fix: patch RomM's fixture instead of rebuilding it

The RomM export fetched known_bios_files.json and never read it: the
fixture was rebuilt in our own order, so three corrections came out
as a diff of the whole file and a key we had not transcribed was
deleted. The export now starts from RomM's file, keeps its order and
fields, and only changes the values it corrects.

scripts/exporter/romm_exporter.py

@@ -52,12 +52,6 @@ class Exporter(BaseExporter):
     def native_sources() -> dict[str, str]:
         return {"known_bios_files.json": SOURCE_URL}
 
-    def __init__(self) -> None:
-        super().__init__()
-        # Filled by render; empty until then, so writable() answers before it.
-        self._platform_keys: set[tuple[str, str]] = set()
-        self._first_addition: dict[tuple[str, str], int] = {}
-
     @staticmethod
     def _verifiable(fe: NativeFile) -> bool:
         return bool(fe.size()) and any(
@@ -102,7 +96,15 @@ class Exporter(BaseExporter):
         originals: dict[str, str],
         scraped: dict | None = None,
     ) -> dict[str, str]:
-        output: OrderedDict[str, dict] = OrderedDict()
+        # RomM's own fixture is the base: its key order stays, and a key our
+        # transcription lacks (added after the scrape, or filtered from the
+        # model) is kept. Rebuilt from the model in our order, three hash
+        # corrections came out as a diff of the whole file.
+        original = originals.get(self.native_filename(), "")
+        output: OrderedDict[str, OrderedDict] = (
+            json.loads(original, object_pairs_hook=OrderedDict)
+            if original else OrderedDict()
+        )
         self._platform_keys = {
             (system.native_id, fe.name)
             for system in systems.values()
@@ -125,7 +127,9 @@ class Exporter(BaseExporter):
             for fe in sorted(system.files, key=lambda f: f.name):
                 if not self.writable(fe):
                     continue
-                entry: OrderedDict[str, str] = OrderedDict()
+                key = f"{system.native_id}:{fe.name}"
+                # Their entry keeps its fields and their order; ours fill in.
+                entry: OrderedDict[str, str] = OrderedDict(output.get(key, {}))
                 # The fixture states every value as a string, size included.
                 # Where the model settles no size, RomM's own value stands;
                 # str(None) wrote "None", and RomM's int() on it raised.
@@ -134,16 +138,11 @@ class Exporter(BaseExporter):
                     size = fe.platform.get("size")
                 if size is not None:
                     entry["size"] = str(int(size))
-                crc = fe.record_hash("crc32")
-                if crc:
-                    entry["crc"] = crc
-                md5 = fe.record_hash("md5")
-                if md5:
-                    entry["md5"] = md5
-                sha1 = fe.record_hash("sha1")
-                if sha1:
-                    entry["sha1"] = sha1
-                output[f"{system.native_id}:{fe.name}"] = entry
+                for field, hash_name in (("crc", "crc32"), ("md5", "md5"), ("sha1", "sha1")):
+                    value = fe.record_hash(hash_name)
+                    if value:
+                        entry[field] = value
+                output[key] = entry
 
         text = json.dumps(output, indent=2, ensure_ascii=False) + "\n"
         return {self.native_filename(): text}

tests/test_exporters.py

@@ -14,6 +14,7 @@ and skip when it has not been populated (python scripts/export_native.py
 
 from __future__ import annotations
 
+import difflib
 import json
 import os
 import re
@@ -22,6 +23,7 @@ import sys
 import tempfile
 import unittest
 import xml.etree.ElementTree as ET
+from collections import OrderedDict
 from pathlib import Path
 
 REPO_ROOT = Path(__file__).resolve().parent.parent
@@ -782,6 +784,46 @@ class RommExport(unittest.TestCase):
         self.assertTrue(all(k.startswith("psx:") for k in data))
 
 
+class RommKeepsItsFixture(unittest.TestCase):
+    """The export rebuilt known_bios_files.json from our model in our order:
+    3 hash corrections came out as a 3116-line diff, and a key RomM had that
+    our transcription lacked was deleted."""
+
+    def test_order_unknown_keys_and_untouched_entries_survive(self):
+        theirs = OrderedDict([
+            ("psx:zzz_first.bin", OrderedDict([("size", "1024"), ("md5", "a" * 32)])),
+            ("psx:only_theirs.bin", OrderedDict([("size", "16"), ("crc", "deadbeef")])),
+            ("psx:scph5501.bin", OrderedDict([("size", "524288"), ("md5", "b" * 32)])),
+        ])
+        original = json.dumps(theirs, indent=2) + "\n"
+        scraped = {"systems": {"sony-playstation": {"native_id": "psx", "files": [
+            {"name": "zzz_first.bin", "destination": "psx/zzz_first.bin",
+             "native_system": "psx", "md5": "a" * 32, "size": 1024},
+            {"name": "scph5501.bin", "destination": "psx/scph5501.bin",
+             "native_system": "psx", "md5": "b" * 32, "size": 524288},
+        ]}}}
+        truth = {"systems": {"sony-playstation": {"files": [{
+            "name": "scph5501.bin", "path": "scph5501.bin", "md5": "c" * 32,
+            "size": 524288, "required": True, "_cores": ["beetle_psx"],
+        }]}}}
+        systems, report = build_native_model(truth, scraped)
+        produced = discover_exporters()["romm"]().render(
+            systems, report, {"known_bios_files.json": original}, None
+        )["known_bios_files.json"]
+        data = json.loads(produced, object_pairs_hook=OrderedDict)
+        self.assertEqual(
+            list(data), ["psx:zzz_first.bin", "psx:only_theirs.bin", "psx:scph5501.bin"]
+        )
+        self.assertEqual(data["psx:only_theirs.bin"], theirs["psx:only_theirs.bin"])
+        self.assertEqual(data["psx:scph5501.bin"]["md5"], "c" * 32)
+        changed = [
+            line for line in difflib.unified_diff(
+                original.splitlines(), produced.splitlines(), lineterm="", n=0
+            ) if line[:1] in "+-" and line[:3] not in ("+++", "---")
+        ]
+        self.assertEqual(len(changed), 2, changed)
+
+
 class SystemDatKeepsSpaceIndentedLines(unittest.TestCase):
     """libretro indents 39 rom lines with four spaces. Read with a tab only,
     they left the base and came back last in their section, re-indented."""