xref: /openbsd-src/gnu/usr.bin/perl/dist/SelfLoader/lib/SelfLoader.pm (revision d13be5d47e4149db2549a9828e244d59dbc43f15)
1package SelfLoader;
2use 5.008;
3use strict;
4our $VERSION = "1.17";
5
6# The following bit of eval-magic is necessary to make this work on
7# perls < 5.009005.
8use vars qw/$AttrList/;
9BEGIN {
10  if ($] > 5.009004) {
11    eval <<'NEWERPERL';
12use 5.009005; # due to new regexp features
13# allow checking for valid ': attrlist' attachments
14# see also AutoSplit
15$AttrList = qr{
16    \s* : \s*
17    (?:
18	# one attribute
19	(?> # no backtrack
20	    (?! \d) \w+
21	    (?<nested> \( (?: [^()]++ | (?&nested)++ )*+ \) ) ?
22	)
23	(?: \s* : \s* | \s+ (?! :) )
24    )*
25}x;
26
27NEWERPERL
28  }
29  else {
30    eval <<'OLDERPERL';
31# allow checking for valid ': attrlist' attachments
32# (we use 'our' rather than 'my' here, due to the rather complex and buggy
33# behaviour of lexicals with qr// and (??{$lex}) )
34our $nested;
35$nested = qr{ \( (?: (?> [^()]+ ) | (??{ $nested }) )* \) }x;
36our $one_attr = qr{ (?> (?! \d) \w+ (?:$nested)? ) (?:\s*\:\s*|\s+(?!\:)) }x;
37$AttrList = qr{ \s* : \s* (?: $one_attr )* }x;
38OLDERPERL
39  }
40}
41use Exporter;
42our @ISA = qw(Exporter);
43our @EXPORT = qw(AUTOLOAD);
44sub Version {$VERSION}
45sub DEBUG () { 0 }
46
47my %Cache;      # private cache for all SelfLoader's client packages
48
49# in croak and carp, protect $@ from "require Carp;" RT #40216
50
51sub croak { { local $@; require Carp; } goto &Carp::croak }
52sub carp { { local $@; require Carp; } goto &Carp::carp }
53
54AUTOLOAD {
55    our $AUTOLOAD;
56    print STDERR "SelfLoader::AUTOLOAD for $AUTOLOAD\n" if DEBUG;
57    my $SL_code = $Cache{$AUTOLOAD};
58    my $save = $@; # evals in both AUTOLOAD and _load_stubs can corrupt $@
59    unless ($SL_code) {
60        # Maybe this pack had stubs before __DATA__, and never initialized.
61        # Or, this maybe an automatic DESTROY method call when none exists.
62        $AUTOLOAD =~ m/^(.*)::/;
63        SelfLoader->_load_stubs($1) unless exists $Cache{"${1}::<DATA"};
64        $SL_code = $Cache{$AUTOLOAD};
65        $SL_code = "sub $AUTOLOAD { }"
66            if (!$SL_code and $AUTOLOAD =~ m/::DESTROY$/);
67        croak "Undefined subroutine $AUTOLOAD" unless $SL_code;
68    }
69    print STDERR "SelfLoader::AUTOLOAD eval: $SL_code\n" if DEBUG;
70
71    {
72	no strict;
73	eval $SL_code;
74    }
75    if ($@) {
76        $@ =~ s/ at .*\n//;
77        croak $@;
78    }
79    $@ = $save;
80    defined(&$AUTOLOAD) || die "SelfLoader inconsistency error";
81    delete $Cache{$AUTOLOAD};
82    goto &$AUTOLOAD
83}
84
85sub load_stubs { shift->_load_stubs((caller)[0]) }
86
87sub _load_stubs {
88    # $endlines is used by Devel::SelfStubber to capture lines after __END__
89    my($self, $callpack, $endlines) = @_;
90    no strict "refs";
91    my $fh = \*{"${callpack}::DATA"};
92    use strict;
93    my $currpack = $callpack;
94    my($line,$name,@lines, @stubs, $protoype);
95
96    print STDERR "SelfLoader::load_stubs($callpack)\n" if DEBUG;
97    croak("$callpack doesn't contain an __DATA__ token")
98        unless defined fileno($fh);
99    # Protect: fork() shares the file pointer between the parent and the kid
100    if(sysseek($fh, tell($fh), 0)) {
101      open my $nfh, '<&', $fh or croak "reopen: $!";# dup() the fd
102      close $fh or die "close: $!";                 # autocloses, but be paranoid
103      open $fh, '<&', $nfh or croak "reopen2: $!";  # dup() the fd "back"
104      close $nfh or die "close after reopen: $!";   # autocloses, but be paranoid
105    }
106    $Cache{"${currpack}::<DATA"} = 1;   # indicate package is cached
107
108    local($/) = "\n";
109    while(defined($line = <$fh>) and $line !~ m/^__END__/) {
110	if ($line =~ m/^\s*sub\s+([\w:]+)\s*((?:\([\\\$\@\%\&\*\;]*\))?(?:$AttrList)?)/) {
111            push(@stubs, $self->_add_to_cache($name, $currpack, \@lines, $protoype));
112            $protoype = $2;
113            @lines = ($line);
114            if (index($1,'::') == -1) {         # simple sub name
115                $name = "${currpack}::$1";
116            } else {                            # sub name with package
117                $name = $1;
118                $name =~ m/^(.*)::/;
119                if (defined(&{"${1}::AUTOLOAD"})) {
120                    \&{"${1}::AUTOLOAD"} == \&SelfLoader::AUTOLOAD ||
121                        die 'SelfLoader Error: attempt to specify Selfloading',
122                            " sub $name in non-selfloading module $1";
123                } else {
124                    $self->export($1,'AUTOLOAD');
125                }
126            }
127        } elsif ($line =~ m/^package\s+([\w:]+)/) { # A package declared
128            push(@stubs, $self->_add_to_cache($name, $currpack, \@lines, $protoype));
129            $self->_package_defined($line);
130            $name = '';
131            @lines = ();
132            $currpack = $1;
133            $Cache{"${currpack}::<DATA"} = 1;   # indicate package is cached
134            if (defined(&{"${1}::AUTOLOAD"})) {
135                \&{"${1}::AUTOLOAD"} == \&SelfLoader::AUTOLOAD ||
136                    die 'SelfLoader Error: attempt to specify Selfloading',
137                        " package $currpack which already has AUTOLOAD";
138            } else {
139                $self->export($currpack,'AUTOLOAD');
140            }
141        } else {
142            push(@lines,$line);
143        }
144    }
145    if (defined($line) && $line =~ /^__END__/) { # __END__
146        unless ($line =~ /^__END__\s*DATA/) {
147            if ($endlines) {
148                # Devel::SelfStubber would like us to capture the lines after
149                # __END__ so it can write out the entire file
150                @$endlines = <$fh>;
151            }
152            close($fh);
153        }
154    }
155    push(@stubs, $self->_add_to_cache($name, $currpack, \@lines, $protoype));
156    no strict;
157    eval join('', @stubs) if @stubs;
158}
159
160
161sub _add_to_cache {
162    my($self,$fullname,$pack,$lines, $protoype) = @_;
163    return () unless $fullname;
164    carp("Redefining sub $fullname")
165      if exists $Cache{$fullname};
166    $Cache{$fullname} = join('', "\n\#line 1 \"sub $fullname\"\npackage $pack; ", @$lines);
167    #$Cache{$fullname} = join('', "package $pack; ",@$lines);
168    print STDERR "SelfLoader cached $fullname: $Cache{$fullname}" if DEBUG;
169    # return stub to be eval'd
170    defined($protoype) ? "sub $fullname $protoype;" : "sub $fullname;"
171}
172
173sub _package_defined {}
174
1751;
176__END__
177
178=head1 NAME
179
180SelfLoader - load functions only on demand
181
182=head1 SYNOPSIS
183
184    package FOOBAR;
185    use SelfLoader;
186
187    ... (initializing code)
188
189    __DATA__
190    sub {....
191
192
193=head1 DESCRIPTION
194
195This module tells its users that functions in the FOOBAR package are to be
196autoloaded from after the C<__DATA__> token.  See also
197L<perlsub/"Autoloading">.
198
199=head2 The __DATA__ token
200
201The C<__DATA__> token tells the perl compiler that the perl code
202for compilation is finished. Everything after the C<__DATA__> token
203is available for reading via the filehandle FOOBAR::DATA,
204where FOOBAR is the name of the current package when the C<__DATA__>
205token is reached. This works just the same as C<__END__> does in
206package 'main', but for other modules data after C<__END__> is not
207automatically retrievable, whereas data after C<__DATA__> is.
208The C<__DATA__> token is not recognized in versions of perl prior to
2095.001m.
210
211Note that it is possible to have C<__DATA__> tokens in the same package
212in multiple files, and that the last C<__DATA__> token in a given
213package that is encountered by the compiler is the one accessible
214by the filehandle. This also applies to C<__END__> and main, i.e. if
215the 'main' program has an C<__END__>, but a module 'require'd (_not_ 'use'd)
216by that program has a 'package main;' declaration followed by an 'C<__DATA__>',
217then the C<DATA> filehandle is set to access the data after the C<__DATA__>
218in the module, _not_ the data after the C<__END__> token in the 'main'
219program, since the compiler encounters the 'require'd file later.
220
221=head2 SelfLoader autoloading
222
223The B<SelfLoader> works by the user placing the C<__DATA__>
224token I<after> perl code which needs to be compiled and
225run at 'require' time, but I<before> subroutine declarations
226that can be loaded in later - usually because they may never
227be called.
228
229The B<SelfLoader> will read from the FOOBAR::DATA filehandle to
230load in the data after C<__DATA__>, and load in any subroutine
231when it is called. The costs are the one-time parsing of the
232data after C<__DATA__>, and a load delay for the _first_
233call of any autoloaded function. The benefits (hopefully)
234are a speeded up compilation phase, with no need to load
235functions which are never used.
236
237The B<SelfLoader> will stop reading from C<__DATA__> if
238it encounters the C<__END__> token - just as you would expect.
239If the C<__END__> token is present, and is followed by the
240token DATA, then the B<SelfLoader> leaves the FOOBAR::DATA
241filehandle open on the line after that token.
242
243The B<SelfLoader> exports the C<AUTOLOAD> subroutine to the
244package using the B<SelfLoader>, and this loads the called
245subroutine when it is first called.
246
247There is no advantage to putting subroutines which will _always_
248be called after the C<__DATA__> token.
249
250=head2 Autoloading and package lexicals
251
252A 'my $pack_lexical' statement makes the variable $pack_lexical
253local _only_ to the file up to the C<__DATA__> token. Subroutines
254declared elsewhere _cannot_ see these types of variables,
255just as if you declared subroutines in the package but in another
256file, they cannot see these variables.
257
258So specifically, autoloaded functions cannot see package
259lexicals (this applies to both the B<SelfLoader> and the Autoloader).
260The C<vars> pragma provides an alternative to defining package-level
261globals that will be visible to autoloaded routines. See the documentation
262on B<vars> in the pragma section of L<perlmod>.
263
264=head2 SelfLoader and AutoLoader
265
266The B<SelfLoader> can replace the AutoLoader - just change 'use AutoLoader'
267to 'use SelfLoader' (though note that the B<SelfLoader> exports
268the AUTOLOAD function - but if you have your own AUTOLOAD and
269are using the AutoLoader too, you probably know what you're doing),
270and the C<__END__> token to C<__DATA__>. You will need perl version 5.001m
271or later to use this (version 5.001 with all patches up to patch m).
272
273There is no need to inherit from the B<SelfLoader>.
274
275The B<SelfLoader> works similarly to the AutoLoader, but picks up the
276subs from after the C<__DATA__> instead of in the 'lib/auto' directory.
277There is a maintenance gain in not needing to run AutoSplit on the module
278at installation, and a runtime gain in not needing to keep opening and
279closing files to load subs. There is a runtime loss in needing
280to parse the code after the C<__DATA__>. Details of the B<AutoLoader> and
281another view of these distinctions can be found in that module's
282documentation.
283
284=head2 __DATA__, __END__, and the FOOBAR::DATA filehandle.
285
286This section is only relevant if you want to use
287the C<FOOBAR::DATA> together with the B<SelfLoader>.
288
289Data after the C<__DATA__> token in a module is read using the
290FOOBAR::DATA filehandle. C<__END__> can still be used to denote the end
291of the C<__DATA__> section if followed by the token DATA - this is supported
292by the B<SelfLoader>. The C<FOOBAR::DATA> filehandle is left open if an
293C<__END__> followed by a DATA is found, with the filehandle positioned at
294the start of the line after the C<__END__> token. If no C<__END__> token is
295present, or an C<__END__> token with no DATA token on the same line, then
296the filehandle is closed.
297
298The B<SelfLoader> reads from wherever the current
299position of the C<FOOBAR::DATA> filehandle is, until the
300EOF or C<__END__>. This means that if you want to use
301that filehandle (and ONLY if you want to), you should either
302
3031. Put all your subroutine declarations immediately after
304the C<__DATA__> token and put your own data after those
305declarations, using the C<__END__> token to mark the end
306of subroutine declarations. You must also ensure that the B<SelfLoader>
307reads first by  calling 'SelfLoader-E<gt>load_stubs();', or by using a
308function which is selfloaded;
309
310or
311
3122. You should read the C<FOOBAR::DATA> filehandle first, leaving
313the handle open and positioned at the first line of subroutine
314declarations.
315
316You could conceivably do both.
317
318=head2 Classes and inherited methods.
319
320For modules which are not classes, this section is not relevant.
321This section is only relevant if you have methods which could
322be inherited.
323
324A subroutine stub (or forward declaration) looks like
325
326  sub stub;
327
328i.e. it is a subroutine declaration without the body of the
329subroutine. For modules which are not classes, there is no real
330need for stubs as far as autoloading is concerned.
331
332For modules which ARE classes, and need to handle inherited methods,
333stubs are needed to ensure that the method inheritance mechanism works
334properly. You can load the stubs into the module at 'require' time, by
335adding the statement 'SelfLoader-E<gt>load_stubs();' to the module to do
336this.
337
338The alternative is to put the stubs in before the C<__DATA__> token BEFORE
339releasing the module, and for this purpose the C<Devel::SelfStubber>
340module is available.  However this does require the extra step of ensuring
341that the stubs are in the module. If this is done I strongly recommend
342that this is done BEFORE releasing the module - it should NOT be done
343at install time in general.
344
345=head1 Multiple packages and fully qualified subroutine names
346
347Subroutines in multiple packages within the same file are supported - but you
348should note that this requires exporting the C<SelfLoader::AUTOLOAD> to
349every package which requires it. This is done automatically by the
350B<SelfLoader> when it first loads the subs into the cache, but you should
351really specify it in the initialization before the C<__DATA__> by putting
352a 'use SelfLoader' statement in each package.
353
354Fully qualified subroutine names are also supported. For example,
355
356   __DATA__
357   sub foo::bar {23}
358   package baz;
359   sub dob {32}
360
361will all be loaded correctly by the B<SelfLoader>, and the B<SelfLoader>
362will ensure that the packages 'foo' and 'baz' correctly have the
363B<SelfLoader> C<AUTOLOAD> method when the data after C<__DATA__> is first
364parsed.
365
366=head1 AUTHOR
367
368C<SelfLoader> is maintained by the perl5-porters. Please direct
369any questions to the canonical mailing list. Anything that
370is applicable to the CPAN release can be sent to its maintainer,
371though.
372
373Author and Maintainer: The Perl5-Porters <perl5-porters@perl.org>
374
375Maintainer of the CPAN release: Steffen Mueller <smueller@cpan.org>
376
377=head1 COPYRIGHT AND LICENSE
378
379This package has been part of the perl core since the first release
380of perl5. It has been released separately to CPAN so older installations
381can benefit from bug fixes.
382
383This package has the same copyright and license as the perl core:
384
385             Copyright (C) 1993, 1994, 1995, 1996, 1997, 1998, 1999,
386        2000, 2001, 2002, 2003, 2004, 2005, 2006 by Larry Wall and others
387
388			    All rights reserved.
389
390    This program is free software; you can redistribute it and/or modify
391    it under the terms of either:
392
393	a) the GNU General Public License as published by the Free
394	Software Foundation; either version 1, or (at your option) any
395	later version, or
396
397	b) the "Artistic License" which comes with this Kit.
398
399    This program is distributed in the hope that it will be useful,
400    but WITHOUT ANY WARRANTY; without even the implied warranty of
401    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See either
402    the GNU General Public License or the Artistic License for more details.
403
404    You should have received a copy of the Artistic License with this
405    Kit, in the file named "Artistic".  If not, I'll be glad to provide one.
406
407    You should also have received a copy of the GNU General Public License
408    along with this program in the file named "Copying". If not, write to the
409    Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
410    02111-1307, USA or visit their web page on the internet at
411    http://www.gnu.org/copyleft/gpl.html.
412
413    For those of you that choose to use the GNU General Public License,
414    my interpretation of the GNU General Public License is that no Perl
415    script falls under the terms of the GPL unless you explicitly put
416    said script under the terms of the GPL yourself.  Furthermore, any
417    object code linked with perl does not automatically fall under the
418    terms of the GPL, provided such object code only adds definitions
419    of subroutines and variables, and does not otherwise impair the
420    resulting interpreter from executing any standard Perl script.  I
421    consider linking in C subroutines in this manner to be the moral
422    equivalent of defining subroutines in the Perl language itself.  You
423    may sell such an object file as proprietary provided that you provide
424    or offer to provide the Perl source, as specified by the GNU General
425    Public License.  (This is merely an alternate way of specifying input
426    to the program.)  You may also sell a binary produced by the dumping of
427    a running Perl script that belongs to you, provided that you provide or
428    offer to provide the Perl source as specified by the GPL.  (The
429    fact that a Perl interpreter and your code are in the same binary file
430    is, in this case, a form of mere aggregation.)  This is my interpretation
431    of the GPL.  If you still have concerns or difficulties understanding
432    my intent, feel free to contact me.  Of course, the Artistic License
433    spells all this out for your protection, so you may prefer to use that.
434
435=cut
436