Coverage for airflow/operators/check_operator.py : 27%

# -*- coding: utf-8 -*- 

# 

# Licensed under the Apache License, Version 2.0 (the "License"); 

# you may not use this file except in compliance with the License. 

# You may obtain a copy of the License at 

# 

# http://www.apache.org/licenses/LICENSE-2.0 

# 

# Unless required by applicable law or agreed to in writing, software 

# distributed under the License is distributed on an "AS IS" BASIS, 

# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 

# See the License for the specific language governing permissions and 

# limitations under the License. 

from builtins import zip 

from builtins import str 

import logging 

from airflow.exceptions import AirflowException 

from airflow.hooks.base_hook import BaseHook 

from airflow.models import BaseOperator 

from airflow.utils.decorators import apply_defaults 

class CheckOperator(BaseOperator): 

""" 

Performs checks against a db. The ``CheckOperator`` expects 

a sql query that will return a single row. Each value on that 

first row is evaluated using python ``bool`` casting. If any of the 

values return ``False`` the check is failed and errors out. 

Note that Python bool casting evals the following as ``False``: 

* False 

* 0 

* Empty string (``""``) 

* Empty list (``[]``) 

* Empty dictionary or set (``{}``) 

Given a query like ``SELECT COUNT(*) FROM foo``, it will fail only if 

the count ``== 0``. You can craft much more complex query that could, 

for instance, check that the table has the same number of rows as 

the source table upstream, or that the count of today's partition is 

greater than yesterday's partition, or that a set of metrics are less 

than 3 standard deviation for the 7 day average. 

This operator can be used as a data quality check in your pipeline, and 

depending on where you put it in your DAG, you have the choice to 

stop the critical path, preventing from 

publishing dubious data, or on the side and receive email alerts 

without stopping the progress of the DAG. 

Note that this is an abstract class and get_db_hook 

needs to be defined. Whereas a get_db_hook is hook that gets a 

single record from an external source. 

:param sql: the sql to be executed 

:type sql: string 

""" 

template_fields = ('sql',) 

template_ext = ('.hql', '.sql',) 

ui_color = '#fff7e6' 

@apply_defaults 

def __init__( 

self, sql, 

conn_id=None, 

*args, **kwargs): 

super(CheckOperator, self).__init__(*args, **kwargs) 

self.conn_id = conn_id 

self.sql = sql 

def execute(self, context=None): 

logging.info('Executing SQL check: ' + self.sql) 

records = self.get_db_hook().get_first(self.sql) 

logging.info("Record: " + str(records)) 

if not records: 

raise AirflowException("The query returned None") 

elif not all([bool(r) for r in records]): 

exceptstr = "Test failed.\nQuery:\n{q}\nResults:\n{r!s}" 

raise AirflowException(exceptstr.format(q=self.sql, r=records)) 

logging.info("Success.") 

def get_db_hook(self): 

return BaseHook.get_hook(conn_id=self.conn_id) 

def _convert_to_float_if_possible(s): 

''' 

A small helper function to convert a string to a numeric value 

if appropriate 

:param s: the string to be converted 

:type s: str 

''' 

try: 

ret = float(s) 

except (ValueError, TypeError): 

ret = s 

return ret 

class ValueCheckOperator(BaseOperator): 

""" 

Performs a simple value check using sql code. 

Note that this is an abstract class and get_db_hook 

needs to be defined. Whereas a get_db_hook is hook that gets a 

single record from an external source. 

:param sql: the sql to be executed 

:type sql: string 

""" 

__mapper_args__ = { 

'polymorphic_identity': 'ValueCheckOperator' 

} 

template_fields = ('sql',) 

template_ext = ('.hql', '.sql',) 

ui_color = '#fff7e6' 

@apply_defaults 

def __init__( 

self, sql, pass_value, tolerance=None, 

conn_id=None, 

*args, **kwargs): 

super(ValueCheckOperator, self).__init__(*args, **kwargs) 

self.sql = sql 

self.conn_id = conn_id 

self.pass_value = _convert_to_float_if_possible(pass_value) 

tol = _convert_to_float_if_possible(tolerance) 

self.tol = tol if isinstance(tol, float) else None 

self.is_numeric_value_check = isinstance(self.pass_value, float) 

self.has_tolerance = self.tol is not None 

def execute(self, context=None): 

logging.info('Executing SQL check: ' + self.sql) 

records = self.get_db_hook().get_first(self.sql) 

if not records: 

raise AirflowException("The query returned None") 

test_results = [] 

except_temp = ("Test failed.\nPass value:{self.pass_value}\n" 

"Query:\n{self.sql}\nResults:\n{records!s}") 

if not self.is_numeric_value_check: 

tests = [str(r) == self.pass_value for r in records] 

elif self.is_numeric_value_check: 

try: 

num_rec = [float(r) for r in records] 

except (ValueError, TypeError) as e: 

cvestr = "Converting a result to float failed.\n" 

raise AirflowException(cvestr+except_temp.format(**locals())) 

if self.has_tolerance: 

tests = [ 

r / (1 + self.tol) <= self.pass_value <= r / (1 - self.tol) 

for r in num_rec] 

else: 

tests = [r == self.pass_value for r in num_rec] 

if not all(tests): 

raise AirflowException(except_temp.format(**locals())) 

def get_db_hook(self): 

return BaseHook.get_hook(conn_id=self.conn_id) 

class IntervalCheckOperator(BaseOperator): 

""" 

Checks that the values of metrics given as SQL expressions are within 

a certain tolerance of the ones from days_back before. 

Note that this is an abstract class and get_db_hook 

needs to be defined. Whereas a get_db_hook is hook that gets a 

single record from an external source. 

:param table: the table name 

:type table: str 

:param days_back: number of days between ds and the ds we want to check 

against. Defaults to 7 days 

:type days_back: int 

:param metrics_threshold: a dictionary of ratios indexed by metrics 

:type metrics_threshold: dict 

""" 

__mapper_args__ = { 

'polymorphic_identity': 'IntervalCheckOperator' 

} 

template_fields = ('sql1', 'sql2') 

template_ext = ('.hql', '.sql',) 

ui_color = '#fff7e6' 

@apply_defaults 

def __init__( 

self, table, metrics_thresholds, 

date_filter_column='ds', days_back=-7, 

conn_id=None, 

*args, **kwargs): 

super(IntervalCheckOperator, self).__init__(*args, **kwargs) 

self.table = table 

self.metrics_thresholds = metrics_thresholds 

self.metrics_sorted = sorted(metrics_thresholds.keys()) 

self.date_filter_column = date_filter_column 

self.days_back = -abs(days_back) 

self.conn_id = conn_id 

sqlexp = ', '.join(self.metrics_sorted) 

sqlt = ("SELECT {sqlexp} FROM {table}" 

" WHERE {date_filter_column}=").format(**locals()) 

self.sql1 = sqlt + "'{{ ds }}'" 

self.sql2 = sqlt + "'{{ macros.ds_add(ds, "+str(self.days_back)+") }}'" 

def execute(self, context=None): 

hook = self.get_db_hook() 

logging.info('Executing SQL check: ' + self.sql2) 

row2 = hook.get_first(self.sql2) 

logging.info('Executing SQL check: ' + self.sql1) 

row1 = hook.get_first(self.sql1) 

if not row2: 

raise AirflowException("The query {q} returned None").format(q=self.sql2) 

if not row1: 

raise AirflowException("The query {q} returned None").format(q=self.sql1) 

current = dict(zip(self.metrics_sorted, row1)) 

reference = dict(zip(self.metrics_sorted, row2)) 

ratios = {} 

test_results = {} 

rlog = "Ratio for {0}: {1} \n Ratio threshold : {2}" 

fstr = "'{k}' check failed. {r} is above {tr}" 

estr = "The following tests have failed:\n {0}" 

countstr = "The following {j} tests out of {n} failed:" 

for m in self.metrics_sorted: 

if current[m] == 0 or reference[m] == 0: 

ratio = None 

else: 

ratio = float(max(current[m], reference[m])) / \ 

min(current[m], reference[m]) 

logging.info(rlog.format(m, ratio, self.metrics_thresholds[m])) 

ratios[m] = ratio 

test_results[m] = ratio < self.metrics_thresholds[m] 

if not all(test_results.values()): 

failed_tests = [it[0] for it in test_results.items() if not it[1]] 

j = len(failed_tests) 

n = len(self.metrics_sorted) 

logging.warning(countstr.format(**locals())) 

for k in failed_tests: 

logging.warning(fstr.format(k=k, r=ratios[k], 

tr=self.metrics_thresholds[k])) 

raise AirflowException(estr.format(", ".join(failed_tests))) 

logging.info("All tests have passed") 

def get_db_hook(self): 

return BaseHook.get_hook(conn_id=self.conn_id)