fix: baseline stars_count first bucket against startDate, not 0 (CM-1819)

[themarolt]

Summary

Fixes the Stars widget's "New" tab showing an inflated spike in the first bucket of any date range - stars_count.pipe now baselines the first bucket against the real cumulative total as of startDate instead of 0.

Changes

• Added stars_count_baseline node: computes cumulative stars total as of startDate (mirrors stars_count_summary's existing github_at_start/gitlab_new logic).
• stars_count_bucketed: pass the baseline as lagInFrame's own default argument instead of via coalesce() after the call - ClickHouse's lagInFrame already returns 0 (not NULL) for a missing prior row, so the previous coalesce(lagInFrame(...), baseline) pattern never triggered the fallback.
• Wrapped the baseline scalar subquery in coalesce((SELECT baseline FROM stars_count_baseline), 0) to avoid a ClickHouse type mismatch (Nullable(Int64) vs Int64) when passed as lagInFrame's default.

Verified live on production (lfx_insights) against two independent projects (k8s, heroic-games-launcher) - first bucket now matches expected deltas instead of the pre-range cumulative total.

Type of change

• Bug fix

JIRA ticket

https://linuxfoundation.atlassian.net/browse/CM-1819

[cursor]

PR Summary

Low Risk
Tinybird analytics-only change to bucketed star deltas; no app or API contract changes beyond corrected chart values.

Overview
Fixes CM-1819: bucketed “new stars” in stars_count.pipe no longer treats the first period as if cumulative stars started at zero, which inflated the first bucket with all pre-range history (e.g. Stars widget “New” tab spike).

Adds stars_count_baseline, a scalar cumulative total as of startDate (GitHub snapshot totals + GitLab star events through startDate, aligned with stars_count_summary). stars_count_bucketed subtracts each bucket’s cumulative total from the prior bucket using lagInFrame, with that baseline as the default for the first row—replacing coalesce(lagInFrame(...), 0), which never applied because ClickHouse returns 0 rather than NULL when there is no prior row. Pipe docs now describe startDate baselining and note that incomplete snapshot backfill can still mis-attribute pre-range stars for affected repos.

^{Reviewed by Cursor Bugbot for commit fbbff7f. Bugbot is set up for automated code reviews on this repo. Configure here.}

[Copilot]

Copilot review overview

🟡 Changes recommended

The documentation incorrectly claims the fix works regardless of snapshot backfill completeness.

Review effort: Balanced
Findings: 1

Open (1)

• Limit new-count guarantee to repositories with pre-start snapshots · New

What changed in this PR

Fixes inflated first-bucket star deltas by baselining against cumulative stars at startDate.

Changes:

• Adds a GitHub and non-GitHub baseline node.
• Uses the baseline as lagInFrame’s default value.
• Updates pipe documentation.

File	Description
services/​libs/​tinybird/​pipes/​stars_count.pipe	Corrects first-bucket star calculations and documents the behavior.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Copilot review overview

🟢 Approval recommended

The baseline logic matches existing summary semantics, and the previous documentation concern is addressed.

Review effort: Balanced
Findings: None

Resolved since last review (1)

• Limit new-count guarantee to repositories with pre-start snapshots