arskom / spynepi Goto Github PK

We should adopt ConfigParser for configuration files.

While it's not that much of a problem, we could just switch to a lightweight WSGI server if we're not going to use twisted features; dropping a big dependency along the way.

There has to be a background thread subscribed to pypi's twitter feed that updates cached packages as they are released.

... whatsoever. Must add it!

Now that we have package caching enabled, proper packages with big description tests are not uncommon in a local repository. So we need to convert these to proper html.

    @rpc(Unicode, Unicode, Unicode, Unicode, File, Unicode, Unicode, Unicode,
         Unicode, Unicode, Unicode, Unicode, Unicode, Unicode, Unicode,
         Unicode, Unicode, String, _patterns=[HttpPattern("/",verb="POST")])
    def register(ctx, name, license, author, home_page, content, comment,
            download_url, platform, description, metadata_version, author_email,
            md5_digest, filetype, pyversion, summary, version, protcol_version):

The type declarations in the above block must be revised with max_len and pattern constraints.

Regexes and other validation code could be stolen from: https://bitbucket.org/loewis/pypi/src/7ffa1f091b87dd58eb3770abc6af30ec37943ba1/verlib.py?at=default

One could define, e.g.:

PackageName = Unicode(64, pattern='[a-zA-Z_][0-9A-Za-z_]*')

And use it in database declarations, rpc decorators and ComplexModel subclasses.

We need tests! ASAP!

When a not-yet-cached package is requested twice, two download sessions start. Thread-level locking should be used to prevent this.

as it does not daemonize.