dutradda / myreco

92%

master: 93%

push

travis-ci

Moved from numpy to cython:
 - Removed the IndicesMap tests from travis;
 - Changed the way how the filters are applied. Now is used cython with redis sets instead numpy array multiplications;
 - Created a class to manage the filters called FiltersSet;
 - Changed the top seller recommendations to use cython;
 - Removed the IndicesMap;
 - Changed the MyrecoAPI to receive the cython redis client;
 - Added a the method _del_old_keys to EngineObjectBase class;
 - Removed the filters by indices;
 - Rewrited the filters to use redis sets instead numpy vectors;
 - Changed the EngineStrategyBase to use heapq instead numpy.argpartition for sorting;
 - Moved the filter applying from EngineStrategyBase to be done inside the specific engine strategy (see recommender.pyx);
 - Changed the class TopSellerArray to TopSellerMap; now it uses a redis hash instead put a numpy vector on redis;
 - Changed the placements get_items to build FiltersSets instead a dict of filters;
 - Added credis to requirements and removed numpy/scipy;
 - Added the cython modules compilation on setup;
 - Fixed the tests.

386 of 491 branches covered (78.62%)

1506 of 1631 relevant lines covered (92.34%)

0.92 hits per line