<- Back

# webdump HTML to plain-text converter

Last modification on 2024-07-17

webdump is (yet another) HTML to plain-text converter tool.

It reads HTML in UTF-8 from stdin and writes plain-text to stdout.


## Goals and scope

The main goal of this tool for me is to use it for converting HTML mails to
plain-text and to convert HTML content in RSS feeds to plain-text.

The tool will only convert HTML to stdout, similarly to links -dump or lynx
-dump but simpler and more secure.

* HTML and XHTML will be supported.
* There will be some workarounds and quirks for broken and legacy HTML code.
* It will be usable and secure for reading HTML from mails and RSS/Atom feeds.
* No remote resources which are part of the HTML will be downloaded:
  images, video, audio, etc. But these may be visible as a link reference.
* Data will be written to stdout. Intended for plain-text or a text terminal.
* No support for Javascript, CSS, frame rendering or form processing.
* No HTTP or network protocol handling: HTML data is read from stdin.
* Listings for references and some options to extract them in a list that is
  usable for scripting. Some references are: link anchors, images, audio, video,
  HTML (i)frames, etc.
* Security: on OpenBSD it uses pledge("stdio", NULL).
* Keep the code relatively small, simple and hackable.


## Features

* Support for word-wrapping.
* A mode to enable basic markup: bold, underline, italic and blink ;)
* Indentation of headers, paragraphs, pre and list items.
* Basic support to query elements or hide them.
* Show link references.
* Show link references and resources such as img, video, audio, subtitles.
* Export link references and resources to a TAB-separated format.


## Usage examples

        url='https://codemadness.org/sfeed.html'
        
        curl -s "$url" | webdump -r -b "$url" | less
        
        curl -s "$url" | webdump -8 -a -i -l -r -b "$url" | less -R
        
        curl -s "$url" | webdump -s 'main' -8 -a -i -l -r -b "$url" | less -R

Yes, all these option flags look ugly, a shellscript wrapper could be used :)


## Practical examples

To use webdump as a HTML to text filter for example in the mutt mail client,
change in ~/.mailcap:

        text/html; webdump -i -l -r < %s; needsterminal; copiousoutput

In mutt you should then add:

        auto_view text/html


Using webdump as a HTML to text filter for sfeed_curses (otherwise the default is lynx):

        SFEED_HTMLCONV="webdump -d -8 -r -i -l -a" sfeed_curses ~/.sfeed/feeds/*


# Query/selector examples

The query syntax using the -s option is a bit inspired by CSS (but much more limited).

To get the title from a HTML page:

        url='https://codemadness.org/sfeed.html'
        
        title=$(curl -s "$url" | webdump -s 'title')
        printf '%s\n' "$title"

List audio and video-related content from a HTML page, redirect fd 3 to fd 1 (stdout):

        url="https://media.ccc.de/v/051_Recent_features_to_OpenBSD-ntpd_and_bgpd"
        curl -s "$url" | webdump -x -s 'audio,video' -b "$url" 3>&1 >/dev/null | cut -f 2


## Clone

        git clone git://git.codemadness.org/webdump


## Browse

You can browse the source-code at:
* https://git.codemadness.org/webdump/
* gopher://codemadness.org/1/git/webdump

## Build and install

        $ make
        # make install


## Dependencies

* C compiler.
* libc + some BSDisms.


## Trade-offs

All software has trade-offs.

webdump processes HTML in a single-pass. It does not buffer the full DOM tree.
Although due to the nature of HTML/XML some parts like attributes need to be
buffered.

Rendering tables in webdump is very limited. Twibright Links has really nice
table rendering. However implementing a similar feature in the current design of
webdump would make the code much more complex. Twibright links
processes a full DOM tree and processes the tables in multiple passes (to
measure the table cells) etc.  Of course tables can be nested also, or HTML tables
that are used for creating layouts (these are mostly older webpages).

These trade-offs and preferences are chosen for now. It may change in the
future.  Fortunately there are the usual good suspects for HTML to plain-text
conversion, each with their own chosen trade-offs of course:
* twibright links: »http://links.twibright.com/«
* lynx: »https://lynx.invisible-island.net/«
* w3m: »https://w3m.sourceforge.net/«
* xmllint (part of libxml2): »https://gitlab.gnome.org/GNOME/libxml2/-/wikis/home«
* xmlstarlet: »https://xmlstar.sourceforge.net/«