benchmark suite for lix

[samuelstroschein]

Context

We want lix to handle "larger" apps. SeeMESDK-137 for example.

Proposal

Implement a benchmark suite that is called in the test script to:

• surface performance regressions
• surface optimization potential

Metrics

• initial load less than 1 second
  • "1.0 second is about the time limit for the user's flow of thought to stay uninterrupted"
• in-app get queries of change, file, take less than 80ms
  • 100ms is the threshold for "perceiving lag"
  • 80ms to have 20ms buffer for app state management, etc.
  • Only get operations because create update can be optimized with optimistic UI

[samuelstroschein]

We need to add a test for changes too. Prob saying something along the lines of each entity has 10 (?) changes to test scalability of lix.

250k variants * 10 changes each = 2.5 million changes

[samuelstroschein]

if we save changes in lix directly, avoiding diffing, no performance issues should arise because it's… just sqlite. The differ is the expensive operation.

the inlang sdk can write directly to the lix database. No need to diff :)

[samuelstroschein]

I changed the title of this issue to benchmark suite to:

• ensure that we have recurring benchmarks we test against
• de-couple lix from inlang

[samuelstroschein]

@martin.lysk1 did you/will you merge the bench testing you've implemented?

[martin-lysk]

we cherrypick the getLeafChange (recursive) and its benchmarks - as examples