feat: orphan pages and merged CMS issues with fix routes

- orphan_page for listed indexable pages no crawled page links to, with
  pages mentioning the title or focus keyphrase as suggestions (D-82)
- CMS-data image_alt and h1_in_content from the content descriptors
  (TopHeading 2 only), stored with the crawl as source cms (D-85, D-50)
- seo-page names Tools edit seo_update_page; every stored issue carries
  a tool or fix code with details.what (P33-38)
- TestIssueMerge (ignored fingerprints survive the next crawl, fixed
  pages lose their issues) and TestStagingCrawl (Pitfall 12)

classes/crawl/checks.go

@@ -125,6 +125,8 @@ type Env struct {
 	Staging bool
 	// GenericAnchors are the generic link texts (any case).
 	GenericAnchors []string
+	// Keyphrases maps an address to its SEO entry's focus keyphrase.
+	Keyphrases map[string]string
 }
 
 func (e *Env) record(key string) *Record {
@@ -198,6 +200,7 @@ func Check(res *Result, env *Env) []Finding {
 		out = append(out, checkRenderedPage(res, p, rec)...)
 	}
 	out = append(out, checkLinks(res, env)...)
+	out = append(out, checkOrphans(res, env)...)
 	out = append(out, checkSitemaps(res, env)...)
 	out = append(out, checkDuplicates(res, env)...)
 	return out

classes/crawl/job.go

@@ -221,7 +221,11 @@ func finish(ctx context.Context, svc *classes.Service, db *gorm.DB, row models.C
 		if err := SavePages(tx, row.ID, res); err != nil {
 			return err
 		}
-		findings := Check(res, &Env{Lookup: Lookup(ctx, svc, tx), Staging: svc.Staging(), GenericAnchors: DefaultGenericAnchors})
+		keyphrases, err := Keyphrases(tx)
+		if err != nil {
+			return err
+		}
+		findings := Check(res, &Env{Lookup: Lookup(ctx, svc, tx), Staging: svc.Staging(), GenericAnchors: DefaultGenericAnchors, Keyphrases: keyphrases})
 		cms, err := svc.Issues(classes.WithTx(ctx, tx))
 		if err != nil {
 			return err
@@ -325,6 +329,22 @@ func Lookup(ctx context.Context, svc *classes.Service, db *gorm.DB) func(key str
 	}
 }
 
+// Keyphrases maps each address with an SEO entry to its focus keyphrase.
+func Keyphrases(db *gorm.DB) (map[string]string, error) {
+	var rows []struct {
+		URL       string
+		Keyphrase string
+	}
+	if err := db.Raw(`SELECT url, keyphrase FROM golem15_seo_entries WHERE url <> '' AND keyphrase <> ''`).Scan(&rows).Error; err != nil {
+		return nil, err
+	}
+	out := make(map[string]string, len(rows))
+	for _, r := range rows {
+		out[r.URL] = r.Keyphrase
+	}
+	return out, nil
+}
+
 // errCancelledCrawl refuses cancelling a crawl that is not queued or
 // running.
 var errCancelledCrawl = errors.New("golem15.seo: the crawl is not running")

classes/crawl/orphans.go

@@ -0,0 +1,79 @@
+package crawl
+
+import (
+	"strings"
+
+	"git.golem15.com/golem15/summercms/modules/pact"
+)
+
+// maxSuggestions caps the pages suggested to link to an orphan.
+const maxSuggestions = 10
+
+// minPhrase is the shortest phrase a suggestion matches (in bytes).
+const minPhrase = 3
+
+// checkOrphans reports orphan_page (D-82): an indexable page a sitemap
+// lists that no crawled page links to (the home page is the crawl's
+// entry and never an orphan). Its suggestions are the pages whose main
+// text mentions the orphan's title or its focus keyphrase; the first
+// suggestion's record names the tool that adds the link, else the fix is
+// a link in the site's code.
+func checkOrphans(res *Result, env *Env) []Finding {
+	links := inbound(res)
+	var out []Finding
+	for _, key := range res.Order {
+		p := res.Pages[key]
+		if key == HomePath || !p.InSitemap || !env.indexable(res, p) || len(links[key]) > 0 {
+			continue
+		}
+		rec := env.record(key)
+		var phrases []string
+		add := func(s string) {
+			if s = normal(s); len(s) >= minPhrase {
+				for _, have := range phrases {
+					if have == s {
+						return
+					}
+				}
+				phrases = append(phrases, s)
+			}
+		}
+		if rec != nil {
+			add(rec.Title)
+		}
+		if h1 := p.Facts.H1(); len(h1) > 0 {
+			add(h1[0])
+		}
+		add(p.Facts.Title)
+		if env.Keyphrases != nil {
+			add(env.Keyphrases[key])
+		}
+		var suggestions []map[string]any
+		var first *Record
+		for _, other := range res.Order {
+			if len(suggestions) >= maxSuggestions {
+				break
+			}
+			op := res.Pages[other]
+			if other == key || !op.OK() {
+				continue
+			}
+			text := normal(op.Facts.Text)
+			for _, phrase := range phrases {
+				if strings.Contains(text, phrase) {
+					suggestions = append(suggestions, map[string]any{"url": other, "phrase": phrase})
+					if first == nil {
+						first = env.record(other)
+					}
+					break
+				}
+			}
+		}
+		f := Finding{Kind: KindOrphanPage, URL: key, Locale: locale(rec), Record: first, Details: map[string]any{"count": len(suggestions), "suggestions": suggestions}}
+		if first != nil {
+			f.Content = []string{pact.ContentToolEdit}
+		}
+		out = append(out, f)
+	}
+	return out
+}

classes/issues.go

@@ -7,6 +7,7 @@ import (
 	"strings"
 
 	"git.golem15.com/golem15/sm-seo-plugin/models"
+	"git.golem15.com/golem15/summercms/modules/pact"
 	"gorm.io/gorm"
 )
 
@@ -26,6 +27,13 @@ const (
 	// keeps Googlebot from (D-24): an error, since the sitemap asks for a
 	// page the crawler may not fetch.
 	IssueSitemapBlocked = "sitemap_blocked"
+	// IssueImageAlt is a record image without alternative text that is
+	// not marked decorative (D-85), from the content descriptors.
+	IssueImageAlt = "image_alt"
+	// IssueH1InContent is an H1 in the body of a record whose page renders
+	// the title as the only H1 (descriptor TopHeading 2, D-50; Journal's
+	// H1 setting on).
+	IssueH1InContent = "h1_in_content"
 )
 
 // The severities an issue may name.
@@ -92,7 +100,92 @@ func (s *Service) Issues(ctx context.Context) ([]Issue, error) {
 		data.listed = append(data.listed, address)
 	}
 	sort.Strings(data.listed)
-	return s.issuesOf(ctx, data, s.canonicalProblem)
+	out, err := s.issuesOf(ctx, data, s.canonicalProblem)
+	if err != nil {
+		return nil, err
+	}
+	content, err := s.ContentIssues(ctx)
+	if err != nil {
+		return nil, err
+	}
+	return append(out, content...), nil
+}
+
+// contentIssuesPage is the page size of the content walk.
+const contentIssuesPage = 200
+
+// ContentIssues are the CMS-data checks on the content descriptors of
+// every published record (D-85, D-50): image_alt for images with an empty
+// alt that are not decorative, and h1_in_content for an H1 in the body of
+// a record whose descriptor sets TopHeading 2 (off when the type sets no
+// rule, as Journal does with its H1 setting off). Each names the record
+// type's fitting tool (ContentDescriptor.Tools: image_alt or headings,
+// else edit); "" when the type names none.
+func (s *Service) ContentIssues(ctx context.Context) ([]Issue, error) {
+	reg := s.Registry()
+	if reg == nil {
+		return nil, nil
+	}
+	var out []Issue
+	for _, typ := range reg.ContentTypes() {
+		if typ.Content == nil || typ.Content.Records == nil {
+			continue
+		}
+		tool := func(ops ...string) string {
+			for _, op := range ops {
+				if t := strings.TrimSpace(typ.Content.Tools[op]); t != "" {
+					return t
+				}
+			}
+			return ""
+		}
+		after := ""
+		for {
+			page, err := reg.Records(ctx, typ.Key, after, contentIssuesPage, s.UnprefixedLocale())
+			if err != nil {
+				return nil, err
+			}
+			for _, rec := range page.Records {
+				if !rec.Published {
+					continue
+				}
+				for _, l := range rec.Locales {
+					if strings.TrimSpace(l.URL) == "" {
+						continue
+					}
+					var missing []string
+					for _, img := range l.Images {
+						if !img.Decorative && strings.TrimSpace(img.Alt) == "" && img.URL != "" {
+							missing = append(missing, img.URL)
+						}
+					}
+					if len(missing) > 0 {
+						out = append(out, Issue{Kind: IssueImageAlt, URL: l.URL, Locale: l.Locale, Tool: tool(pact.ContentToolImageAlt, pact.ContentToolEdit), Details: map[string]any{
+							"count": len(missing), "images": missing, "type": typ.Key, "id": rec.ID,
+						}})
+					}
+					if rec.TopHeading == 2 {
+						var h1 []string
+						for _, h := range l.Headings {
+							if h.Level == 1 {
+								h1 = append(h1, h.Text)
+							}
+						}
+						if len(h1) > 0 {
+							out = append(out, Issue{Kind: IssueH1InContent, URL: l.URL, Locale: l.Locale, Tool: tool(pact.ContentToolHeadings, pact.ContentToolEdit), Details: map[string]any{
+								"count": len(h1), "h1": h1, "type": typ.Key, "id": rec.ID,
+							}})
+						}
+					}
+				}
+			}
+			if page.Next == "" || page.Next == after || len(page.Records) == 0 {
+				break
+			}
+			after = page.Next
+		}
+	}
+	return out, nil
 }
 
 // issueData is what the issue report reads from the database: the SEO

classes/registry.go

@@ -247,6 +247,9 @@ func PageMenuItemType(service func() *Service) pact.MenuItemType {
 			Controller: PagesController,
 			Model:      PageModel,
 			Kind:       pact.ContentKindPage,
+			// A registry page's fields are changed through seo_update_page,
+			// the tool the site audit names for its content (P33-38).
+			Tools: map[string]string{pact.ContentToolEdit: "seo_update_page"},
 			Records: func(ctx context.Context, in pact.ContentList) (pact.ContentPage, error) {
 				s, err := svc()
 				if err != nil {

crawler_test.go

@@ -59,7 +59,7 @@ var checkedKinds = map[string]bool{
 	"h1_title_mismatch": true, "jsonld_invalid": true, "jsonld_multiple": true, "jsonld_required": true,
 	"jsonld_recommended": true, "hreflang_reciprocal": true, "hreflang_error": true, "generic_anchor": true,
 	"broken_link": true, "redirected_link": true, "image_alt": true, "image_size": true, "featured_lazy": true,
-	"content_js": true,
+	"content_js": true, "orphan_page": true,
 }
 
 // onlyChecked keeps the "kind url" entries of checkedKinds.
@@ -175,6 +175,10 @@ func TestCrawlerChecks(t *testing.T) {
 			t.Fatalf("content_js %s = %+v, want check %s naming %q", url, i, check, what)
 		}
 	}
+	detail("orphan_page", "/p/orphan", "count", 1)
+	if i := findIssue(t, list, "orphan_page", "/p/orphan"); !strings.Contains(fmt.Sprint(i.Details["suggestions"]), "/p/mentions") {
+		t.Fatalf("orphan suggestions = %v", i.Details)
+	}
 	content("/c/js-h1", "cms", "Record heading")
 	content("/c/js-words", "cms", "of 400 words")
 	content("/c/js-links", "cms", "/c/parent")
@@ -241,3 +245,123 @@ func TestRenderSkipped(t *testing.T) {
 	}
 	findIssue(t, env.issues(t, token, ""), "h1_multiple", "/p/two-h1")
 }
+
+// TestIssueMerge checks the CMS-data issues stored with the crawl and the
+// fix routes (SC-3, P33-38): before any crawl the list holds the CMS
+// issues; after it they are stored with the crawl as source cms; image_alt
+// names the record type's image_alt tool; h1_in_content appears for a post
+// and a category whose descriptor sets TopHeading 2 and not for TopHeading
+// 0, naming the headings tool; a skipped level in a category's pillar text
+// (a type without an edit tool) is fix code; no issue lacks a route. An
+// ignored fingerprint stays ignored in the next crawl, and a fixed page's
+// issue is gone from it.
+func TestIssueMerge(t *testing.T) {
+	env := bootCrawlEnv(t)
+	token := env.admin(t, true)
+	setCrawlRecords(t, map[string][]crawlRecord{
+		crawlPostType: {
+			{id: "1", path: "/c/post-a", title: "Post A", topHeading: 2, headings: []pact.ContentHeading{{Level: 1, Text: "Body H1"}, {Level: 2, Text: "Part"}},
+				images: []pact.ContentImage{{URL: "/img/a.jpg"}, {URL: "/img/deco.png", Decorative: true}}},
+			{id: "2", path: "/c/post-b", title: "Post B", topHeading: 0, headings: []pact.ContentHeading{{Level: 1, Text: "Body H1 of B"}}},
+		},
+		crawlCategoryType: {
+			{id: "10", path: "/cat/news", title: "News", topHeading: 2, headings: []pact.ContentHeading{{Level: 1, Text: "Pillar H1"}, {Level: 2, Text: "Intro"}, {Level: 4, Text: "Deep"}}},
+		},
+	})
+	site := newFakeSite()
+	site.sitemap("/", "/c/post-a", "/c/post-b", "/cat/news", "/p/fixme")
+	site.plant(
+		doc{path: "/", title: "Home", links: []string{"/c/post-a", "/c/post-b", "/cat/news", "/p/fixme"}},
+		doc{path: "/c/post-a", title: "Post A", body: `<h2>Part</h2><img src="/img/a.jpg" alt="Photo" width="1" height="1">`},
+		doc{path: "/c/post-b", title: "Post B"},
+		doc{path: "/cat/news", title: "News", body: "<h2>Intro</h2><h4>Deep</h4>"},
+		doc{path: "/p/fixme", title: "Fix me", noDesc: true},
+	)
+
+	before := env.issues(t, token, "")
+	if i := findIssue(t, before, "image_alt", "/c/post-a"); i.Source != "cms" || i.CrawlID != nil || i.Tool != "acme_set_alt" {
+		t.Fatalf("CMS image_alt before any crawl = %+v", i)
+	}
+
+	ctx := siteCtx(t, site)
+	first := env.crawlOnce(t, token, ctx)
+	list := env.issues(t, token, "")
+	alt := findIssue(t, list, "image_alt", "/c/post-a")
+	if alt.Source != "cms" || alt.CrawlID == nil || *alt.CrawlID != first || alt.Tool != "acme_set_alt" || alt.Fix != "tool" {
+		t.Fatalf("image_alt = %+v", alt)
+	}
+	if i := findIssue(t, list, "h1_in_content", "/c/post-a"); i.Tool != "acme_fix_headings" || i.Source != "cms" || i.Severity != "warning" {
+		t.Fatalf("h1_in_content post = %+v", i)
+	}
+	if i := findIssue(t, list, "h1_in_content", "/cat/news"); i.Tool != "acme_category_headings" {
+		t.Fatalf("h1_in_content category = %+v", i)
+	}
+	for _, i := range list {
+		if i.Kind == "h1_in_content" && i.URL == "/c/post-b" {
+			t.Fatalf("h1_in_content for TopHeading 0: %+v", i)
+		}
+		if i.Fix == "tool" && i.Tool == "" || i.Fix != "tool" && i.Fix != "code" {
+			t.Fatalf("issue without a route: %+v", i)
+		}
+		if i.Fix == "code" && i.Details["what"] == nil {
+			t.Fatalf("fix code without what: %+v", i)
+		}
+	}
+	skipped := findIssue(t, list, "heading_skipped", "/cat/news")
+	if skipped.Fix != "code" || skipped.Tool != "" {
+		t.Fatalf("heading_skipped in pillar text = %+v", skipped)
+	}
+	findIssue(t, list, "missing_description", "/p/fixme")
+
+	// Ignore the skipped level, fix the description, crawl again.
+	var fp string
+	if err := env.db.Raw(`SELECT fingerprint FROM golem15_seo_issues WHERE id = ?`, skipped.ID).Scan(&fp).Error; err != nil || fp == "" {
+		t.Fatalf("fingerprint: %q %v", fp, err)
+	}
+	if err := env.db.Exec(`INSERT INTO golem15_seo_issue_ignores (fingerprint, ignored_at) VALUES (?, NOW())`, fp).Error; err != nil {
+		t.Fatal(err)
+	}
+	site.plant(doc{path: "/p/fixme", title: "Fix me"})
+	second := env.crawlOnce(t, token, ctx)
+	list = env.issues(t, token, "")
+	again := findIssue(t, list, "heading_skipped", "/cat/news")
+	if !again.Ignored || again.CrawlID == nil || *again.CrawlID != second {
+		t.Fatalf("ignored issue in the next crawl = %+v", again)
+	}
+	for _, i := range list {
+		if i.URL == "/p/fixme" {
+			t.Fatalf("fixed page still has %+v", i)
+		}
+	}
+}
+
+// TestStagingCrawl checks RESEARCH Pitfall 12: on a non-production
+// environment every page carries the staging lock's noindex, and no
+// noindex-derived issue is reported; the pages still count as indexable.
+func TestStagingCrawl(t *testing.T) {
+	env := bootCrawlEnv(t, "SUMMER_ENV=staging")
+	if !env.seoPlugin(t).Service().Staging() {
+		t.Fatal("the environment is not staging")
+	}
+	token := env.admin(t, true)
+	site := newFakeSite()
+	site.sitemap("/", "/a", "/b")
+	for _, d := range []doc{
+		{path: "/", title: "Home", links: []string{"/a", "/b", "/c"}},
+		{path: "/a", title: "Same"},
+		{path: "/b", title: "Same"},
+		{path: "/c", title: "Unlisted"},
+	} {
+		d.robots = "noindex, nofollow"
+		site.set(d.path, 200, d.render(), "X-Robots-Tag", "noindex")
+	}
+	env.crawlOnce(t, token, siteCtx(t, site))
+	list := env.issues(t, token, "")
+	for _, i := range list {
+		if i.Kind == "sitemap_noindex" || i.Kind == "canonical_noindex" || i.Kind == "noindex_blocked" {
+			t.Fatalf("staging noindex reported: %+v", i)
+		}
+	}
+	findIssue(t, list, "duplicate_title", "/a")
+	findIssue(t, list, "sitemap_missing", "/c")
+}

lang/en/lang.yaml

@@ -526,6 +526,10 @@ crawls:
     already_running: A crawl is already running (started :time). Wait for it, or cancel it in SEO › Crawls.
     no_site_url: The site address for crawls is not configured (golem15.seo.site_url). Ask the administrator to set it.
 issues:
+    kind_orphan_page: "Orphan page (no internal links)"
+    sum_orphan_page: ":count suggestions"
+    kind_h1_in_content: "H1 in the content"
+    sum_h1_in_content: ":count H1 headings"
     kind_jsonld_invalid: "Structured data cannot be read"
     sum_jsonld_invalid: ":error"
     kind_jsonld_multiple: "More than one JSON-LD graph"

lang/pl/lang.yaml

@@ -540,6 +540,10 @@ crawls:
     already_running: Skanowanie już trwa (rozpoczęte :time). Poczekaj na nie lub anuluj je w SEO › Skanowania.
     no_site_url: Adres witryny do skanowania nie jest skonfigurowany (golem15.seo.site_url). Poproś administratora o jego ustawienie.
 issues:
+    kind_orphan_page: "Strona osierocona (brak linków wewnętrznych)"
+    sum_orphan_page: "Propozycje: :count"
+    kind_h1_in_content: "H1 w treści"
+    sum_h1_in_content: "Nagłówki H1: :count"
     kind_jsonld_invalid: "Nie można odczytać danych strukturalnych"
     sum_jsonld_invalid: ":error"
     kind_jsonld_multiple: "Więcej niż jeden graf JSON-LD"

menu_item_types.go

@@ -8,7 +8,8 @@ import (
 var _ pact.HasMenuItemTypes = (*Plugin)(nil)
 
 // MenuItemTypes registers the page registry's type seo-page (D-33): its
-// content descriptor (controller golem15.seo.pages, Kind page) lists,
+// content descriptor (controller golem15.seo.pages, Kind page, Tools edit
+// seo_update_page for the site audit's fix routes) lists,
 // describes and matches the registry pages, so every page gets the SEO
 // tab, resolves ok and is listed in the sitemaps grouped by its kind.
 func (p *Plugin) MenuItemTypes() []pact.MenuItemType {