fix: show inner HTML in place of non-existent sounds

server/app/dicts/mdict/html_cleaner.py

@@ -1,10 +1,22 @@
 import os
+import re
 import shutil
 from pathlib import Path
-import re
+from html.parser import HTMLParser
 
 
 class HTMLCleaner:
+	class _Concatenator(HTMLParser):
+		def __init__(self) -> None:
+			super().__init__()
+			self._buf = []
+
+		def handle_data(self, data: str) -> None:
+			self._buf.append(data)
+
+		def get_text(self) -> str:
+			return ''.join(self._buf)
+
 	_re_non_printing_chars = re.compile(r'[\x00-\x08\x0b\x0c\x0e-\x1f\x7f-\x9f]')
 	_re_compact_html_index = re.compile(r'`(\d+)`')
 	_re_single_quotes = re.compile(r"=\'([^']*)\'(?=[ >])")
@@ -180,27 +192,21 @@ def _fix_internal_href(self, definition_html: str) -> str:
 		# That is, links like entry://#81305a5747ca42b28f2b50de9b762963_nav2
 		return definition_html.replace('entry://#', '#')
 
-	def _flatten_nested_a(self, definition_html: str, depth: int) -> str:
+	def _flatten_nested_a(self, definition_html: str) -> str:
 		# Sometimes there're multiple inner elements inside the <a> element, which should be removed
 		# For example, in my Fr-En En-Fr Collins Dictionary, there's a <span> element inside the <a> element
 		# The text within the <span> should be preserved, though
 		# <a class="ref" href="/lookup/collinse22f/badly" title="Translation of badly"><span class="orth">badly</span></a>
-		if depth == 0:
-			return definition_html
-		else:
-			a_closing_tag_pos = 0
-			while (a_tag_start_pos := definition_html.find('<a', a_closing_tag_pos)) != -1:
-				a_tag_end_pos = definition_html.find('>', a_tag_start_pos)
-				inner_html_start_pos = definition_html.find('>', a_tag_end_pos + 1) + 1
-				if (a_closing_tag_pos := definition_html.find('</a>', a_tag_end_pos, inner_html_start_pos)) != -1:
-					continue
-				inner_html_end_pos = definition_html.find('</', inner_html_start_pos)
-				inner_html = definition_html[inner_html_start_pos:inner_html_end_pos]
-				a_closing_tag_pos = definition_html.find('</a>', inner_html_end_pos)
-				if definition_html.find('href', a_tag_start_pos, a_tag_end_pos) != -1:
-					definition_html = definition_html[:a_tag_end_pos + 1] +\
-						inner_html + definition_html[a_closing_tag_pos:]
-			return self._flatten_nested_a(definition_html, depth - 1)
+		a_closing_tag_pos = 0
+		while (a_tag_start_pos := definition_html.find('<a', a_closing_tag_pos)) != -1:
+			a_tag_end_pos = definition_html.find('>', a_tag_start_pos)
+			a_closing_tag_pos = definition_html.find('</a>', a_tag_end_pos)
+			a_inner_html = definition_html[a_tag_end_pos+1:a_closing_tag_pos]
+			concatenator = self._Concatenator()
+			concatenator.feed(a_inner_html)
+			a_inner_text = concatenator.get_text()
+			definition_html = definition_html[:a_tag_end_pos+1] + a_inner_text + definition_html[a_closing_tag_pos:]
+		return definition_html
 
 	def _fix_entry_cross_ref(self, definition_html: str) -> str:
 		if definition_html.startswith('@@@LINK='): # strange special case
@@ -211,27 +217,34 @@ def _fix_entry_cross_ref(self, definition_html: str) -> str:
 			return f'<a href="{self._lookup_url_root + entry_linked}">{entry_linked}</a>'
 		else:
 			definition_html = definition_html.replace('entry://', self._lookup_url_root)
-			# fingers crossed there are no more than three layers
-			return self._flatten_nested_a(definition_html, 3)
+			return self._flatten_nested_a(definition_html)
 
 	def _fix_sound_link(self, definition_html: str) -> str:
 		# Use HTML sound element instead of the original <a> element, which looks like this:
 		# <a class="hwd_sound sound audio_play_button icon-volume-up ptr fa fa-volume-up" data-lang="en_GB" data-src-mp3="https://www.collinsdictionary.com/sounds/hwd_sounds/EN-GB-W0020530.mp3" href="sound://audio/ef/7650.mp3" title="Pronunciation for "><img class="soundpng" src="/api/cache/collinse22f/img/sound.png"></a>
 		autoplay_string = 'autoplay'
 		sound_element_template = '<audio controls %s src="%s">%s</audio>'
+		not_found_fallback = '<span>%s</span>'
 		while (sound_link_start_pos := definition_html.find('sound://')) != -1:
 			sound_link_end_pos = definition_html.find('"', sound_link_start_pos)
 			original_sound_link = definition_html[sound_link_start_pos:sound_link_end_pos]
 			sound_link = original_sound_link.replace('sound://', self._href_root_dir)
+			sound_path = os.path.join(self._resources_dir, sound_link[len(self._href_root_dir):])
+
 			inner_html_start_pos = definition_html.find('>', sound_link_end_pos) + 1
 			inner_html_end_pos = definition_html.find('</a>', inner_html_start_pos)
 			inner_html = definition_html[inner_html_start_pos:inner_html_end_pos]
 			outer_html_start_pos = definition_html.rfind('<a', 0, sound_link_start_pos)
 			outer_html_end_pos = definition_html.find('</a>', inner_html_end_pos) + len('</a>')
-			definition_html = definition_html[:outer_html_start_pos] +\
-				sound_element_template % (autoplay_string, sound_link, inner_html) +\
+
+			if os.path.isfile(sound_path):
+				sound_html = sound_element_template % (autoplay_string, sound_link, inner_html)
+				autoplay_string = ''
+			else:
+				sound_html = not_found_fallback % inner_html
+
+			definition_html = definition_html[:outer_html_start_pos] + sound_html +\
 				definition_html[outer_html_end_pos:]
-			autoplay_string = ''
 
 		return definition_html