Doctable

NOTE! This is the last version used by hugedwarves.

Added cpickle and subsubdoc column types.

subsubdocs will store two levels of sub-doc organization. Perfect for doc->paragraph->sentence->token organization!

In this version, I'm releasing DocTable2. It was built using the sqlalchemy core functionality, and is more object-oriented than the first DocTable. Note that the DocTable class remains unchanged. I may update to support legacy systems in the future.

* Now will pull full schema as a dataframe and simplifies schema checking.

* Added separate "constraints" arg to constructor for passing commands like "UNIQUE(name,year)", which usually look like columns.

* Now constructor takes column names as lists (i.e. "name str"), for easier parsing.

* Fixed issue with maintaining connections. If you choose to create persistent connection, you may run "update" commands in the middle of a "get" loop. If you do not use persistent connection, you can work with the same file from multiple scripts at once.

This release allows for dynamic database connections. Instead of starting a database connection upon __init__, this version will give the option to create database connections every time a query is performed.

This was inspired by an attempt to steam sentences into gensim word2vec. Because of the way it parallelized processes, it needed to make a new connection for each query.