NAME¶
HTML::SimpleParse - a bare-bones HTML parser
SYNOPSIS¶
use HTML::SimpleParse;
# Parse the text into a simple tree
my $p = new HTML::SimpleParse( $html_text );
$p->output; # Output the HTML verbatim
$p->text( $new_text ); # Give it some new HTML to chew on
$p->parse # Parse the new HTML
$p->output;
my %attrs = HTML::SimpleParse->parse_args('A="xx" B=3');
# %attrs is now ('A' => 'xx', 'B' => '3')
DESCRIPTION¶
This module is a simple HTML parser. It is similar in concept to HTML::Parser,
but it differs from HTML::TreeBuilder in a couple of important ways.
First, HTML::TreeBuilder knows which tags can contain other tags, which start
tags have corresponding end tags, which tags can exist only in the
<HEAD> portion of the document, and so forth. HTML::SimpleParse does not
know any of these things. It just finds tags and text in the HTML you give it,
it does not care about the specific content of these tags (though it does
distiguish between different _types_ of tags, such as comments, starting tags
like <b>, ending tags like </b>, and so on).
Second, HTML::SimpleParse does not create a hierarchical tree of HTML content,
but rather a simple linear list. It does not pay any attention to balancing
start tags with corresponding end tags, or which pairs of tags are inside
other pairs of tags.
Because of these characteristics, you can make a very effective HTML filter by
sub-classing HTML::SimpleParse. For example, to remove all comments from HTML:
package NoComment;
use HTML::SimpleParse;
@ISA = qw(HTML::SimpleParse);
sub output_comment {}
package main;
NoComment->new($some_html)->output;
Historically, I started the HTML::SimpleParse project in part because of a
misunderstanding about HTML::Parser's functionality. Many aspects of these two
modules actually overlap. I continue to maintain the HTML::SimpleParse module
because people seem to be depending on it, and because beginners sometimes
find HTML::SimpleParse to be simpler than HTML::Parser's more powerful
interface. People also seem to get a fair amount of usage out of the
"parse_args()" method directly.
Methods¶
- •
- new
$p = new HTML::SimpleParse( $some_html );
Creates a new HTML::SimpleParse object. Optionally takes one argument, a
string containing some HTML with which to initialize the object. If you
give it a non-empty string, the HTML will be parsed into a tree and ready
for outputting.
Can also take a list of attributes, such as
$p = new HTML::SimpleParse( $some_html, 'fix_case' => -1);
See the "parse_args()" method below for an explanation of this
attribute.
- •
- text
$text = $p->text;
$p->text( $new_text );
Get or set the contents of the HTML to be parsed.
- •
- tree
foreach ($p->tree) { ... }
Returns a list of all the nodes in the tree, in case you want to step
through them manually or something. Each node in the tree is an anonymous
hash with (at least) three data members, $node->{type} (is this a
comment, a start tag, an end tag, etc.), $node->{content} (all the text
between the angle brackets, verbatim), and $node->{offset} (number of
bytes from the beginning of the string).
The possible values of $node->{type} are "text",
"starttag", "endtag", "ssi", and
"markup".
- •
- parse
$p->parse;
Once an object has been initialized with some text, call $p->parse and a
tree will be created. After the tree is created, you can call
$p->output. If you feed some text to the new() method, parse
will be called automatically during your object's construction.
- •
- parse_args
%hash = $p->parse_args( $arg_string );
This routine is handy for parsing the contents of an HTML tag into key=value
pairs. For instance:
$text = 'type=checkbox checked name=flavor value="chocolate or strawberry"';
%hash = $p->parse_args( $text );
# %hash is ( TYPE=>'checkbox', CHECKED=>undef, NAME=>'flavor',
# VALUE=>'chocolate or strawberry' )
Note that the position of the last m//g search on the string (the value
returned by Perl's pos() function) will be altered by the
parse_args function, so make sure you take that into account if (in the
above example) you do "$text =~ m/something/g".
The parse_args() method can be run as either an object method or as a
class method, i.e. as either $p->parse_args(...) or
HTML::SimpleParse->parse_args(...).
HTML attribute lists are supposed to be case-insensitive with respect to
attribute names. To achieve this behavior, parse_args() respects
the 'fix_case' flag, which can be set either as a package global
$FIX_CASE, or as a class member datum 'fix_case'. If set to 0, no case
conversion is done. If set to 1, all keys are converted to upper case. If
set to -1, all keys are converted to lower case. The default is 1, i.e.
all keys are uppercased.
If an attribute takes no value (like "checked" in the above
example) then it will still have an entry in the returned hash, but its
value will be "undef". For example:
%hash = $p->parse_args('type=checkbox checked name=banana value=""');
# $hash{CHECKED} is undef, but $hash{VALUE} is ""
This method actually returns a list (not a hash), so duplicate attributes
and order will be preserved if you want them to be:
@hash = $p->parse_args("name=family value=gwen value=mom value=pop");
# @hash is qw(NAME family VALUE gwen VALUE mom VALUE pop)
- •
- output
$p->output;
This will output the contents of the HTML, passing the real work off to the
output_text, output_comment, etc. functions. If you do not override any of
these methods, this module will output the exact text that it parsed into
a tree in the first place.
- •
- get_output
print $p->get_output
Similar to $p-> output(), but returns its result instead of
printing it.
- •
- execute
foreach ($p->tree) {
print $p->execute($_);
}
Executes a single node in the HTML parse tree. Useful if you want to loop
through the nodes and output them individually.
The following methods do the actual outputting of the various parts of the HTML.
Override some of them if you want to change the way the HTML is output. For
instance, to strip comments from the HTML, override the output_comment method
like so:
# In subclass:
sub output_comment { } # Does nothing
- •
- output_text
- •
- output_comment
- •
- output_endtag
- •
- output_starttag
- •
- output_markup
- •
- output_ssi
CAVEATS¶
Please do not assume that the interface here is stable. This is a first pass,
and I'm still trying to incorporate suggestions from the community. If you
employ this module somewhere, make doubly sure before upgrading that none of
your code breaks when you use the newer version.
BUGS¶
- •
- Embedded >s are broken
Won't handle tags with embedded >s in them, like <input name=expr
value="x > y">. This will be fixed in a future version,
probably by using the parse_args method. Suggestions are welcome.
TO DO¶
- •
- extensibility
Based on a suggestion from Randy Harmon (thanks), I'd like to make it easier
for subclasses of SimpleParse to pick out other kinds of HTML blocks, i.e.
extend the set {text, comment, endtag, starttag, markup, ssi} to include
more members. Currently the only easy way to do that is by overriding the
"parse" method:
sub parse { # In subclass
my $self = $_[0];
$self->SUPER::parse(@_);
foreach ($self->tree) {
if ($_->{content} =~ m#^a\s+#i) {
$_->{type} = 'anchor_start';
}
}
}
sub output_anchor_start {
# Whatever you want...
}
Alternatively, this feature might be implemented by hanging attatchments
onto the parsing loop, like this:
my $parser = new SimpleParse( $html_text );
$regex = '<(a\s+.*?)>';
$parser->watch_for( 'anchor_start', $regex );
sub SimpleParse::output_anchor_start {
# Whatever you want...
}
I think I like that idea better. If you wanted to, you could make a subclass
with output_anchor_start as one of its methods, and put the ->watch_for
stuff in the constructor.
- •
- reading from filehandles
It would be nice if you could initialize an object by giving it a filehandle
or filename instead of the text itself.
- •
- tests
I need to write a few tests that run under "make test".
AUTHOR¶
Ken Williams <ken@forum.swarthmore.edu>
COPYRIGHT¶
Copyright 1998 Swarthmore College. All rights reserved.
This library is free software; you can redistribute it and/or modify it under
the same terms as Perl itself.