Start adding a mechanism for grouping references configurably

Add types `RefGroup` and `RefGrouper` that will allow reference to be
grouped into multiple configurable groups, and to tally up the count
of references by group.

Author: mhagger

git-sizer.go

@@ -478,7 +478,11 @@ func mainImplementation(args []string) error {
 		filter = showRefFilter{filter}
 	}
 
-	historySize, err := sizes.ScanRepositoryUsingGraph(repo, filter, nameStyle, progress)
+	rg := refGrouper{
+		filter: filter,
+	}
+
+	historySize, err := sizes.ScanRepositoryUsingGraph(repo, &rg, nameStyle, progress)
 	if err != nil {
 		return fmt.Errorf("error scanning repository: %s", err)
 	}
@@ -505,6 +509,18 @@ func mainImplementation(args []string) error {
 	return nil
 }
 
+type refGrouper struct {
+	filter git.ReferenceFilter
+}
+
+func (rg *refGrouper) Categorize(refname string) (bool, []sizes.RefGroupSymbol) {
+	return rg.filter.Filter(refname), nil
+}
+
+func (rg *refGrouper) Groups() []sizes.RefGroup {
+	return nil
+}
+
 type showRefFilter struct {
 	f git.ReferenceFilter
 }

git_sizer_test.go

@@ -285,6 +285,16 @@ func pow(x uint64, n int) uint64 {
 	return p
 }
 
+type refGrouper struct{}
+
+func (rg refGrouper) Categorize(refname string) (bool, []sizes.RefGroupSymbol) {
+	return true, nil
+}
+
+func (rg refGrouper) Groups() []sizes.RefGroup {
+	return nil
+}
+
 func TestBomb(t *testing.T) {
 	t.Parallel()
 
@@ -295,7 +305,7 @@ func TestBomb(t *testing.T) {
 
 	h, err := sizes.ScanRepositoryUsingGraph(
 		repo.Repository(t),
-		git.AllReferencesFilter, sizes.NameStyleFull, false,
+		refGrouper{}, sizes.NameStyleFull, false,
 	)
 	require.NoError(t, err)
 
@@ -368,7 +378,7 @@ func TestTaggedTags(t *testing.T) {
 
 	h, err := sizes.ScanRepositoryUsingGraph(
 		repo.Repository(t),
-		git.AllReferencesFilter, sizes.NameStyleNone, false,
+		refGrouper{}, sizes.NameStyleNone, false,
 	)
 	require.NoError(t, err, "scanning repository")
 	assert.Equal(t, counts.Count32(3), h.MaxTagDepth, "tag depth")
@@ -390,7 +400,7 @@ func TestFromSubdir(t *testing.T) {
 
 	h, err := sizes.ScanRepositoryUsingGraph(
 		repo.Repository(t),
-		git.AllReferencesFilter, sizes.NameStyleNone, false,
+		refGrouper{}, sizes.NameStyleNone, false,
 	)
 	require.NoError(t, err, "scanning repository")
 	assert.Equal(t, counts.Count32(2), h.MaxPathDepth, "max path depth")
@@ -443,7 +453,7 @@ func TestSubmodule(t *testing.T) {
 	// Analyze the main repo:
 	h, err := sizes.ScanRepositoryUsingGraph(
 		mainRepo.Repository(t),
-		git.AllReferencesFilter, sizes.NameStyleNone, false,
+		refGrouper{}, sizes.NameStyleNone, false,
 	)
 	require.NoError(t, err, "scanning repository")
 	assert.Equal(t, counts.Count32(2), h.UniqueBlobCount, "unique blob count")
@@ -456,7 +466,7 @@ func TestSubmodule(t *testing.T) {
 	}
 	h, err = sizes.ScanRepositoryUsingGraph(
 		submRepo2.Repository(t),
-		git.AllReferencesFilter, sizes.NameStyleNone, false,
+		refGrouper{}, sizes.NameStyleNone, false,
 	)
 	require.NoError(t, err, "scanning repository")
 	assert.Equal(t, counts.Count32(2), h.UniqueBlobCount, "unique blob count")

sizes/graph.go

@@ -13,8 +13,36 @@ import (
 	"github.com/github/git-sizer/meter"
 )
 
+type RefGroupSymbol string
+
+// RefGroup is a group of references, for example "branches" or
+// "tags". Reference groups might overlap.
+type RefGroup struct {
+	// Symbol is the unique string by which this `RefGroup` is
+	// identified and configured. It consists of dot-separated
+	// components, which implicitly makes a nested tree-like
+	// structure.
+	Symbol RefGroupSymbol
+
+	// Name is the name for this `ReferenceGroup` to be presented
+	// in user-readable output.
+	Name string
+}
+
+type RefGrouper interface {
+	// Categorize tells whether `refname` should be walked at all,
+	// and if so, the symbols of the reference groups to which it
+	// belongs.
+	Categorize(refname string) (bool, []RefGroupSymbol)
+
+	// Groups returns the list of `ReferenceGroup`s, in the order
+	// that they should be presented. The return value might
+	// depend on which references have been seen so far.
+	Groups() []RefGroup
+}
+
 func ScanRepositoryUsingGraph(
-	repo *git.Repository, filter git.ReferenceFilter, nameStyle NameStyle, progress bool,
+	repo *git.Repository, rg RefGrouper, nameStyle NameStyle, progress bool,
 ) (HistorySize, error) {
 	graph := NewGraph(nameStyle)
 	var progressMeter meter.Progress
@@ -62,7 +90,9 @@ func ScanRepositoryUsingGraph(
 			if !ok {
 				break
 			}
-			if !filter.Filter(ref.Refname) {
+
+			walk, _ := rg.Categorize(ref.Refname)
+			if !walk {
 				continue
 			}
 			refs = append(refs, ref)