165 lines
5.3 KiB
Python
Raw Normal View History

# Copyright 2021 Collate
# 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.
"""
Define Median function
"""
# Keep SQA docs style defining custom constructs
# pylint: disable=consider-using-f-string,duplicate-code
from sqlalchemy.ext.compiler import compiles
from sqlalchemy.sql.functions import FunctionElement
from metadata.profiler.metrics.core import CACHE
from metadata.profiler.orm.registry import Dialects
from metadata.utils.logger import profiler_logger
logger = profiler_logger()
class MedianFn(FunctionElement):
inherit_cache = CACHE
@compiles(MedianFn)
def _(elements, compiler, **kwargs): # pylint: disable=unused-argument
col = compiler.process(elements.clauses.clauses[0])
percentile = elements.clauses.clauses[2].value
return "percentile_cont(%.2f) WITHIN GROUP (ORDER BY %s ASC)" % (percentile, col)
@compiles(MedianFn, Dialects.BigQuery)
def _(elements, compiler, **kwargs):
col, _, percentile = [
compiler.process(element, **kwargs) for element in elements.clauses
]
return "percentile_cont(%s , %s) OVER()" % (col, percentile)
@compiles(MedianFn, Dialects.ClickHouse)
def _(elements, compiler, **kwargs):
col, _, percentile = [
compiler.process(element, **kwargs) for element in elements.clauses
]
return "quantile(%s)(%s)" % (percentile, col)
# pylint: disable=unused-argument
@compiles(MedianFn, Dialects.Athena)
@compiles(MedianFn, Dialects.Trino)
@compiles(MedianFn, Dialects.Presto)
def _(elements, compiler, **kwargs):
col = compiler.process(elements.clauses.clauses[0])
percentile = elements.clauses.clauses[2].value
return 'approx_percentile("%s", %.2f)' % (col, percentile)
@compiles(MedianFn, Dialects.MSSQL)
def _(elements, compiler, **kwargs):
"""Median computation for MSSQL"""
col = compiler.process(elements.clauses.clauses[0])
percentile = elements.clauses.clauses[2].value
return "percentile_cont(%.2f) WITHIN GROUP (ORDER BY %s ASC) OVER()" % (
percentile,
col,
)
@compiles(MedianFn, Dialects.Hive)
def _(elements, compiler, **kwargs):
"""Median computation for Hive"""
col, _, percentile = [
compiler.process(element, **kwargs) for element in elements.clauses
]
return "percentile(cast(%s as BIGINT), %s)" % (col, percentile)
@compiles(MedianFn, Dialects.Impala)
def _(elements, compiler, **kwargs):
"""Median computation for Impala
Median compution for Impala uses the appx_median function.
OM uses this median function to also compute first and third quartiles.
These calculations are not supported with a simple function inside Impala.
The if statement returns null when we are not looking for the .5 precentile
In Impala to get the first quartile a full SQL statement like this is necessary:
with ntiles as
(
select filesize, ntile(4) over (order by filesize) as quarter
from hdfs_files
)
, quarters as
(
select 1 as grp, max(filesize) as quartile_value, quarter
from ntiles
group by quarter
)
select max(case when quarter = 1 then quartile_value end) as first_q
, max(case when quarter = 2 then quartile_value end) as second_q
, max(case when quarter = 3 then quartile_value end) as third_q
, max(case when quarter = 4 then quartile_value end) as fourth_q
from quarters
group by grp
;
"""
col, _, percentile = [
compiler.process(element, **kwargs) for element in elements.clauses
]
Impalaconnection 0.2.1 + string datatypes enabled in profile (#11364) * updated metadata to work with the impala query engine. Uses the describe function to grab column names, data types, and comments. * added the ordinalPosition data point into the Column constructor. * renamed variable to better describe its usage. * updated profile errors. Hive connections now comment columns by default. * removed print statements * Cleaned up code by pulling check into its own function * Updated median function to return null when it is being used for first and third quartiles. * updated metadata to work with the impala query engine. Uses the describe function to grab column names, data types, and comments. * added the ordinalPosition data point into the Column constructor. * renamed variable to better describe its usage. * updated profile errors. Hive connections now comment columns by default. * removed print statements * Cleaned up code by pulling check into its own function * Updated median function to return null when it is being used for first and third quartiles. * removed print statements and ran make py_format * updated to fix some pylint errors. imported Dialects to remove string compare to "impala" engine * moved huge comment into function docstring. This comment shows us the sql to get quartiles in Impala * added cast to decimal for column when running average in mean.py * fixed lint error * fixed ui ordering of precision and scale. Precision should be ordred in front of scale since the precision is set first in decimal data types * Fixed overflow error when converting large numbers to bigint Fixed error for CHAR datatype missing. * Fixed NaN issues with Impala Profile * py formatting * Fixed warnings from SqlAlchemy The GenericFunction 'max' is already registered and is going to be overridden. The GenericFunction 'min' is already registered and is going to be overridden. Updated Min/Max to handle strings by getting they length. * Updated profiler to handle strings by using the string length as the parameter to compute the profile * py_format updates * fix: ran linting * fix: Mysql hardcoded table alias --------- Co-authored-by: Chirag Madlani <12962843+chirag-madlani@users.noreply.github.com> Co-authored-by: Teddy Crepineau <teddy.crepineau@gmail.com>
2023-04-30 03:03:56 -05:00
return f"if({percentile} = .5, appx_median(if(is_nan({col}) or is_inf({col}), null, {col})), null)"
@compiles(MedianFn, Dialects.MySQL)
def _(elements, compiler, **kwargs): # pylint: disable=unused-argument
"""Median computation for MySQL"""
col = compiler.process(elements.clauses.clauses[0])
table = elements.clauses.clauses[1].value
percentile = elements.clauses.clauses[2].value
return """
(SELECT
{col}
FROM (
SELECT
Impalaconnection 0.2.1 + string datatypes enabled in profile (#11364) * updated metadata to work with the impala query engine. Uses the describe function to grab column names, data types, and comments. * added the ordinalPosition data point into the Column constructor. * renamed variable to better describe its usage. * updated profile errors. Hive connections now comment columns by default. * removed print statements * Cleaned up code by pulling check into its own function * Updated median function to return null when it is being used for first and third quartiles. * updated metadata to work with the impala query engine. Uses the describe function to grab column names, data types, and comments. * added the ordinalPosition data point into the Column constructor. * renamed variable to better describe its usage. * updated profile errors. Hive connections now comment columns by default. * removed print statements * Cleaned up code by pulling check into its own function * Updated median function to return null when it is being used for first and third quartiles. * removed print statements and ran make py_format * updated to fix some pylint errors. imported Dialects to remove string compare to "impala" engine * moved huge comment into function docstring. This comment shows us the sql to get quartiles in Impala * added cast to decimal for column when running average in mean.py * fixed lint error * fixed ui ordering of precision and scale. Precision should be ordred in front of scale since the precision is set first in decimal data types * Fixed overflow error when converting large numbers to bigint Fixed error for CHAR datatype missing. * Fixed NaN issues with Impala Profile * py formatting * Fixed warnings from SqlAlchemy The GenericFunction 'max' is already registered and is going to be overridden. The GenericFunction 'min' is already registered and is going to be overridden. Updated Min/Max to handle strings by getting they length. * Updated profiler to handle strings by using the string length as the parameter to compute the profile * py_format updates * fix: ran linting * fix: Mysql hardcoded table alias --------- Co-authored-by: Chirag Madlani <12962843+chirag-madlani@users.noreply.github.com> Co-authored-by: Teddy Crepineau <teddy.crepineau@gmail.com>
2023-04-30 03:03:56 -05:00
{col},
ROW_NUMBER() OVER () AS row_num
FROM
Impalaconnection 0.2.1 + string datatypes enabled in profile (#11364) * updated metadata to work with the impala query engine. Uses the describe function to grab column names, data types, and comments. * added the ordinalPosition data point into the Column constructor. * renamed variable to better describe its usage. * updated profile errors. Hive connections now comment columns by default. * removed print statements * Cleaned up code by pulling check into its own function * Updated median function to return null when it is being used for first and third quartiles. * updated metadata to work with the impala query engine. Uses the describe function to grab column names, data types, and comments. * added the ordinalPosition data point into the Column constructor. * renamed variable to better describe its usage. * updated profile errors. Hive connections now comment columns by default. * removed print statements * Cleaned up code by pulling check into its own function * Updated median function to return null when it is being used for first and third quartiles. * removed print statements and ran make py_format * updated to fix some pylint errors. imported Dialects to remove string compare to "impala" engine * moved huge comment into function docstring. This comment shows us the sql to get quartiles in Impala * added cast to decimal for column when running average in mean.py * fixed lint error * fixed ui ordering of precision and scale. Precision should be ordred in front of scale since the precision is set first in decimal data types * Fixed overflow error when converting large numbers to bigint Fixed error for CHAR datatype missing. * Fixed NaN issues with Impala Profile * py formatting * Fixed warnings from SqlAlchemy The GenericFunction 'max' is already registered and is going to be overridden. The GenericFunction 'min' is already registered and is going to be overridden. Updated Min/Max to handle strings by getting they length. * Updated profiler to handle strings by using the string length as the parameter to compute the profile * py_format updates * fix: ran linting * fix: Mysql hardcoded table alias --------- Co-authored-by: Chirag Madlani <12962843+chirag-madlani@users.noreply.github.com> Co-authored-by: Teddy Crepineau <teddy.crepineau@gmail.com>
2023-04-30 03:03:56 -05:00
{table},
(SELECT @counter := COUNT(*) FROM {table}) t_count
ORDER BY {col}
) temp
WHERE temp.row_num = ROUND({percentile} * @counter)
)
""".format(
col=col, table=table, percentile=percentile
)
@compiles(MedianFn, Dialects.SQLite)
def _(elements, compiler, **kwargs): # pylint: disable=unused-argument
col = compiler.process(elements.clauses.clauses[0])
table = elements.clauses.clauses[1].value
percentile = elements.clauses.clauses[2].value
return """
(SELECT
{col}
FROM {table}
WHERE {col} IS NOT NULL
ORDER BY {col}
LIMIT 1
OFFSET (
SELECT ROUND(COUNT(*) * {percentile} -1)
FROM {table}
WHERE {col} IS NOT NULL
)
)
""".format(
col=col, table=table, percentile=percentile
)