User:Datagrok/A simple wiki

A little CGI wiki powered mostly by Apache .htaccess directives

Introduction

Yet another wiki/cms thingy. It's small, fast, and simple. If you're a programmer, you can hack on it in your favorite CGI-capable language to add whatever features you like while still keeping the core "let apache do the work" design.

I built this to achieve this combination of features, which I could not find elsewhere:

• Optimized for speed under many reads, few writes.
  • All wiki pages are saved as static files and served by apache.
  • The only time a script is run is when a page is edited.

Thus, the only time a database (if any!) is queried (if at all!) is when a page is edited.

• Nice, clean URLs with multi-level heiarchy.
• Wiki pages may exist along side any other content, including HTML authored by other tools, or CGI scripts.
• No dependence on a database.
• No dependence on mod_rewrite.
• Apache does most of the work.
  • Apache serves all the flat files.
  • Apache provides (somewhat weak) authentication via mod_auth or mod_auth_db.
  • Options +MultiViews in .htaccess gives nice clean URLs.

Core idea

• Configure the 404 error page for the wiki-ized folder to point instead to a "create new page" script.

Thus, links into the wiki will result in a 404 which has a "create this page now" form at the bottom, or a redirect to an editor script.

That's enough for wiki-like functionality. Other features are orthogonal to this design. For my purposes, I'll also add:

• Multiple content types for modular, specialized editors. (Wiki pages, images, etc.)
• History tracked with standard version control tools.
  • I like Subversion because I already use it a lot on Dreamhost.
  • I might use Mercurial because I think I can "flatten" parts of the history, to completely eliminate any overhead caused by spam.
  • I hope to make this modular, so the user may use the VC they like.
• Smart markup using my own twist on Python's reStructuredText/Markdown

But what about...

• Common look-and-feel: Editor script wraps content in template or shtml includes.
  • Wrapping in a template would require re-rendering the entire site in the case of template change.
  • SSI incurs a small CPU/disk performance hit.
• Dynamic site map: Sitemap data written in JSON to base directory, HTML generated by javascript.
  • Care must be taken to make the site still navigable for Google and non-javascript browsers.

Details

/wiki/.htaccess contains:

Options +MultiViews
ErrorDocument 404 /wiki/edit/

/wiki/edit/.htaccess contains:

DirectoryIndex editor.py
# This is for password-protecting it:
AuthUserFile /full/path/to/htpasswd
AuthName "web editor"
AuthType Basic
<Limit POST>
require valid-user
</Limit>

/wiki/edit/editor.py performs a few tricks with CGI variables to figure out which file to edit:

import os
class Editor(object):
    def __init__(self):
        e = os.environ.get
        self.user = e('REMOTE_USER')
        self.host = e('HTTP_HOST')
        self.root = e('DOCUMENT_ROOT')
        edit_dir, dummy = os.path.split(e('SCRIPT_NAME'))
        parent, dummy = os.path.split(edit_dir[1:])
        path = None
        for prefix in [e('SCRIPT_NAME')+'/', e('REDIRECT_URL'), '/'+parent+'/']:
            if e('SCRIPT_URL').startswith(prefix):
                path = e('SCRIPT_URL')[len(prefix):]
                break
        path, file = os.path.split(path)
        if not file:
            file = "contents"
        self.parent = parent
        self.path = path
        self.file = file
e = Editor()
print 'Content-type:text/plain'
print
print repr((e.root, e.parent, e.path, e.file))

I'm still building this for my purposes, but that should be a good start...