支持 H32 DLogCapture（MSOP+DIFOP）导出到 V1 中间格式

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-05 08:58:31 +08:00
parent 4ff176d184
commit 30f7e66db3
14 changed files with 993 additions and 43 deletions
@@ -0,0 +1,179 @@
+"""Index and read Medulla DObject recordings (dobject/ + dobject_recording/)."""
+
+from __future__ import annotations
+
+import re
+import struct
+from dataclasses import dataclass
+from pathlib import Path
+from typing import BinaryIO, Iterator
+
+
+RECORD_RE = re.compile(
+    r"^\[(?P<log_time>[^]]+)\].*?DObject `(?P<name>[^`]+)` post "
+    r"len=(?P<len>\d+)B, id:(?P<id>[0-9A-Fa-f]+), tic:(?P<tic>\d+), "
+    r"@(?P<file>[^:]+):(?P<offset>\d+)"
+)
+
+
+@dataclass(frozen=True)
+class RecordRef:
+    sequence: int
+    object_name: str
+    log_time: str
+    source_log: str
+    source_dorec: str
+    source_offset: int
+    payload_length: int
+    log_record_id: str
+    dotnet_ticks: int
+
+
+def resolve_dlog_root(value: Path | str) -> Path:
+    root = Path(value).expanduser().resolve()
+    if (root / "dobject").is_dir() and (root / "dobject_recording").is_dir():
+        return root
+    child = root / "dlog"
+    if (child / "dobject").is_dir() and (child / "dobject_recording").is_dir():
+        return child
+    raise FileNotFoundError(f"{root} does not contain dobject and dobject_recording")
+
+
+def discover_records(dlog_root: Path, object_name: str) -> list[RecordRef]:
+    pending: list[tuple[str, str, str, int, int, str, int, str]] = []
+    for log_path in sorted((dlog_root / "dobject").rglob("*.log")):
+        relative_log = log_path.relative_to(dlog_root).as_posix()
+        with log_path.open("r", encoding="utf-8", errors="replace") as stream:
+            for line in stream:
+                match = RECORD_RE.search(line)
+                if not match or match.group("name").casefold() != object_name.casefold():
+                    continue
+                pending.append(
+                    (
+                        match.group("name"),
+                        match.group("log_time"),
+                        relative_log,
+                        int(match.group("offset")),
+                        int(match.group("len")),
+                        match.group("id").upper(),
+                        int(match.group("tic")),
+                        match.group("file"),
+                    )
+                )
+    pending.sort(key=lambda item: (item[6], item[7].casefold(), item[3]))
+    seen: set[tuple[str, int, int]] = set()
+    records: list[RecordRef] = []
+    for item in pending:
+        key = (item[7].casefold(), item[3], item[6])
+        if key in seen:
+            continue
+        seen.add(key)
+        records.append(
+            RecordRef(
+                sequence=len(records),
+                object_name=item[0],
+                log_time=item[1],
+                source_log=item[2],
+                source_dorec=item[7],
+                source_offset=item[3],
+                payload_length=item[4],
+                log_record_id=item[5],
+                dotnet_ticks=item[6],
+            )
+        )
+    return records
+
+
+def index_dorec_files(dlog_root: Path) -> dict[str, list[Path]]:
+    result: dict[str, list[Path]] = {}
+    for path in (dlog_root / "dobject_recording").rglob("*.dorec"):
+        result.setdefault(path.name.casefold(), []).append(path)
+    return result
+
+
+def choose_dorec(index: dict[str, list[Path]], name: str) -> Path:
+    matches = index.get(Path(name).name.casefold(), [])
+    if not matches:
+        raise FileNotFoundError(f"missing recording file: {name}")
+    if len(matches) > 1:
+        raise RuntimeError(f"ambiguous recording file {name}: {matches}")
+    return matches[0]
+
+
+def read_exact(stream: BinaryIO, size: int) -> bytes:
+    data = stream.read(size)
+    if len(data) != size:
+        raise EOFError(f"expected {size} bytes, got {len(data)}")
+    return data
+
+
+def read_record_payload(path: Path, record: RecordRef) -> bytes:
+    with path.open("rb") as stream:
+        stream.seek(record.source_offset)
+        name_length = read_exact(stream, 1)[0]
+        name = read_exact(stream, name_length).decode("ascii")
+        ticks = struct.unpack("<q", read_exact(stream, 8))[0]
+        id_length = read_exact(stream, 1)[0]
+        id_bytes = read_exact(stream, id_length)
+        payload_length = struct.unpack("<i", read_exact(stream, 4))[0]
+        payload = read_exact(stream, payload_length)
+
+    try:
+        record_id = id_bytes.decode("ascii")
+    except UnicodeDecodeError:
+        record_id = id_bytes.hex().upper()
+    if name != record.object_name:
+        raise ValueError(f"name mismatch: log={record.object_name}, dorec={name}")
+    if ticks != record.dotnet_ticks:
+        raise ValueError(f"tick mismatch: log={record.dotnet_ticks}, dorec={ticks}")
+    if payload_length != record.payload_length:
+        raise ValueError(f"payload mismatch: log={record.payload_length}, dorec={payload_length}")
+    if record_id.upper() != record.log_record_id.upper():
+        raise ValueError(f"record id mismatch: log={record.log_record_id}, dorec={record_id}")
+    return payload
+
+
+def iter_payloads(dlog_root: Path, object_name: str) -> Iterator[tuple[RecordRef, bytes]]:
+    root = resolve_dlog_root(dlog_root)
+    records = discover_records(root, object_name)
+    if not records:
+        return
+    dorec_index = index_dorec_files(root)
+    open_files: dict[str, tuple[Path, BinaryIO]] = {}
+    try:
+        for record in records:
+            key = record.source_dorec.casefold()
+            handle = open_files.get(key)
+            if handle is None:
+                path = choose_dorec(dorec_index, record.source_dorec)
+                handle = (path, path.open("rb"))
+                open_files[key] = handle
+            path, stream = handle
+            stream.seek(record.source_offset)
+            name_length = read_exact(stream, 1)[0]
+            name = read_exact(stream, name_length).decode("ascii")
+            ticks = struct.unpack("<q", read_exact(stream, 8))[0]
+            id_length = read_exact(stream, 1)[0]
+            id_bytes = read_exact(stream, id_length)
+            payload_length = struct.unpack("<i", read_exact(stream, 4))[0]
+            payload = read_exact(stream, payload_length)
+            try:
+                record_id = id_bytes.decode("ascii")
+            except UnicodeDecodeError:
+                record_id = id_bytes.hex().upper()
+            if name != record.object_name:
+                raise ValueError(f"name mismatch: log={record.object_name}, dorec={name}")
+            if ticks != record.dotnet_ticks:
+                raise ValueError(f"tick mismatch: log={record.dotnet_ticks}, dorec={ticks}")
+            if payload_length != record.payload_length:
+                raise ValueError(
+                    f"payload mismatch: log={record.payload_length}, dorec={payload_length}"
+                )
+            if record_id.upper() != record.log_record_id.upper():
+                raise ValueError(
+                    f"record id mismatch: log={record.log_record_id}, dorec={record_id}"
+                )
+            yield record, payload
+    finally:
+        for _path, stream in open_files.values():
+            stream.close()