perf(scanner): size stat scanners via the shared parallel worker pool

docs/architecture.md

@@ -45,7 +45,7 @@ type Scanner interface {
 }
 ```
 
-Scanners report progress via context-attached callbacks for real-time UI updates; the walk batch reports under a single "projects" label, stat scanners under their own names. Sizing collects two figures per artifact via an in-process walk (`scanner.Measure`): **disk** (allocated blocks, `st_blocks×512` — sparse-accurate and matching `du`) and **apparent** (sum of logical file sizes). Directories contribute their own blocks to disk (real on ext4, ~0 on APFS); symlinks are never followed. Hard-linked inodes (`Nlink>1`) are counted once per artifact and recorded (keyed by `(dev, ino)`) so shared blocks can be netted out across artifacts. The walk engine does not size inline — it collects every matched artifact during the single pass, then sizes them across a bounded worker pool (`min(NumCPU, 8)`) so the traversals and their I/O overlap. Disk is the primary figure (sorting, `--min-size`, totals); apparent surfaces only when a file is materially sparse.
+Scanners report progress via context-attached callbacks for real-time UI updates; the walk batch reports under a single "projects" label, stat scanners under their own names. Sizing collects two figures per artifact via an in-process walk (`scanner.Measure`): **disk** (allocated blocks, `st_blocks×512` — sparse-accurate and matching `du`) and **apparent** (sum of logical file sizes). Directories contribute their own blocks to disk (real on ext4, ~0 on APFS); symlinks are never followed. Hard-linked inodes (`Nlink>1`) are counted once per artifact and recorded (keyed by `(dev, ino)`) so shared blocks can be netted out across artifacts. The walk engine does not size inline — it collects every matched artifact during the single pass, then sizes them across a bounded worker pool (`min(NumCPU, 8)`) so the traversals and their I/O overlap. The stat scanners (global/xcode/llm) size the same way: they discover artifacts via `stat` first, then defer to that same `sizePending` pool instead of measuring each inline. Disk is the primary figure (sorting, `--min-size`, totals); apparent surfaces only when a file is materially sparse.
 
 ### Walk engine
 

internal/scanner/global.go

@@ -176,7 +176,7 @@ func (s *GlobalScanner) Scan(ctx context.Context, root string) ([]model.ScanResu
 			continue
 		}
 
-		results = append(results, sized(model.ScanResult{
+		results = append(results, model.ScanResult{
 			Path:           full,
 			Ecosystem:      model.EcoGlobal,
 			Category:       c.category,
@@ -185,7 +185,7 @@ func (s *GlobalScanner) Scan(ctx context.Context, root string) ([]model.ScanResu
 			ProjectRoot:    filepath.Dir(full),
 			Label:          c.description,
 			Recommendation: c.rec,
-		}))
+		})
 		ReportProgress(ctx, len(results))
 	}
 
@@ -197,6 +197,12 @@ func (s *GlobalScanner) Scan(ctx context.Context, root string) ([]model.ScanResu
 	if runtime.GOOS == "darwin" && isUnderRoot(home, absRoot) {
 		results = append(results, s.scanCodeSignClones(ctx, len(results))...)
 	}
+
+	// Size all caches (and any code-sign clones) in one bounded worker pool
+	// rather than serially as each is discovered.
+	if err := sizePending(ctx, results); err != nil {
+		return results, err
+	}
 	return results, nil
 }
 
@@ -299,7 +305,7 @@ func (s *GlobalScanner) scanCodeSignClones(ctx context.Context, found int) []mod
 			}
 		}
 
-		out = append(out, sized(model.ScanResult{
+		out = append(out, model.ScanResult{
 			Path:           m,
 			Ecosystem:      model.EcoGlobal,
 			Category:       model.CatCache,
@@ -308,7 +314,7 @@ func (s *GlobalScanner) scanCodeSignClones(ctx context.Context, found int) []mod
 			ProjectRoot:    filepath.Dir(m),
 			Label:          codeSignCloneLabel(m, name),
 			Recommendation: rec,
-		}))
+		})
 		ReportProgress(ctx, found+len(out))
 	}
 	return out

internal/scanner/llm.go

@@ -53,7 +53,7 @@ func (s *LLMScanner) Scan(ctx context.Context, root string) ([]model.ScanResult,
 			return false
 		default:
 		}
-		results = append(results, sized(model.ScanResult{
+		results = append(results, model.ScanResult{
 			Path:           path,
 			Ecosystem:      model.EcoLLM,
 			Category:       model.CatCache,
@@ -63,7 +63,7 @@ func (s *LLMScanner) Scan(ctx context.Context, root string) ([]model.ScanResult,
 			Label:          label,
 			Recommendation: rec,
 			LastUsedAt:     ModTime(path),
-		}))
+		})
 		ReportProgress(ctx, len(results))
 		return true
 	}
@@ -120,6 +120,11 @@ func (s *LLMScanner) Scan(ctx context.Context, root string) ([]model.ScanResult,
 		}
 	}
 
+	// Size every discovered store in one bounded worker pool rather than
+	// serially as each is found (one tree walk per store overlaps the rest).
+	if err := sizePending(ctx, results); err != nil {
+		return results, err
+	}
 	return results, nil
 }
 

internal/scanner/scanner.go

@@ -227,9 +227,10 @@ func DirSize(path string) int64 {
 	return Measure(path).Disk
 }
 
-// sized fills Size, ApparentSize and Links on r by measuring r.Path. Convenience
-// for the stat-based scanners (xcode/global/llm) that build results one at a
-// time, mirroring what the walk engine's sizePending does in bulk.
+// sized fills Size, ApparentSize and Links on r by measuring r.Path — the
+// single-result form of what the walk engine's sizePending does in bulk (the
+// stat-based scanners now defer to sizePending; this remains for callers that
+// size one result at a time).
 func sized(r model.ScanResult) model.ScanResult {
 	st := Measure(r.Path)
 	r.Size = st.Disk

internal/scanner/xcode.go

@@ -120,16 +120,22 @@ func (s *XcodeScanner) Scan(ctx context.Context, root string) ([]model.ScanResul
 			continue
 		}
 
-		results = append(results, sized(model.ScanResult{
+		results = append(results, model.ScanResult{
 			Path:        full,
 			Ecosystem:   model.EcoXcode,
 			Category:    a.category,
 			LastMod:     ModTime(full),
 			Safety:      a.safety,
 			ProjectRoot: filepath.Dir(full),
-		}))
+		})
 		ReportProgress(ctx, len(results))
 	}
+
+	// Size every artifact (including expanded children) in one bounded worker
+	// pool rather than serially as each is discovered.
+	if err := sizePending(ctx, results); err != nil {
+		return results, err
+	}
 	return results, nil
 }
 
@@ -151,14 +157,14 @@ func expandChildren(ctx context.Context, parent string, a xcodeArtifact) []model
 		default:
 		}
 		child := filepath.Join(parent, e.Name())
-		out = append(out, sized(model.ScanResult{
+		out = append(out, model.ScanResult{
 			Path:        child,
 			Ecosystem:   model.EcoXcode,
 			Category:    a.category,
 			LastMod:     ModTime(child),
 			Safety:      a.safety,
 			ProjectRoot: parent,
-		}))
+		})
 		ReportProgress(ctx, len(out))
 	}
 	return out