Skip to content

feat(website): final benchmark run and per-library code on the benchmark page#278

Merged
zantvoort merged 2 commits into
mainfrom
feat/benchmarks-final-run
Jul 16, 2026
Merged

feat(website): final benchmark run and per-library code on the benchmark page#278
zantvoort merged 2 commits into
mainfrom
feat/benchmarks-final-run

Conversation

@zantvoort

Copy link
Copy Markdown
Collaborator

Summary

Updates the benchmark page and homepage to the final re-run, and makes every workload (and the entity model) show all seven implementations rather than Storm alone.

Numbers

  • All eight workloads refreshed to the final run (measured 2026-07-16); round-trip baseline ~155 µs.
  • Homepage per-library speed figures and the three highlight cards updated; matrix summary reworded for the new data (Exposed no longer edges Storm on the 1,000-row join).

Per-library code and SQL

  • Every workload and the model now expose Storm, JDBC, Hibernate, jOOQ, Exposed, Exposed DAO and Jimmer behind the variant selector.
  • editor() gains per-variant sql, tag and file, so Show SQL and the tab chrome (language + filename) switch with the selected library. Backward compatible with single-SQL editors.
  • Show SQL shows the exact statement each library issues, including the divergences: jOOQ's multi-row insert and MULTISET, Hibernate's sequence pre-allocation and DISTINCT join fetch, Exposed DAO and Jimmer's batched secondary queries.
  • The entity model shows City, Owner and Pet in full for every library. Code is taken from the storm-benchmarks suite.

UI

  • Clone-bar copy button flips to a green check on click.
  • Homepage "vs" cards always show a comparison (no deselect).

…he cards

Update the benchmark page and homepage to the final 2026-07-16 re-run:
all eight workloads, the ~155 us round trip, and the measured date.

- Benchmarks: rewrite the three highlight cards (workloads won, worst-case
  gap, overhead vs the next-fastest ORM) and refresh the matrix summary for
  the new numbers (Exposed no longer edges Storm on the 1,000-row join).
- Homepage: the vs cards always show a comparison (no deselect, feature
  fronts removed) and carry the updated per-ORM speed figures.
- Clone-bar copy button flips to a green check on click.
Extend the benchmark page so each workload and the model expose all seven
implementations behind the variant selector, not just Storm.

- editor() gains per-variant sql, tag and filename, so Show SQL and the tab
  chrome switch with the selected library (backward compatible).
- Each workload carries the real code from the benchmark suite, and the
  Show SQL panel shows the exact statement each library issues: jOOQ's
  multi-row insert and MULTISET, Hibernate's sequence pre-allocation and
  DISTINCT join fetch, Exposed DAO and Jimmer's batched secondary queries.
- The model shows City, Owner and Pet in full for every library.
- Copy: reworked matrix summary and stat cards; "On average across 8
  workloads" on the homepage.
@zantvoort
zantvoort merged commit 60e1700 into main Jul 16, 2026
6 checks passed
@zantvoort
zantvoort deleted the feat/benchmarks-final-run branch July 16, 2026 23:08
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant