Update with additional safety measures

ja-readings-fix/ja-readings.py

@@ -6,12 +6,21 @@ import sys
 import pywikibot
 import regex as re
 
+
 ROMAJI_TRANSLITERATION_PATTERN = re.compile(r"\(\w+?\)")
+MULTIPLE_SPACE_PATTERN = re.compile(r" {2,}")
+UNEXPECTED_JA_READINGS_SYNTAX_PATTERN = re.compile(r"[^\u3040-\u309F\u30A0-\u30FF\-.,<\s]")
+
+
+def wikicode_is_safe(param: mwparserfromhell.wikicode.Wikicode) -> bool:
+    # Considered "safe" if it contains only text and maybe plain wikilinks
+    return all(type(node) in (mwparserfromhell.nodes.text.Text, mwparserfromhell.wikicode.Wikilink) for node in param.nodes)
 
 def fix_reading_str(reading_str: str) -> str:
     all_readings = [each.strip() for each in reading_str.split(",")]
-    all_readings = [kovachevbot.links_to_plaintext(ROMAJI_TRANSLITERATION_PATTERN.sub("", each)).strip() for each in all_readings]
+    all_readings = [MULTIPLE_SPACE_PATTERN.sub(" ", ROMAJI_TRANSLITERATION_PATTERN.sub("", kovachevbot.links_to_plaintext(each))).strip() for each in all_readings]
     return ", ".join(all_readings)
+    # return ROMAJI_TRANSLITERATION_PATTERN.sub("", kovachevbot.links_to_plaintext(reading_str))
 
 def fix_page(page: pywikibot.Page):
     kanji = page.title()
@@ -25,16 +34,45 @@ def fix_page(page: pywikibot.Page):
     ja_readingses: list[mwparserfromhell.wikicode.Template] = japanese_section.filter(forcetype=mwparserfromhell.wikicode.Template, matches="ja-readings")
 
     for ja_reading_template in ja_readingses:
+        if str(ja_reading_template.name).strip() != "ja-readings":
+            print("Mistakenly captured template", ja_reading_template.name, file=sys.stderr)
+            continue
+
+        params_on_newlines = ja_reading_template.name == "ja-readings\n"
         params_to_remove = list()
         for param in ja_reading_template.params:
             param: mwparserfromhell.nodes.extras.Parameter
+            # Places where a manual transliteration has distinguished a syllable boundary, e.g.
+            # utsukushii instead of utsukushī, are too complicated, and I opt to fix these by hand.
+            # Hence we just keep track of them by making a persistent file.
+            if "ii" in param.value or "aa" in param.value or "ee" in param.value or "oo" in param.value or "uu" in param.value:
+                print("Warning: potential transliteration variance due to doubled vowel", file=sys.stderr)
+                with open(f"READINGS_EXCEPTION_{kanji}", "w") as f:
+                    f.write(str(param.value))
 
             # Can't delete params while iterating, so we need to store them to delete later
             if param.value == "":  # Delete parameters that are supplied but not populated, e.g. "|nanori="
                 params_to_remove.append(param)
             else:
-                param.value = fix_reading_str(str(param.value))
+                if not wikicode_is_safe(param.value):
-        
+                    print("CRITICAL WARNING! NON-TEXT ELEMENT DETECTED", kanji, file=sys.stderr)
+                    with open(f"READINGS_EXCEPTION_{kanji}", "w") as f:
+                        f.write(str(param.value))
+                        continue
+
+                fixed = fix_reading_str(str(param.value)) + ("\n" if params_on_newlines else "")
+
+                # If there are non-textual elements, e.g. comments, blah blah, then the parameter cannot be trusted at all
+                # Ensure only the expected family of symbols (kana, comma, <, full stop, hyphen) are present
+
+                if UNEXPECTED_JA_READINGS_SYNTAX_PATTERN.match(fixed):
+                    print("CRITICAL WARNING! INVALID SYNTAX DETECTED", kanji, file=sys.stderr)
+                    with open(f"READINGS_EXCEPTION_{kanji}", "w") as f:
+                        f.write(fixed)
+                        continue
+
+                param.value = fixed
+
         for param in params_to_remove:
             ja_reading_template.remove(param)
 
@@ -44,12 +82,10 @@ def main():
     with open("ja-readings-to-fix.txt") as f:
         kanji_to_fix = f.read()
 
-
     pages = (kovachevbot.wikt_page(kanji) for kanji in kanji_to_fix)
     checked_pages_iter: Iterator[pywikibot.Page]  = kovachevbot.iterate_safe(pages)
     try:
         for i, page in enumerate(checked_pages_iter):
-            print(page.title())
             fix_page(page)
             page.save("Remove redundant ja-readings markup (manual transliterations; manual links; empty params)")
     except: