Datasets:
File size: 8,484 Bytes
e1ced61 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 | """SQLite fingerprint index for decontamination (``docs/02`` §5, ``docs/07`` §5).
Two indexes are produced by the ``fingerprint`` command:
- ``eval.sqlite`` — fingerprints of the frozen evaluation registry
- ``train_candidates.sqlite`` — fingerprints of every normalized train source
The schema is intentionally simple and deterministic: one ``fingerprints`` row
per base_id plus one ``images`` row per (base_id, image) so exact-image and
pHash near-duplicate joins are plain indexed SQL. The ``decontaminate`` command
reads both indexes and never mutates them.
OCR MinHash is stored as a nullable column populated later by the P5 OCR pass
(:func:`update_ocr_minhash`); at P3 it is empty for every row.
"""
from __future__ import annotations
import json
import sqlite3
from collections.abc import Iterable, Iterator, Mapping, Sequence
from dataclasses import dataclass
from pathlib import Path
from typing import Any
from .fingerprints import Fingerprints
_SCHEMA = """
CREATE TABLE IF NOT EXISTS fingerprints (
base_id TEXT PRIMARY KEY,
source TEXT NOT NULL,
source_revision TEXT NOT NULL,
config TEXT NOT NULL,
split TEXT NOT NULL,
native_id TEXT NOT NULL,
policy TEXT NOT NULL,
question_sha256 TEXT NOT NULL,
choices_sha256 TEXT NOT NULL,
question_canonical_sha256 TEXT NOT NULL,
question_minhash TEXT NOT NULL,
choice_minhash TEXT NOT NULL,
ocr_minhash TEXT NOT NULL DEFAULT '[]',
derived_from_eval INTEGER NOT NULL DEFAULT 0,
raw_json TEXT NOT NULL
);
CREATE TABLE IF NOT EXISTS images (
base_id TEXT NOT NULL,
image_sha256 TEXT NOT NULL,
phash TEXT NOT NULL,
PRIMARY KEY (base_id, image_sha256)
);
CREATE INDEX IF NOT EXISTS idx_images_sha ON images(image_sha256);
CREATE INDEX IF NOT EXISTS idx_images_phash ON images(phash);
CREATE INDEX IF NOT EXISTS idx_fp_source ON fingerprints(source);
"""
def _connect(db_path: str | Path) -> sqlite3.Connection:
path = Path(db_path)
path.parent.mkdir(parents=True, exist_ok=True)
conn = sqlite3.connect(str(path))
conn.execute("PRAGMA journal_mode=WAL")
conn.executescript(_SCHEMA)
return conn
def _signature_json(sig: Sequence[int]) -> str:
return json.dumps(list(sig))
def write_fingerprints(db_path: str | Path, fingerprints: Iterable[Fingerprints]) -> int:
"""(Re)build the index from ``fingerprints``; return the row count.
The index is a derived artifact, so this overwrites any prior contents.
"""
conn = _connect(db_path)
try:
with conn:
conn.execute("DELETE FROM fingerprints")
conn.execute("DELETE FROM images")
count = 0
for fp in fingerprints:
conn.execute(
"""INSERT INTO fingerprints
(base_id, source, source_revision, config, split, native_id,
policy, question_sha256, choices_sha256, question_canonical_sha256,
question_minhash, choice_minhash, ocr_minhash, derived_from_eval,
raw_json)
VALUES (?,?,?,?,?,?,?,?,?,?,?,?,?,?,?)""",
(
fp.base_id,
fp.source,
fp.source_revision,
fp.config,
fp.split,
fp.native_id,
fp.policy,
fp.question_sha256,
fp.choices_sha256,
fp.question_canonical_sha256,
_signature_json(fp.question_minhash),
_signature_json(fp.choice_minhash),
_signature_json(fp.ocr_minhash),
int(fp.derived_from_eval),
json.dumps(fp.raw, sort_keys=True),
),
)
for sha, phash in zip(fp.image_sha256, fp.image_phash, strict=True):
conn.execute(
"INSERT OR REPLACE INTO images (base_id, image_sha256, phash) "
"VALUES (?,?,?)",
(fp.base_id, sha, phash),
)
# Text-only items still get one placeholder image row keyed by their
# (single) sha so nothing is silently dropped; text items have none.
count += 1
return count
finally:
conn.close()
def update_ocr_minhash(db_path: str | Path, base_id: str, ocr_signature: Sequence[int]) -> bool:
"""Set the OCR MinHash for one base_id (P5 OCR enrichment). Returns whether updated."""
conn = _connect(db_path)
try:
with conn:
cur = conn.execute(
"UPDATE fingerprints SET ocr_minhash = ? WHERE base_id = ?",
(_signature_json(ocr_signature), base_id),
)
return cur.rowcount > 0
finally:
conn.close()
@dataclass(frozen=True)
class LoadedFingerprints:
base_id: str
source: str
source_revision: str
config: str
split: str
native_id: str
policy: str
question_sha256: str
choices_sha256: str
question_canonical_sha256: str
image_sha256: tuple[str, ...]
image_phash: tuple[str, ...]
question_minhash: tuple[int, ...]
choice_minhash: tuple[int, ...]
ocr_minhash: tuple[int, ...]
derived_from_eval: bool
def _load_images(conn: sqlite3.Connection, base_id: str) -> tuple[tuple[str, ...], tuple[str, ...]]:
rows = conn.execute(
"SELECT image_sha256, phash FROM images WHERE base_id = ?", (base_id,)
).fetchall()
shas = tuple(r[0] for r in rows)
phashes = tuple(r[1] for r in rows)
return shas, phashes
def _parse_sig(value: str) -> tuple[int, ...]:
if not value:
return ()
return tuple(int(x) for x in json.loads(value))
def load_all(db_path: str | Path) -> list[LoadedFingerprints]:
"""Load every fingerprint from the index, keyed by base_id (sorted)."""
conn = sqlite3.connect(str(db_path))
try:
rows = conn.execute(
"""SELECT base_id, source, source_revision, config, split, native_id,
policy, question_sha256, choices_sha256, question_canonical_sha256,
question_minhash, choice_minhash, ocr_minhash, derived_from_eval
FROM fingerprints ORDER BY base_id"""
).fetchall()
result: list[LoadedFingerprints] = []
for r in rows:
shas, phashes = _load_images(conn, r[0])
result.append(
LoadedFingerprints(
base_id=r[0],
source=r[1],
source_revision=r[2],
config=r[3],
split=r[4],
native_id=r[5],
policy=r[6],
question_sha256=r[7],
choices_sha256=r[8],
question_canonical_sha256=r[9],
image_sha256=shas,
image_phash=phashes,
question_minhash=_parse_sig(r[10]),
choice_minhash=_parse_sig(r[11]),
ocr_minhash=_parse_sig(r[12]),
derived_from_eval=bool(r[13]),
)
)
return result
finally:
conn.close()
def eval_image_shas(db_path: str | Path) -> set[str]:
"""All image SHA-256 present in an index (for exact-image reject)."""
conn = sqlite3.connect(str(db_path))
try:
return {row[0] for row in conn.execute("SELECT DISTINCT image_sha256 FROM images")}
finally:
conn.close()
def row_count(db_path: str | Path) -> int:
conn = sqlite3.connect(str(db_path))
try:
return int(conn.execute("SELECT COUNT(*) FROM fingerprints").fetchone()[0])
finally:
conn.close()
def iter_rows(db_path: str | Path) -> Iterator[Mapping[str, Any]]:
"""Yield raw source rows (the ``raw_json`` column) for audit/rebuild."""
conn = sqlite3.connect(str(db_path))
try:
for raw in conn.execute("SELECT raw_json FROM fingerprints ORDER BY base_id"):
yield json.loads(raw[0])
finally:
conn.close()
|