wikiundump saves all Wikipedia pages from a Wikipedia XML dump to individual files.

Download a Wikipedia dump from https://dumps.wikimedia.org/backup-index.html

$ wikiundump -dir ./simple -keep ,Template simplewiki-*pages-articles-multistream.xml

or:

$ bzcat simplewiki-*.xml.bz2 | wikiundump -dir ./simple -keep ,Template

By default all pages from all namespaces are saved. -keep is used to store only pages from the specified namespaces (comma separated). Note that the main namespace is an empty string, so if you want to keep the main namespace only use wikiundump -keep "," :(

Files are stored with 3 levels of subdirectories, with the namespace as an extra leading subdirectory if there is any. The directory names are the first three letters of the page title, lowercased, with each non ASCII alphanum replaced with _. If the name of the page is shorter than 3 characters it's padded with _s to length 3. The title of the page will have the first letter uppercased, according to the namespace's Case (we support only "first-letter").

e.g.:

Accordion                   -> /a/c/c/Accordion
a cappella                  -> /a/_/c/A cappella
Açaí Palm                   -> /a/_/a/Açaí Palm
-1                          -> /_/1/_/-1
A∴A∴                        -> /a/_/a/A∴A∴
Template:Abbreviations      -> /Template/a/b/b/Template:Abbreviations
Template:AA                 -> /Template/a/a/_/Template:AA
Templater:AA                -> /t/e/m/Templater:AA

The list of namespaces is saved in .../namespaces.json. Later, when finding files, it is needed to know what pages have namespace prefixes and what pages happen to have a ':' in their name. You also need to know the case rules for each namespace.

Pages which redirect to another page are symlinked. Use wikiundump -symlink false if that's not wanted (redirected pages will be ignored).

You need a case sensitive filesystem, and it must not choke on Unicode filenames.