[en] don't extract converted topic data as form

tatuylonen · Nov 21, 2024 · 6802294 · 6802294
1 parent f92d800
commit 6802294
Show file tree

Hide file tree

Showing 3 changed files with 29 additions and 1 deletion.
diff --git a/src/wiktextract/extractor/en/form_descriptions.py b/src/wiktextract/extractor/en/form_descriptions.py
@@ -2441,6 +2441,8 @@ def strokes_repl(m: re.Match) -> str:
                     ):
                         if alt_related is not None:
                             break
+                        if len(topics) > 0 and len(tagsets) > 0:
+                            break
                         continue
                     if (
                         i > 1

diff --git a/src/wiktextract/extractor/en/page.py b/src/wiktextract/extractor/en/page.py
@@ -94,7 +94,7 @@
 
 # Matches head tag
 HEAD_TAG_RE = re.compile(
-    r"^(head|Han char|arabic-noun|arabic-noun-form|"
+    r"^(head|Han char|arabic-noun|arabic-noun-form|term-label|tlb|"
     r"hangul-symbol|syllable-hangul)$|"
     + r"^(latin|"
     + "|".join(lang_code for lang_code, *_ in get_all_names("en"))

diff --git a/tests/test_en_head.py b/tests/test_en_head.py
@@ -777,3 +777,29 @@ def test_two_head_lines(self):
             )[0]["senses"][0]["tags"],
             ["feminine", "dialectal", "masculine"],
         )
+
+    def test_converted_topic_is_not_form(self):
+        # GH issue #906
+        # "fandom slang" -> "slang lifestyle" in "tags.py"
+        self.wxr.wtp.add_page(
+            "Template:term-label",
+            10,
+            """<span class="usage-label-term"><span class="ib-brac">(</span><span class="ib-content">[[fandom]] [[slang]]<span class="ib-comma">,</span>&#32;sometimes&#32;[[derogatory]]</span><span class="ib-brac">)</span></span>""",
+        )
+        self.wxr.wtp.add_page(
+            "Template:en-noun",
+            10,
+            """<span class="headword-line"><strong class="Latn headword" lang="en">chuunibyou</strong> (<i>[[Appendix:Glossary#countable|countable]] and [[Appendix:Glossary#uncountable|uncountable]]</i>, <i>plural</i> <b class="Latn form-of lang-en p-form-of" lang="en">[[chuunibyou#English|chuunibyou]]</b>)</span>""",
+        )
+        self.assertEqual(
+            parse_page(
+                self.wxr,
+                "chuunibyou",
+                """==English==
+===Noun===
+{{en-noun|~|chuunibyou}} {{term-label|en|fandom slang|sometimes|derogatory}}
+
+# gloss""",
+            )[0]["forms"],
+            [{"form": "chuunibyou", "tags": ["plural"]}],
+        )