1 # Pod::Text -- Convert POD data to formatted ASCII text.
3 # Copyright 1999, 2000, 2001, 2002, 2004, 2006, 2008
4 # Russ Allbery <rra@stanford.edu>
6 # This program is free software; you may redistribute it and/or modify it
7 # under the same terms as Perl itself.
9 # This module converts POD to formatted text. It replaces the old Pod::Text
10 # module that came with versions of Perl prior to 5.6.0 and attempts to match
11 # its output except for some specific circumstances where other decisions
12 # seemed to produce better output. It uses Pod::Parser and is designed to be
13 # very easy to subclass.
15 # Perl core hackers, please note that this module is also separately
16 # maintained outside of the Perl core as part of the podlators. Please send
17 # me any patches at the address above in addition to sending them to the
18 # standard Perl mailing lists.
20 ##############################################################################
21 # Modules and declarations
22 ##############################################################################
29 use vars qw(@ISA @EXPORT %ESCAPES $VERSION);
31 use Carp qw(carp croak);
35 @ISA = qw(Pod::Simple Exporter);
37 # We have to export pod2text for backward compatibility.
38 @EXPORT = qw(pod2text);
42 ##############################################################################
44 ##############################################################################
46 # This function handles code blocks. It's registered as a callback to
47 # Pod::Simple and therefore doesn't work as a regular method call, but all it
48 # does is call output_code with the line.
50 my ($line, $number, $parser) = @_;
51 $parser->output_code ($line . "\n");
54 # Initialize the object and set various Pod::Simple options that we need.
55 # Here, we also process any additional options passed to the constructor or
56 # set up defaults if none were given. Note that all internal object keys are
57 # in all-caps, reserving all lower-case object keys for Pod::Simple and user
61 my $self = $class->SUPER::new;
63 # Tell Pod::Simple to handle S<> by automatically inserting .
64 $self->nbsp_for_S (1);
66 # Tell Pod::Simple to keep whitespace whenever possible.
67 if ($self->can ('preserve_whitespace')) {
68 $self->preserve_whitespace (1);
70 $self->fullstop_space_harden (1);
73 # The =for and =begin targets that we accept.
74 $self->accept_targets (qw/text TEXT/);
76 # Ensure that contiguous blocks of code are merged together. Otherwise,
77 # some of the guesswork heuristics don't work right.
78 $self->merge_text (1);
80 # Pod::Simple doesn't do anything useful with our arguments, but we want
81 # to put them in our object as hash keys and values. This could cause
82 # problems if we ever clash with Pod::Simple's own internal class
85 my @opts = map { ("opt_$_", $opts{$_}) } keys %opts;
86 %$self = (%$self, @opts);
88 # Send errors to stderr if requested.
89 if ($$self{opt_stderr}) {
90 $self->no_errata_section (1);
91 $self->complain_stderr (1);
92 delete $$self{opt_stderr};
95 # Initialize various things from our parameters.
96 $$self{opt_alt} = 0 unless defined $$self{opt_alt};
97 $$self{opt_indent} = 4 unless defined $$self{opt_indent};
98 $$self{opt_margin} = 0 unless defined $$self{opt_margin};
99 $$self{opt_loose} = 0 unless defined $$self{opt_loose};
100 $$self{opt_sentence} = 0 unless defined $$self{opt_sentence};
101 $$self{opt_width} = 76 unless defined $$self{opt_width};
103 # Figure out what quotes we'll be using for C<> text.
104 $$self{opt_quotes} ||= '"';
105 if ($$self{opt_quotes} eq 'none') {
106 $$self{LQUOTE} = $$self{RQUOTE} = '';
107 } elsif (length ($$self{opt_quotes}) == 1) {
108 $$self{LQUOTE} = $$self{RQUOTE} = $$self{opt_quotes};
109 } elsif ($$self{opt_quotes} =~ /^(.)(.)$/
110 || $$self{opt_quotes} =~ /^(..)(..)$/) {
114 croak qq(Invalid quote specification "$$self{opt_quotes}");
117 # If requested, do something with the non-POD text.
118 $self->code_handler (\&handle_code) if $$self{opt_code};
120 # Return the created object.
124 ##############################################################################
126 ##############################################################################
128 # This is the glue that connects the code below with Pod::Simple itself. The
129 # goal is to convert the event stream coming from the POD parser into method
130 # calls to handlers once the complete content of a tag has been seen. Each
131 # paragraph or POD command will have textual content associated with it, and
132 # as soon as all of a paragraph or POD command has been seen, that content
133 # will be passed in to the corresponding method for handling that type of
134 # object. The exceptions are handlers for lists, which have opening tag
135 # handlers and closing tag handlers that will be called right away.
137 # The internal hash key PENDING is used to store the contents of a tag until
138 # all of it has been seen. It holds a stack of open tags, each one
139 # represented by a tuple of the attributes hash for the tag and the contents
142 # Add a block of text to the contents of the current node, formatting it
143 # according to the current formatting instructions as we do.
145 my ($self, $text) = @_;
146 my $tag = $$self{PENDING}[-1];
150 # Given an element name, get the corresponding method name.
151 sub method_for_element {
152 my ($self, $element) = @_;
154 $element =~ tr/A-Z/a-z/;
155 $element =~ tr/_a-z0-9//cd;
159 # Handle the start of a new element. If cmd_element is defined, assume that
160 # we need to collect the entire tree for this element before passing it to the
161 # element method, and create a new tree into which we'll collect blocks of
162 # text and nested elements. Otherwise, if start_element is defined, call it.
163 sub _handle_element_start {
164 my ($self, $element, $attrs) = @_;
165 my $method = $self->method_for_element ($element);
167 # If we have a command handler, we need to accumulate the contents of the
168 # tag before calling it.
169 if ($self->can ("cmd_$method")) {
170 push (@{ $$self{PENDING} }, [ $attrs, '' ]);
171 } elsif ($self->can ("start_$method")) {
172 my $method = 'start_' . $method;
173 $self->$method ($attrs, '');
177 # Handle the end of an element. If we had a cmd_ method for this element,
178 # this is where we pass along the text that we've accumulated. Otherwise, if
179 # we have an end_ method for the element, call that.
180 sub _handle_element_end {
181 my ($self, $element) = @_;
182 my $method = $self->method_for_element ($element);
184 # If we have a command handler, pull off the pending text and pass it to
185 # the handler along with the saved attribute hash.
186 if ($self->can ("cmd_$method")) {
187 my $tag = pop @{ $$self{PENDING} };
188 my $method = 'cmd_' . $method;
189 my $text = $self->$method (@$tag);
191 if (@{ $$self{PENDING} } > 1) {
192 $$self{PENDING}[-1][1] .= $text;
194 $self->output ($text);
197 } elsif ($self->can ("end_$method")) {
198 my $method = 'end_' . $method;
203 ##############################################################################
205 ##############################################################################
207 # Wrap a line, indenting by the current left margin. We can't use Text::Wrap
208 # because it plays games with tabs. We can't use formline, even though we'd
209 # really like to, because it screws up non-printing characters. So we have to
210 # do the wrapping ourselves.
215 my $spaces = ' ' x $$self{MARGIN};
216 my $width = $$self{opt_width} - $$self{MARGIN};
217 while (length > $width) {
218 if (s/^([^\n]{0,$width})\s+// || s/^([^\n]{$width})//) {
219 $output .= $spaces . $1 . "\n";
224 $output .= $spaces . $_;
225 $output =~ s/\s+$/\n\n/;
229 # Reformat a paragraph of text for the current margin. Takes the text to
230 # reformat and returns the formatted text.
235 # If we're trying to preserve two spaces after sentences, do some munging
236 # to support that. Otherwise, smash all repeated whitespace.
237 if ($$self{opt_sentence}) {
245 return $self->wrap ($_);
248 # Output text to the output device. Replace non-breaking spaces with spaces
249 # and soft hyphens with nothing.
251 my ($self, $text) = @_;
252 $text =~ tr/\240\255/ /d;
253 print { $$self{output_fh} } $text;
256 # Output a block of code (something that isn't part of the POD text). Called
257 # by preprocess_paragraph only if we were given the code option. Exists here
258 # only so that it can be overridden by subclasses.
259 sub output_code { $_[0]->output ($_[1]) }
261 ##############################################################################
262 # Document initialization
263 ##############################################################################
265 # Set up various things that have to be initialized on a per-document basis.
268 my $margin = $$self{opt_indent} + $$self{opt_margin};
270 # Initialize a few per-document variables.
271 $$self{INDENTS} = []; # Stack of indentations.
272 $$self{MARGIN} = $margin; # Default left margin.
273 $$self{PENDING} = [[]]; # Pending output.
278 ##############################################################################
280 ##############################################################################
282 # This method is called whenever an =item command is complete (in other words,
283 # we've seen its associated paragraph or know for certain that it doesn't have
284 # one). It gets the paragraph associated with the item as an argument. If
285 # that argument is empty, just output the item tag; if it contains a newline,
286 # output the item tag followed by the newline. Otherwise, see if there's
287 # enough room for us to output the item tag in the margin of the text or if we
288 # have to put it on a separate line.
290 my ($self, $text) = @_;
291 my $tag = $$self{ITEM};
292 unless (defined $tag) {
293 carp "Item called without tag";
298 # Calculate the indentation and margin. $fits is set to true if the tag
299 # will fit into the margin of the paragraph given our indentation level.
300 my $indent = $$self{INDENTS}[-1];
301 $indent = $$self{opt_indent} unless defined $indent;
302 my $margin = ' ' x $$self{opt_margin};
303 my $fits = ($$self{MARGIN} - $indent >= length ($tag) + 1);
305 # If the tag doesn't fit, or if we have no associated text, print out the
306 # tag separately. Otherwise, put the tag in the margin of the paragraph.
307 if (!$text || $text =~ /^\s+$/ || !$fits) {
308 my $realindent = $$self{MARGIN};
309 $$self{MARGIN} = $indent;
310 my $output = $self->reformat ($tag);
311 $output =~ s/^$margin /$margin:/ if ($$self{opt_alt} && $indent > 0);
312 $output =~ s/\n*$/\n/;
314 # If the text is just whitespace, we have an empty item paragraph;
315 # this can result from =over/=item/=back without any intermixed
316 # paragraphs. Insert some whitespace to keep the =item from merging
317 # into the next paragraph.
318 $output .= "\n" if $text && $text =~ /^\s*$/;
320 $self->output ($output);
321 $$self{MARGIN} = $realindent;
322 $self->output ($self->reformat ($text)) if ($text && $text =~ /\S/);
324 my $space = ' ' x $indent;
325 $space =~ s/^$margin /$margin:/ if $$self{opt_alt};
326 $text = $self->reformat ($text);
327 $text =~ s/^$margin /$margin:/ if ($$self{opt_alt} && $indent > 0);
328 my $tagspace = ' ' x length $tag;
329 $text =~ s/^($space)$tagspace/$1$tag/ or warn "Bizarre space in item";
330 $self->output ($text);
334 # Handle a basic block of text. The only tricky thing here is that if there
335 # is a pending item tag, we need to format this as an item paragraph.
337 my ($self, $attrs, $text) = @_;
339 if (defined $$self{ITEM}) {
340 $self->item ($text . "\n");
342 $self->output ($self->reformat ($text . "\n"));
347 # Handle a verbatim paragraph. Just print it out, but indent it according to
350 my ($self, $attrs, $text) = @_;
351 $self->item if defined $$self{ITEM};
352 return if $text =~ /^\s*$/;
353 $text =~ s/^(\n*)(\s*\S+)/$1 . (' ' x $$self{MARGIN}) . $2/gme;
354 $text =~ s/\s*$/\n\n/;
355 $self->output ($text);
359 # Handle literal text (produced by =for and similar constructs). Just output
360 # it with the minimum of changes.
362 my ($self, $attrs, $text) = @_;
364 $text =~ s/\n{0,2}$/\n/;
365 $self->output ($text);
369 ##############################################################################
371 ##############################################################################
373 # The common code for handling all headers. Takes the header text, the
374 # indentation, and the surrounding marker for the alt formatting method.
376 my ($self, $text, $indent, $marker) = @_;
377 $self->item ("\n\n") if defined $$self{ITEM};
379 if ($$self{opt_alt}) {
380 my $closemark = reverse (split (//, $marker));
381 my $margin = ' ' x $$self{opt_margin};
382 $self->output ("\n" . "$margin$marker $text $closemark" . "\n\n");
384 $text .= "\n" if $$self{opt_loose};
385 my $margin = ' ' x ($$self{opt_margin} + $indent);
386 $self->output ($margin . $text . "\n");
391 # First level heading.
393 my ($self, $attrs, $text) = @_;
394 $self->heading ($text, 0, '====');
397 # Second level heading.
399 my ($self, $attrs, $text) = @_;
400 $self->heading ($text, $$self{opt_indent} / 2, '== ');
403 # Third level heading.
405 my ($self, $attrs, $text) = @_;
406 $self->heading ($text, $$self{opt_indent} * 2 / 3 + 0.5, '= ');
409 # Fourth level heading.
411 my ($self, $attrs, $text) = @_;
412 $self->heading ($text, $$self{opt_indent} * 3 / 4 + 0.5, '- ');
415 ##############################################################################
417 ##############################################################################
419 # Handle the beginning of an =over block. Takes the type of the block as the
420 # first argument, and then the attr hash. This is called by the handlers for
421 # the four different types of lists (bullet, number, text, and block).
422 sub over_common_start {
423 my ($self, $attrs) = @_;
424 $self->item ("\n\n") if defined $$self{ITEM};
426 # Find the indentation level.
427 my $indent = $$attrs{indent};
428 unless (defined ($indent) && $indent =~ /^\s*[-+]?\d{1,4}\s*$/) {
429 $indent = $$self{opt_indent};
432 # Add this to our stack of indents and increase our current margin.
433 push (@{ $$self{INDENTS} }, $$self{MARGIN});
434 $$self{MARGIN} += ($indent + 0);
438 # End an =over block. Takes no options other than the class pointer. Output
439 # any pending items and then pop one level of indentation.
440 sub over_common_end {
442 $self->item ("\n\n") if defined $$self{ITEM};
443 $$self{MARGIN} = pop @{ $$self{INDENTS} };
447 # Dispatch the start and end calls as appropriate.
448 sub start_over_bullet { $_[0]->over_common_start ($_[1]) }
449 sub start_over_number { $_[0]->over_common_start ($_[1]) }
450 sub start_over_text { $_[0]->over_common_start ($_[1]) }
451 sub start_over_block { $_[0]->over_common_start ($_[1]) }
452 sub end_over_bullet { $_[0]->over_common_end }
453 sub end_over_number { $_[0]->over_common_end }
454 sub end_over_text { $_[0]->over_common_end }
455 sub end_over_block { $_[0]->over_common_end }
457 # The common handler for all item commands. Takes the type of the item, the
458 # attributes, and then the text of the item.
460 my ($self, $type, $attrs, $text) = @_;
461 $self->item if defined $$self{ITEM};
463 # Clean up the text. We want to end up with two variables, one ($text)
464 # which contains any body text after taking out the item portion, and
465 # another ($item) which contains the actual item text. Note the use of
466 # the internal Pod::Simple attribute here; that's a potential land mine.
469 if ($type eq 'bullet') {
471 } elsif ($type eq 'number') {
472 $item = $$attrs{'~orig_content'};
475 $item =~ s/\s*\n\s*/ /g;
478 $$self{ITEM} = $item;
480 # If body text for this item was included, go ahead and output that now.
488 # Dispatch the item commands to the appropriate place.
489 sub cmd_item_bullet { my $self = shift; $self->item_common ('bullet', @_) }
490 sub cmd_item_number { my $self = shift; $self->item_common ('number', @_) }
491 sub cmd_item_text { my $self = shift; $self->item_common ('text', @_) }
492 sub cmd_item_block { my $self = shift; $self->item_common ('block', @_) }
494 ##############################################################################
496 ##############################################################################
499 sub cmd_b { return $_[0]{alt} ? "``$_[2]''" : $_[2] }
500 sub cmd_f { return $_[0]{alt} ? "\"$_[2]\"" : $_[2] }
501 sub cmd_i { return '*' . $_[2] . '*' }
502 sub cmd_x { return '' }
504 # Apply a whole bunch of messy heuristics to not quote things that don't
505 # benefit from being quoted. These originally come from Barrie Slaymaker and
506 # largely duplicate code in Pod::Man.
508 my ($self, $attrs, $text) = @_;
510 # A regex that matches the portion of a variable reference that's the
511 # array or hash index, separated out just because we want to use it in
512 # several places in the following regex.
513 my $index = '(?: \[.*\] | \{.*\} )?';
515 # Check for things that we don't want to quote, and if we find any of
516 # them, return the string with just a font change and no quoting.
520 ( [\'\`\"] ) .* \1 # already quoted
521 | \` .* \' # `quoted'
522 | \$+ [\#^]? \S $index # special ($^Foo, $")
523 | [\$\@%&*]+ \#? [:\'\w]+ $index # plain var or func
524 | [\$\@%&*]* [:\'\w]+ (?: -> )? \(\s*[^\s,]\s*\) # 0/1-arg func call
525 | [+-]? ( \d[\d.]* | \.\d+ ) (?: [eE][+-]?\d+ )? # a number
526 | 0x [a-fA-F\d]+ # a hex constant
531 # If we didn't return, go ahead and quote the text.
532 return $$self{opt_alt}
534 : "$$self{LQUOTE}$text$$self{RQUOTE}";
537 # Links reduce to the text that we're given, wrapped in angle brackets if it's
540 my ($self, $attrs, $text) = @_;
541 return $$attrs{type} eq 'url' ? "<$text>" : $text;
544 ##############################################################################
545 # Backwards compatibility
546 ##############################################################################
548 # The old Pod::Text module did everything in a pod2text() function. This
549 # tries to provide the same interface for legacy applications.
553 # This is really ugly; I hate doing option parsing in the middle of a
554 # module. But the old Pod::Text module supported passing flags to its
555 # entry function, so handle -a and -<number>.
556 while ($_[0] =~ /^-/) {
558 if ($flag eq '-a') { push (@args, alt => 1) }
559 elsif ($flag =~ /^-(\d+)$/) { push (@args, width => $1) }
566 # Now that we know what arguments we're using, create the parser.
567 my $parser = Pod::Text->new (@args);
569 # If two arguments were given, the second argument is going to be a file
570 # handle. That means we want to call parse_from_filehandle(), which means
571 # we need to turn the first argument into a file handle. Magic open will
572 # handle the <&STDIN case automagically.
576 unless (open (IN, $fhs[0])) {
577 croak ("Can't open $fhs[0] for reading: $!\n");
581 $parser->output_fh ($fhs[1]);
582 my $retval = $parser->parse_file ($fhs[0]);
583 my $fh = $parser->output_fh ();
587 return $parser->parse_file (@_);
591 # Reset the underlying Pod::Simple object between calls to parse_from_file so
592 # that the same object can be reused to convert multiple pages.
593 sub parse_from_file {
597 # Fake the old cutting option to Pod::Parser. This fiddings with internal
598 # Pod::Simple state and is quite ugly; we need a better approach.
599 if (ref ($_[0]) eq 'HASH') {
601 if (defined ($$opts{-cutting}) && !$$opts{-cutting}) {
603 $$self{last_was_blank} = 1;
608 my $retval = $self->Pod::Simple::parse_from_file (@_);
610 # Flush output, since Pod::Simple doesn't do this. Ideally we should also
611 # close the file descriptor if we had to open one, but we can't easily
613 my $fh = $self->output_fh ();
614 my $oldfh = select $fh;
623 # Pod::Simple failed to provide this backward compatibility function, so
624 # implement it ourselves. File handles are one of the inputs that
625 # parse_from_file supports.
626 sub parse_from_filehandle {
628 $self->parse_from_file (@_);
631 ##############################################################################
632 # Module return value and documentation
633 ##############################################################################
640 Pod::Text - Convert POD data to formatted ASCII text
643 alt stderr Allbery Sean Burke's
648 my $parser = Pod::Text->new (sentence => 0, width => 78);
650 # Read POD from STDIN and write to STDOUT.
651 $parser->parse_from_filehandle;
653 # Read POD from file.pod and write to file.txt.
654 $parser->parse_from_file ('file.pod', 'file.txt');
658 Pod::Text is a module that can convert documentation in the POD format (the
659 preferred language for documenting Perl) into formatted ASCII. It uses no
660 special formatting controls or codes whatsoever, and its output is therefore
661 suitable for nearly any device.
663 As a derived class from Pod::Simple, Pod::Text supports the same methods and
664 interfaces. See L<Pod::Simple> for all the details; briefly, one creates a
665 new parser with C<< Pod::Text->new() >> and then normally calls parse_file().
667 new() can take options, in the form of key/value pairs, that control the
668 behavior of the parser. The currently recognized options are:
674 If set to a true value, selects an alternate output format that, among other
675 things, uses a different heading style and marks C<=item> entries with a
676 colon in the left margin. Defaults to false.
680 If set to a true value, the non-POD parts of the input file will be included
681 in the output. Useful for viewing code documented with POD blocks with the
682 POD rendered and the code left intact.
686 The number of spaces to indent regular text, and the default indentation for
687 C<=over> blocks. Defaults to 4.
691 If set to a true value, a blank line is printed after a C<=head1> heading.
692 If set to false (the default), no blank line is printed after C<=head1>,
693 although one is still printed after C<=head2>. This is the default because
694 it's the expected formatting for manual pages; if you're formatting
695 arbitrary text documents, setting this to true may result in more pleasing
700 The width of the left margin in spaces. Defaults to 0. This is the margin
701 for all text, including headings, not the amount by which regular text is
702 indented; for the latter, see the I<indent> option. To set the right
703 margin, see the I<width> option.
707 Sets the quote marks used to surround CE<lt>> text. If the value is a
708 single character, it is used as both the left and right quote; if it is two
709 characters, the first character is used as the left quote and the second as
710 the right quoted; and if it is four characters, the first two are used as
711 the left quote and the second two as the right quote.
713 This may also be set to the special value C<none>, in which case no quote
714 marks are added around CE<lt>> text.
718 If set to a true value, Pod::Text will assume that each sentence ends in two
719 spaces, and will try to preserve that spacing. If set to false, all
720 consecutive whitespace in non-verbatim paragraphs is compressed into a
721 single space. Defaults to true.
725 Send error messages about invalid POD to standard error instead of
726 appending a POD ERRORS section to the generated output.
730 The column at which to wrap text on the right-hand side. Defaults to 76.
734 The standard Pod::Simple method parse_file() takes one argument, the file or
735 file handle to read from, and writes output to standard output unless that
736 has been changed with the output_fh() method. See L<Pod::Simple> for the
737 specific details and for other alternative interfaces.
743 =item Bizarre space in item
745 =item Item called without tag
747 (W) Something has gone wrong in internal C<=item> processing. These
748 messages indicate a bug in Pod::Text; you should never see them.
750 =item Can't open %s for reading: %s
752 (F) Pod::Text was invoked via the compatibility mode pod2text() interface
753 and the input file it was given could not be opened.
755 =item Invalid quote specification "%s"
757 (F) The quote specification given (the quotes option to the constructor) was
758 invalid. A quote specification must be one, two, or four characters long.
764 This is a replacement for an earlier Pod::Text module written by Tom
765 Christiansen. It has a revamped interface, since it now uses Pod::Simple,
766 but an interface roughly compatible with the old Pod::Text::pod2text()
767 function is still available. Please change to the new calling convention,
770 The original Pod::Text contained code to do formatting via termcap
771 sequences, although it wasn't turned on by default and it was problematic to
772 get it to work at all. This rewrite doesn't even try to do that, but a
773 subclass of it does. Look for L<Pod::Text::Termcap>.
777 L<Pod::Simple>, L<Pod::Text::Termcap>, L<pod2text(1)>
779 The current version of this module is always available from its web site at
780 L<http://www.eyrie.org/~eagle/software/podlators/>. It is also part of the
781 Perl core distribution as of 5.6.0.
785 Russ Allbery <rra@stanford.edu>, based I<very> heavily on the original
786 Pod::Text by Tom Christiansen <tchrist@mox.perl.com> and its conversion to
787 Pod::Parser by Brad Appleton <bradapp@enteract.com>. Sean Burke's initial
788 conversion of Pod::Man to use Pod::Simple provided much-needed guidance on
789 how to use Pod::Simple.
791 =head1 COPYRIGHT AND LICENSE
793 Copyright 1999, 2000, 2001, 2002, 2004, 2006, 2008 Russ Allbery
796 This program is free software; you may redistribute it and/or modify it
797 under the same terms as Perl itself.