JSON report aggregate tool

README.md

@@ -48,6 +48,10 @@ All scripts will become available to use in your terminal with the `reccmp-` pre
 
 * [`decomplint`](/reccmp/tools/decomplint.py): Checks the decompilation annotations (see above)
     * e.g. `reccmp-decomplint --module LEGO1 LEGO1`
+* [`denoise`](/reccmp/tools/denoise.py): Combines JSON reports into a single aggregate file.
+    * Create aggregate report: `reccmp-denoise --samples ./sample0.json ./sample1.json ./sample2.json --output ./combined.json`
+    * Diff two saved reports: `reccmp-denoise --diff ./before.json ./after.json`
+    * Diff against the aggregate: `reccmp-denoise --samples ./sample0.json ./sample1.json ./sample2.json --diff ./before.json`
 * [`reccmp`](/reccmp/tools/asmcmp.py): Compares an original binary with a recompiled binary, provided a PDB file. For example:
     * Display the diff for a single function: `reccmp-reccmp --target LEGO1 --verbose 0x100ae1a0`
     * Generate an HTML report: `reccmp-reccmp --target LEGO1 --html output.html`

reccmp/assets/template.html

@@ -180,7 +180,10 @@
         margin-bottom: 0;
       }
     </style>
-    <script>var data = {{{data}}};</script>
+    <script>
+      var report = {{{report}}};
+      var data = report["data"];
+    </script>
     <script>{{{reccmp_js}}}</script>
     </script>
   </head>

reccmp/isledecomp/compare/diff.py

@@ -1,6 +1,5 @@
 from difflib import SequenceMatcher
-from typing import TypedDict
-from typing_extensions import NotRequired
+from typing_extensions import NotRequired, TypedDict
 
 CombinedDiffInput = list[tuple[str, str]]
 

reccmp/isledecomp/compare/report.py

@@ -0,0 +1,185 @@
+from datetime import datetime
+from dataclasses import dataclass
+from typing import Literal, Iterable, Iterator
+from pydantic import BaseModel, ValidationError
+from pydantic_core import from_json
+from .diff import CombinedDiffOutput
+
+
+class ReccmpReportDeserializeError(Exception):
+    """The given file is not a serialized reccmp report file"""
+
+
+@dataclass
+class ReccmpComparedEntity:
+    orig_addr: str
+    name: str
+    accuracy: float
+    recomp_addr: str | None = None
+    is_effective_match: bool = False
+    is_stub: bool = False
+    diff: CombinedDiffOutput | None = None
+
+
+class ReccmpStatusReport:
+    # The filename of the original binary.
+    # This is here to avoid comparing reports derived from different files.
+    # TODO: in the future, we may want to use the hash instead
+    filename: str
+
+    # Creation date of the report file.
+    timestamp: datetime
+
+    # Using orig addr as the key.
+    entities: dict[str, ReccmpComparedEntity]
+
+    def __init__(self, filename: str, timestamp: datetime | None = None) -> None:
+        self.filename = filename
+        if timestamp is not None:
+            self.timestamp = timestamp
+        else:
+            self.timestamp = datetime.now().replace(microsecond=0)
+
+        self.entities = {}
+
+
+def _get_entity_for_addr(
+    samples: Iterable[ReccmpStatusReport], addr: str
+) -> Iterator[ReccmpComparedEntity]:
+    """Helper to return entities from xreports that have the given address."""
+    for sample in samples:
+        if addr in sample.entities:
+            yield sample.entities[addr]
+
+
+def _accuracy_sort_key(entity: ReccmpComparedEntity) -> float:
+    """Helper to sort entity samples by accuracy score.
+    100% match is preferred over effective match.
+    Effective match is preferred over any accuracy.
+    Stubs rank lower than any accuracy score."""
+    if entity.is_stub:
+        return -1.0
+
+    if entity.accuracy == 1.0:
+        if not entity.is_effective_match:
+            return 1000.0
+
+    if entity.is_effective_match:
+        return 1.0
+
+    return entity.accuracy
+
+
+def combine_reports(samples: list[ReccmpStatusReport]) -> ReccmpStatusReport:
+    """Combines the sample reports into a single report.
+    The current strategy is to use the entity with the highest
+    accuracy score from any report."""
+    assert len(samples) > 0
+
+    # TODO: hack
+    assert all(samples[0].filename == s.filename for s in samples)
+
+    output = ReccmpStatusReport(filename=samples[0].filename)
+
+    # Combine every orig addr used in any of the reports.
+    orig_addr_set = {key for sample in samples for key in sample.entities.keys()}
+
+    all_orig_addrs = sorted(list(orig_addr_set))
+
+    for addr in all_orig_addrs:
+        e_list = list(_get_entity_for_addr(samples, addr))
+        assert len(e_list) > 0
+
+        # Our aggregate accuracy score is the highest from any report.
+        e_list.sort(key=_accuracy_sort_key, reverse=True)
+
+        output.entities[addr] = e_list[0]
+
+        # Recomp addr will most likely vary between samples, so clear it
+        output.entities[addr].recomp_addr = None
+
+    return output
+
+
+#### JSON schemas and conversion functions ####
+
+
+@dataclass
+class JSONEntityVersion1:
+    address: str
+    name: str
+    matching: float
+    # Optional fields
+    recomp: str | None = None
+    stub: bool = False
+    effective: bool = False
+    diff: CombinedDiffOutput | None = None
+
+
+class JSONReportVersion1(BaseModel):
+    file: str
+    format: Literal[1]
+    timestamp: float
+    data: list[JSONEntityVersion1]
+
+
+def _serialize_version_1(
+    report: ReccmpStatusReport, diff_included: bool = False
+) -> JSONReportVersion1:
+    """The HTML file needs the diff data, but it is omitted from the JSON report."""
+    entities = [
+        JSONEntityVersion1(
+            address=addr,  # prefer dict key over redundant value in entity
+            name=e.name,
+            matching=e.accuracy,
+            recomp=e.recomp_addr,
+            stub=e.is_stub,
+            effective=e.is_effective_match,
+            diff=e.diff if diff_included else None,
+        )
+        for addr, e in report.entities.items()
+    ]
+
+    return JSONReportVersion1(
+        file=report.filename,
+        format=1,
+        timestamp=report.timestamp.timestamp(),
+        data=entities,
+    )
+
+
+def _deserialize_version_1(obj: JSONReportVersion1) -> ReccmpStatusReport:
+    report = ReccmpStatusReport(
+        filename=obj.file, timestamp=datetime.fromtimestamp(obj.timestamp)
+    )
+
+    for e in obj.data:
+        report.entities[e.address] = ReccmpComparedEntity(
+            orig_addr=e.address,
+            name=e.name,
+            accuracy=e.matching,
+            recomp_addr=e.recomp,
+            is_stub=e.stub,
+            is_effective_match=e.effective,
+        )
+
+    return report
+
+
+def deserialize_reccmp_report(json_str: str) -> ReccmpStatusReport:
+    try:
+        obj = JSONReportVersion1.model_validate(from_json(json_str))
+        return _deserialize_version_1(obj)
+    except ValidationError as ex:
+        raise ReccmpReportDeserializeError from ex
+
+
+def serialize_reccmp_report(
+    report: ReccmpStatusReport, diff_included: bool = False
+) -> str:
+    """Create a JSON string for the report so it can be written to a file."""
+    now = datetime.now().replace(microsecond=0)
+    report.timestamp = now
+    obj = _serialize_version_1(report, diff_included=diff_included)
+
+    return obj.model_dump_json(exclude_defaults=True)