Profile ORM queries against bounded table windows (#52) #64

Merged
michielbdejong merged 4 commits from codex/issue52-orm-query-profile into main 2026-10-28 13:05:11 +01:00

Direct ORM paging is not a drop-in replacement for the current backend-query plus bounded live window: query semantics differ, and direct ordering blocks the renderer on a large graph. Add a reproducible diagnostic, API/semantics audit and raw measurements without changing production query behavior.

Updated against current main, resolving the package script conflict after the experimental-app move. Add pnpm typecheck:orm-query for repeatable helper validation. Preserve the original 21 October evidence and add fresh 28 October cloud samples with their host/measurement limitations.

Current validation:

  • Seven paging/window tests pass.
  • Profiling helper TypeScript check passes.
  • Native Chromium diagnostics at 1,000 and 100,000 rows pass: twelve samples at each size verify IDs, ordering and retained row counts.
  • Fresh 100k medians: direct ORM readiness about 26.7s ascending / 35.4s descending; existing backend plus window about 1.7s / 1.9s. These are local diagnostics, not production latency guarantees; do not compare hosts as a controlled regression study.
  • git diff --check passes.

Ready for review as the completed profiling investigation. Related to #52; deliberately leaves that issue open for query latency, membership/search costs, renderer blocking and representative performance acceptance evidence. This PR does not claim to solve those performance requirements.

Direct ORM paging is not a drop-in replacement for the current backend-query plus bounded live window: query semantics differ, and direct ordering blocks the renderer on a large graph. Add a reproducible diagnostic, API/semantics audit and raw measurements without changing production query behavior. Updated against current main, resolving the package script conflict after the experimental-app move. Add `pnpm typecheck:orm-query` for repeatable helper validation. Preserve the original 21 October evidence and add fresh 28 October cloud samples with their host/measurement limitations. Current validation: - Seven paging/window tests pass. - Profiling helper TypeScript check passes. - Native Chromium diagnostics at 1,000 and 100,000 rows pass: twelve samples at each size verify IDs, ordering and retained row counts. - Fresh 100k medians: direct ORM readiness about 26.7s ascending / 35.4s descending; existing backend plus window about 1.7s / 1.9s. These are local diagnostics, not production latency guarantees; do not compare hosts as a controlled regression study. - git diff --check passes. Ready for review as the completed profiling investigation. Related to #52; deliberately leaves that issue open for query latency, membership/search costs, renderer blocking and representative performance acceptance evidence. This PR does not claim to solve those performance requirements.
michielbdejong changed title from WIP: Profile ORM queries against bounded table windows (#52) to Profile ORM queries against bounded table windows (#52) 2026-10-28 12:20:56 +01:00
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
ELFA/database-apps!64
No description provided.