ai-resume/internal/site/site_test.go

package site

import (
	"os"
	"path/filepath"
	"strings"
	"testing"

	"gitea.lab.audasmedia.com.au/sam/ai-resume/internal/content"
	"gitea.lab.audasmedia.com.au/sam/ai-resume/internal/search"
)

// newSite loads the repository's own content. If this fails, every other test
// here would be testing nothing.
func newSite(t *testing.T) *Site {
	t.Helper()
	root, err := filepath.Abs("../..")
	if err != nil {
		t.Fatal(err)
	}
	s, err := New(Config{Root: root, OutDir: filepath.Join(t.TempDir(), "dist")})
	if err != nil {
		t.Fatalf("load site: %v", err)
	}
	return s
}

// TestSixAreasHaveLandingPages is the DoD line "L1 (6) and L2 (6) pages
// complete": every area has a published level 1 page and at least one level 2
// page, and every area's L1 fragment is the default stream.
func TestSixAreasHaveLandingPages(t *testing.T) {
	s := newSite(t)
	if len(content.Areas) != 6 {
		t.Fatalf("%d areas, want 6", len(content.Areas))
	}
	if got := len(s.StreamKeys()); got != 6 {
		t.Fatalf("stream keys = %d, want the six area landing fragments", got)
	}
	for i, a := range content.Areas {
		l1, ok := s.Content.AreaLanding(a.Key)
		if !ok {
			t.Errorf("%s: no level 1 page", a.Key)
			continue
		}
		if !l1.Public {
			t.Errorf("%s: level 1 page %s is not public", a.Key, l1.ID)
		}
		if f, ok := s.Fragment(s.StreamKeys()[i]); !ok {
			t.Errorf("%s: stream key %s is not a fragment", a.Key, s.StreamKeys()[i])
		} else if f.DocID != l1.ID {
			t.Errorf("%s: stream fragment %s belongs to %s, want %s", a.Key, f.ID, f.DocID, l1.ID)
		}
		levels := map[int]int{}
		for _, d := range s.PublicInArea(a.Key) {
			levels[d.Level]++
		}
		if levels[2] == 0 {
			t.Errorf("%s: no level 2 pages", a.Key)
		}
		if levels[3] == 0 {
			t.Errorf("%s: no level 3 pages", a.Key)
		}
	}
}

// TestEveryArtefactIsReachableWithinThreeClicks is acceptance test C4 against
// the view models: landing → area → artefact page is three clicks, and every
// artefact has a working link.
func TestEveryArtefactIsReachableWithinThreeClicks(t *testing.T) {
	s := newSite(t)
	for _, area := range content.Areas {
		l1, ok := s.Content.AreaLanding(area.Key)
		if !ok {
			continue
		}
		linkedFromL1 := map[string]bool{}
		for _, l := range s.Doc(l1).SubPages {
			linkedFromL1[l.Href] = true
		}
		for _, l := range s.Doc(l1).Evidence {
			linkedFromL1[l.Href] = true
		}

		for _, d := range s.PublicInArea(area.Key) {
			if d.Level == 1 {
				continue
			}
			if d.Level == 2 {
				if !linkedFromL1[d.URL] {
					t.Errorf("%s/level 2 page %s is not linked from the area landing page", area.Key, d.URL)
				}
				continue
			}
			// Level 3: linked from an L2 page's drill-down or evidence table.
			reachable := false
			for _, l2 := range s.PublicInArea(area.Key) {
				if l2.Level != 2 {
					continue
				}
				view := s.Doc(l2)
				for _, l := range view.SubPages {
					if l.Href == d.URL {
						reachable = true
					}
				}
				for _, l := range view.Evidence {
					if l.Href == d.URL {
						reachable = true
					}
				}
			}
			if !reachable {
				t.Errorf("%s/level 3 page %s is not linked from any level 2 page", area.Key, d.URL)
			}
		}
	}
}

// TestEveryArtefactHasAStageFragment keeps the panel honest: if an evidence row
// offers hx-get, the fragment it names must exist.
func TestEveryArtefactHasAStageFragment(t *testing.T) {
	s := newSite(t)
	for _, d := range s.PublicDocs() {
		for _, l := range s.evidence(d) {
			if l.HXGet == "" {
				t.Errorf("%s: evidence %q has no stage fragment", d.ID, l.Label)
				continue
			}
			id := strings.TrimPrefix(l.HXGet, "/stage/")
			if _, ok := s.Fragment(id); !ok {
				t.Errorf("%s: evidence %q points at missing fragment %q", d.ID, l.Label, id)
			}
		}
	}
}

// TestBuildWritesEveryPublishedPage runs the real build once and checks the
// files the server depends on.
func TestBuildWritesEveryPublishedPage(t *testing.T) {
	s := newSite(t)
	if err := s.Build(); err != nil {
		t.Fatalf("Build: %v", err)
	}
	for _, url := range s.PublishedPages() {
		file := s.filePathForURL(url)
		if !strings.HasPrefix(file, s.Config.OutDir) {
			t.Errorf("%s: output escapes the build directory", url)
		}
		if _, err := os.ReadFile(file); err != nil {
			t.Errorf("%s: %v", url, err)
		}
	}
	for _, name := range []string{IndexFile, ManifestFile, BuildInfoFile, "static/site.css", "static/htmx.min.js", "static/chroma.css"} {
		if _, err := os.ReadFile(filepath.Join(s.Config.OutDir, filepath.FromSlash(name))); err != nil {
			t.Errorf("build output %s: %v", name, err)
		}
	}
	info, err := LoadBuildInfo(s.Config.OutDir)
	if err != nil {
		t.Fatalf("build info: %v", err)
	}
	if info.Pages == 0 || info.Fragments == 0 || info.Indexed == 0 {
		t.Errorf("build info is empty: %+v", info)
	}
	// The manifest order is the server's default stream: six document
	// fragments first, and they are the six area landings.
	if len(s.Manifest.Entries) < 6 {
		t.Fatalf("manifest has %d entries", len(s.Manifest.Entries))
	}
	for _, e := range s.Manifest.Entries[:6] {
		if !strings.HasPrefix(e.ID, "doc-") {
			t.Errorf("manifest entry %s is not a document fragment; the default stream must be the six areas", e.ID)
		}
	}
}

// copyTree copies a directory tree. The freshness test needs its own content
// tree so it can change a page without touching the repository.
func copyTree(src, dst string) error {
	return filepath.WalkDir(src, func(p string, d os.DirEntry, err error) error {
		if err != nil {
			return err
		}
		rel, err := filepath.Rel(src, p)
		if err != nil {
			return err
		}
		target := filepath.Join(dst, rel)
		if d.IsDir() {
			return os.MkdirAll(target, 0o755)
		}
		raw, err := os.ReadFile(p)
		if err != nil {
			return err
		}
		return os.WriteFile(target, raw, 0o644)
	})
}

// TestSearchIndexFollowsContent is the AUD-151 freshness line: a content change
// followed by a build is reflected in search. The deploy path is exactly this —
// deploy/deploy.sh runs `make dist` before it stages a release, and the build
// rewrites dist/search.db from content/ every time (search.Build removes the old
// file), so search can never serve an index from a previous content tree.
func TestSearchIndexFollowsContent(t *testing.T) {
	root, err := filepath.Abs("../..")
	if err != nil {
		t.Fatal(err)
	}
	dir := t.TempDir()
	contentDir := filepath.Join(dir, "content")
	if err := copyTree(filepath.Join(root, "content"), contentDir); err != nil {
		t.Fatalf("copy content: %v", err)
	}
	out := filepath.Join(dir, "dist")

	build := func() {
		t.Helper()
		s, err := New(Config{Root: root, ContentDir: contentDir, OutDir: out})
		if err != nil {
			t.Fatalf("load site: %v", err)
		}
		if err := s.Build(); err != nil {
			t.Fatalf("build: %v", err)
		}
	}
	query := func(q string) []search.Result {
		t.Helper()
		idx, err := search.Open(filepath.Join(out, IndexFile))
		if err != nil {
			t.Fatalf("open index: %v", err)
		}
		defer idx.Close()
		results, err := idx.Query(q, 20)
		if err != nil {
			t.Fatalf("query %q: %v", q, err)
		}
		return results
	}

	const token = "kumquatprotocol"
	build()
	if got := query(token); len(got) != 0 {
		t.Fatalf("token is already indexed in %d results; the test proves nothing", len(got))
	}

	page := filepath.Join(contentDir, "edge", "edge-fleet.md")
	raw, err := os.ReadFile(page)
	if err != nil {
		t.Fatal(err)
	}
	added := string(raw) + "\n\nA " + token + " harness keeps the nodes in step.\n"
	if err := os.WriteFile(page, []byte(added), 0o644); err != nil {
		t.Fatal(err)
	}

	build()
	results := query(token)
	if len(results) == 0 {
		t.Fatal("content change is not in the rebuilt index")
	}
	if results[0].URL != "/edge/edge-fleet/" {
		t.Errorf("token found at %s, want /edge/edge-fleet/", results[0].URL)
	}
}

// TestImagesReserveTheirSpace is the layout-shift guard: every image a content
// page carries must declare its intrinsic size and load lazily, so a page of
// diagrams and screenshots never reflows as the files arrive.
func TestImagesReserveTheirSpace(t *testing.T) {
	s := newSite(t)
	seen := 0
	for _, d := range s.Content.Docs {
		html := d.HTML
		for i := 0; ; i++ {
			start := strings.Index(html, "<img ")
			if start < 0 {
				break
			}
			end := strings.Index(html[start:], ">")
			if end < 0 {
				break
			}
			tag := html[start : start+end+1]
			html = html[start+end+1:]
			seen++
			for _, want := range []string{`width="`, `height="`, `loading="lazy"`, `alt="`} {
				if !strings.Contains(tag, want) {
					t.Errorf("%s: image tag missing %s: %s", d.Path, want, tag)
				}
			}
		}
	}
	if seen == 0 {
		t.Fatal("no images found in content; the guard is not exercising anything")
	}
}