Html2Wiki
This extension to MediaWiki is used to import HTML content into the wiki.
Contents
Requirements or Dependencies[edit | edit source]
This extension was built on MediaWiki version 1.25alpha, and is likely not compatible with earlier releases since there are a number of external libraries such as jQuery which have changed over time
It may depend on a Parsoid service, which is used to transform an HTML DOM into wikitext. More at http://www.mediawiki.org/wiki/Parsoid
Since parsing the DOM is problematic when using PHP's native DOM manipulation (which is itself based on libxml), we use the QueryPath project to provide a more flexible parsing platform.
Html2Wiki can import entire document sets and maintain a hierarchy of those documents. The $wgNamespacesWithSubpages variable will allow you to create a hierarchy in your wiki's 'main' namespace; and even automatically create navigation links to parent article content. Taking this further, the SubPageList extension creates navigation blocks for subpages.
The document sets we were importing were based on generated source code documentation (coming from an open source documentation generator called Natural Docs) which creates DHTML "mouseovers" for glossary terms. To create similar functionality in the wiki environment, we will rely on the Lingo extension to create a Glossary of terms.
Usage[edit | edit source]
Select a file using the import HTML form found on the Special:Html2Wiki page (similar to Special:Upload for regular images). When Special:Html2Wiki is installed, it adds a link to the 'Tools' section of your wiki, for quick easy access to the importer.
Enter a comment in the Comment field, which is logged in the 'Recent Changes' content as well as the Special:Log area.
The upload is automatically categorized according to the content provided.
You can optionally specify a "Collection Name" for your content. The Collection Name represents where this content is coming from (e.g. The book, or website). Any unique identifier will do. The "Collection Name" is used to tag (categorize) all content that is part of that collection. And, all content that is part of a Collection will be organized "under" that Collection Name in a hierarchy. This lets you have 2 or more articles in your wiki named "Introduction" if they belong to separate Collections. Specifying an existing Collection Name + article title will update the existing content. In fact, to reimport a single file and maintain it's 'position' in a collection, you would specify the full path to the file.
Uploads in MediaWiki[edit | edit source]
How does upload work?
Parsoid.hooks.php has a hook onFileUpload, which refers to filerepo/file/LocalFile.php
If I'm reading it correctly, LocalFile.php is mostly called from a Repo object.
LocalFile::upload() and recordUpload() are informative.
Zip archive handling[edit | edit source]
In order to handle the zip upload, we'll have to traverse all files and index hrefs as they exist. We'll need to map those to safe titles and rewrite the source to use those safe URLs. This has to be done for both anchors and images.
Practically speaking MW is probably more flexible than we need; but we'll want to check
[legaltitlechars] => %!"$&'()*,\-.\/0-9:;=?@A-Z\\^_`a-z~\x80-\xFF+
Since we're using First-letter capitals, we'll need to account for that in rewriting all hrefs.
Styles and Scripts[edit | edit source]
Cascading Style Sheets (CSS) as well as JavaScript (js) are not kept as part of the transformation. Although, we are working on including CSS
Wiki Text markup[edit | edit source]
The fundamental requirement for this extension is to transform input (HTML) into Wiki Text (see http://www.mediawiki.org/wiki/Help:Formatting) because that is the format stored by the MediaWiki system. Originally, it was envisioned that we would make API calls to the Parsoid service which is used by the Visual Editor extension. However, Parsoid is not very flexible in the HTML that it will handle. To get a more flexible converter, we use the Pandoc project which is able to (read and) write to MediaWiki Text format.
For each source type (ie. UVM, InfoHub) we will need to survey the content to identify the essential content, and remove navigation, JavaScript, presentational graphics, etc. We should have a "fingerprint" that we can use to sniff out the type of document set that the user is uploading to the wiki.
As a result of sniffing the source type, we can properly index and import content only, while discarding the dross. We can likewise apply the correct transformation to the source. For example, there is a bunch of Verilog source in UVM, that should be converted to GeSHi <source> tags while maybe there isn't any in the InfoHubs.
Form file content is saved to server (tmp), and that triggers conversion attempt. A Title is proposed from text (checked in the db), and user can override naming HTML is converted to wiki text for the content of the article.
Image references are extracted from source
grep -P -r --only-matching '(?<=<img src=["'\''])([^"'\'']*)' ./data/uvm-1.1d/docs/html/files/
or you could use PHPDOM, but it's awfully finicky and under-documented
For each of those images, the image asset is retrieved, uploaded with automatic file naming based on the "prefix" + src attribute.
Also, each image is tagged with the collection name for easier identification
Once all the images are contained in the wiki, the wiki markup for the article can be updated to reference those images. IOW, it may be possible to upload an HTML source file, and batch a job to import all images AND update the article source to use the images... Or, since we know ahead of time what the image file name will be, we can just reference the non-existant images in the article. They will exist in the wiki after a short delay required to fetch and process the image files.
(Eliminate duplicate images based on checksum?)
What, if any, additional tables do we need in the database? [1]
We may need to store checksums for zip uploads, because we don't want to store the zip itself, but we may want to recognize a re-upload attempt?
Logging is provided at Special:Log/html2wiki The facility for logging will tap into LogEntry as outlined at https://www.mediawiki.org/wiki/Manual:Logging_to_Special:Log
Interestingly, SpecialUpload must call LogEntry from it's hooks SpecialImport calls LogPage which itself invokes LogEntry (see includes/logging).
More information can be found at https://www.mediawiki.org/wiki/Extension:Html2Wiki
The code is simultaneously updated on MediaWiki servers and GitHub, so feel free to fork, or pull it from either location.
- @todo publish the extension upstream. first with the @link http://www.mediawiki.org/wiki/Template:Extension
git clone https://gerrit.wikimedia.org/r/p/mediawiki/
or (with gerrit auth)
git clone ssh://USERNAME@gerrit.wikimedia.org:29418/mediawiki/services/parsoid
In order to use Parsoid at all, we need to have the content conform to the
MediaWikiDOMspec, which is based on HTML5 and RDFa https://www.mediawiki.org/wiki/Parsoid/MediaWiki_DOM_spec#Ref_and_References.
We need to parse the incoming content, validate and possibly transform the document type to HTML5 (<!DOCTYPE html>) and then transform the HTML5 to MediaWiki DOMspec http://www.w3.org/TR/html5/syntax.html#html-parser
UVM Content is assumed to be HTML4 Transitional (as some of it is expressly) @todo verify this with Tidy
The InfoHubs content is produced by Quadralay WebWorks AutoMap 2003 for FrameMaker 8.0.2.1385 and outputs XHTML 1.0
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
@link http://www.w3.org/TR/html5/obsolete.html
Parsoid offers an API with basically two actions: POST and GET You can test the API at http://parsoid-lb.eqiad.wikimedia.org/_html/
You can also test it locally on the vm through port 8000
We're converting Mentor Graphics Documentation System
- mgc_html v3.2_2.03
- infohub_core v3.2_2.02
Variables we care about[edit | edit source]
- We probably want a variable that can interact with the max upload size
- $wgMaxUploadSize[*] = 104857600 bytes (100 MB)
- $wgFileBlacklist we don't care about because we use our own file upload and mime detection
- $wgVisualEditorParsoidURL we can use for API requests to Parsoid
- $wgLegalTitleChars we use to check for valid file naming
- $wgMaxArticleSize default is 2048 KB, which may be too small?
- $wgMimeInfoFile we don't yet use
- Also, how do imagelimits come into play? http://localhost:8080/w/api.php?action=query&meta=siteinfo&format=txt
Features[edit | edit source]
Add a link to the sidebar for our extension. $wgUploadNavigationUrl is for overriding the regular 'upload' link (not what we want).
Instead, we have to edit MediaWiki:Common.js see https://www.mediawiki.org/wiki/Manual:Interface/Sidebar
Internationalization[edit | edit source]
http://localhost:8080/wiki/Special:Html2Wiki?uselang=qqx shows the interface messages You can see most of the messages in Special:AllMessages if you filter by the prefix 'Html2Wiki'
Error handling[edit | edit source]
- submitting the form with no file There was an error handling the file upload: No file sent.
- choosing a file that is too big (limit is set to 100 MB, so far haven't tested)
- choosing a file of the wrong type There was an error handling the file upload: Invalid file format.
- choosing a file that has completely broken HTML
Developing[edit | edit source]
The project code is hosted on both GitHub and WikiMedia Foundation servers on theHtml2Wiki Extension page. You should use git to clone the project and submit pull requests.
The best way to setup a full development environment is to use MediaWiki Vagrant. This handy bit of wizardry will create a full LAMP stack for you and package it into a VirtualBox container (among others).
