kuhumcst / cuphic Goto Github PK

So far, most of the functionality has been focused on finding and replacing (transforming), but finding should also be available stand-alone (needed for e.g. finding metadata in TEI-XML documents produced by Dorte).

The crucial thing here is to make it a single-pass thing (just like transformations) to preserve performance. It makes sense to make the API take 1 or more patterns, possibly using named args and then simply return a map from keyword->list of each found pattern (plus metadata most likely) within the hiccup.

Subtle errors such as the one fixed in a9d6a59 keep popping up way too often. Fixing the errors are hard since the code has become fairly complex.

I think this complexity that came about as a result of having to support fragment patterns. Suddenly, the algorithm went from being very simple to grasp, to being extremely complex with multiple routes to follow. However, fragments are quite essential to the workings of the algorithm.

Perhaps I need to look at incorporating more third party libs for doing the matching. Core.unify? Or perhaps one of the other data transformation libs as a base?