dev_instructions.txt

Welcome as a developer of DLATK! 

DLATK is an open source package. Our typical development is motivated by rapid prototyping 
of new techniques. However, there are several core pieces of DLATK that need to be reliable: 
running DLA on words, topic extraction, and running ridge regression, logistic regression 
and/or extra tree classifiers. We rely on developers making clear commit statements documenting 
their process, as well as contributing to the overall documentation for more significant 
updates. We try to balance not getting bogged down with protocols with also trying to make 
it easy to update and insure reliability of our core functionality. 


Protocols for editing:
======================
1. Create a new branch for your edit: git checkout -b new_branch_name

2. Commit often (i.e. every 30 minutes): git commit -am "ADDED FEATURE: double tokenizer"; git push origin new_branch_name

3. Make sure to include a commit message that breifly describes the update:
 -- if it's a bug fix, prepend message with "FIXED BUG:"
 -- if it's a feature addition, prepend message with "ADDED FEATURE:"
 These can be short but to the point, e.g. "BUG FIX: regressionPredictor working with sklearn v0.22)

4. Add to the package documentation for new options that have been added (i.e. anytime there is a new -- option). 

5. Send a pull request to the public branch when ready for main-stream. 
 -- first make sure to push to github: git push origin new_branch_name
 -- go to https://github.com/dlatk/dlatk, switch to your branch, and click "Compare & pull request" -> "Create pull request"
 

A majority of this package is maintained at the level of prototyping new techniques; The "Suggested Unit Tests" below are the pieces we plan to maintain consistency. With current resources, it should not be considered on a "mature" update cycle. 
The process that the maintainers take to update public (and pip..etc...) consists of:
   1. Running unit tests that cover word DLA, topic DLA, and regression and classification
   2. A look through of the commit log to pull out key peices for change log
   3. An increment of the version number
   4. Make sure setup.py runs
   5. Make sure dlatk webserver get updated with source

Suggested Unit Tests (using age and occupation as outcomes):
--ngram extraction, feat_combine, p_occ
--topic extraction
--rmatrix --print_csv --tagclouds (ngrams)  --cat_to_bin occupation)
--topic_tagclouds --topic_lex
--nfold_test_regression --model ridgecv
--nfold_test_classif --model lr --cat_to_bin occupation