Optimizing Stitch Schema Transforms

[RemLampa]

Hello everyone!

We are trying to implement schema stitching on a couple of remote schemas.

One remote schema is quite massive. Introspection result alone is 11MB. We apply a few transforms on it to fit our needs, as the schema is autogenerated (Drupal).

Problem is the whole thing results in a startup time of about 20-30secs, which we found about 90% is due to stitchSchemas. Not to mention the huge spike in processing power.

We're planning on just combining the transforms into a single massive one since we're thinking each transform = one schema traversal, many transforms = many schema traversals. Are we right in taking this approach?

Also curious about how others tackle similar situations.

Thanks, everyone!

[Urigo]

@RemLampa sorry for the late reply.
Do you think you could create a sample with fake schemas that would help us benchmark ?