internal/labels: clean issue body

Remove HTML comments and some headings that come
from the issue template.

This fixes the classification of 69040, an empty issue (see
internal/devtools/cmd/labeleval). Previously, the LLM was
confused by the comments and the presence of some headers.

For #64.

Change-Id: I82a68bba3f3aac9365b4d14035ed976b405651dc
Reviewed-on: https://go-review.googlesource.com/c/oscar/+/635875
Reviewed-by: Tatiana Bradley <tatianabradley@google.com>
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

internal/labels/labels.go

@@ -14,9 +14,12 @@ import (
 	"fmt"
 	"html/template"
 	"log"
+	"regexp"
+	"strings"
 
 	"golang.org/x/oscar/internal/github"
 	"golang.org/x/oscar/internal/llm"
+	"rsc.io/markdown"
 )
 
 // A Category is a classification for an issue.
@@ -38,11 +41,12 @@ func IssueCategoryFromList(ctx context.Context, cgen llm.ContentGenerator, iss *
 		return Category{}, "", errors.New("issue is a pull request")
 	}
 
+	body := cleanIssueBody(iss.Body)
 	// Extract issue text into a string.
 	var issueText bytes.Buffer
-	err = template.Must(template.New("body").Parse(body)).Execute(&issueText, bodyArgs{
+	err = template.Must(template.New("body").Parse(bodyTemplate)).Execute(&issueText, bodyArgs{
 		Title: iss.Title,
-		Body:  iss.Body,
+		Body:  body,
 	})
 	if err != nil {
 		return Category{}, "", err
@@ -73,6 +77,57 @@ func IssueCategoryFromList(ctx context.Context, cgen llm.ContentGenerator, iss *
 	return Category{}, "", fmt.Errorf("no category matches LLM response %q", jsonRes)
 }
 
+// TODO(jba): this is approximate.
+// See https://developer.mozilla.org/en-US/docs/Web/HTML/Comments for the exact syntax.
+var htmlCommentRegexp = regexp.MustCompile(`<!--(\n|.)*?-->`)
+
+// cleanIssueBody adjusts the issue body to improve the odds that it will be properly
+// labeled.
+func cleanIssueBody(text string) string {
+	// TODO(jba): These settings are also used in fix.go to parse bodies. Factor out.
+	p := &markdown.Parser{
+		AutoLinkText:  true,
+		Strikethrough: true,
+		HeadingIDs:    true,
+		Emoji:         true,
+	}
+	doc := p.Parse(text)
+
+	var cleanBlock func(markdown.Block)
+	cleanBlock = func(x markdown.Block) {
+		switch x := x.(type) {
+		case *markdown.Document:
+			for _, sub := range x.Blocks {
+				cleanBlock(sub)
+			}
+		case *markdown.HTMLBlock:
+			// Delete comments.
+			// Each Text is a line.
+			t := strings.Join(x.Text, "\n")
+			t = htmlCommentRegexp.ReplaceAllString(t, "")
+			x.Text = strings.Split(t, "\n")
+		case *markdown.Quote:
+			for _, sub := range x.Blocks {
+				cleanBlock(sub)
+			}
+		case *markdown.List:
+			for _, sub := range x.Items {
+				cleanBlock(sub)
+			}
+		case *markdown.Item:
+			for _, sub := range x.Blocks {
+				cleanBlock(sub)
+			}
+		case *markdown.Heading:
+			cleanBlock(x.Text)
+		case *markdown.Paragraph:
+			cleanBlock(x.Text)
+		}
+	}
+	cleanBlock(doc)
+	return markdown.Format(doc)
+}
+
 // response is the response that should generated by the LLM.
 // It must match [responseSchema].
 type response struct {
@@ -102,7 +157,7 @@ Report the category of the issue and an explanation of your decision.
 Each category and its description are listed below.
 
 `
-const body = `
+const bodyTemplate = `
 The title of the issue is: {{.Title}}
 The body of the issue is: {{.Body}}
 `

internal/labels/labels_test.go

@@ -39,3 +39,30 @@ func kindTestGenerator() llm.ContentGenerator {
 			return `{"CategoryName":"other","Explanation":"whatever"}`, nil
 		})
 }
+
+func TestCleanIssueBody(t *testing.T) {
+	for _, tc := range []struct {
+		in   string
+		want string
+	}{
+		{"", ""},
+		{"# H\nword\nword2\n", "# H\n\nword\nword2\n"},
+		{
+			"<!-- comment -->\n### H3\n<!-- another --> done",
+			"\n\n### H3\n\n done\n",
+		},
+		{
+			"<!--\ncomment\n-->\n### H3\n<!-- another -->\ndone",
+			"\n\n### H3\n\n\n\ndone\n",
+		},
+		{
+			"<!-- a --> b -->",
+			" b -->\n",
+		},
+	} {
+		got := cleanIssueBody(tc.in)
+		if got != tc.want {
+			t.Errorf("%q:\ngot  %q\nwant %q", tc.in, got, tc.want)
+		}
+	}
+}