This repository has been archived by the owner on May 28, 2024. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 3
/
metric.py
426 lines (359 loc) · 15.8 KB
/
metric.py
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
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
import copy
import fnmatch
import re
from collections import defaultdict
from enum import Enum
from textwrap import dedent
from typing import TYPE_CHECKING, Any, Dict, List, Optional, Union
import attr
import jinja2
from mozilla_nimbus_schemas.jetstream import AnalysisBasis
from metric_config_parser.errors import DefinitionNotFound
if TYPE_CHECKING:
from .analysis import AnalysisSpec
from .config import ConfigCollection
from .experiment import ExperimentConfiguration
from .definition import DefinitionSpecSub
from .project import ProjectConfiguration
from .data_source import DataSource, DataSourceReference
from .parameter import ParameterDefinition
from .pre_treatment import PreTreatmentReference
from .statistic import Statistic
from .util import converter, is_valid_slug
class AnalysisPeriod(Enum):
DAY = "day"
WEEK = "week"
DAYS_28 = "days28"
OVERALL = "overall"
PREENROLLMENT_WEEK = "preenrollment_week"
PREENROLLMENT_DAYS_28 = "preenrollment_days28"
@property
def mozanalysis_label(self) -> str:
d = {
"day": "daily",
"week": "weekly",
"days28": "28_day",
"overall": "overall",
"preenrollment_week": "preenrollment_weekly",
"preenrollment_days28": "preenrollment_days28",
}
return d[self.value]
@property
def table_suffix(self) -> str:
d = {
"day": "daily",
"week": "weekly",
"days28": "days28",
"overall": "overall",
"preenrollment_week": "preenrollment_weekly",
"preenrollment_days28": "preenrollment_days28",
}
return d[self.value]
class MetricLevel(Enum):
GOLD = "gold"
SILVER = "silver"
BRONZE = "bronze"
@attr.s(auto_attribs=True)
class Summary:
"""Represents a metric with a statistical treatment."""
metric: "Metric"
statistic: "Statistic"
pre_treatments: List[PreTreatmentReference] = attr.Factory(list)
@attr.s(auto_attribs=True, frozen=True, slots=True)
class Metric:
"""
Metric representation.
Metrics are supersets of mozanalysis metrics with additional
metadata required for analysis.
"""
name: str
data_source: Optional[DataSource]
select_expression: Optional[str]
friendly_name: Optional[str] = None
description: Optional[str] = None
bigger_is_better: bool = True
analysis_bases: List[AnalysisBasis] = [AnalysisBasis.ENROLLMENTS, AnalysisBasis.EXPOSURES]
type: str = "scalar"
category: Optional[str] = None
depends_on: Optional[List[Summary]] = None
owner: Optional[List[str]] = None
deprecated: bool = False
level: Optional[MetricLevel] = None
@attr.s(auto_attribs=True)
class MetricReference:
name: str
def resolve(
self,
spec: "AnalysisSpec",
conf: Union["ExperimentConfiguration", "ProjectConfiguration"],
configs: "ConfigCollection",
) -> List[Summary]:
if self.name in spec.metrics.definitions:
return spec.metrics.definitions[self.name].resolve(spec, conf, configs)
metric_definition = configs.get_metric_definition(self.name, conf.app_name)
if metric_definition:
return metric_definition.resolve(spec, conf, configs=configs)
raise DefinitionNotFound(f"Could not locate metric {self.name}")
# These are bare strings in the configuration file.
converter.register_structure_hook(MetricReference, lambda obj, _type: MetricReference(name=obj))
converter.register_structure_hook(Union[str, List[str], None], lambda obj, _type: obj)
@attr.s(auto_attribs=True)
class MetricDefinition:
"""Describes the interface for defining a metric in configuration.
The `select_expression` of the metric may use Jinja2 template syntax to refer to the
aggregation helper functions defined in `mozanalysis.metrics`, like
'{{agg_any("payload.processes.scalars.some_boolean_thing")}}'
"""
name: str # implicit in configuration
statistics: Optional[Dict[str, Dict[str, Any]]] = None
select_expression: Optional[str] = None
data_source: Optional[DataSourceReference] = None
friendly_name: Optional[str] = None
description: Optional[str] = None
bigger_is_better: bool = True
analysis_bases: Optional[List[AnalysisBasis]] = None
type: Optional[str] = None
category: Optional[str] = None
depends_on: Optional[List[MetricReference]] = None
owner: Optional[Union[str, List[str]]] = None
deprecated: bool = False
level: Optional[MetricLevel] = None
@staticmethod
def generate_select_expression(
param_definitions: Dict[str, ParameterDefinition],
select_expr_template: Union[str, jinja2.nodes.Template],
configs: "ConfigCollection",
) -> str:
"""
Takes in param configuration and converts it to a select statement string
"""
if "parameters" not in str(select_expr_template):
return configs.get_env().from_string(select_expr_template).render()
formatted_params: Dict[str, Any] = defaultdict()
for param_name, param_definition in param_definitions.items():
if param_definition.distinct_by_branch and isinstance(param_definition.value, dict):
formatted_params.update(
{
param_name: "CASE e.branch "
+ " ".join(
[
f'WHEN "{branch}" THEN "{value}"'
for branch, value in param_definition.value.items()
]
)
+ " END"
}
)
else:
formatted_params.update({param_name: param_definition.value})
return (
configs.get_env().from_string(select_expr_template).render(parameters=formatted_params)
)
def resolve(
self,
spec: "DefinitionSpecSub",
conf: Union["ExperimentConfiguration", "ProjectConfiguration"],
configs: "ConfigCollection",
) -> List[Summary]:
metric_summary = None
metric = None
upstream_metrics = None
# check if metric depends on other metrics
if self.depends_on:
upstream_metrics = []
# resolve upstream metrics
for metric_ref in self.depends_on:
# check if upstream metric is defined externally as a "definition"
upstream_metric = configs.get_metric_definition(metric_ref.name, conf.app_name)
if upstream_metric is None:
# check if upstream metric is part of the analysis spec
upstream_metric = spec.metrics.definitions.get(metric_ref.name, None)
if upstream_metric is None:
raise DefinitionNotFound(
f"No definition found for referenced upstream metric {metric_ref}"
)
upstream_metrics += upstream_metric.resolve(spec, conf, configs)
if self.select_expression is None or self.data_source is None:
# checks if a metric from mozanalysis was referenced
metric_definition = configs.get_metric_definition(self.name, conf.app_name)
if metric_definition is None and upstream_metrics is None:
raise DefinitionNotFound(
f"No default definition found for referenced metric {self.name}"
)
elif upstream_metrics:
metric = Metric(
name=self.name,
data_source=None,
select_expression=None,
friendly_name=(
dedent(self.friendly_name) if self.friendly_name else self.friendly_name
),
description=dedent(self.description) if self.description else self.description,
bigger_is_better=self.bigger_is_better,
analysis_bases=self.analysis_bases
or [AnalysisBasis.ENROLLMENTS, AnalysisBasis.EXPOSURES],
type=self.type or "scalar",
category=self.category,
depends_on=upstream_metrics,
owner=[self.owner] if isinstance(self.owner, str) else self.owner,
deprecated=self.deprecated,
level=self.level,
)
elif metric_definition:
metric_definition.analysis_bases = self.analysis_bases or [
AnalysisBasis.ENROLLMENTS,
AnalysisBasis.EXPOSURES,
]
metric_definition.statistics = self.statistics
metric_summary = metric_definition.resolve(spec, conf, configs)
else:
select_expression = self.generate_select_expression(
spec.parameters.definitions,
select_expr_template=self.select_expression,
configs=configs,
)
metric = Metric(
name=self.name,
data_source=self.data_source.resolve(spec, conf, configs),
select_expression=select_expression,
friendly_name=(
dedent(self.friendly_name) if self.friendly_name else self.friendly_name
),
description=dedent(self.description) if self.description else self.description,
bigger_is_better=self.bigger_is_better,
analysis_bases=self.analysis_bases
or [AnalysisBasis.ENROLLMENTS, AnalysisBasis.EXPOSURES],
type=self.type or "scalar",
category=self.category,
depends_on=upstream_metrics,
owner=[self.owner] if isinstance(self.owner, str) else self.owner,
deprecated=self.deprecated,
level=self.level,
)
metrics_with_treatments = []
if metric_summary:
if self.statistics:
for statistic_name, params in self.statistics.items():
stats_params = copy.deepcopy(params)
pre_treatments = []
for pt in stats_params.pop("pre_treatments", []):
if isinstance(pt, str):
ref = PreTreatmentReference(pt, {})
else:
name = pt.pop("name")
ref = PreTreatmentReference(name, pt)
pre_treatments.append(ref.resolve(spec))
metrics_with_treatments.append(
Summary(
metric=metric_summary[0].metric,
statistic=Statistic(statistic_name, stats_params),
pre_treatments=pre_treatments,
)
)
else:
metrics_with_treatments += metric_summary
elif metric:
if self.statistics is None:
raise ValueError(f"No statistical treatment defined for metric '{self.name}'")
for statistic_name, params in self.statistics.items():
stats_params = copy.deepcopy(params)
pre_treatments = []
for pt in stats_params.pop("pre_treatments", []):
if isinstance(pt, str):
ref = PreTreatmentReference(pt, {})
else:
name = pt.pop("name")
ref = PreTreatmentReference(name, pt)
pre_treatments.append(ref.resolve(spec))
metrics_with_treatments.append(
Summary(
metric=metric,
statistic=Statistic(statistic_name, stats_params),
pre_treatments=pre_treatments,
)
)
if len(metrics_with_treatments) == 0:
raise ValueError(f"Metric {self.name} has no statistical treatment defined.")
return metrics_with_treatments
def merge(self, other: "MetricDefinition"):
"""Merge with another metric definition."""
for key in attr.fields_dict(type(self)):
setattr(self, key, getattr(other, key) or getattr(self, key))
MetricsConfigurationType = Dict[AnalysisPeriod, List[Summary]]
@attr.s(auto_attribs=True)
class MetricsSpec:
"""Describes the interface for the metrics section in configuration."""
daily: List[MetricReference] = attr.Factory(list)
weekly: List[MetricReference] = attr.Factory(list)
days28: List[MetricReference] = attr.Factory(list)
overall: List[MetricReference] = attr.Factory(list)
preenrollment_weekly: List[MetricReference] = attr.Factory(list)
preenrollment_days28: List[MetricReference] = attr.Factory(list)
definitions: Dict[str, MetricDefinition] = attr.Factory(dict)
@classmethod
def from_dict(cls, d: dict) -> "MetricsSpec":
params: Dict[str, Any] = {}
known_keys = {f.name for f in attr.fields(cls)}
for k in known_keys:
if k == "days28":
v = d.get("28_day", [])
else:
v = d.get(k, [])
if not isinstance(v, list):
raise ValueError(f"metrics.{k} should be a list of metrics")
params[k] = [MetricReference(m) for m in v]
params["definitions"] = {
k: converter.structure(
{"name": k, **dict((kk.lower(), vv) for kk, vv in v.items())}, MetricDefinition
)
for k, v in d.items()
if k not in known_keys and k != "28_day"
}
return cls(**params)
def resolve(
self,
spec: "AnalysisSpec",
conf: Union["ExperimentConfiguration", "ProjectConfiguration"],
configs: "ConfigCollection",
) -> MetricsConfigurationType:
result = {}
for period in AnalysisPeriod:
summaries = [
summary
for ref in getattr(self, period.table_suffix)
for summary in ref.resolve(spec, conf, configs)
]
unique_summaries = []
seen_summaries = set()
# summaries needs to be reversed to make sure merged configs overwrite existing ones
summaries.reverse()
for summary in summaries:
if (summary.metric.name, summary.statistic.name) not in seen_summaries:
seen_summaries.add((summary.metric.name, summary.statistic.name))
unique_summaries.append(summary)
result[period] = unique_summaries
return result
def merge(self, other: "MetricsSpec"):
"""
Merges another metrics spec into the current one.
The `other` MetricsSpec overwrites existing metrics.
"""
self.daily = other.daily + self.daily
self.weekly = other.weekly + self.weekly
self.days28 = other.days28 + self.days28
self.overall = other.overall + self.overall
self.preenrollment_weekly = other.preenrollment_weekly + self.preenrollment_weekly
self.preenrollment_days28 = other.preenrollment_days28 + self.preenrollment_days28
seen = set()
for key, _ in self.definitions.items():
for other_key in other.definitions:
# support wildcard characters in `other`
other_key_regex = re.compile(fnmatch.translate(other_key))
if other_key_regex.fullmatch(key):
self.definitions[key].merge(other.definitions[other_key])
seen.add(other_key)
seen.add(key)
for key, definition in other.definitions.items():
if key not in seen and is_valid_slug(key):
self.definitions[key] = definition
converter.register_structure_hook(MetricsSpec, lambda obj, _type: MetricsSpec.from_dict(obj))