inital commit

comment_parser.go

@@ -0,0 +1,131 @@
+package fa
+
+import (
+	"strconv"
+	"strings"
+
+	"github.com/PuerkitoBio/goquery"
+
+	"git.anthrove.art/public/go-fa-api/internal/urls"
+)
+
+// parseComments walks a submission or journal page's comment section and
+// returns the comments in document order. Depth and Parent are inferred from
+// data-attributes when present, otherwise from the legacy "width: NN%" style
+// FA still emits on threaded replies.
+func parseComments(doc *goquery.Document) []*Comment {
+	var out []*Comment
+
+	doc.Find("div.comment-container, div[id^='cid:'], div[id^='comment-']").Each(func(_ int, sel *goquery.Selection) {
+		idAttr, _ := sel.Attr("id")
+		idStr := strings.TrimPrefix(idAttr, "cid:")
+		idStr = strings.TrimPrefix(idStr, "comment-")
+		id, _ := parseID[CommentID](idStr)
+
+		c := &Comment{ID: id}
+
+		// Deleted comments: class on the container.
+		if class, _ := sel.Attr("class"); strings.Contains(class, "comment-deleted") ||
+			strings.Contains(class, "deleted-comment") {
+			c.Deleted = true
+		}
+
+		// Depth: prefer data-depth, then class "c-1/c-2/...", then style width %.
+		if d, ok := sel.Attr("data-depth"); ok {
+			if n, err := strconv.Atoi(strings.TrimSpace(d)); err == nil {
+				c.Depth = n
+			}
+		} else if class, _ := sel.Attr("class"); class != "" {
+			for _, tok := range strings.Fields(class) {
+				if strings.HasPrefix(tok, "c-") {
+					if n, err := strconv.Atoi(strings.TrimPrefix(tok, "c-")); err == nil {
+						c.Depth = n
+						break
+					}
+				}
+			}
+		}
+		if c.Depth == 0 {
+			if style, ok := sel.Attr("style"); ok {
+				c.Depth = depthFromWidthStyle(style)
+			}
+		}
+
+		// Parent: from data-parent or replyto-* class.
+		if p, ok := sel.Attr("data-parent"); ok {
+			if n, err := parseID[CommentID](strings.TrimSpace(p)); err == nil {
+				c.Parent = n
+			}
+		} else if class, _ := sel.Attr("class"); class != "" {
+			for _, tok := range strings.Fields(class) {
+				if strings.HasPrefix(tok, "replyto-") {
+					if n, err := parseID[CommentID](strings.TrimPrefix(tok, "replyto-")); err == nil {
+						c.Parent = n
+						break
+					}
+				}
+			}
+		}
+
+		// Author.
+		authorSel := sel.Find("a.iconusername, .comment-username a, .c-usernameBlock a").First()
+		if authorSel.Length() > 0 {
+			href, _ := authorSel.Attr("href")
+			c.Author = UserRef{
+				DisplayName: trimText(authorSel),
+				AvatarURL:   urls.AbsoluteCDN(trimAttr(authorSel.Find("img").First(), "src")),
+			}
+			if parts := strings.Split(strings.Trim(href, "/"), "/"); len(parts) >= 2 {
+				c.Author.Name = parts[1]
+			}
+		}
+
+		// Date.
+		dateSel := sel.Find("span.popup_date").First()
+		if t, err := ParseFADate(firstNonEmpty(trimAttr(dateSel, "title"), trimText(dateSel))); err == nil {
+			c.PostedAt = t
+		}
+
+		// Body.
+		body := sel.Find("div.comment-user-text, div.user-text, .no_overflow").First()
+		c.BodyHTML = htmlOf(body)
+		c.BodyText = strings.TrimSpace(body.Text())
+
+		if c.ID != 0 || c.Author.DisplayName != "" || c.BodyText != "" {
+			out = append(out, c)
+		}
+	})
+
+	return out
+}
+
+// depthFromWidthStyle reads a legacy FA inline style like
+// "width: 96%; padding: ..." and maps it to a depth level. FA used to shrink
+// each reply by 3% per level, which is how earlier scrapers detected depth.
+// Returns 0 if no usable width found.
+func depthFromWidthStyle(style string) int {
+	low := strings.ToLower(style)
+	i := strings.Index(low, "width:")
+	if i == -1 {
+		return 0
+	}
+	rest := low[i+len("width:"):]
+	end := strings.Index(rest, "%")
+	if end == -1 {
+		return 0
+	}
+	numStr := strings.TrimSpace(rest[:end])
+	num, err := strconv.Atoi(numStr)
+	if err != nil {
+		return 0
+	}
+	// 100% or 99% -> depth 0; each 3% step is one level deeper.
+	switch {
+	case num >= 99:
+		return 0
+	case num <= 0:
+		return 0
+	default:
+		return (99 - num) / 3
+	}
+}