This commit is contained in:
Alexandre Flament 2022-09-18 11:13:08 +02:00
parent b3708e4137
commit efc728b4ed
3 changed files with 65 additions and 52 deletions

View file

@ -2,14 +2,13 @@
# lint: pylint
# pylint: disable=missing-module-docstring
import typing
from typing import Dict, TypedDict, List, Optional, Any
import math
import contextlib
from timeit import default_timer
from operator import itemgetter
from searx.engines import engines
from .models import HistogramStorage, CounterStorage, VoidHistogram, VoidCounterStorage
from .models import Histogram, HistogramStorage, CounterStorage, VoidHistogram, VoidCounterStorage
from .error_recorder import count_error, count_exception, errors_per_engines
__all__ = [
@ -27,11 +26,8 @@ __all__ = [
]
ENDPOINTS = {'search'}
histogram_storage: typing.Optional[HistogramStorage] = None
counter_storage: typing.Optional[CounterStorage] = None
histogram_storage: Optional[HistogramStorage] = None
counter_storage: Optional[CounterStorage] = None
@contextlib.contextmanager
@ -50,7 +46,7 @@ def histogram_observe(duration, *args):
histogram_storage.get(*args).observe(duration)
def histogram(*args, raise_on_not_found=True):
def histogram(*args, raise_on_not_found=True) -> Histogram:
h = histogram_storage.get(*args)
if raise_on_not_found and h is None:
raise ValueError("histogram " + repr((*args,)) + " doesn't not exist")
@ -65,7 +61,7 @@ def counter_add(value, *args):
counter_storage.add(value, *args)
def counter(*args):
def counter(*args) -> int:
return counter_storage.get(*args)
@ -110,7 +106,21 @@ def initialize(engine_names=None, enabled=True):
histogram_storage.configure(histogram_width, histogram_size, 'engine', engine_name, 'time', 'total')
def get_engine_errors(engline_name_list):
class EngineError(TypedDict):
"""Describe an engine error. To do : check the types"""
filename: str
function: str
line_no: int
code: str
exception_classname: str
log_message: str
log_parameters: List[str]
secondary: bool
percentage: int
def get_engine_errors(engline_name_list) -> Dict[str, List[EngineError]]:
result = {}
engine_names = list(errors_per_engines.keys())
engine_names.sort()
@ -141,7 +151,15 @@ def get_engine_errors(engline_name_list):
return result
def get_reliabilities(engline_name_list, checker_results):
class EngineReliability(TypedDict):
"""Describe the engine reliability. To do: update the checker field type"""
reliability: int
errors: List[EngineError]
checker: Optional[Any]
def get_reliabilities(engline_name_list, checker_results) -> Dict[str, EngineReliability]:
reliabilities = {}
engine_errors = get_engine_errors(engline_name_list)
@ -184,11 +202,11 @@ def get_engines_stats(engine_name_list):
if sent_count == 0:
continue
result_count = histogram('engine', engine_name, 'result', 'count').percentage(50)
result_count = histogram('engine', engine_name, 'result', 'count').percentile(50)
result_count_sum = histogram('engine', engine_name, 'result', 'count').sum
successful_count = counter('engine', engine_name, 'search', 'count', 'successful')
time_total = histogram('engine', engine_name, 'time', 'total').percentage(50)
time_total = histogram('engine', engine_name, 'time', 'total').percentile(50)
max_time_total = max(time_total or 0, max_time_total or 0)
max_result_count = max(result_count or 0, max_result_count or 0)
@ -214,13 +232,13 @@ def get_engines_stats(engine_name_list):
stats['score'] = score
stats['score_per_result'] = score / float(result_count_sum)
time_http = histogram('engine', engine_name, 'time', 'http').percentage(50)
time_http = histogram('engine', engine_name, 'time', 'http').percentile(50)
time_http_p80 = time_http_p95 = 0
if time_http is not None:
time_http_p80 = histogram('engine', engine_name, 'time', 'http').percentage(80)
time_http_p95 = histogram('engine', engine_name, 'time', 'http').percentage(95)
time_http_p80 = histogram('engine', engine_name, 'time', 'http').percentile(80)
time_http_p95 = histogram('engine', engine_name, 'time', 'http').percentile(95)
stats['http'] = round(time_http, 1)
stats['http_p80'] = round(time_http_p80, 1)
@ -228,8 +246,8 @@ def get_engines_stats(engine_name_list):
if time_total is not None:
time_total_p80 = histogram('engine', engine_name, 'time', 'total').percentage(80)
time_total_p95 = histogram('engine', engine_name, 'time', 'total').percentage(95)
time_total_p80 = histogram('engine', engine_name, 'time', 'total').percentile(80)
time_total_p95 = histogram('engine', engine_name, 'time', 'total').percentile(95)
stats['total'] = round(time_total, 1)
stats['total_p80'] = round(time_total_p80, 1)

View file

@ -1,7 +1,9 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
import decimal
from numbers import Number
import threading
from typing import Dict, List, Optional, Tuple
from searx import logger
@ -15,15 +17,19 @@ class Histogram:
_slots__ = '_lock', '_size', '_sum', '_quartiles', '_count', '_width'
def __init__(self, width=10, size=200):
def __init__(self, width: int = 10, size: int = 200):
"""
* width: quantile width
* size: number of quantiles
"""
self._lock = threading.Lock()
self._width = width
self._size = size
self._quartiles = [0] * size
self._count = 0
self._sum = 0
self._count: int = 0
self._sum: int = 0
def observe(self, value):
def observe(self, value: Number):
q = int(value / self._width)
if q < 0:
"""Value below zero is ignored"""
@ -37,19 +43,19 @@ class Histogram:
self._sum += value
@property
def quartiles(self):
def quartiles(self) -> List[int]:
return list(self._quartiles)
@property
def count(self):
def count(self) -> int:
return self._count
@property
def sum(self):
def sum(self) -> int:
return self._sum
@property
def average(self):
def average(self) -> float:
with self._lock:
if self._count != 0:
return self._sum / self._count
@ -57,31 +63,20 @@ class Histogram:
return 0
@property
def quartile_percentage(self):
'''Quartile in percentage'''
def quartile_percentages(self) -> List[int]:
"""Quartile in percentage"""
with self._lock:
if self._count > 0:
return [int(q * 100 / self._count) for q in self._quartiles]
else:
return self._quartiles
@property
def quartile_percentage_map(self):
result = {}
# use Decimal to avoid rounding errors
x = decimal.Decimal(0)
width = decimal.Decimal(self._width)
width_exponent = -width.as_tuple().exponent
with self._lock:
if self._count > 0:
for y in self._quartiles:
yp = int(y * 100 / self._count)
if yp != 0:
result[round(float(x), width_exponent)] = yp
x += width
return result
def percentile(self, percentage: Number) -> Optional[decimal.Decimal]:
"""
Return the percentile.
def percentage(self, percentage):
* percentage from 0 to 100
"""
# use Decimal to avoid rounding errors
x = decimal.Decimal(0)
width = decimal.Decimal(self._width)
@ -109,14 +104,14 @@ class HistogramStorage:
self.histogram_class = histogram_class
def clear(self):
self.measures = {}
self.measures: Dict[Tuple[str], Histogram] = {}
def configure(self, width, size, *args):
measure = self.histogram_class(width, size)
self.measures[args] = measure
return measure
def get(self, *args):
def get(self, *args) -> Optional[Histogram]:
return self.measures.get(args, None)
def dump(self):
@ -136,13 +131,13 @@ class CounterStorage:
def clear(self):
with self.lock:
self.counters = {}
self.counters: Dict[Tuple[str], int] = {}
def configure(self, *args):
with self.lock:
self.counters[args] = 0
def get(self, *args):
def get(self, *args) -> int:
return self.counters[args]
def add(self, value, *args):

View file

@ -984,9 +984,9 @@ def preferences():
max_rate95 = 0
for _, e in filtered_engines.items():
h = histogram('engine', e.name, 'time', 'total')
median = round(h.percentage(50), 1) if h.count > 0 else None
rate80 = round(h.percentage(80), 1) if h.count > 0 else None
rate95 = round(h.percentage(95), 1) if h.count > 0 else None
median = round(h.percentile(50), 1) if h.count > 0 else None # type: ignore
rate80 = round(h.percentile(80), 1) if h.count > 0 else None # type: ignore
rate95 = round(h.percentile(95), 1) if h.count > 0 else None # type: ignore
max_rate95 = max(max_rate95, rate95 or 0)