benchmarks-website/server/src/api/charts.rs::collect_group_charts re-runs the entire collect_groups discovery pass per call before fetching each chart, which makes the landing-page render O(groups * charts_per_group) DB queries plus the discovery scan.
Fine for the current dataset; revisit when chart counts grow or the cold-page render starts feeling slow.
Sketch of the fix:
- Walk the fact tables once per page render.
- Materialize
(group_key, chart_links, chart_payloads) in a single pass keyed by group_key, so the chart-card collection no longer needs the discovery pass.
- Or: take a
Vec<ChartLink> directly from the caller (the landing-page handler already has them) and skip the discovery rerun.
This is a pure performance refactor; the wire format stays identical.
See the doc comment on collect_group_charts in
benchmarks-website/server/src/api/charts.rs for the existing breadcrumb.
benchmarks-website/server/src/api/charts.rs::collect_group_chartsre-runs the entirecollect_groupsdiscovery pass per call before fetching each chart, which makes the landing-page renderO(groups * charts_per_group)DB queries plus the discovery scan.Fine for the current dataset; revisit when chart counts grow or the cold-page render starts feeling slow.
Sketch of the fix:
(group_key, chart_links, chart_payloads)in a single pass keyed bygroup_key, so the chart-card collection no longer needs the discovery pass.Vec<ChartLink>directly from the caller (the landing-page handler already has them) and skip the discovery rerun.This is a pure performance refactor; the wire format stays identical.
See the doc comment on
collect_group_chartsinbenchmarks-website/server/src/api/charts.rsfor the existing breadcrumb.