fix: list archive members in accepted_hashes

scripts/generate_site.py

@@ -505,6 +505,21 @@ def _accepted_hash_rows(platform_files: list[dict], emulator_items: list[dict])
                         "name": entry.get("name", ""), "path": entry.get("path", "") or "",
                         "algorithm": algorithm, "value": value,
                     })
+            # Members of an archive: MAME and FBNeo hash the ROMs inside the
+            # set, never the set. Left out, a loose nmk004.bin (crc32
+            # 83b6f611) found no row while the site search found it.
+            archive = entry.get("name", "")
+            for member in entry.get("contents") or []:
+                if not isinstance(member, dict) or not member.get("name"):
+                    continue
+                for algorithm in HASH_ALGORITHMS:
+                    for value in _hash_values(member.get(algorithm)):
+                        rows.append({
+                            "layer": "emulator", "owner": item["id"],
+                            "system": _json_text(entry.get("system")),
+                            "name": str(member["name"]), "path": f"{archive}/{member['name']}",
+                            "algorithm": algorithm, "value": value,
+                        })
     rows.sort(key=lambda r: (r["layer"], r["owner"], r["name"], r["algorithm"], r["value"]))
     return rows
 
@@ -1123,7 +1138,9 @@ def generate_data_page(exports: list[dict]) -> str:
         "To find who accepts a hash, query `accepted_hashes` (or "
         "`accepted-hashes.csv`): one lowercase row per accepted value, per "
         "algorithm, for platform declarations and emulator profiles alike, "
-        "indexed on `(algorithm, value)`. The hash columns of `platform_files` "
+        "indexed on `(algorithm, value)`. A ROM inside a MAME or FBNeo set "
+        "has its own row, its `path` naming the archive and the member "
+        "(`nmk004.zip/nmk004.bin`). The hash columns of `platform_files` "
         "and `emulator_files` keep each source's own form: a Recalbox md5 may "
         "list several revisions separated by commas, and an emulator field "
         "holding several accepted values is a JSON array.",

tests/test_site_exports.py

@@ -458,5 +458,24 @@ class ASystemCardLandsOnThePlatformsSection(unittest.TestCase):
         )
 
 
+
+class ArchiveMembersAreAcceptedHashes(unittest.TestCase):
+    """accepted_hashes read only an entry's own hashes: 4,673 member crc32
+    values of MAME and FBNeo sets were absent from the table the Data page
+    calls the one to ask."""
+
+    def test_member_rows(self):
+        from generate_site import _accepted_hash_rows
+
+        item = {"id": "fbalpha2019", "profile": {"files": [
+            {"name": "nmk004.zip", "system": "arcade",
+             "contents": [{"name": "nmk004.bin", "crc32": "83b6f611"}]}]}}
+        rows = _accepted_hash_rows([], [item])
+        self.assertIn(
+            ("nmk004.bin", "nmk004.zip/nmk004.bin", "crc32", "83b6f611"),
+            {(r["name"], r["path"], r["algorithm"], r["value"]) for r in rows},
+        )
+
+
 if __name__ == "__main__":
     unittest.main()