fix: use UTF-16 offsets for description links

spec/invidious/videos/description_spec.cr

@@ -0,0 +1,56 @@
+require "../../spec_helper"
+require "../../../src/invidious/videos/description"
+
+Spectator.describe "Video description parser" do
+  describe "#parse_description" do
+    it "uses UTF-16 offsets when an emoji precedes a command" do
+      description = JSON.parse(<<-JSON)
+        {
+          "content": "🚀 Visit example now",
+          "commandRuns": [{
+            "startIndex": 3,
+            "length": 5,
+            "onTap": {
+              "innertubeCommand": {
+                "urlEndpoint": {"url": "https://example.com"}
+              }
+            }
+          }]
+        }
+        JSON
+
+      expect(parse_description(description, "video-id")).to eq(
+        %(🚀 <a href="https://example.com">Visit</a> example now)
+      )
+    end
+
+    it "uses UTF-16 lengths when an emoji is inside a command" do
+      description = JSON.parse(<<-JSON)
+        {
+          "content": "Open 😀 docs",
+          "commandRuns": [{
+            "startIndex": 0,
+            "length": 7,
+            "onTap": {
+              "innertubeCommand": {
+                "urlEndpoint": {"url": "https://example.com"}
+              }
+            }
+          }]
+        }
+        JSON
+
+      expect(parse_description(description, "video-id")).to eq(
+        %(<a href="https://example.com">Open 😀</a> docs)
+      )
+    end
+
+    it "escapes complete descriptions without command runs" do
+      description = JSON.parse(%({"content":"🚀 <safe> & complete"}))
+
+      expect(parse_description(description, "video-id")).to eq(
+        %(🚀 &lt;safe&gt; &amp; complete)
+      )
+    end
+  end
+end

src/invidious/videos/description.cr

@@ -1,9 +1,9 @@
 require "json"
 require "uri"
 
-private def copy_string(str : String::Builder, iter : Iterator, count : Int) : Int
+private def copy_string(str : String::Builder, iter : Iterator, count : Int? = nil) : Int
   copied = 0
-  while copied < count
+  while count.nil? || copied < count
     cp = iter.next
     break if cp.is_a?(Iterator::Stop)
 
@@ -21,7 +21,9 @@ private def copy_string(str : String::Builder, iter : Iterator, count : Int) : I
       str << cp.chr
     end
 
-    copied += 1
+    # YouTube expresses command offsets in UTF-16 code units. Astral
+    # codepoints therefore count as two units instead of one.
+    copied += cp > 0xFFFF ? 2 : 1
   end
 
   return copied
@@ -38,7 +40,7 @@ def parse_description(desc, video_id : String) : String?
     # Slightly faster than HTML.escape, as we're only doing one pass on
     # the string instead of five for the standard library
     return String.build do |str|
-      copy_string(str, content.each_codepoint, content.size)
+      copy_string(str, content.each_codepoint)
     end
   end
 
@@ -70,7 +72,6 @@ def parse_description(desc, video_id : String) : String?
     end
 
     # Copy the end of the string (past the last command).
-    remaining_length = content.size - index
+    copy_string(str, iter)
-    copy_string(str, iter, remaining_length) if remaining_length > 0
   end
 end