strazto/contextractr

contextractr's Issues

Feature Parity with CLACK - Allow suffix and prefix to specified

TODO (for this description):
Further extrapolate on this - List CLACK interface fields that still need to be implemented

Consider changing groups to factors in processing

Enhancement - Use optional per-entry function handler in order to define lookahead/lookback

This should be part of the suffix/prefix implementation.

The goal would be to allow users to supply a particular function handler for a given entry, such that users can customize the amount of characters/mechanism for lookahead and look back

eg, a lookahead function that says "count 4 instances of whitespace, and then take the string until that point", or a lookback function that says "look for anything until 'Past Medical History'",
etc

(Decide on the inputs and outputs later (Whether it takes strings, indices, etc, and what it returns)
(Indices might be a good choice for input since multiple occurences of the same keyword can be present)

Also need to decide how to serialize this - Probably best to use a scoped function name, or perhaps could allow for parsing of a function

Documentation: Fix Documentation, add @usage field for all docstrings

Need at least one value for approx.match set for fuzzy map to work, and set correct default values

approx.match should be full optional

Allow conversion from unnested csv to serialized.

Description

Presently, contextractr accepts serialized formats like json and yaml, as well as dataframes that are presented with the correct list-cols.

Allow an unnested dataframe to be given, s.t:

"title" has a 1:many relationship with "keywords" + "approx.match"
and "keywords" has 1:1 with "approx.match"

ie, a single val for title may span 1 or more rows, and for each of these, a unique value of "keyword" and corresponding (or missing) spec for approx is given.