This is a basic set of python3ools to analyze chatlogs and manipulate that 
data. As I learn more about the Natural Language ToolKit (NLTK) I will add more
of that functionality. At the moment most functionality is statistical time 
series analysis. The tools use SQLite databases for persistence and to avoid
expensive parsing when performing multiple operations on the same dataset.

Getting started:

First you will want to create a master database containing all messages from 
the sources you want to analyze. There are a few tools to do this included 
with the prefix parse_* that are self documenting. The first argument should
be a database to append to (or create) and the remaining depend on the program.

For example, after downloading your data from Google Hangouts from Google 
Takeout https://myaccount.google.com/privacy#takeout the archive can be used
to create a database as follows:

  ./parse_hangouts gchat.db takeout-20151210T032131Z.tgz

Analysis:

The basic functionality starts with iterating over or counting sqlite 
queries against a relational database containing the fields:

  (timestamp,src,dest,msg)

The exact meanings of src,dest,msg varry from source to source, but the
timestamp is UNIX UTC seconds since Jan 1, 1970. The database class
provides a function to iterate over querried results returned as messages:

  [msg for msg in db.get_iter('src LIKE ?','BenLand100%')]

This will efficiently build a list of messages where the source satisfies the
optional SQLite constraint. The function count will return the number of rows
if further inspection is not required.

A few simple programs with simple output are included.
 * test.py - prints statistics about a database with an optional src pattern
 * crunch.py - dumps a CSV timeseries of message lengths to stdout
 * wordprofile.py - dumps word statistics as CSV to stdout

There is additionally markov.py which will train a simple ngram markov chain
and generate random text based on a seed.