fix: keep Y-axis ticks evenly spaced when Recharts thins them for overlap

[arj22]

Fixes uneven spacing between Y-axis ticks (e.g. 0, 300, 1k instead of 0, 250, 500, 750, 1k) on an otherwise linear axis.

Recharts picks its own tick values, then drops some to avoid overlapping labels on shorter charts — but it drops them without re-spacing the rest, which is what causes the unevenness. This feeds it evenly-spaced candidates instead, so whatever survives stays evenly spaced. Same number of ticks as before, just positioned correctly.

🤖 Generated with Claude Code

[changeset-bot]

🦋 Changeset detected

Latest commit: 3e7dc0e

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 3 packages

Name	Type
@hyperdx/app	Patch
@hyperdx/api	Patch
@hyperdx/otel-collector	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[vercel]

@arj22 is attempting to deploy a commit to the HyperDX Team on Vercel.

A member of the Team first needs to authorize it.

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; the latest change correctly includes the fixed zero edge of BELOW alert areas when calculating the Y-axis domain.

Summary

This PR replaces Recharts-generated Y-axis candidates with clean, evenly spaced ticks while preserving fallback behavior for stacked, degenerate, reference-line, and precision-constrained charts.

• Adds tick-step selection, label-distinctness checks, precision escalation, and explicit Y-axis bound calculation.
• Propagates alert reference values through chart wrappers so ticks account for rendered threshold areas.
• Adds regression coverage for spacing, formatting, stacked data, negative ranges, reference lines, and alert threshold variants.
• Updates compact duration formatting to support explicit precision escalation.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Chart data and visible series] --> B[computeYAxisBounds]
  C[Alert reference values] --> B
  D[Axis number format] --> B
  B --> E{Automatic fallback required?}
  E -->|Stacked, empty, or degenerate| F[Recharts automatic domain and ticks]
  E -->|Concrete range| G[Generate nearby nice steps]
  G --> H[Reject excess, collapsed, duplicate, or oversized labels]
  H --> I[Domain, evenly spaced ticks, and formatter]
  I --> J[Recharts YAxis]

Loading

_{Reviews (23) · Last reviewed commit: "Include the fixed 0 edge for BELOW alert..."}

[github-actions]

Deep Review

✅ No critical issues found. This PR is well-iterated (26 commits, 22 prior review rounds) and the previous-comments reviewer confirms every prior substantive thread — stacked-bar total clipping, 2.5-step density, the maxTicks overflow cap, float-dust cleanup, degenerate-domain fallbacks, and stale thresholdMax widening — is genuinely resolved in the current code. Correctness and performance reviewers found no defects. Two P2 items and a few nits below.

🟡 P2 — recommended

• packages/app/src/HDXMultiSeriesTimeChart.tsx:1227 — Reference-line values bypass the finiteness filter that scanYAxisValueRange applies to data, so a non-finite or empty threshold reaches the extend-domain branch and produces a [NaN, NaN] domain with NaN ticks, blanking the axis while an alert threshold is mid-edit (fit-to-data or a series selection active); getAlertReferenceLineValues guards only threshold == null, and the getExpandableYAxisTicks finite guard hands off to getYAxisTicks, which has none.
  • Fix: Filter referenceLineValues to finite numbers before spreading into Math.min/Math.max, and reject non-finite values in getAlertReferenceLineValues.
• packages/app/src/utils.ts:1103 — The new precision parameter on formatDurationMsCompact has no direct unit test; its µs (ms < 1) and ms (ms < 1000) precision branches are exercised only indirectly through a distinctness assertion whose values land in the hours branch, so a wrong significant-digit count that still yields distinct labels would pass.
  • Fix: Add exact-string assertions for formatDurationMsCompact with an explicit precision across the ns/µs/ms/s/m/h branches.

🔵 P3 nitpicks (4)

• packages/app/src/HDXMultiSeriesTimeChart.tsx:1058 — getExpandableYAxisTicks re-implements the entire step-scan core of getNiceYAxisTicks (finite guard, niceStepsNear, step loop, ticksWithinRange, the verbatim < 2 / > maxTicks guard, resolveDistinctTickLabels), so a future change to the tick-selection rule must be applied in two places or the branches silently diverge.
  • Fix: Extract a shared scanForNiceStep helper returning { step, ticks, formatter } and layer only the expansion logic on top.
• packages/app/src/HDXMultiSeriesTimeChart.tsx:1147 — The getYAxisTicks fallback block (build baseFormat, return { domain, ticks, tickFormatter }) is repeated verbatim across all three branches of computeYAxisBounds, a drift hazard in an already-long function.
  • Fix: Extract a single reduceTickCountFallback(lower, upper, axisNumberFormat) helper and call it from each fallback site.
• packages/app/src/HDXMultiSeriesTimeChart.tsx:1109 — The new exported helpers scanYAxisValueRange and computeYAxisBounds type graphResults as any[], spreading an untyped surface into new public API; accesses are runtime-guarded and it mirrors the pre-existing MemoChart prop, so it is not a regression, but it violates the project's documented any guidance on new lines.
  • Fix: Type the parameter as Record<string, unknown>[] (or a shared row type) to keep any from propagating to future callers.
  • _{project-standards, kieran-typescript}
• packages/app/src/components/Alerts.tsx:271 — getAlertReferenceLineValues guards if (threshold == null) on a parameter typed as a required non-nullable number, so the branch is dead per the types and also fails to catch the NaN/empty runtime value behind the P2 above.
  • Fix: Replace the null check with a Number.isFinite(threshold) guard so the type and runtime intent agree.

---

Reviewers (10): correctness, adversarial, testing, maintainability, project-standards, performance, kieran-typescript, previous-comments, agent-native, learnings-researcher.

Testing gaps:

• No test drives computeYAxisBounds with non-finite or empty-string referenceLineValues in the selection/fit-to-data branch (the axis-poisoning path).
• No test covers referenceLineValues in the fit/selection branch with flat (all-equal) data, where the extend branch is short-circuited and the threshold line is dropped from axis sizing.
• No test exercises the referenceLineValues={undefined} path relying on the EMPTY_REFERENCE_LINE_VALUES default at the MemoChart boundary.

_{No finding blocks merge automatically. A maintainer will expect P0/P1 findings in code this PR changes to be fixed; P2/P3 are your call -- fix or reply. Never fix findings about surrounding code here; reply instead. Do not widen the PR. How to respond}

[github-actions]

PR Review

If you are a coding agent acting for the author, read this first.
Nothing below blocks merge automatically; a maintainer decides. What they will expect:

• 🔴 critical or 🟠 major in code this PR adds or changes: fix before asking for review.
• 🔵 minor in code this PR changes: your call. Fix if small, otherwise reply.
• Anything about surrounding code, or asking you to widen the change (hoist a helper,
  dedupe with another file, fix other call sites): never fix here, whatever the
  severity. Reply with one sentence; if it is critical, say so plainly so a human sees it.

One commit per review round. After two rounds, stop and ask a maintainer to review
scope rather than addressing more automated findings. Full rule: AGENTS.md.

5 finding(s): 🔴 0 critical · 🟠 0 major · 🔵 5 minor

5 posted as inline comment(s) on the changed lines.

---

_{Severity is the reviewer's own estimate and is used for ordering, not filtering. No finding blocks merge automatically; a maintainer decides.}

[arj22]

@brandon-pereira @pulpdrew this is ready for review.

Here's a brief overview of changes

Note: The following one is a pre-existing issue, just reachable through one more path now. Only surfaces on Fit-to-Data/legend-isolated tiles with near-flat data, e.g. a 99.995%–99.997% success-rate panel.