fix: use UTF-16 code units for description link offsets

YouTube provides startIndex and length values in UTF-16 code units,
but the code was iterating Unicode codepoints. This mismatch caused
links to be misaligned when the description contained emoji or other
supplementary characters before a link.

Fixes #5821

src/invidious/videos/description.cr

@@ -0,0 +1,129 @@
+require "json"
+require "uri"
+
+private def copy_string(str : String::Builder, iter : Iterator, count : Int) : Int
+  copied = 0
+  while copied < count
+    cp = iter.next
+    break if cp.is_a?(Iterator::Stop)
+
+    if cp == 0x26 # Ampersand (&)
+      str << "&amp;"
+    elsif cp == 0x27 # Single quote (')
+      str << "&#39;"
+    elsif cp == 0x22 # Double quote (")
+      str << "&quot;"
+    elsif cp == 0x3C # Less-than (<)
+      str << "&lt;"
+    elsif cp == 0x3E # Greater than (>)
+      str << "&gt;"
+    else
+      str << cp.chr
+    end
+
+    copied += 1
+  end
+
+  return copied
+end
+
+# YouTube provides startIndex and length values in UTF-16 code units
+# (where supplementary characters like emoji count as 2), but Crystal's
+# String#each_codepoint iterates Unicode codepoints (where each emoji
+# counts as 1). This mismatch causes links to be misaligned when the
+# description contains emoji or other supplementary characters before a link.
+#
+# This iterator yields one unit per UTF-16 code unit, so that the existing
+# copy_string logic works correctly with YouTube's indices. Each BMP character
+# yields itself once; each supplementary character (codepoint >= 0x10000)
+# yields its high surrogate then its low surrogate.
+private class UTF16CodeUnitIterator
+  include Iterator(UInt32)
+
+  @codepoints : Array(UInt32)
+  @index : Int32 = 0
+  @pending : UInt32? = nil
+
+  def initialize(@codepoints)
+  end
+
+  def next
+    if pending = @pending
+      @pending = nil
+      return pending
+    end
+
+    if @index >= @codepoints.size
+      return Iterator::Stop::INSTANCE
+    end
+
+    cp = @codepoints[@index]
+    @index += 1
+
+    if cp >= 0x10000
+      # Encode as UTF-16 surrogate pair
+      cp -= 0x10000
+      high = 0xD800_u32 + (cp >> 10)
+      low = 0xDC00_u32 + (cp & 0x3FF)
+      @pending = low
+      return high
+    end
+
+    return cp
+  end
+end
+
+private def utf16_codeunit_iter(content : String)
+  UTF16CodeUnitIterator.new(content.codepoints)
+end
+
+def parse_description(desc, video_id : String) : String?
+  return "" if desc.nil?
+
+  content = desc["content"].as_s
+  return "" if content.empty?
+
+  commands = desc["commandRuns"]?.try &.as_a
+  if commands.nil?
+    # 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)
+    end
+  end
+
+  # YouTube's startIndex/length are measured in UTF-16 code units.
+  # We must iterate using UTF-16 code units so that indices align correctly,
+  # even when the description contains emoji or other supplementary characters.
+  iter = utf16_codeunit_iter(content)
+
+  index = 0
+
+  return String.build do |str|
+    commands.each do |command|
+      cmd_start = command["startIndex"].as_i
+      cmd_length = command["length"].as_i
+
+      # Copy the text chunk between this command and the previous if needed.
+      length = cmd_start - index
+      index += copy_string(str, iter, length)
+
+      # We need to copy the command's text using the iterator
+      # and the special function defined above.
+      cmd_content = String.build(cmd_length) do |str2|
+        copy_string(str2, iter, cmd_length)
+      end
+
+      link = cmd_content
+      if on_tap = command.dig?("onTap", "innertubeCommand")
+        link = parse_link_endpoint(on_tap, cmd_content, video_id)
+      end
+      str << link
+      index += cmd_length
+    end
+
+    # Copy the end of the string (past the last command).
+    remaining_length = content.size - index
+    copy_string(str, iter, remaining_length) if remaining_length > 0
+  end
+end