search

linkml / prefixmaps

Semantic prefix map registry
https://linkml.io/prefixmaps/
Apache License 2.0
11 stars 3 forks source link

Possible unnecessary error checking loading from included contexts #75

Open ptgolden opened 1 week ago

ptgolden commented 1 week ago

While investigating https://github.com/monarch-initiative/monarch-app/issues/819, I noticed a bottleneck in loading the "merged" context in these lines in curie:

https://github.com/biopragmatics/curies/blob/47d1f22e3aba42ea365cd7e99403eacfede4af5a/src/curies/api.py#L527-L533

Is it necessary to check for duplicate prefixes on known-good prefix maps (i.e. those bundled with this package)? If not, the converter loaded here:

https://github.com/linkml/prefixmaps/blob/c61e812f6de398daf388b996f6dbb691f93addac/src/prefixmaps/io/parser.py#L30-L31

can be initialized with strict=False, resulting in a speedup of about ~15s on the "merged" context. I am happy to create a PR if that sounds okay.

ptgolden commented 1 week ago

To be clear, strict=False would be added as a keyword to the .from_extended_prefix_map method here:

https://github.com/linkml/prefixmaps/blob/c61e812f6de398daf388b996f6dbb691f93addac/src/prefixmaps/datamodel/context.py#L379-L386