From 6fbda7dc7839262441b3383c7c78e1030a03734d Mon Sep 17 00:00:00 2001
From: Shadowfacts <me@shadowfacts.net>
Date: Sat, 15 Jan 2022 19:31:04 -0500
Subject: [PATCH] Fix excerpt generator getting confused by massive img srcs

---
 Reader/ExcerptGenerator.swift | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/Reader/ExcerptGenerator.swift b/Reader/ExcerptGenerator.swift
index 9aa681a..96eb4e9 100644
--- a/Reader/ExcerptGenerator.swift
+++ b/Reader/ExcerptGenerator.swift
@@ -120,9 +120,19 @@ private func elementHandler(element: OpaquePointer!, userData: UnsafeMutableRawP
 
 private func textHandler(chunk: OpaquePointer!, userData: UnsafeMutableRawPointer!) -> lol_html_rewriter_directive_t {
     let userDataPtr = userData.assumingMemoryBound(to: UserData.self)
-    let s = lol_html_text_chunk_content_get(chunk)
-    let content = String(bytesNoCopy: UnsafeMutableRawPointer(mutating: s.data), length: s.len, encoding: .utf8, freeWhenDone: false)!
-    userDataPtr.pointee.paragraphText += content
+    if userDataPtr.pointee.isInParagraph {
+        let s = lol_html_text_chunk_content_get(chunk)
+        let content = String(bytesNoCopy: UnsafeMutableRawPointer(mutating: s.data), length: s.len, encoding: .utf8, freeWhenDone: false)!
+        userDataPtr.pointee.paragraphText += content
+        if userDataPtr.pointee.paragraphText.underestimatedCount >= 1024 {
+            // lol-html seems to get confused by img tags with hundreds of kilobytes of data in their src attributes
+            // and returns that data as text even though it's a tag
+            // if the text is over 1024 characters so far, we assume that's what's happened
+            // and abandon this attempt and try again at the next paragraph
+            userDataPtr.pointee.paragraphText = ""
+            userDataPtr.pointee.isInParagraph = false
+        }
+    }
     return LOL_HTML_CONTINUE
 }