1 # Devel::Peek - A data debugging tool for the XS programmer
2 # The documentation is after the __END__
7 $XS_VERSION = $VERSION;
8 $VERSION = eval $VERSION;
14 @EXPORT = qw(Dump mstat DeadCode DumpArray DumpWithOP DumpProg
15 fill_mstats mstats_fillhash mstats2hash runops_debug debug_flags);
16 @EXPORT_OK = qw(SvREFCNT CvGV);
17 %EXPORT_TAGS = ('ALL' => [@EXPORT, @EXPORT_OK]);
23 my $ops_rx = qr/^:opd(=[stP]*)?\b/;
24 my @db = grep m/$ops_rx/, @_;
25 @_ = grep !m/$ops_rx/, @_;
27 die "Too many :opd options" if @db > 1;
29 my $flags = ($db[0] =~ m/$ops_rx/ and $1);
30 $flags = 'st' unless defined $flags;
32 $f |= 2 if $flags =~ /s/;
33 $f |= 8 if $flags =~ /t/;
34 $f |= 64 if $flags =~ /P/;
38 goto &Exporter::import;
41 sub DumpWithOP ($;$) {
42 local($Devel::Peek::dump_ops)=1;
43 my $depth = @_ > 1 ? $_[1] : 4 ;
47 $D_flags = 'psltocPmfrxuLHXDSTR';
49 sub debug_flags (;$) {
51 for my $i (0 .. length($D_flags)-1) {
52 $out .= substr $D_flags, $i, 1 if $^D & (1<<$i);
56 if (defined $arg and $arg =~ /\D/) {
57 die "unknown flags in debug_flags()" if $arg =~ /[^-$D_flags]/;
58 my ($on,$off) = split /-/, "$arg-";
60 $num |= (1<<index($D_flags, $_)) for split //, $on;
61 $num &= ~(1<<index($D_flags, $_)) for split //, $off;
63 $^D = $num if defined $arg;
67 sub B::Deparse::pp_Devel_Peek_Dump {
68 my ($deparse,$op,$cx) = @_;
69 my @kids = $deparse->deparse($op->first, 6);
70 my $sib = $op->first->sibling;
71 if (ref $sib ne 'B::NULL') {
72 push @kids, $deparse->deparse($sib, 6);
74 return "Devel::Peek::Dump(" . join(", ", @kids) . ")";
82 Devel::Peek - A data debugging tool for the XS programmer
91 DumpArray( 5, $a, $b, ... );
94 use Devel::Peek ':opd=st';
98 Devel::Peek contains functions which allows raw Perl datatypes to be
99 manipulated from a Perl script. This is used by those who do XS programming
100 to check that the data they are sending from C to Perl looks as they think
101 it should look. The trick, then, is to know what the raw datatype is
102 supposed to look like when it gets to Perl. This document offers some tips
103 and hints to describe good and bad raw data.
105 It is very possible that this document will fall far short of being useful
106 to the casual reader. The reader is expected to understand the material in
107 the first few sections of L<perlguts>.
109 Devel::Peek supplies a C<Dump()> function which can dump a raw Perl
110 datatype, and C<mstat("marker")> function to report on memory usage
111 (if perl is compiled with corresponding option). The function
112 DeadCode() provides statistics on the data "frozen" into inactive
113 C<CV>. Devel::Peek also supplies C<SvREFCNT()> which can query reference
114 counts on SVs. This document will take a passive, and safe, approach
115 to data debugging and for that it will describe only the C<Dump()>
118 The C<Dump()> function takes one or two arguments: something to dump, and
119 an optional limit for recursion and array elements (default is 4). The
120 first argument is evaluted in rvalue scalar context, with exceptions for
121 @array and %hash, which dump the array or hash itself. So C<Dump @array>
122 works, as does C<Dump $foo>. And C<Dump pos> will call C<pos> in rvalue
123 context, whereas C<Dump ${\pos}> will call it in lvalue context.
125 Function C<DumpArray()> allows dumping of multiple values (useful when you
126 need to analyze returns of functions).
128 The global variable $Devel::Peek::pv_limit can be set to limit the
129 number of character printed in various string values. Setting it to 0
132 If C<use Devel::Peek> directive has a C<:opd=FLAGS> argument,
133 this switches on debugging of opcode dispatch. C<FLAGS> should be a
134 combination of C<s>, C<t>, and C<P> (see B<-D> flags in L<perlrun>).
135 C<:opd> is a shortcut for C<:opd=st>.
137 =head2 Runtime debugging
139 C<CvGV($cv)> return one of the globs associated to a subroutine reference $cv.
141 debug_flags() returns a string representation of C<$^D> (similar to
142 what is allowed for B<-D> flag). When called with a numeric argument,
143 sets $^D to the corresponding value. When called with an argument of
144 the form C<"flags-flags">, set on/off bits of C<$^D> corresponding to
145 letters before/after C<->. (The returned value is for C<$^D> before
148 runops_debug() returns true if the current I<opcode dispatcher> is the
149 debugging one. When called with an argument, switches to debugging or
150 non-debugging dispatcher depending on the argument (active for
151 newly-entered subs/etc only). (The returned value is for the dispatcher before the modification.)
153 =head2 Memory footprint debugging
155 When perl is compiled with support for memory footprint debugging
156 (default with Perl's malloc()), Devel::Peek provides an access to this API.
158 Use mstat() function to emit a memory state statistic to the terminal.
159 For more information on the format of output of mstat() see
160 L<perldebguts/Using $ENV{PERL_DEBUG_MSTATS}>.
162 Three additional functions allow access to this statistic from Perl.
163 First, use C<mstats_fillhash(%hash)> to get the information contained
164 in the output of mstat() into %hash. The field of this hash are
166 minbucket nbuckets sbrk_good sbrk_slack sbrked_remains sbrks
167 start_slack topbucket topbucket_ev topbucket_odd total total_chain
170 Two additional fields C<free>, C<used> contain array references which
171 provide per-bucket count of free and used chunks. Two other fields
172 C<mem_size>, C<available_size> contain array references which provide
173 the information about the allocated size and usable size of chunks in
174 each bucket. Again, see L<perldebguts/Using $ENV{PERL_DEBUG_MSTATS}>
178 Keep in mind that only the first several "odd-numbered" buckets are
179 used, so the information on size of the "odd-numbered" buckets which are
180 not used is probably meaningless.
184 mem_size available_size minbucket nbuckets
186 is the property of a particular build of perl, and does not depend on
187 the current process. If you do not provide the optional argument to
188 the functions mstats_fillhash(), fill_mstats(), mstats2hash(), then
189 the information in fields C<mem_size>, C<available_size> is not
192 C<fill_mstats($buf)> is a much cheaper call (both speedwise and
193 memory-wise) which collects the statistic into $buf in
194 machine-readable form. At a later moment you may need to call
195 C<mstats2hash($buf, %hash)> to use this information to fill %hash.
197 All three APIs C<fill_mstats($buf)>, C<mstats_fillhash(%hash)>, and
198 C<mstats2hash($buf, %hash)> are designed to allocate no memory if used
199 I<the second time> on the same $buf and/or %hash.
201 So, if you want to collect memory info in a cycle, you may call
204 fill_mstats($_) for @buf;
205 mstats_fillhash(%report, 1); # Static info too
209 fill_mstats $_; # Collect statistic
212 mstats2hash($_, %report); # Preserve static info
213 # Do something with %report
218 The following examples don't attempt to show everything as that would be a
219 monumental task, and, frankly, we don't want this manpage to be an internals
220 document for Perl. The examples do demonstrate some basics of the raw Perl
221 datatypes, and should suffice to get most determined people on their way.
222 There are no guidewires or safety nets, nor blazed trails, so be prepared to
223 travel alone from this point and on and, if at all possible, don't fall into
224 the quicksand (it's bad for business).
226 Oh, one final bit of advice: take L<perlguts> with you. When you return we
227 expect to see it well-thumbed.
229 =head2 A simple scalar string
231 Let's begin by looking a simple scalar which is holding a string.
234 $a = 42; $a = "hello";
239 SV = PVIV(0xbc288) at 0xbe9a8
243 PV = 0xb2048 "hello"\0
247 This says C<$a> is an SV, a scalar. The scalar type is a PVIV, which is
248 capable of holding an integer (IV) and/or a string (PV) value. The scalar's
249 head is allocated at address 0xbe9a8, while the body is at 0xbc288.
250 Its reference count is 1. It has the C<POK> flag set, meaning its
251 current PV field is valid. Because POK is set we look at the PV item
252 to see what is in the scalar. The \0 at the end indicate that this
253 PV is properly NUL-terminated.
254 Note that the IV field still contains its old numeric value, but because
255 FLAGS doesn't have IOK set, we must ignore the IV item.
256 CUR indicates the number of characters in the PV. LEN indicates the
257 number of bytes allocated for the PV (at least one more than CUR, because
258 LEN includes an extra byte for the end-of-string marker, then usually
259 rounded up to some efficient allocation unit).
261 =head2 A simple scalar number
263 If the scalar contains a number the raw SV will be leaner.
271 SV = IV(0xbc818) at 0xbe9a8
276 This says C<$a> is an SV, a scalar. The scalar is an IV, a number. Its
277 reference count is 1. It has the C<IOK> flag set, meaning it is currently
278 being evaluated as a number. Because IOK is set we look at the IV item to
279 see what is in the scalar.
281 =head2 A simple scalar with an extra reference
283 If the scalar from the previous example had an extra reference:
292 SV = IV(0xbe860) at 0xbe9a8
297 Notice that this example differs from the previous example only in its
298 reference count. Compare this to the next example, where we dump C<$b>
301 =head2 A reference to a simple scalar
303 This shows what a reference looks like when it references a simple scalar.
312 SV = IV(0xf041c) at 0xbe9a0
316 SV = IV(0xbe860) at 0xbe9a8
321 Starting from the top, this says C<$b> is an SV. The scalar is an IV,
322 which is capable of holding an integer or reference value.
323 It has the C<ROK> flag set, meaning it is a reference (rather than an
324 integer or string). Notice that Dump
325 follows the reference and shows us what C<$b> was referencing. We see the
326 same C<$a> that we found in the previous example.
328 Note that the value of C<RV> coincides with the numbers we see when we
329 stringify $b. The addresses inside IV() are addresses of
330 C<X***> structures which hold the current state of an C<SV>. This
331 address may change during lifetime of an SV.
333 =head2 A reference to an array
335 This shows what a reference to an array looks like.
343 SV = IV(0xc85998) at 0xc859a8
347 SV = PVAV(0xc71e10) at 0xc70de8
356 SV = IV(0xc70f88) at 0xc70f98
361 This says C<$a> is a reference (ROK), which points to
362 another SV which is a PVAV, an array. The array has one element,
363 element zero, which is another SV. The field C<FILL> above indicates
364 the last element in the array, similar to C<$#$a>.
366 If C<$a> pointed to an array of two elements then we would see the
369 use Devel::Peek 'Dump';
375 SV = IV(0x158c998) at 0x158c9a8
379 SV = PVAV(0x1578e10) at 0x1577de8
388 SV = IV(0x1577f88) at 0x1577f98
393 SV = IV(0x158be88) at 0x158be98
398 Note that C<Dump> will not report I<all> the elements in the array,
399 only several first (depending on how deep it already went into the
402 =head2 A reference to a hash
404 The following shows the raw form of a reference to a hash.
412 SV = IV(0x8177858) at 0x816a618
416 SV = PVHV(0x8167768) at 0x814fc10
419 ARRAY = 0x816c5b8 (0:7, 1:1)
420 hash quality = 100.0%
426 Elt "hello" HASH = 0xc8fd181b
427 SV = IV(0x816c030) at 0x814fcf4
432 This shows C<$a> is a reference pointing to an SV. That SV is a PVHV, a
433 hash. Fields RITER and EITER are used by C<L<perlfunc/each>>.
435 The "quality" of a hash is defined as the total number of comparisons needed
436 to access every element once, relative to the expected number needed for a
437 random hash. The value can go over 100%.
439 The total number of comparisons is equal to the sum of the squares of the
440 number of entries in each bucket. For a random hash of C<<n>> keys into
441 C<<k>> buckets, the expected value is:
445 =head2 Dumping a large array or hash
447 The C<Dump()> function, by default, dumps up to 4 elements from a
448 toplevel array or hash. This number can be increased by supplying a
449 second argument to the function.
452 $a = [10,11,12,13,14];
455 Notice that C<Dump()> prints only elements 10 through 13 in the above code.
456 The following code will print all of the elements.
458 use Devel::Peek 'Dump';
459 $a = [10,11,12,13,14];
462 =head2 A reference to an SV which holds a C pointer
464 This is what you really need to know as an XS programmer, of course. When
465 an XSUB returns a pointer to a C structure that pointer is stored in an SV
466 and a reference to that SV is placed on the XSUB stack. So the output from
467 an XSUB which uses something like the T_PTROBJ map might look something like
470 SV = IV(0xf381c) at 0xc859a8
474 SV = PVMG(0xbb3c8) at 0xc859a0
476 FLAGS = (OBJECT,IOK,pIOK)
480 STASH = 0xc1d10 "CookBookB::Opaque"
482 This shows that we have an SV which is a reference, which points at another
483 SV. In this case that second SV is a PVMG, a blessed scalar. Because it is
484 blessed it has the C<OBJECT> flag set. Note that an SV which holds a C
485 pointer also has the C<IOK> flag set. The C<STASH> is set to the package
486 name which this SV was blessed into.
488 The output from an XSUB which uses something like the T_PTRREF map, which
489 doesn't bless the object, might look something like this:
491 SV = IV(0xf381c) at 0xc859a8
495 SV = PVMG(0xbb3c8) at 0xc859a0
502 =head2 A reference to a subroutine
506 SV = IV(0x24d2dd8) at 0x24d2de8
510 SV = PVCV(0x24e5798) at 0x24e79d8
513 COMP_STASH = 0x22c9c50 "main"
514 START = 0x22eed60 ===> 0
516 GVGV::GV = 0x22de9d8 "MY" :: "top_targets"
522 PADNAME = 0x22e9ec0(0x22eed00) PAD = 0x22e9ea8(0x22eecd0)
523 OUTSIDE = 0x22c9fb0 (MAIN)
532 the subroutine is not an XSUB (since C<START> and C<ROOT> are
533 non-zero, and C<XSUB> is not listed, and is thus null);
537 that it was compiled in the package C<main>;
541 under the name C<MY::top_targets>;
545 inside a 5th eval in the program;
549 it is not currently executed (see C<DEPTH>);
553 it has no prototype (C<PROTOTYPE> field is missing).
559 C<Dump>, C<mstat>, C<DeadCode>, C<DumpArray>, C<DumpWithOP> and
560 C<DumpProg>, C<fill_mstats>, C<mstats_fillhash>, C<mstats2hash> by
561 default. Additionally available C<SvREFCNT>, C<SvREFCNT_inc> and
566 Readers have been known to skip important parts of L<perlguts>, causing much
571 Ilya Zakharevich ilya@math.ohio-state.edu
573 Copyright (c) 1995-98 Ilya Zakharevich. All rights reserved.
574 This program is free software; you can redistribute it and/or
575 modify it under the same terms as Perl itself.
577 Author of this software makes no claim whatsoever about suitability,
578 reliability, edability, editability or usability of this product, and
579 should not be kept liable for any damage resulting from the use of
580 it. If you can use it, you are in luck, if not, I should not be kept
581 responsible. Keep a handy copy of your backup tape at hand.
585 L<perlguts>, and L<perlguts>, again.