Rewriting Neo4j interface from py2neo to neomodel

BackofenLab / AxoWise

An integrated graph system of biological entities, functional terms, and publications.

Apache License 2.0

4 stars 0 forks source link

Rewriting Neo4j interface from py2neo to neomodel #4

Closed bornabesic closed 5 years ago

bornabesic commented 5 years ago

neomodel

Pros:

Get rid of hardcoded Cypher queries in the code
Change the schema more easily

Cons:

The whole cypher_queries.py module needs to be rewritten
Tests need to be rewritten

bornabesic commented 5 years ago

Writing with neomodel is slower right now. Maybe customized queries could help. Stack Overflow

bornabesic commented 5 years ago

Listing proteins with a custom query (just id, name and species_id fields) takes around 0.25 s. Getting all proteins and inflating them to their neomodel class takes a little bit more than 1 s.

I guess inflate function/operations is what takes most of the time.

bornabesic commented 5 years ago

In order to connect two nodes with RelationshipManager.connect(), (e.g. protein1.associations.connect(protein2)), both of them need to be fully inflated (i.e. contain all the information from the database).

bornabesic commented 5 years ago

Next step: Use neomodel in build_graph_db.py and compare the performance to the original implementation.