Update CLI to match design

[hmm34]

We want to modify the current behavior of the command line to support training and classifying of specific directories or paths in order to match our design given in the progress presentation .

Possible usages include:

• karat-sleuth train [data-set|path] [ham|spam]
  • If a data-set name is passed (easy, hard, unknown) then the messages will be loaded from the default training directory
  • If a file / directory path is specified it will learn from all messages specified in the path
• karat-sleuth classify [data-set|path]
  • Can re-classify an entire data-set
  • Can classify one or more messages from a specified path

For example,

• karat-sleuth train easy
  • This will train the classifier using the easy data set located in the default data directory
• karat-sleuth classify *.eml
  • Will classify all messages in the current working directory. Outputting a list of file names and spam/ham flags

[evanpurkhiser]

I'm thinking we'll add a 3rd option called reclassify which accepts parameters the same way that train does. This will be different than classify in that it will do some fancy statistic-y things to the classifications (how many were correctly classified vs how many weren't, how many it got right out of the hard set, stuff like that). Then we can do that in the demo and it will be super impressive.

[evanpurkhiser]

Most of the hard parts of this have been done in 11c7daf.

[evanpurkhiser]

I'm going to go ahead and close this since the basic CLI parsing is now there, and the previous 'train' functionality from the prior CLI implementation is good now since 297d7b1975dd2a34e4ecb534abaa78b2ce868d6b.