Summarize joined GPS route results

This commit is contained in:
2026-06-20 21:33:08 -05:00
parent 54f014b74d
commit bf72dcf6e3
3 changed files with 320 additions and 47 deletions
+72 -27
View File
@@ -4,12 +4,12 @@ Current date/time context: 2026-06-20/21, local timezone America/Chicago.
## Current Installed State
- Latest commit: `64ddf7a Trace GPS provider filters`.
- Latest installed probe commit: `54f014b Trace full GPS query results`.
- Installed RED4ext plugin:
`Cyberpunk 2077/red4ext/plugins/EdgeWeightGPS/EdgeWeightGPS.dll`.
- Installed build is read-only. It should not alter route behavior.
- Plugin log is cleared and empty before the next test:
`Cyberpunk 2077/red4ext/plugins/EdgeWeightGPS/EdgeWeightGPS.log`.
- Latest archived test log:
`logs/EdgeWeightGPS_query_route_probe_2308.log`.
- Game is cold as of the latest user message.
## New Toolbox Tools
@@ -32,6 +32,7 @@ In `red4ext/EdgeWeightGPS/src/Main.cpp`:
- `kEnableGpsTraceHooks = true`
- `kEnableGpsQueryLifecycleHooks = true`
- `kEnableGpsRouteJobLifecycleHooks = false`
- `kEnableGpsRouteJobStartTrace = false`
- `kEnableGpsLocalSearchHooks = true`
- `kEnableRouteUiTraceHooks = false`
@@ -113,37 +114,81 @@ Route result records:
- `h00` in those records matches `worldTrafficLanePersistent.nodeRefHash`.
- Route records can be joined to `work/raw-segment-json/all.traffic_persistent.json`.
- The low byte of route-record `u08` is not the same as the search point class.
- The log's human wall timestamp has a millisecond rollover bug around second
boundaries. Use the `+Nms` elapsed field for ordering and durations.
## Next User Test
## Latest 23:08 Test Result
Ask the user to run the usual routine:
The user clicked:
1. Launch game.
2. Hit Continue on main menu.
3. Hit Space to Continue.
4. Wait for minimap/GPS to settle.
5. Open map.
6. Plot the usual three routes plus one custom highway-ish target.
7. Close game.
- 23:08 continue on main menu
- 23:30 space to continue
- 23:46 map open
- 23:56 side job
- 24:06 Sinnerman
- 24:14 Claire
- 24:21 custom
- 24:30 Aldecaldo camp
- 24:47 far gig east
- 25:03 Violence
- 25:10 cyberpsycho unfinished highway
Ask for timestamps for:
Use:
- Space-to-Continue
- map open
- each route click
- game cold
```bash
python3 tools/summarize_gps_route_records.py \
--traffic-json work/raw-segment-json/all.traffic_persistent.json \
logs/EdgeWeightGPS_query_route_probe_2308.log
```
After the run:
Important result table:
1. Archive the log to `logs/EdgeWeightGPS_provider_filter_probe_<time>.log`.
2. Inspect `GPSQuerySubmit`/`GPSQueryResultFetch` lines for tracked route IDs
from submit return RVA `0x8d20d4`, especially successful `0x7094b8`
records with `gpsResult_ptr28_rec40`.
3. Inspect `GPSSearch result` lines for `filterPassClasses`,
`filterFailClasses`, `filterPassMaskBits`, and `filterFailMaskBits`.
4. Compare filter pass/fail classes to `edgeSrcClasses` and route output
records. The key question is whether the class/mask filter is selecting the
graph subset before cost weighting can matter.
- qid 1: startup/minimap route, 18 records, 50 ms, 1216.5 pavement length.
- qid 2: side job, 12 records, 32 ms, 578.6 pavement length.
- qid 3: Sinnerman, 23 records, 45 ms, 897.4 pavement, 160.4 highway,
89.8 road.
- qid 4: Claire, 18 records, 34 ms, 1216.5 pavement.
- qid 5: custom, 17 records, 44 ms, 1006.3 pavement, 241.0 highway,
38.0 road.
- qid 6: Aldecaldo camp, 80 records, 325 ms, 5853.5 road, 1492.9 highway,
1111.5 pavement, 24.6 gpsonly.
- qid 7: far east gig, 81 records, 144 ms, 5086.2 road, 1523.0 highway,
1111.5 pavement, 24.6 gpsonly.
- qid 8: Violence, 55 records, 97 ms, 1991.5 pavement, 358.5 highway,
86.1 road.
- qid 9: cyberpsycho unfinished highway, 58 records, 147 ms, 2657.4 pavement,
519.0 highway, 113.2 road, 56.7 gpsonly.
All 362 route-record handles matched traffic lane hashes in the extracted
resource.
Local search/provider aggregate from this run:
- 159 searches, 4036 provider filter calls, only 3 filter failures.
- Filter pass classes: `3:3114`, `5:451`, `15:310`, `1:116`, `4:42`.
- Edge source classes: `3:3244`, `5:487`, `15:318`, `1:117`, `4:58`.
- Edge return RVAs: `0x44f457:4033`, `0x44f629:151`,
`0x8d74d3:30`, `0x8d76b0:10`.
Inference:
- Provider vtable slot `+0x08` filtering is almost not pruning these route
searches, so it is not the dominant reason class-cost patches failed to
visibly steer GPS.
- Search-side edge classes are only `{1,3,4,5,15}` in this run. Final emitted
route-record classes include `{0,1,2,3,4,5,7,8,9}` and class 7/8/9 are
highway-heavy. Therefore final route-record `u08` low byte and search point
`point+0x13` class are different domains.
## Next Work
- Static disassembly around result-drain/producer RVAs `0x520783`,
`0x41be2c`, `0x41be94`, `0x70a42c`, and `0x8d20d4`.
- The plausible patch point is upstream of result copy: find where search
output is converted to 0x28 route records or where candidate lanes are ranked
with enough lane-hash/resource context to prefer highway/major-road lanes.
- Post-`0x7094b8` rewriting would likely only change displayed path records and
risks desynchronizing instructions from the real planner.
## Useful Commands
+31
View File
@@ -264,6 +264,37 @@ Route-record/resource correlation:
provider initialization site or a way to join search-side route points back to
`nodeRefHash`/lane flags before the edge cost is returned.
Latest full-route join:
- `logs/EdgeWeightGPS_query_route_probe_2308.log` contains nine solved route
results from the 23:08 user test. All 362 emitted route records matched
`worldTrafficLanePersistent.nodeRefHash` in
`work/raw-segment-json/all.traffic_persistent.json`.
- The log's wall timestamp currently wraps milliseconds without incrementing
the displayed second. Use the `+Nms` elapsed field when comparing submit and
result times.
- Representative solved route composition:
- side job: 12 records, 32 ms, 578.6 pavement length.
- Sinnerman: 23 records, 45 ms, 897.4 pavement, 160.4 highway, 89.8 road.
- Claire: 18 records, 34 ms, 1216.5 pavement.
- custom highway-ish pin: 17 records, 44 ms, 1006.3 pavement,
241.0 highway, 38.0 road.
- Aldecaldo camp: 80 records, 325 ms, 5853.5 road, 1492.9 highway,
1111.5 pavement, 24.6 gpsonly.
- far east gig: 81 records, 144 ms, 5086.2 road, 1523.0 highway,
1111.5 pavement, 24.6 gpsonly.
- Violence: 55 records, 97 ms, 1991.5 pavement, 358.5 highway,
86.1 road.
- cyberpsycho unfinished highway: 58 records, 147 ms, 2657.4 pavement,
519.0 highway, 113.2 road, 56.7 gpsonly.
- Search/provider aggregate in the same log: 159 searches, 4036 provider
filter calls, only 3 filter failures. Filter slot `+0x08` is therefore
almost never pruning this test graph.
- Search-side point classes observed by `0x44f838` are `{1,3,4,5,15}`. Final
route-record low-byte classes are `{0,1,2,3,4,5,7,8,9}`, and route-record
classes 7/8/9 are highway-heavy. These are different domains, which explains
why provider point-class tuning did not map cleanly to highway preference.
Current static producer lead:
- `0x70a908` packages the route endpoints and query settings, then calls
+217 -20
View File
@@ -5,15 +5,21 @@ from __future__ import annotations
import argparse
import collections
import dataclasses
import json
import re
from pathlib import Path
from typing import Any
RE_TIMESTAMP = re.compile(r"^(\d{4}-\d\d-\d\d \d\d:\d\d:\d\d\.\d{3})")
RE_ELAPSED_MS = re.compile(r" \+(\d+)ms ")
RE_RECORD_BLOCK = re.compile(r"gpsResult_ptr28_rec40=\[(.*?)\]")
RE_RECORD = re.compile(r"(\d+):([0-9a-f]+),0x([0-9a-f]{8}),([^;\]]+)")
RE_JOB_COUNT = re.compile(r"GPSRouteJobBuild result call=(\d+).*?job_count50=(\d+)/0x([0-9a-f]+)")
RE_FIELD_DEC_HEX = re.compile(r"([A-Za-z0-9_]+)=(-?\d+)/0x([0-9a-fA-F]+)")
RE_FIELD_HEX = re.compile(r"([A-Za-z0-9_]+)=0x([0-9a-fA-F]+)")
RE_FIELD_DEC = re.compile(r"([A-Za-z0-9_]+)=(-?\d+)(?= |$)")
FLAGS = {
"FromRoadSpline": 1,
@@ -35,6 +41,39 @@ FLAGS = {
}
@dataclasses.dataclass
class SubmitInfo:
line_no: int
timestamp: str
elapsed_ms: int | None
call: int
fields: dict[str, int]
@dataclasses.dataclass
class RouteRecord:
index: int
handle: int
packed: int
parts: list[str]
@property
def low_class(self) -> int:
return self.packed & 0xFF
@dataclasses.dataclass
class RouteResult:
line_no: int
timestamp: str
elapsed_ms: int | None
query_id: int | None
submit_call: int | None
submit_ret_rva: int | None
fields: dict[str, int]
records: list[RouteRecord]
def bytes_le(value: int) -> tuple[int, int, int, int]:
return tuple((value >> (8 * index)) & 0xFF for index in range(4))
@@ -57,6 +96,27 @@ def lane_category(flags: int) -> str:
return "other"
def parse_timestamp(line: str) -> str:
match = RE_TIMESTAMP.search(line)
return match.group(1) if match else "<unknown>"
def parse_elapsed_ms(line: str) -> int | None:
match = RE_ELAPSED_MS.search(line)
return int(match.group(1)) if match else None
def parse_fields(line: str) -> dict[str, int]:
fields: dict[str, int] = {}
for match in RE_FIELD_DEC_HEX.finditer(line):
fields[match.group(1)] = int(match.group(2))
for match in RE_FIELD_HEX.finditer(line):
fields.setdefault(match.group(1), int(match.group(2), 16))
for match in RE_FIELD_DEC.finditer(line):
fields.setdefault(match.group(1), int(match.group(2)))
return fields
def load_lane_lookup(path: Path | None) -> dict[int, dict[str, Any]]:
if not path:
return {}
@@ -79,7 +139,158 @@ def load_lane_lookup(path: Path | None) -> dict[int, dict[str, Any]]:
return lookup
def parse_route_records(block: str) -> list[RouteRecord]:
records: list[RouteRecord] = []
for record_match in RE_RECORD.finditer(block):
pieces = record_match.group(4).split(",")
if len(pieces) < 4:
continue
records.append(
RouteRecord(
index=int(record_match.group(1)),
handle=int(record_match.group(2), 16),
packed=int(record_match.group(3), 16),
parts=pieces,
)
)
return records
def iter_route_results(path: Path) -> tuple[dict[int, SubmitInfo], list[RouteResult], list[int]]:
submits: dict[int, SubmitInfo] = {}
results: list[RouteResult] = []
job_counts: list[int] = []
for line_no, line in enumerate(path.read_text(encoding="utf-8", errors="replace").splitlines(), start=1):
fields = parse_fields(line)
if "hook GPSQuerySubmit 0x70a42c" in line:
call = fields.get("call")
if call is not None:
submits[call] = SubmitInfo(
line_no=line_no,
timestamp=parse_timestamp(line),
elapsed_ms=parse_elapsed_ms(line),
call=call,
fields=fields,
)
job_match = RE_JOB_COUNT.search(line)
if job_match:
job_counts.append(int(job_match.group(2)))
block_match = RE_RECORD_BLOCK.search(line)
if not block_match:
continue
records = parse_route_records(block_match.group(1))
if not records:
continue
results.append(
RouteResult(
line_no=line_no,
timestamp=parse_timestamp(line),
elapsed_ms=parse_elapsed_ms(line),
query_id=fields.get("queryId"),
submit_call=fields.get("submitCall"),
submit_ret_rva=fields.get("submitRetRva"),
fields=fields,
records=records,
)
)
return submits, results, job_counts
def summarize_route_result(route: RouteResult, lane_lookup: dict[int, dict[str, Any]]) -> tuple[dict[str, Any], int]:
class_counts = collections.Counter(record.low_class for record in route.records)
category_counts: collections.Counter[str] = collections.Counter()
category_lengths: collections.Counter[str] = collections.Counter()
speed_lengths: collections.Counter[int] = collections.Counter()
unmatched = 0
for record in route.records:
lane = lane_lookup.get(record.handle)
if not lane:
unmatched += 1
continue
category = lane["category"]
length = float(lane.get("length") or 0.0)
category_counts[category] += 1
category_lengths[category] += length
speed_lengths[int(lane["speed"])] += length
total_length = sum(category_lengths.values())
summary = {
"class_counts": class_counts,
"category_counts": category_counts,
"category_lengths": category_lengths,
"speed_lengths": speed_lengths,
"total_length": total_length,
"unmatched": unmatched,
}
return summary, len(route.records) - unmatched
def format_counter(counter: collections.Counter[Any], limit: int | None = None, digits: int | None = None) -> str:
items = counter.most_common(limit)
parts = []
for key, value in items:
if digits is None:
parts.append(f"{key}:{value}")
else:
parts.append(f"{key}:{value:.{digits}f}")
return "{" + ", ".join(parts) + "}"
def print_per_route(
path: Path,
submits: dict[int, SubmitInfo],
route_results: list[RouteResult],
lane_lookup: dict[int, dict[str, Any]],
) -> None:
print(" per-route results:")
for ordinal, route in enumerate(route_results, start=1):
submit = submits.get(route.submit_call) if route.submit_call is not None else None
summary, matched = summarize_route_result(route, lane_lookup)
span = route.fields.get("gpsResult_u20")
submit_fields = submit.fields if submit else {}
ret_rva = route.submit_ret_rva
ret_text = f"0x{ret_rva:x}" if ret_rva is not None else "?"
submit_ms = submit.elapsed_ms if submit else None
result_ms = route.elapsed_ms
duration_ms = result_ms - submit_ms if submit_ms is not None and result_ms is not None else None
submit_marker = f"+{submit_ms}ms" if submit_ms is not None else "?"
result_marker = f"+{result_ms}ms" if result_ms is not None else "?"
duration_marker = f"{duration_ms}ms" if duration_ms is not None else "?"
print(
f" route#{ordinal} qid={route.query_id} submitCall={route.submit_call} "
f"submit={submit_marker} result={result_marker} duration={duration_marker} line={route.line_no} "
f"ret={ret_text} records={len(route.records)} spanU20={span} matched={matched}"
)
if submit_fields:
print(
" query "
f"f08=0x{submit_fields.get('query_f08', 0):x} "
f"f0c={submit_fields.get('query_f0c')} "
f"fcc={submit_fields.get('query_fcc')}"
)
if lane_lookup:
category_lengths = summary["category_lengths"]
total = summary["total_length"]
print(
" resource "
f"segments={format_counter(summary['category_counts'])} "
f"length={format_counter(category_lengths, digits=1)} "
f"total={total:.1f} "
f"speedLength={format_counter(summary['speed_lengths'], limit=6, digits=1)} "
f"unmatched={summary['unmatched']}"
)
print(f" routeClasses={format_counter(summary['class_counts'])}")
def summarize(path: Path, lane_lookup: dict[int, dict[str, Any]]) -> None:
submits, route_results, job_counts = iter_route_results(path)
routes = 0
records = 0
first_byte = collections.Counter()
@@ -91,37 +302,22 @@ def summarize(path: Path, lane_lookup: dict[int, dict[str, Any]]) -> None:
class_speed: dict[int, collections.Counter[int]] = collections.defaultdict(collections.Counter)
matched_handles = 0
unmatched_handles = 0
job_counts: list[int] = []
for line in path.read_text(encoding="utf-8", errors="replace").splitlines():
job_match = RE_JOB_COUNT.search(line)
if job_match:
job_counts.append(int(job_match.group(2)))
block_match = RE_RECORD_BLOCK.search(line)
if not block_match:
continue
for route in route_results:
route_records = 0
for record_match in RE_RECORD.finditer(block_match.group(1)):
handle = record_match.group(2)
packed = int(record_match.group(3), 16)
pieces = record_match.group(4).split(",")
if len(pieces) < 4:
continue
for record in route.records:
route_records += 1
records += 1
bytes_tuple = bytes_le(packed)
bytes_tuple = bytes_le(record.packed)
class_id = bytes_tuple[0]
patterns[bytes_tuple] += 1
first_byte[class_id] += 1
for index, byte in enumerate(bytes_tuple):
byte_positions[index][byte] += 1
segments[(handle, class_id)] += 1
segments[(record.handle, class_id)] += 1
if lane_lookup:
lane = lane_lookup.get(int(handle, 16))
lane = lane_lookup.get(record.handle)
if lane:
matched_handles += 1
class_categories[class_id][lane["category"]] += 1
@@ -155,6 +351,7 @@ def summarize(path: Path, lane_lookup: dict[int, dict[str, Any]]) -> None:
speeds = dict(class_speed[class_id].most_common(8))
flags = dict(class_flags[class_id].most_common(8))
print(f" {class_id}: categories={categories} speeds={speeds} flags={flags}")
print_per_route(path, submits, route_results, lane_lookup)
for index, counter in enumerate(byte_positions):
print(f" byte{index} counts:", dict(counter.most_common(12)))
print(" packed metadata patterns:")