xmltraintest.py for POPFile

Joseph's xmltraintest.py for POPFile 0.22

xmltraintest.py was designed as a replacement for the unsupported POPFile utility traintest. The original traintest has not been updated and no longer works with recent versions.

Its purpose is to simulate use of PF by running messages from your saved archive. It is mainly used for testing, but may be used to rebuild your corpus from scratch.

My version is written in Python and requires you to install POPFile's XMLRPC module.

Download - v0.03 beta (August 22, 2005)

The current version only works with POPFile 0.22 (and for now the 0.23 development version). If you want an older version for POPFile 0.21 you can still download 0.02. There are only minor changes between versions but the name of POPFile's API changed.

New in v0.03 beta (August 22, 2005):

This version has not been well tested yet. It may not work with the archive names created by 0.22 (I don't have one to test at the moment).
Option to write messages to POPFile's history. Reclassifications done by the script will not be reflected in history.
Add login and password options in preperation for 0.23.

New in v0.02 (March 23, 2004):

New testing only option. This allows your to see how your current corpus would do on your archive without making any changes to your corpus.
New detailed info option to display word counts and messages classified for each bucket, also indicates buckets with magnets.
Display how long (in CPU time) it took too complete.
Output errors.txt similar to original traintest.
Better handling of buckets that weren't in the archive. An existing bucket could still cause classification errors in other buckets if buckets aren't cleared.
Much improved output when redirecting to a file.

Usage: python xmltraintest.py [options] <archive dir>

Options:
  -a   Set training mode to Train Always (default is Train Only Errors)
  -c   Clear exising corpus if it exists
  -d   Display detailed information about buckets
  -r#  Display row accuracy every # rows (0 turns off row accuracy display)
  -t   Test archive against current corpus (no changes made to corpus)
  -u   Set user login (default is admin)
  -p   Set user password (default is blank)
  -h   Write archive messages to PF history (reclassifications will not show)
  -?   Display this help message

Warning: You should backup your corpus before using this program.
         It will modify or replace your existing corpus.

To start testing with a blank corpus, rename your current popfile.db, PF will automatically creates a new empty one. Then the script will create the buckets from what it finds in the archive directory.

If you want to start with your existing popfile.db you should copy it to another name also as a backup. There is an option to clear words from existing buckets (leaving magnets), otherwise it will just act like the original Traintest and add words to your existing bucket.

If you get tired of running hitting Ctrl-C will print out the stats up to that point and quit.