[alex-fedotyev]

Summary

Closes #1825

• Field classification functions in deltaChartUtils.ts:
  • stripTypeWrappers() — removes LowCardinality(...) and Nullable(...) wrappers from ClickHouse type strings
  • isIdField() — detects structural ID fields (top-level String or Array(String) ending in "Id"/"ID")
  • isTimestampArrayField() — detects per-index timestamp array elements from Array(DateTime64(...)) columns
  • isDenylisted() — combined: ID fields + timestamp arrays
  • isHighCardinality() — detects fields where min(outlierUniqueness, inlierUniqueness) > 0.9 with combined sample > 20
• getPropertyStatistics now also returns propertyOccurences and valueOccurences (needed for cardinality detection)
• Two-section layout in DBDeltaChart.tsx:
  • Properties split into visibleProperties (primary) and hiddenProperties (denylisted + high-cardinality)
  • Rendered in two separate CSS grids with a "Lower-priority fields (N)" divider
  • Row-based pagination: hidden fields always start from column 1 (not filling incomplete visible rows)
• Tests: 29 unit tests covering all classification functions + integration test

Test plan

• TraceId, SpanId, ParentSpanId appear in "Lower-priority fields" section
• Events.Timestamp[N] elements appear in lower-priority section
• High-cardinality fields (>90% unique values, sample > 20) appear in lower-priority section
• Low-cardinality fields like http.method stay in primary section
• Hidden fields start from column 1 on their grid rows
• Pagination correctly accounts for both sections' row counts
• Small sample sizes (≤ 20 combined) keep fields visible regardless of uniqueness
• Run npx jest src/components/__tests__/deltaChartFieldClassification.test.ts — 29 tests pass

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
hyperdx-oss	Ready	Preview, Comment	Mar 3, 2026 11:47pm

[changeset-bot]

🦋 Changeset detected

Latest commit: 4e8b4d6

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

[github-actions]

E2E Test Results

✅ All tests passed • 77 passed • 4 skipped • 905s

Status	Count
✅ Passed	77
❌ Failed	0
⚠️ Flaky	1
⏭️ Skipped	4

Tests ran across 4 shards in parallel.

View full report →

[github-actions]

Claude finished @teeohhem's task in 1m 55s —— View job

---

PR Review: feat: field filtering and priority classification

✅ No critical issues found. The implementation is clean, well-tested, and follows project patterns.

A few minor observations:

• ⚠️ Code duplication in deltaChartUtils.ts — the same two-line pattern for extracting colName from a key (regex + ternary) is copy-pasted into both isIdField (lines 143–144) and isTimestampArrayField (lines 167–168). Consider extracting a getBaseColumnName(key) helper to reduce duplication. Fix this →

• ⚠️ Cardinality blind spot: isHighCardinality uses outlierRawValueOccurences (from valueOccurences), which only tracks fields that appear ≥ MIN_PROPERTY_OCCURENCES (5) times. If a field appears 4 times in the outlier group and 17 in the inlier group (combined 21 > threshold), the outlier group's value map will be absent, resulting in outlierUniqueness = 0, suppressing detection. This is a conservative false-negative (field stays visible) so not harmful, but worth noting.

• ℹ️ propertyOccurences / valueOccurences (one 'r') typo is consistent with the existing codebase — no change needed.

[github-actions]

Claude finished @alex-fedotyev's task in 30s —— View job

---

PR Review: feat: field filtering and priority classification

✅ No critical issues found. The previous review's code duplication concern (getBaseColumnName helper) has already been addressed in the latest commit.

Minor observations:

• ⚠️ isHighCardinality uses outlierRawValueOccurences (raw counts) but receives percentageOccurences for outlier/inlier value maps — Looking at DBDeltaChart.tsx:267-274, outlierRawValueOccurences and inlierRawValueOccurences are passed correctly. The naming in the useMemo shadowing (outlierValueOccurences for percentages vs outlierRawValueOccurences for raw) is a bit confusing at first read, but the logic is correct.

• ℹ️ Pagination edge case: When visibleProperties is empty and only hiddenProperties exist, showHiddenHeader will be true and showDivider false — the divider/label still renders correctly. Works as intended.

• ℹ️ getBaseColumnName regex (/^([^[]+)\[(\d+)\]$/) returns null for Events[0][1] but the test at line 44 confirms this is intentional. The "malformed" case is actually a nested array access, correctly excluded.

• ℹ️ The propertyOccurences/valueOccurences (one 'r') typo noted in a prior review remains consistent with the existing codebase — no change needed.

The implementation is clean, well-tested (29 unit tests), and the two-section layout logic with row-based pagination is correct.

[github-actions]

Claude finished @alex-fedotyev's task in 58s —— View job

---

PR Review: feat: field filtering and priority classification

• Read changed files
• Review code

---

✅ No critical issues found. The implementation is clean and well-tested.

One minor observation:

• ℹ️ isHighCardinality false-negative edge case (carried over from prior reviews, still applies): valueOccurences only tracks keys with ≥ MIN_PROPERTY_OCCURENCES (5) occurrences. If a field appears < 5 times in one group but ≥ 5 in the other, that group's unique value count will be 0, making Math.min(0, highCardinality) = 0, suppressing detection. This is a conservative false-negative (field stays visible) — no harm done, but worth a comment in the code if future maintainers hit it.

Everything else looks good: getBaseColumnName helper correctly avoids the earlier code duplication, row-based pagination math handles edge cases correctly (empty visible section, page boundaries), and the 29-test suite covers the classification functions well.

[github-actions]

Claude finished @alex-fedotyev's task in 1m 17s —— View job

---

• Read changed files
• Review code

---

✅ No critical issues found.

The getBaseColumnName refactor (from prior review) is in place. Pagination math is correct across edge cases (empty visible section, page boundaries, cross-section pages). Classification functions are sound.

One observation carried over from prior reviews that still applies:

• ℹ️ isHighCardinality conservative false-negative: valueOccurences only tracks keys with ≥ 5 occurrences (MIN_PROPERTY_OCCURENCES). A field with < 5 outlier observations will have outlierUniqueValues = 0, making outlierUniqueness = 0/total = 0 — Math.min(0, highInlierUniqueness) = 0, suppressing detection even if both groups are genuinely high-cardinality. Field stays visible (false-negative), so no harm done. Worth a one-line code comment if future maintainers hit it.

The two-section layout, row-based pagination, and 29-test suite are all solid.