The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.


HTML::FormatExternal - HTML to text formatting using external programs


This is a collection of formatter modules which turn HTML into plain text by dumping it through the respective external programs.


The module interfaces are compatible with HTML::Formatter modules such as HTML::FormatText, but the external programs do all the work.

Common formatting options are used where possible, such as leftmargin and rightmargin. So just by switching the class you can use a different program (or the plain HTML::FormatText) according to personal preference, or strengths and weaknesses, or what you've got.

There's nothing particularly difficult about piping through these programs, but a unified interface hides details like how to set margins and how to force input or output charsets.


Each of the classes above provide the following functions. The XXX in the class names here is a placeholder for any of Elinks, Lynx, etc as above.

See examples/ in the HTML-FormatExternal sources for a complete sample program.

Formatter Compatible Functions

$text = HTML::FormatText::XXX->format_file ($filename, key=>value,...)
$text = HTML::FormatText::XXX->format_string ($html_string, key=>value,...)

Run the formatter program over a file or string with the given options and return the formatted result as a string. See "OPTIONS" below for possible key/value options. For example,

    $text = HTML::FormatText::Lynx->format_file ('/my/file.html');

    $text = HTML::FormatText::W3m->format_string
      ('<html><body> <p> Hello world! </p </body></html>');

format_file() ensures any $filename is interpreted as a filename (by escaping as necessary against however the programs interpret command line arguments).

$formatter = HTML::FormatText::XXX->new (key=>value, ...)

Create a formatter object with the given options. In the current implementation an object doesn't do much more than remember the options for future use.

    $formatter = HTML::FormatText::Elinks->new(rightmargin => 60);
$text = $formatter->format ($tree_or_string)

Run the $formatter program on a HTML::TreeBuilder tree or a string, using the options in $formatter, and return the result as a string.

A TreeBuilder argument (ie. a HTML::Element) is accepted for compatibility with HTML::Formatter. The tree is simply turned into a string with $tree->as_HTML to pass to the program, so if you've got a string already then give that instead of a tree.

HTML::Element itself has a format() method (see "format" in HTML::Element) which runs a given $formatter. A HTML::FormatExternal object can be used for $formatter.

    $text = $tree->format($formatter);

    # which dispatches to
    $text = $formatter->format($tree);

Extra Functions

The following are extra methods not available in the plain HTML::FormatText.

HTML::FormatText::XXX->program_version ()
HTML::FormatText::XXX->program_full_version ()
$formatter->program_version ()
$formatter->program_full_version ()

Return the version number of the formatter program as reported by its --version or similar option. If the formatter program is not available then return undef.

program_version() is the bare version number, perhaps with "beta" or similar indication. program_full_version() is the entire version output, which may include build options, copyright notice, etc.

    $str = HTML::FormatText::Lynx->program_version();
    # eg. "2.8.7dev.10"

    $str = HTML::FormatText::W3m->program_full_version();
    # eg. "w3m version w3m/0.5.2, options lang=en,m17n,image,..."

The version number of the respective Perl module itself is available in the usual way (see "VERSION" in UNIVERSAL).

    $modulever = HTML::FormatText::Netrik->VERSION;
    $modulever = $formatter->VERSION


File or byte string input is by default interpreted by the programs in their usual ways. This should mean HTML Latin-1 but user configurations might override that and some programs recognise a <meta> charset declaration or a Unicode BOM. The input_charset option below can force the input charset.

Perl wide-character input string is encoded and passed to the program in whatever way it best understands. Usually this is UTF-8 but in some cases it is entitized instead. The input_charset option can force the input charset to use if for some reason UTF-8 is not best.

The output string is either bytes or wide chars. By default output is the same as input, so wide char string input gives wide output and byte input string or file input gives byte output. The output_wide option can force the output type (and is the way to get wide chars back from format_file()).

Byte output is whatever the program produces. Its default might be the locale charset or other user configuration which suits direct display to the user's terminal. The output_charset option can force the output to be certain or to be ready for further processing.

Wide char output is done by choosing the best output charset the program can do and decoding its output. Usually this means UTF-8 but some of the programs may only have less. The output_charset option can force the charset used and decoded. If it's something less than UTF-8 then some programs might for example give ASCII art approximations of otherwise unrepresentable characters.

Byte input is usual for HTML downloaded from a HTTP server or from a MIME email and the headers have the input_charset which applies. Byte output is good to go straight out to a tty or back to more MIME etc. The input and output charsets could differ if a server gives something other than what you want for final output.

Wide chars are most convenient for crunching text within Perl. The default wide input giving wide output is designed to be transparent for this.

For reference, if a HTML::Element tree contains wide char strings then its usual as_HTML() method, which is used by format() above, produces wide char HTML so the formatters here give wide char text. Actually as_HTML() produces all ASCII because its default behaviour is to entitize anything "unsafe", but it's still a wide char string so the formatted output text is wide.


The following options can be given to the constructor or to the formatting methods. The defaults are whatever the respective programs do. The programs generally read their config files when dumping so the defaults and formatting details may follow the user's personal preferences. Usually this is a good thing.

leftmargin => INTEGER
rightmargin => INTEGER

The column numbers for the left and right hand ends of the text. leftmargin 0 means no padding on the left. rightmargin is the text width, so for instance 60 would mean the longest line is 60 characters (inclusive of any leftmargin). These options are compatible with HTML::FormatText.

rightmargin is not necessarily a hard limit. Some of the programs will exceed it in a HTML literal <pre>, or a run of &nbsp; or similar.

input_charset => STRING

Force the HTML input to be interpreted as bytes of the given charset, irrespective of locale, user configuration, <meta> in the HTML, etc.

output_charset => STRING

Force the text output to be encoded as the given charset. The default varies among the programs, but usually defaults to the locale.

output_wide => 0,1,"as_input"

Select output string as wide characters rather than bytes. The default is "as_input" which means a wide char input string results in a wide char output string and a byte input or file input is byte output. See "CHARSETS" above for how wide characters work.

Bytes or wide chars output can be forced by 0 or 1 respectively. For example to get wide char output when formatting a file,

    $wide_char_text = HTML::FormatText::W3m->format_file
                       ('/my/file.html', output_wide => 1);
base => STRING

Set the base URL for any relative links within the HTML (similar to HTML::FormatText::WithLinks). Usually this should be the location the HTML was downloaded from.

If the document contains its own <base> setting then currently the document takes precedence. Only Lynx and Elinks display absolutized link targets and the option has no effect on the other programs.


The formatter modules can be used under perl -T taint mode. They run external programs so it's necessary to untaint $ENV{PATH} in the usual way per "Cleaning Up Your Path" in perlsec.

The formatted text strings returned are always tainted, on the basis that they use or include data from outside the Perl program. The program_version() and program_full_version() strings are tainted too.


leftmargin is implemented by adding spaces to the program output. For byte output it this is ASCII spaces and that will be badly wrong for unusual output like UTF-16 which is not a byte superset of ASCII. For wide char output the margin is applied after decoding to wide chars so is correct. It'd be better to ask the programs to do the margin but their options for that are poor.

There's nothing done with errors or warning messages from the programs. Generally they make a best effort on doubtful HTML, but fatal errors like bad options or missing libraries ought to be somehow trapped.


elinks (from Aug 2008 onwards) and netrik can produce ANSI escapes for colours, underline, etc, and html2text and lynx can produce tty style backspace overstriking. This might be good for text destined for a tty or further crunching. Perhaps an ansi or tty option could enable this, where possible, but for now it's deliberately turned off in those programs to keep the default as plain text.


HTML::FormatText::Elinks, HTML::FormatText::Html2text, HTML::FormatText::Links, HTML::FormatText::Netrik, HTML::FormatText::Lynx, HTML::FormatText::Vilistextum, HTML::FormatText::W3m, HTML::FormatText::Zen

HTML::FormatText, HTML::FormatText::WithLinks, HTML::FormatText::WithLinks::AndTables



Copyright 2008, 2009, 2010, 2011, 2012, 2013, 2015 Kevin Ryde

HTML-FormatExternal is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3, or (at your option) any later version.

HTML-FormatExternal is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with HTML-FormatExternal. If not, see