codehygiene / server /telemetry.py
brianbellx's picture
Remove Most improved, score every ranked file, and feed ranked files into stats
419a527 verified
Raw History Blame Contribute Delete
15.4 kB
"""Aggregate, code-free scoring counters for the Space. See space/TELEMETRY.md."""
from __future__ import annotations
import asyncio
from collections import OrderedDict
from collections.abc import Iterable, Mapping
from datetime import datetime, timezone
import hashlib
import hmac
import json
import logging
from pathlib import Path
import secrets
import time
from typing import Annotated, Any, Literal, Protocol
from pydantic import BaseModel, ConfigDict, Field
log = logging.getLogger("space.telemetry")
DIGEST_HEX = 16
SEEN_LIMIT = 100_000
STATS_VERSION = 2
STATS_FILE = "stats.json"
SQUIGGLE_KINDS = (
"deep_nesting",
"poor_naming",
"noisy_comment",
"swallowed_error",
"dense_line",
"special_casing",
"duplicate_block",
"cramped_layout",
)
LOC_BUCKETS = ("0-50", "50-100", "100-300", "300-1000", "1000+")
CODE_EXTENSIONS = frozenset({
"py", "pyi", "js", "mjs", "cjs", "ts", "tsx", "jsx", "c", "h", "cc", "cpp", "cxx",
"hpp", "rs", "go", "zig", "java", "cs", "lua", "rb", "php", "swift", "kt", "scala", "sh",
})
OTHER_EXT = "other"
SCORE_KEYS = frozenset(str(points) for points in range(101))
MS_STEPS_PER_DECADE = 8
MS_DECADES = 4
def _two_figures(value: float) -> int:
"""Round to two significant figures so bucket edges read as plain numbers."""
digits = len(str(int(value)))
return int(round(value, 2 - digits))
MS_EDGES = tuple(sorted({
_two_figures(10 ** (step / MS_STEPS_PER_DECADE))
for step in range(MS_STEPS_PER_DECADE * MS_DECADES + 1)
}))
MS_OVERFLOW = f"{MS_EDGES[-1]}+"
MS_KEYS = frozenset(str(edge) for edge in MS_EDGES) | {MS_OVERFLOW}
SourceKind = Literal["examples", "repo", "folder", "snippet"]
SquiggleKind = Literal[
"deep_nesting", "poor_naming", "noisy_comment", "swallowed_error",
"dense_line", "special_casing", "duplicate_block", "cramped_layout",
]
Count = Annotated[int, Field(ge=0, le=10_000)]
COUNTERS = ("files_scored", "scan_batches", "score_scored_sum")
TALLIES = (
"scored_by_ext",
"score_sum_by_ext",
"scored_by_loc_bucket",
"squiggles_by_kind",
"score_hist",
"score_ms_hist",
)
TALLY_KEYS: dict[str, frozenset[str]] = {
"scored_by_ext": CODE_EXTENSIONS | {OTHER_EXT},
"score_sum_by_ext": CODE_EXTENSIONS | {OTHER_EXT},
"scored_by_loc_bucket": frozenset(LOC_BUCKETS),
"squiggles_by_kind": frozenset(SQUIGGLE_KINDS),
"score_hist": SCORE_KEYS,
"score_ms_hist": MS_KEYS,
}
class ScanFile(BaseModel):
model_config = ConfigDict(extra="forbid")
digest16: str = Field(pattern=r"^[0-9a-f]{16}$")
ext: str = Field("", max_length=8)
loc_bucket: Literal["0-50", "50-100", "100-300", "300-1000", "1000+"]
score: int = Field(ge=0, le=100)
squiggles: dict[SquiggleKind, Count] = Field(default_factory=dict)
score_ms: int | None = Field(None, ge=0, le=60_000)
class ScanBody(BaseModel):
model_config = ConfigDict(extra="forbid")
source_kind: SourceKind
files: list[ScanFile] = Field(max_length=200)
def extension_of(filename: str) -> str:
"""Lower-case extension of the base name, matching the browser computation."""
base = filename.rsplit("/", 1)[-1]
stem, dot, ext = base.rpartition(".")
return ext.lower() if dot and stem else ""
def known_ext(ext: str) -> str:
"""Bounded tally key: an extension from the fixed code set, or `other`."""
return ext if ext in CODE_EXTENSIONS else OTHER_EXT
def content_digest(ext: str, content: str) -> str:
"""First 64 bits of SHA-256(ext + NUL + content), matching the browser digest."""
payload = f"{ext}\0{content}".encode("utf-8", errors="replace")
return hashlib.sha256(payload).hexdigest()[:DIGEST_HEX]
def loc_bucket_of(content: str) -> Literal["0-50", "50-100", "100-300", "300-1000", "1000+"]:
lines = content.count("\n") + 1
if lines < 50:
return "0-50"
if lines < 100:
return "50-100"
if lines < 300:
return "100-300"
return "300-1000" if lines < 1000 else "1000+"
def scan_file_from(
filename: str,
content: str,
score: float,
squiggles: Mapping[str, int],
score_ms: int | None = None,
) -> ScanFile:
raw_ext = extension_of(filename)
ext = raw_ext if len(raw_ext) <= 8 else ""
points = min(100, max(0, round(score * 100)))
allowed = TALLY_KEYS["squiggles_by_kind"]
clean_squiggles = {
kind: min(10_000, count)
for kind, count in squiggles.items()
if kind in allowed and count > 0
}
clamped_ms = min(60_000, max(0, score_ms)) if score_ms is not None else None
return ScanFile.model_construct(
digest16=content_digest(ext, content),
ext=ext,
loc_bucket=loc_bucket_of(content),
score=points,
squiggles=clean_squiggles,
score_ms=clamped_ms,
)
class DigestSet:
"""Rolling set of the most recent 64-bit digests, oldest dropped first."""
def __init__(self, limit: int = SEEN_LIMIT) -> None:
self.limit = limit
self._items: OrderedDict[int, None] = OrderedDict()
def add(self, digest: str) -> bool:
key = int(digest, 16)
if key in self._items:
self._items.move_to_end(key)
return False
self._items[key] = None
if len(self._items) > self.limit:
self._items.popitem(last=False)
return True
def __len__(self) -> int:
return len(self._items)
def empty_stats() -> dict[str, Any]:
stats: dict[str, Any] = {"version": STATS_VERSION}
stats.update({name: 0 for name in COUNTERS})
stats.update({name: {} for name in TALLIES})
return stats
def _clean_count(value: Any) -> int:
return value if isinstance(value, int) and not isinstance(value, bool) and value >= 0 else 0
def sanitize_stats(raw: Any) -> dict[str, Any]:
"""Keep only known scoring counters and bounded tally keys from a stats file."""
stats = empty_stats()
if not isinstance(raw, dict):
return stats
for name in COUNTERS:
stats[name] = _clean_count(raw.get(name))
for name in TALLIES:
tally = raw.get(name)
if name == "squiggles_by_kind" and not isinstance(tally, dict):
tally = raw.get("squiggles_before_by_kind")
if isinstance(tally, dict):
allowed = TALLY_KEYS[name]
stats[name] = {key: _clean_count(val) for key, val in tally.items() if key in allowed}
return stats
def _bump(tally: dict[str, int], key: str, amount: int = 1) -> None:
if amount > 0:
tally[key] = tally.get(key, 0) + amount
def _mean(total: int, count: int) -> float | None:
return round(total / count, 1) if count else None
def ms_bucket(ms: int) -> str:
"""Histogram key for a scoring time: the smallest edge at or above it."""
edge = next((edge for edge in MS_EDGES if ms <= edge), None)
return str(edge) if edge is not None else MS_OVERFLOW
def _median_key(tally: Mapping[str, int], keys: Iterable[str]) -> str | None:
"""Key holding the middle sample when keys are walked in order."""
total = sum(tally.values())
seen = 0
for key in keys:
seen += tally.get(key, 0)
if total and seen * 2 >= total:
return key
return None
def median_score(tally: Mapping[str, int]) -> int | None:
key = _median_key(tally, (str(points) for points in range(101)))
return int(key) if key is not None else None
def median_ms(tally: Mapping[str, int]) -> int | None:
key = _median_key(tally, [*(str(edge) for edge in MS_EDGES), MS_OVERFLOW])
return int(key.rstrip("+")) if key is not None else None
class Telemetry:
"""Aggregate scoring counters. Mutated only on the event loop; zero code, names or IPs."""
def __init__(self, excluded: Iterable[str] = ()) -> None:
self.stats = empty_stats()
self.dirty = False
self.excluded = frozenset(excluded)
self.seen_scored = DigestSet()
def merge(self, raw: Any) -> None:
loaded = sanitize_stats(raw)
for name in COUNTERS:
self.stats[name] += loaded[name]
for name in TALLIES:
for key, value in loaded[name].items():
_bump(self.stats[name], key, value)
def record_files(self, files: Iterable[ScanFile]) -> int:
counted = 0
for item in files:
if item.digest16 in self.excluded or not self.seen_scored.add(item.digest16):
continue
counted += 1
ext = known_ext(item.ext)
self.stats["score_scored_sum"] += item.score
_bump(self.stats["scored_by_ext"], ext)
_bump(self.stats["score_sum_by_ext"], ext, item.score)
_bump(self.stats["scored_by_loc_bucket"], item.loc_bucket)
_bump(self.stats["score_hist"], str(item.score))
if item.score_ms is not None:
_bump(self.stats["score_ms_hist"], ms_bucket(item.score_ms))
for kind, amount in item.squiggles.items():
_bump(self.stats["squiggles_by_kind"], kind, amount)
self.stats["files_scored"] += counted
self.stats["scan_batches"] += 1
self.dirty = True
return counted
def record_scan(self, body: ScanBody) -> int:
if body.source_kind == "examples":
return 0
return self.record_files(body.files)
def snapshot(self) -> str:
stats = {**self.stats}
stats["updated_at"] = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
return json.dumps(stats, indent=2, sort_keys=True) + "\n"
def _by_ext(self) -> dict[str, dict[str, Any]]:
stats = self.stats
rows = {}
for ext in sorted(stats["scored_by_ext"]):
scored = stats["scored_by_ext"].get(ext, 0)
rows[ext] = {
"scored": scored,
"avg_score": _mean(stats["score_sum_by_ext"].get(ext, 0), scored),
}
return rows
def public(self) -> dict[str, Any]:
stats = self.stats
squiggles = stats["squiggles_by_kind"]
return {
"files_scored": stats["files_scored"],
"issues_found": sum(squiggles.values()),
"avg_score_scored": _mean(stats["score_scored_sum"], stats["files_scored"]),
"median_score": median_score(stats["score_hist"]),
"score_ms_median": median_ms(stats["score_ms_hist"]),
"score_ms_samples": sum(stats["score_ms_hist"].values()),
"by_ext": self._by_ext(),
"by_loc_bucket": dict(stats["scored_by_loc_bucket"]),
"squiggles_by_kind": dict(squiggles),
}
def preset_digests(presets_path: Path | None) -> set[str]:
"""Digests of the bundled examples, which never count."""
if presets_path is None or not presets_path.is_file():
return set()
rows = json.loads(presets_path.read_text(encoding="utf-8"))
return {content_digest(extension_of(row["filename"]), row["before"]) for row in rows}
class RateLimiter:
"""Token bucket keyed by a salted hash of the caller; the address itself is never kept."""
def __init__(self, per_minute: int, idle_seconds: float = 120.0) -> None:
self.per_minute = per_minute
self.idle_seconds = idle_seconds
self._salt = secrets.token_bytes(16)
self._buckets: dict[bytes, tuple[float, float]] = {}
def allow(self, caller: str) -> bool:
key = hmac.new(self._salt, caller.encode(), hashlib.sha256).digest()[:8]
now = time.monotonic()
tokens, updated = self._buckets.get(key, (float(self.per_minute), now))
tokens = min(float(self.per_minute), tokens + (now - updated) * self.per_minute / 60.0)
allowed = tokens >= 1.0
self._buckets[key] = (tokens - 1.0 if allowed else tokens, now)
if len(self._buckets) > 10_000:
self._prune(now)
return allowed
def _prune(self, now: float) -> None:
stale = [key for key, (_, updated) in self._buckets.items() if now - updated > self.idle_seconds]
for key in stale:
del self._buckets[key]
class StatsStore:
"""Reads and writes one JSON file locally and, when configured, in a Hugging Face Dataset repo."""
def __init__(
self, local_path: Path, dataset_repo: str = "", token: str = "", filename: str = STATS_FILE,
) -> None:
self.local_path = local_path
self.dataset_repo = dataset_repo
self.token = token
self.filename = filename
self.remote_loaded = not self.remote
@property
def remote(self) -> bool:
return bool(self.dataset_repo and self.token)
def load(self) -> dict[str, Any] | None:
"""Blocking. Returns the saved counters, or None when there is nothing to load yet."""
if self.remote:
loaded = self._load_remote()
self.remote_loaded = True
return loaded
if self.local_path.is_file():
return json.loads(self.local_path.read_text(encoding="utf-8"))
return None
def _load_remote(self) -> dict[str, Any] | None:
from huggingface_hub import hf_hub_download
from huggingface_hub.errors import EntryNotFoundError
try:
path = hf_hub_download(
self.dataset_repo,
self.filename,
repo_type="dataset",
token=self.token,
force_download=True,
)
except EntryNotFoundError:
return None
return json.loads(Path(path).read_text(encoding="utf-8"))
def save(self, text: str) -> None:
"""Blocking. Never uploads before the remote copy was read, so it cannot be overwritten."""
self.local_path.parent.mkdir(parents=True, exist_ok=True)
self.local_path.write_text(text, encoding="utf-8")
if not self.remote or not self.remote_loaded:
return
from huggingface_hub import HfApi
HfApi(token=self.token).upload_file(
path_or_fileobj=text.encode("utf-8"),
path_in_repo=self.filename,
repo_id=self.dataset_repo,
repo_type="dataset",
commit_message=f"Update {self.filename}",
)
class Persisted(Protocol):
dirty: bool
def merge(self, raw: Any) -> None: ...
def snapshot(self) -> str: ...
class SyncedFile:
"""Loads a model from its store once, then writes it back whenever it changed."""
def __init__(self, store: StatsStore, model: Persisted) -> None:
self.store = store
self.model = model
def _warn(self, action: str, error: Exception) -> None:
log.warning("%s %s failed: %s", self.store.filename, action, type(error).__name__)
async def load(self) -> None:
try:
loaded = await asyncio.to_thread(self.store.load)
except Exception as error:
self._warn("load", error)
return
self.model.merge(loaded)
async def flush(self) -> None:
if not self.store.remote_loaded:
await self.load()
if not self.model.dirty:
return
text = self.model.snapshot()
self.model.dirty = False
try:
await asyncio.to_thread(self.store.save, text)
except Exception as error:
self.model.dirty = True
self._warn("sync", error)