inital commit

system_message.go

@@ -0,0 +1,65 @@
+package fa
+
+import (
+	"strings"
+
+	"github.com/PuerkitoBio/goquery"
+)
+
+// classifySystemMessage looks at a fetched document and, when it is one of
+// FA's gate/error templates rather than a real content page, maps it to the
+// most specific sentinel error. Returns nil for normal content pages so
+// parsing continues.
+//
+// FA emits at least two distinct gate templates:
+//
+//   - <title>System Error</title> generic "not found / no permission /
+//     rate limited" page with the human message in a section-body.
+//   - <title>Login Required …</title> auth-gated content (some submissions
+//     are not viewable anonymously). Always maps to ErrUnauthorized.
+//
+// We classify by <title> first to pick a template, then by message text
+// within the section-body when needed. Anything else is left for the
+// downstream parser to handle.
+func classifySystemMessage(doc *goquery.Document) error {
+	pageTitle := strings.ToLower(strings.TrimSpace(doc.Find("title").First().Text()))
+
+	switch {
+	case strings.HasPrefix(pageTitle, "login required"):
+		return ErrUnauthorized
+	case pageTitle == "system error",
+		strings.HasPrefix(pageTitle, "system error"):
+		// fall through to body classification below
+	default:
+		return nil
+	}
+
+	headerTitle := trimText(doc.Find("section .section-header h2").First())
+	if headerTitle == "" {
+		headerTitle = trimText(doc.Find("h2").First())
+	}
+	body := strings.TrimSpace(doc.Find("section .section-body").First().Text())
+	if body == "" {
+		body = strings.TrimSpace(doc.Find(".section-body").First().Text())
+	}
+
+	low := strings.ToLower(headerTitle + " " + body)
+	switch {
+	case strings.Contains(low, "not in our database"),
+		strings.Contains(low, "submission not found"),
+		strings.Contains(low, "user not found"),
+		strings.Contains(low, "journal not found"),
+		strings.Contains(low, "page not found"),
+		strings.Contains(low, "no such"):
+		return ErrNotFound
+	case strings.Contains(low, "you must be logged in"),
+		strings.Contains(low, "log in to view"),
+		strings.Contains(low, "permission to view"),
+		strings.Contains(low, "permission to access"):
+		return ErrUnauthorized
+	case strings.Contains(low, "rate limit"),
+		strings.Contains(low, "too many requests"):
+		return ErrRateLimited
+	}
+	return &SystemMessageError{Title: headerTitle, Body: body}
+}