[refactor] Semantic function clustering: 8 verified de-duplication opportunities in pkg/cli and pkg/workflow

[github-actions]

🔧 Semantic Function Clustering Analysis

Analysis of repository: github/gh-aw · 2026-07-16

Analyzed the non-test Go sources under pkg/ (~1,098 files; the two dominant packages are pkg/workflow with 452 files and pkg/cli with 356). Function clusters were extracted by naming pattern, then each candidate was verified by reading the actual function bodies — only confirmed, actionable findings are listed below. The codebase is generally well-consolidated (many util helpers are already delegated to correctly), so this focuses on the concrete remaining duplication.

Overview

Metric	Value
Packages analyzed	33 under pkg/
Primary hotspots	pkg/cli, pkg/workflow
Confirmed findings	8 (5 high-confidence)
Detection method	Naming-cluster extraction + body-level verification

Critical actions (high confidence)

1. Collapse the near-identical int parsers in pkg/workflow/engine_config_parser.go — 5 functions, 2 clusters, differ only by a log string / comparison operator.
2. Consolidate 4 reimplementations of "workflow-ID-from-path" in pkg/cli onto the existing stringutil.NormalizeWorkflowName.
3. Unify the 3 copies + 4 inline copies of the ${{ }}-expression predicate in pkg/workflow.
4. De-duplicate the two byte-identical timeline row renderers in pkg/cli/gateway_logs_timeline_render.go.

1. Duplicate — near-identical integer parsers (pkg/workflow/engine_config_parser.go) · high

parseMaxRunsValue (L29) and parseMaxTurnCacheMissesValue (L44) are byte-for-byte identical except the log-message literal:

func parseMaxRunsValue(raw any) int {
    if val, ok := typeutil.ParseIntValue(raw); ok && val > 0 {
        return val
    }
    if rawStr, ok := raw.(string); ok {
        if parsed, err := strconv.Atoi(rawStr); err == nil && parsed > 0 {
            return parsed
        }
        engineLog.Printf("Ignoring invalid max-runs value: %q", rawStr) // only difference
    }
    return 0
}

A second near-identical cluster: parseMaxTurnsValue (L57), parseMaxToolDenialsValue (L104), parseNonNegativeIntOrExpressionValue (L84) — all ParseIntValue → strconv.Atoi(trimmed) → ${{ }} passthrough → log, differing only by the log string and >= 0 vs > 0.

Fix: collapse each cluster into one helper parameterized by (fieldName string, minValue int, allowExpression bool).

2. Scattered helper — "workflow-ID-from-path" reimplemented 4× (pkg/cli) · high

Canonical helper already exists: stringutil.NormalizeWorkflowName (pkg/stringutil/identifiers.go:29). Four local functions each hand-roll "base filename minus workflow extensions":

• pkg/cli/logs_format_compact.go:18 workflowIDFromPath
• pkg/cli/evals_branch.go:66 workflowIDFromRunPath
• pkg/cli/workflows.go:172 extractWorkflowNameFromPath
• pkg/cli/forecast_metadata.go:221 extractWorkflowIDFromName

// logs_format_compact.go:24-29
base = strings.TrimSuffix(base, ".lock.yml")
base = strings.TrimSuffix(base, ".yml")
base = strings.TrimSuffix(base, ".yaml")

Fix: extend NormalizeWorkflowName to also strip .yml/.yaml/.lock and delete the ad-hoc variants. normalizeWorkflowID (workflows.go:512) already delegates correctly — use it as the model.

3. Scattered helper — `${{ }}` expression predicate defined 3× + inlined 4× (pkg/workflow) · high

Same predicate (prefix ${{ + suffix }}, after TrimSpace) in three places:

• isExpression — pkg/workflow/expression_patterns.go:84
• isGitHubActionsExpression — pkg/workflow/observability_otlp.go:104
• isGitHubExpression — pkg/workflow/publish_assets.go:17 (regex form)

Plus inline copies in engine_config_parser.go:72,96,116 and safe_outputs_config_base.go:25.

Fix: keep the canonical isExpression in expression_patterns.go, delete the other two, replace the inline copies.

4. Duplicate — two byte-identical timeline row renderers (pkg/cli/gateway_logs_timeline_render.go) · high

renderAgentAssistantMessageRow (L325) and renderAgentReasoningRow (L338) are identical except one constant:

func renderAgentAssistantMessageRow(evt UnifiedTimelineEvent) []string {
    ts := formatTimelineTime(evt)
    src := timelineSourceLabel(evt.Source)
    kind := timelineEventIcon(TimelineKindAssistantMessage) + " " + timelineEventKindLabel(TimelineKindAssistantMessage)
    detail := stringutil.Truncate(evt.MessageContent, 48)
    return []string{ts, src, kind, detail, ""}
}
// renderAgentReasoningRow: identical, but TimelineKindReasoning

Fix: collapse into one helper taking the TimelineKind as a parameter.

5. Scattered helper — inline os.Stat existence probes instead of fileutil (pkg/cli) · high

pkg/fileutil already provides FileExists and DirExists; several sites re-probe with os.Stat:

• redacted_domains.go:91,101,112 — 3× if _, err := os.Stat(directPath); err == nil (discards error) → fileutil.FileExists
• forecast_metadata.go:117-121 — candidate-path existence loop → fileutil.FileExists
• firewall_policy.go:460 — os.Stat + !info.IsDir() guard → !fileutil.DirExists(...) (file already imports fileutil)
• compile_repository_manifest.go:105 — README existence probe → fileutil.FileExists (medium; the check at :94 legitimately needs os.IsNotExist — leave it)

Fix: replace the pure existence probes with fileutil.FileExists/fileutil.DirExists.

6. Scattered helper — manual map-key collect-and-sort instead of sliceutil.SortedKeys (pkg/workflow) · medium

sliceutil.SortedKeys is already used 97× across 47 files here, yet a few sites hand-roll it:

• lsp_manager.go:36-40 and permissions_operations.go:434-437 — pure collect + sort.Strings (clean SortedKeys cases)
• cache.go:872-875, safe_jobs_needs_validation.go:161-164 — same shape with a per-key transform (SortedKeys + sliceutil.Map)

Fix: use sliceutil.SortedKeys in the clean cases.

7. In-file repetition — timeline row preamble copied across 13 renderers (pkg/cli/gateway_logs_timeline_render.go) · medium

The 3-line preamble (ts/src/kind) appears in 13 render*Row functions, and the "server/tool detail (truncated)" block is copied in renderGatewayToolCallRow:143, renderGatewayDIFCFilteredRow:173, renderGatewayGuardPolicyBlockedRow:198, renderAgentToolStartRow:284, renderAgentToolDoneRow:303.

Fix: extract newTimelineRow(evt, kind) and serverToolDetail(evt, maxLen) helpers in the same file.

8. Duplicate — shared section blocks between compact/verbose log renderers (pkg/cli/logs_format_compact.go) · medium

renderLogsCompact (L56) and renderLogsCompactVerbose (L225) share byte-identical [mcp-failures] (:196↔:357), [missing-tools] (:204↔:365), and [location] (:212↔:382) blocks.

Fix: extract shared renderCompactMCPFailures/MissingTools/Location(data) helpers used by both.

Explicitly rejected (verified as correct by-design, not duplicates)

• containsWorkflowDispatch/containsWorkflowCall already delegate to containsTrigger; dedupeAllowedTools wraps sliceutil.Deduplicate; truncateSHAForLog wraps stringutil.Truncate; parseRolesValue/parseBotsValue delegate to parseStringSliceAny.
• resolveDigestViaBuildx/Crane/Pull, render*DiffMarkdownSection vs render*DiffPrettySection, the sanitize* family, and getRepositorySlugFrom* — all genuinely distinct logic/output.
• map[string]bool set-membership is not replaceable by setutil.Contains (needs map[K]struct{}).

Next actions

• P1 (high, ~1–2h): items 1–5 — pure mechanical de-duplication, low risk, existing tests cover the behavior.
• P2 (medium, ~1–2h): items 6–8 — in-file helper extraction.
• Run go test ./pkg/cli/... ./pkg/workflow/... after each consolidation to confirm no behavior change.

References: §29460289143

Generated by 🔧 Semantic Function Refactoring · 568.1 AIC · ⌖ 16.1 AIC · ⊞ 9.4K · ◷

• expires on Jul 17, 2026, 4:17 PM UTC-08:00

[github-actions]

🍪 Issue Monster selected this for Copilot

I've identified this issue as a good candidate for automated resolution and requested assignment to the Copilot coding agent.

If assignment succeeds, the Copilot coding agent will analyze the issue and create a pull request with the fix.

Om nom nom! 🍪

🍪 Om nom nom by Issue Monster · 0.685 AIC · ⌖ 4.7 AIC · ⊞ 6K · ◷