing in the pattern. In this
1194case, we have to advance C<patcopy> along with C<pat> when we see spaces:
1196 if (isSPACE(datumtype))
1197 continue;
1199needs to become
1201 if (isSPACE(datumtype)) {
1202 patcopy++;
1203 continue;
1204 }
1206OK. That's the C part done. Now we must do two additional things before
1207this patch is ready to go: we've changed the behaviour of Perl, and so
1208we must document that change. We must also provide some more regression
1209tests to make sure our patch works and doesn't create a bug somewhere
1210else along the line.
1212The regression tests for each operator live in F<t/op/>, and so we make
1213a copy of F<t/op/pack.t> to F<t/op/pack.t~>. Now we can add our tests
1214to the end. First, we'll test that the C<U> does indeed create Unicode
1217 print 'not ' unless "1.20.300.4000" eq sprintf "%vd", pack("U*",1,20,300,4000);
1218 print "ok $test\n"; $test++;
1220Now we'll test that we got that space-at-the-beginning business right:
1222 print 'not ' unless "1.20.300.4000" eq
1223 sprintf "%vd", pack(" U*",1,20,300,4000);
1224 print "ok $test\n"; $test++;
1226And finally we'll test that we don't make Unicode strings if C<U> is B<not>
1227the first active format:
1229 print 'not ' unless v1.20.300.4000 ne
1230 sprintf "%vd", pack("C0U*",1,20,300,4000);
1231 print "ok $test\n"; $test++;
b1866b2d 1233Mustn't forget to change the number of tests which appears at the top, or
1234else the automated tester will get confused:
1236 -print "1..156\n";
1237 +print "1..159\n";
1239We now compile up Perl, and run it through the test suite. Our new
1240tests pass, hooray!
1242Finally, the documentation. The job is never done until the paperwork is
1243over, so let's describe the change we've just made. The relevant place
1244is F<pod/perlfunc.pod>; again, we make a copy, and then we'll insert
1245this text in the description of C<pack>:
1247 =item *
1249 If the pattern begins with a C<U>, the resulting string will be treated
1250 as Unicode-encoded. You can force UTF8 encoding on in a string with an
1251 initial C<U0>, and the bytes that follow will be interpreted as Unicode
1252 characters. If you don't want this to happen, you can begin your pattern
1253 with C<C0> (or anything else) to force Perl not to UTF8 encode your
1254 string, and then follow this with a C<U*> somewhere in your pattern.
1256All done. Now let's create the patch. F<Porting/patching.pod> tells us
1257that if we're making major changes, we should copy the entire directory
1258to somewhere safe before we begin fiddling, and then do
1260 diff -ruN old new > patch
1262However, we know which files we've changed, and we can simply do this:
1264 diff -u pp.c~ pp.c > patch
1265 diff -u t/op/pack.t~ t/op/pack.t >> patch
1266 diff -u pod/perlfunc.pod~ pod/perlfunc.pod >> patch
1268We end up with a patch looking a little like this:
1270 --- pp.c~ Fri Jun 02 04:34:10 2000
1271 +++ pp.c Fri Jun 16 11:37:25 2000
1272 @@ -4375,6 +4375,7 @@
1273 register I32 items;
1274 STRLEN fromlen;
1275 register char *pat = SvPVx(*++MARK, fromlen);
1276 + char *patcopy;
1277 register char *patend = pat + fromlen;
1278 register I32 len;
1279 I32 datumtype;
1280 @@ -4405,6 +4406,7 @@
1281 ...
1283And finally, we submit it, with our rationale, to perl5-porters. Job
1286=head2 CONCLUSION
1288We've had a brief look around the Perl source, an overview of the stages
1289F<perl> goes through when it's running your code, and how to use a
1290debugger to poke at the Perl guts. Finally, we took a very simple
1291problem and demonstrated how to solve it fully - with documentation,
1292regression tests, and finally a patch for submission to p5p.
1294I'd now suggest you read over those references again, and then, as soon
1295as possible, get your hands dirty. The best way to learn is by doing,
1298=over 3
1300=item *
1302Subscribe to perl5-porters, follow the patches and try and understand
1303them; don't be afraid to ask if there's a portion you're not clear on -
1304who knows, you may unearth a bug in the patch...
1306=item *
1308Keep up to date with the bleeding edge Perl distributions and get
1309familiar with the changes. Try and get an idea of what areas people are
1310working on and the changes they're making.
1312=item *
1314Find an area of Perl that seems interesting to you, and see if you can
1315work out how it works. Scan through the source, and step over it in the
1316debugger. Play, poke, investigate, fiddle! You'll probably get to
1317understand not just your chosen area but a much wider range of F<perl>'s
1318activity as well, and probably sooner than you'd think.
1322=over 3
1324=item I<The Road goes ever on and on, down from the door where it began.>
1328If you can do these things, you've started on the long road to Perl porting.
1329Thanks for wanting to help make Perl better - and happy hacking!
1331=head1 AUTHOR
1333This document was written by Nathan Torkington, and is maintained by
1334the perl5-porters mailing list.