fix(preview): stop column flicker by carrying enriched columns into preview refetches

[janosmiko]

Fixes the latest comment on #408: "The sort order seems stable now but the columns are not - several surprise columns display in the preview and then after a moment disappear."

Root cause

Preview fetches return unenriched items — no CPU/MEM metrics columns (the drilled list merges those asynchronously via the pod/node list-metrics ticks) and no Service endpoint rollup. Two consequences:

• The rt-level watch tick drops the preview cache fingerprint every interval, so each hover refetch replaced enriched items with unenriched ones. With the metrics columns gone, the auto-detect width budget admits other extras (NODE, QoS, IP, ...), and they vanish again when an enriched payload comes back — the "surprise columns" flicker.
• The preview cache prime (added in fix(preview): follow the list's sorting and column config in list previews #410) overwrote the drilled list's enriched itemCache entry with unenriched data, so the next drill-in flashed the same way.

Fix

Mirror the main list's existing anti-flash mechanism in the preview handler:

• Extract carryOverMetricsColumnsFrom / carryOverServiceEndpointColumnsFrom — source-explicit cores of the existing middleItems-based helpers.
• In updateResourcesLoadedPreview, carry enriched columns from the drill-in cache (fallback: the currently shown preview items of the same kind) into fresh Pod/Node/Service preview loads, before the cache prime so the prime keeps the enrichment instead of clobbering it.

Test plan

• Carry-over from the drill-in cache (Pod, Node) and from current rightItems (Pod)
• Service endpoint rollup carry-over
• Cache prime retains enrichment (no clobber)
• Kind-mismatch guard (Node metrics never leak into a Pod preview)
• go test -race, go vet, golangci-lint (0 issues), gofumpt clean

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 79fd58ae-a4ec-4ab4-be4b-8a8b53b2c329

📥 Commits

Reviewing files that changed from the base of the PR and between f916464 and 71d2331.

📒 Files selected for processing (4)

• internal/app/cursor.go
• internal/app/cursor_carryover_union_test.go
• internal/app/preview_carryover_test.go
• internal/app/update_resources_loaded_preview.go

🚧 Files skipped from review as they are similar to previous changes (3)

• internal/app/cursor_carryover_union_test.go
• internal/app/cursor.go
• internal/app/preview_carryover_test.go

---

Summary by CodeRabbit

• Bug Fixes

  • Preview carry-over now keys preserved metric and endpoint columns by cluster+namespace+name, uses an explicit previous-source for carry-over, preserves non-metric columns, and avoids cross-kind or cross-cluster leakage. An anti-flap step prevents enriched drill-in data from being overwritten during preview updates.

• Tests

  • Added tests validating preview carry-over behavior, cluster isolation, endpoint rollup preservation, and kind-mismatch guardrails.

Walkthrough

This PR refactors preview carry-over into source-explicit helpers and applies them during preview resource loading to prevent column-set flapping. New helpers carryOverMetricsColumnsFrom(oldItems,newItems) and carryOverServiceEndpointColumnsFrom(oldItems,newItems) preserve enriched metric and endpoint rollup columns by copying them from a drilled-in cache entry or, if empty, from current right-side items. The integration runs before priming the item cache for Pod, Node, and Service previews. A new test suite verifies carry-over from cache, fallback behavior, endpoint preservation, cluster-isolation, and kind-mismatch skipping.

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (1 warning, 1 inconclusive)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 33.33% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Description check	❓ Inconclusive	The PR description covers the summary, root cause analysis, fix approach, and completed test plan, but is missing the structured template sections (Type of change, Changes checklist, formal Testing checklist, Checklist items).	Fill out the PR template sections including Type of change checkbox (fix), explicit Changes bullet list, Testing verification checkboxes, and Checklist items to match the repository template.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title directly addresses the main change: fixing column flicker in preview by carrying enriched columns into preview refetches, which is the core issue and solution described in the PR.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

internal/app/cursor.go (1)

141-188: ⚠️ Potential issue | 🟠 Major | ⚡ Quick win

Include ClusterName in carry-over keys to prevent cross-cluster column bleed.

Line 153 and Line 218 key by namespace+name only. In union mode, identical resources across clusters collide, so one cluster’s carried columns can overwrite another’s. Use ClusterName+namespace+name in both helpers.

Proposed fix

-func carryOverMetricsColumnsFrom(oldItems, newItems []model.Item) {
+func carryOverMetricsColumnsFrom(oldItems, newItems []model.Item) {
@@
-	type itemKey struct{ ns, name string }
+	type itemKey struct{ cluster, ns, name string }
@@
-			oldMetrics[itemKey{item.Namespace, item.Name}] = cols
+			oldMetrics[itemKey{item.ClusterName, item.Namespace, item.Name}] = cols
@@
-		key := itemKey{newItems[i].Namespace, newItems[i].Name}
+		key := itemKey{newItems[i].ClusterName, newItems[i].Namespace, newItems[i].Name}
@@
-func carryOverServiceEndpointColumnsFrom(oldItems, newItems []model.Item) {
+func carryOverServiceEndpointColumnsFrom(oldItems, newItems []model.Item) {
@@
-	type itemKey struct{ ns, name string }
+	type itemKey struct{ cluster, ns, name string }
@@
-			old[itemKey{item.Namespace, item.Name}] = cols
+			old[itemKey{item.ClusterName, item.Namespace, item.Name}] = cols
@@
-		key := itemKey{newItems[i].Namespace, newItems[i].Name}
+		key := itemKey{newItems[i].ClusterName, newItems[i].Namespace, newItems[i].Name}

Also applies to: 213-254

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@internal/app/cursor.go` around lines 141 - 188, The carry-over mapping keys
only use namespace+name which causes cross-cluster collisions; update the
itemKey struct used in carryOverMetricsColumnsFrom (and the other carry-over
helper later in the file) to include ClusterName and populate it from the
model.Item (e.g. item.ClusterName) when building oldMetrics and when computing
key for newItems (use the same ClusterName+Namespace+Name tuple), and make the
analogous change in the other carryover function that currently keys by
namespace+name.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Outside diff comments:
In `@internal/app/cursor.go`:
- Around line 141-188: The carry-over mapping keys only use namespace+name which
causes cross-cluster collisions; update the itemKey struct used in
carryOverMetricsColumnsFrom (and the other carry-over helper later in the file)
to include ClusterName and populate it from the model.Item (e.g.
item.ClusterName) when building oldMetrics and when computing key for newItems
(use the same ClusterName+Namespace+Name tuple), and make the analogous change
in the other carryover function that currently keys by namespace+name.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: ff26af8d-4cb5-4db9-a209-8b8b9da16738

📥 Commits

Reviewing files that changed from the base of the PR and between 55f4ca8 and db07496.

📒 Files selected for processing (3)

• internal/app/cursor.go
• internal/app/preview_carryover_test.go
• internal/app/update_resources_loaded_preview.go