Spaces:
Running
Running
File size: 9,862 Bytes
3116157 ac7a28c 3116157 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 | """Sprint A14-S5 — ``MetricRegistry`` instancié explicitement.
Vérifie le contrat critique du S5 : pas de singleton global, pas
de side-effect d'import, association explicite ``MetricSpec ↔
Callable``, sélection par signature de types.
Anti-pattern testé négativement : ``import picarones.evaluation``
ne doit PAS auto-enregistrer de métrique.
"""
from __future__ import annotations
import pytest
from picarones.domain import ArtifactType, MetricSpec
from picarones.evaluation.registry import (
MetricNotFoundError,
MetricRegistrationError,
MetricRegistry,
)
def _cer(reference: str, hypothesis: str) -> float:
"""Stub CER pour les tests."""
return 0.0 if reference == hypothesis else 1.0
def _wer(reference: str, hypothesis: str) -> float:
return 0.0 if reference == hypothesis else 1.0
def _ner_f1(ref_entities: list[dict], hyp_entities: list[dict]) -> float:
return 1.0
# ──────────────────────────────────────────────────────────────────────
# Instanciation et état initial
# ──────────────────────────────────────────────────────────────────────
class TestEmptyRegistry:
def test_starts_empty(self) -> None:
reg = MetricRegistry()
assert len(reg) == 0
assert reg.names() == []
def test_unknown_metric_raises(self) -> None:
reg = MetricRegistry()
with pytest.raises(MetricNotFoundError):
reg.get_spec("cer")
with pytest.raises(MetricNotFoundError):
reg.get_callable("cer")
# ──────────────────────────────────────────────────────────────────────
# Enregistrement
# ──────────────────────────────────────────────────────────────────────
class TestRegistration:
def test_register_one_metric(self) -> None:
reg = MetricRegistry()
spec = MetricSpec(
name="cer",
input_types=(ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT),
)
reg.register(spec, _cer)
assert "cer" in reg
assert len(reg) == 1
assert reg.get_spec("cer") is spec
assert reg.get_callable("cer") is _cer
def test_register_non_callable_raises(self) -> None:
reg = MetricRegistry()
spec = MetricSpec(
name="cer",
input_types=(ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT),
)
with pytest.raises(MetricRegistrationError, match="callable"):
reg.register(spec, "not_a_function") # type: ignore[arg-type]
def test_duplicate_name_with_different_func_raises(self) -> None:
reg = MetricRegistry()
spec = MetricSpec(
name="cer",
input_types=(ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT),
)
reg.register(spec, _cer)
with pytest.raises(MetricRegistrationError, match="déjà enregistrée"):
reg.register(spec, _wer) # même spec, autre callable
def test_idempotent_re_registration(self) -> None:
"""Re-enregistrer la même spec + même callable est silencieux
(utile pour les tests qui re-instancient le service)."""
reg = MetricRegistry()
spec = MetricSpec(
name="cer",
input_types=(ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT),
)
reg.register(spec, _cer)
reg.register(spec, _cer) # ne lève pas
assert len(reg) == 1
# ──────────────────────────────────────────────────────────────────────
# Sélection par signature de types
# ──────────────────────────────────────────────────────────────────────
class TestSelectByTypes:
def _filled_registry(self) -> MetricRegistry:
reg = MetricRegistry()
reg.register(
MetricSpec(name="cer", input_types=(
ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT,
)),
_cer,
)
reg.register(
MetricSpec(name="wer", input_types=(
ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT,
)),
_wer,
)
reg.register(
MetricSpec(name="ner_f1", input_types=(
ArtifactType.ENTITIES, ArtifactType.ENTITIES,
), higher_is_better=True),
_ner_f1,
)
return reg
def test_select_text_text(self) -> None:
reg = self._filled_registry()
selected = reg.select(ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT)
names = sorted(s.name for s in selected)
assert names == ["cer", "wer"]
def test_select_entities(self) -> None:
reg = self._filled_registry()
selected = reg.select(ArtifactType.ENTITIES, ArtifactType.ENTITIES)
assert [s.name for s in selected] == ["ner_f1"]
def test_select_no_match(self) -> None:
reg = self._filled_registry()
selected = reg.select(ArtifactType.IMAGE, ArtifactType.IMAGE)
assert selected == []
def test_select_distinguishes_text_subtypes(self) -> None:
"""Important : RAW_TEXT et CORRECTED_TEXT sont des types distincts.
Une métrique enregistrée pour (RAW_TEXT, RAW_TEXT) ne s'applique
pas automatiquement à (CORRECTED_TEXT, RAW_TEXT)."""
reg = self._filled_registry()
selected = reg.select(ArtifactType.CORRECTED_TEXT, ArtifactType.RAW_TEXT)
assert selected == []
# ──────────────────────────────────────────────────────────────────────
# Calcul
# ──────────────────────────────────────────────────────────────────────
class TestCompute:
def test_compute_named(self) -> None:
reg = MetricRegistry()
reg.register(
MetricSpec(name="cer", input_types=(
ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT,
)),
_cer,
)
assert reg.compute("cer", "hello", "hello") == 0.0
assert reg.compute("cer", "hello", "world") == 1.0
def test_compute_unknown_raises(self) -> None:
reg = MetricRegistry()
with pytest.raises(MetricNotFoundError):
reg.compute("missing", "x", "y")
def test_compute_at_junction_runs_all_applicable(self) -> None:
reg = MetricRegistry()
reg.register(
MetricSpec(name="cer", input_types=(
ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT,
)),
_cer,
)
reg.register(
MetricSpec(name="wer", input_types=(
ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT,
)),
_wer,
)
reg.register(
MetricSpec(name="ner_f1", input_types=(
ArtifactType.ENTITIES, ArtifactType.ENTITIES,
)),
_ner_f1,
)
out = reg.compute_at_junction(
"hello", "hello",
ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT,
)
assert set(out.keys()) == {"cer", "wer"}
assert out["cer"] == 0.0
assert "ner_f1" not in out # mauvaise signature
def test_compute_at_junction_propagates_exceptions(self) -> None:
"""Le S5 ne capture pas les exceptions des métriques.
C'est l'EvaluationViewExecutor (S13) qui décidera quoi en
faire dans son ProjectionReport."""
def _broken(r: str, h: str) -> float:
raise RuntimeError("boom")
reg = MetricRegistry()
reg.register(
MetricSpec(name="broken", input_types=(
ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT,
)),
_broken,
)
with pytest.raises(RuntimeError, match="boom"):
reg.compute_at_junction(
"x", "y",
ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT,
)
# ──────────────────────────────────────────────────────────────────────
# Anti-pattern : pas de singleton global
# ──────────────────────────────────────────────────────────────────────
class TestNoGlobalSingleton:
def test_two_registries_are_independent(self) -> None:
"""Différence cruciale avec
``picarones.evaluation.metric_registry`` qui a un dict global :
deux ``MetricRegistry()`` ne se partagent rien."""
reg_a = MetricRegistry()
reg_b = MetricRegistry()
spec = MetricSpec(name="cer", input_types=(
ArtifactType.RAW_TEXT, ArtifactType.RAW_TEXT,
))
reg_a.register(spec, _cer)
assert "cer" in reg_a
assert "cer" not in reg_b
assert len(reg_a) == 1
assert len(reg_b) == 0
|