Added tests for clone detection framework

Makefile

@@ -14,3 +14,9 @@ dist/metascan: $(wildcard metascan/*.go) $(wildcard metascan/internal/*.go) $(wi
 	@echo "Building metascan"
 
 dist: dist/runtime-packages dist/metascan
+
+.venv/bin/python:
+	python3.10 -m venv .venv
+
+dev: .venv/bin/python
+	.venv/bin/python -m pip install .

go-libs/git/maintainers.go

@@ -118,14 +118,13 @@ func (all Commits) Ended() time.Time {
 	return started
 }
 
-func (all Commits) Filter(prefix string) (out Commits) {
-	// this is a straightforward code ownership detection strategy,
-	// though we can go berserk with prefix tree for paths
+// Filter reduces the history based on the predicate from path
+func (all Commits) Filter(predicate func(pathname string) bool) (out Commits) {
 	for _, c := range all {
 		stats := []NumStat{}
 		for _, ns := range c.Stats {
 			// we don't handle path renames
-			if !strings.HasPrefix(ns.Pathname, prefix) {
+			if !predicate(ns.Pathname) {
 				continue
 			}
 			stats = append(stats, ns)
@@ -144,6 +143,43 @@ func (all Commits) Filter(prefix string) (out Commits) {
 	return out
 }
 
+func (all Commits) LanguageStats() map[string]int {
+	// this is a straightforward code language detection strategy
+	stats := map[string]int{}
+	for _, c := range all {
+		for _, ns := range c.Stats {
+			split := strings.Split(ns.Pathname, ".")
+			ext := split[len(split)-1]
+			stats[ext] += ns.Added + ns.Deleted
+		}
+		if len(stats) == 0 {
+			continue
+		}
+	}
+	return stats
+}
+
+func (all Commits) Language() string {
+	type lang struct {
+		Ext     string
+		Changes int
+	}
+	var out []lang
+	for k, v := range all.LanguageStats() {
+		out = append(out, lang{
+			Ext:     k,
+			Changes: v,
+		})
+	}
+	if len(out) == 0 {
+		return "unknown"
+	}
+	sort.Slice(out, func(i, j int) bool {
+		return out[i].Changes > out[j].Changes
+	})
+	return out[0].Ext
+}
+
 func (all Commits) Authors() (out Authors) {
 	type tmp struct {
 		Author, Email string

metascan/clone/clone.go

@@ -0,0 +1,136 @@
+package clone
+
+import (
+	"context"
+	"fmt"
+	"path"
+	"sort"
+	"strings"
+
+	"github.com/databrickslabs/sandbox/go-libs/fileset"
+	"github.com/databrickslabs/sandbox/go-libs/git"
+	"github.com/databrickslabs/sandbox/go-libs/github"
+	"github.com/databrickslabs/sandbox/metascan/inventory"
+	"github.com/databrickslabs/sandbox/metascan/metadata"
+	"github.com/yuin/goldmark"
+	meta "github.com/yuin/goldmark-meta"
+	"github.com/yuin/goldmark/parser"
+	"github.com/yuin/goldmark/text"
+)
+
+type Clones []*Clone
+
+func (cc Clones) Metadatas(ctx context.Context) (out []metadata.Metadata, err error) {
+	for _, c := range cc {
+		m, err := c.Metadatas(ctx)
+		if err != nil {
+			return nil, err
+		}
+		out = append(out, m...)
+	}
+	sort.Slice(out, func(i, j int) bool {
+		return out[i].LastUpdated.After(out[j].LastUpdated)
+	})
+	return out, nil
+}
+
+type Clone struct {
+	Inventory inventory.Item
+	Git       *git.Checkout
+	Repo      github.Repo
+	FileSet   fileset.FileSet
+}
+
+func (c *Clone) Name() string {
+	return fmt.Sprintf("%s/%s", c.Inventory.Org, c.Repo.Name)
+}
+
+func (c *Clone) Metadatas(ctx context.Context) ([]metadata.Metadata, error) {
+	markdown := goldmark.New(
+		goldmark.WithExtensions(
+			meta.Meta,
+		),
+		goldmark.WithParserOptions(
+			parser.WithAutoHeadingID(),
+		),
+	)
+	if c.Inventory.IsSandbox {
+		history, err := c.Git.History(ctx)
+		if err != nil {
+			return nil, err
+		}
+		out := []metadata.Metadata{}
+		readmes := c.FileSet.Filter(`README.md`)
+		for _, readme := range readmes {
+			folder := path.Dir(readme.Relative)
+			if folder == "." {
+				continue
+			}
+			subHistory := history.Filter(func(pathname string) bool {
+				if strings.HasSuffix(pathname, ".md") {
+					// exclude any documentation
+					return false
+				}
+				return strings.HasPrefix(pathname, folder)
+			})
+			authors := subHistory.Authors()
+			raw, err := readme.Raw()
+			if err != nil {
+				return nil, err
+			}
+			document := markdown.Parser().Parse(text.NewReader(raw))
+			doc := document.OwnerDocument()
+			child := doc.FirstChild()
+			title := string(child.Text(raw))
+			if title == "" {
+				continue
+			}
+			if len(authors) == 0 {
+				continue
+			}
+			// todo: need the rest of the readme file
+			out = append(out, metadata.Metadata{
+				Title:       title,
+				Author:      authors.Primary(),
+				Language:    subHistory.Language(),
+				Date:        subHistory.Started(),
+				LastUpdated: subHistory.Ended(),
+				Maturity:    c.Inventory.Maturity,
+				URL:         fmt.Sprintf("%s/%s", c.Repo.HtmlURL, folder),
+			})
+		}
+		return out, nil
+	}
+	return []metadata.Metadata{{
+		Title:    c.Repo.Description,
+		Tags:     c.Repo.Topics,
+		Language: c.Repo.Langauge,
+		Date:     c.FileSet.LastUpdated(),
+		Maturity: c.Inventory.Maturity,
+		URL:      c.Repo.HtmlURL,
+	}}, nil
+}
+
+func (c *Clone) Maintainers(ctx context.Context) ([]string, error) {
+	history, err := c.Git.History(ctx)
+	if err != nil {
+		return nil, err
+	}
+	authors := history.Authors()
+	atMost := 2
+	if atMost > len(authors) {
+		atMost = len(authors)
+	}
+	// TODO: build up author stats remapper
+	var out []string
+	for _, v := range authors {
+		if v.Email == "action@github.com" {
+			continue
+		}
+		if v.Author == "dependabot[bot]" {
+			continue
+		}
+		out = append(out, v.Author)
+	}
+	return out[:atMost], nil
+}

metascan/clone/clone_test.go

@@ -0,0 +1,45 @@
+package clone
+
+import (
+	"context"
+	"path/filepath"
+	"testing"
+
+	"github.com/databrickslabs/sandbox/go-libs/env"
+	"github.com/databrickslabs/sandbox/go-libs/fileset"
+	"github.com/databrickslabs/sandbox/go-libs/git"
+	"github.com/databrickslabs/sandbox/go-libs/github"
+	"github.com/databrickslabs/sandbox/metascan/inventory"
+	"github.com/stretchr/testify/require"
+)
+
+func TestDiscoversSandbox(t *testing.T) {
+	ctx := context.Background()
+	home, _ := env.UserHomeDir(ctx)
+	dir := filepath.Join(home, ".databricks/labs/metascan/cache/databricks/terraform-databricks-examples")
+
+	fs, err := fileset.RecursiveChildren(dir)
+	require.NoError(t, err)
+
+	checkout, err := git.NewCheckout(ctx, dir)
+	require.NoError(t, err)
+
+	clone := &Clone{
+		Inventory: inventory.Item{
+			Org:       "databricks",
+			Repo:      "terraform-databricks-examples",
+			IsSandbox: true,
+		},
+		Repo: github.Repo{
+			Name:   "terraform-databricks-examples",
+			Topics: []string{"terraform", "modules"},
+		},
+		Git:     checkout,
+		FileSet: fs,
+	}
+
+	metadatas, err := clone.Metadatas(ctx)
+	require.NoError(t, err)
+
+	require.True(t, len(metadatas) > 0)
+}

metascan/internal/clones.go

@@ -5,12 +5,13 @@ import (
 	"fmt"
 	"path"
 	"sort"
+	"strings"
 
 	"github.com/databrickslabs/sandbox/go-libs/fileset"
 	"github.com/databrickslabs/sandbox/go-libs/git"
 	"github.com/databrickslabs/sandbox/go-libs/github"
 	"github.com/yuin/goldmark"
-	"github.com/yuin/goldmark-meta"
+	meta "github.com/yuin/goldmark-meta"
 	"github.com/yuin/goldmark/parser"
 	"github.com/yuin/goldmark/text"
 )
@@ -64,7 +65,13 @@ func (c Clone) Metadatas(ctx context.Context) ([]Metadata, error) {
 			folder := path.Dir(readme.Relative)
 
 			subFileset := c.FileSet.Filter(folder)
-			subHistory := history.Filter(folder)
+			subHistory := history.Filter(func(pathname string) bool {
+				if strings.HasSuffix(pathname, ".md") {
+					// exclude any documentation
+					return false
+				}
+				return strings.HasPrefix(pathname, folder)
+			})
 			authors := subHistory.Authors()
 
 			raw, err := readme.Raw()