[
https://issues.apache.org/jira/browse/SPARK-53440?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Malthe Borch updated SPARK-53440:
---------------------------------
Description:
The [transform
function|https://spark.apache.org/docs/latest/api/python/reference/pyspark.sql/api/pyspark.sql.functions.transform.html]
takes a Python {_}function{_}-argument (unary or binary) which is then
converted into a Catalyst expression using introspection.
In some situations, it's preferable to express such functions using SQL, e.g.
`x -> x + 1` since no introspection happens, just simple parsing.
It should be possible to specify such a function as a string or using some
other construct that ultimately lets the user provide the function simply as a
string.
Note that it _is_ already possible to use `call_function` like so:
{code:python}
from pyspark.sql.functions import array, call_function, expr
column = call_function("transform", array(1, 2, 3), expr("x -> x + 1"))
{code}
But this works on an array, not a scalar column.
was:
The [transform
function|https://spark.apache.org/docs/latest/api/python/reference/pyspark.sql/api/pyspark.sql.functions.transform.html]
takes a Python _function_-argument (unary or binary) which is then converted
into a Catalyst expression using introspection.
In some situations, it's preferable to express such functions using SQL, e.g.
`x -> x + 1` since no introspection happens, just simple parsing.
It should be possible to specify such a function as a string or using some
other construct that ultimately lets the user provide the function simply as a
string.
Note that it _is_ already possible to use `call_function` like so:
{code:python}
from pyspark.sql.functions import array, call_function, expr
column = call_function("transform", array(1, 2, 3), expr("x -> x + 1"))
{code}
> Transform function to accept SQL-defined function
> -------------------------------------------------
>
> Key: SPARK-53440
> URL: https://issues.apache.org/jira/browse/SPARK-53440
> Project: Spark
> Issue Type: Wish
> Components: PySpark
> Affects Versions: 4.0.0
> Reporter: Malthe Borch
> Priority: Major
> Labels: pull-request-available
>
> The [transform
> function|https://spark.apache.org/docs/latest/api/python/reference/pyspark.sql/api/pyspark.sql.functions.transform.html]
> takes a Python {_}function{_}-argument (unary or binary) which is then
> converted into a Catalyst expression using introspection.
> In some situations, it's preferable to express such functions using SQL, e.g.
> `x -> x + 1` since no introspection happens, just simple parsing.
> It should be possible to specify such a function as a string or using some
> other construct that ultimately lets the user provide the function simply as
> a string.
> Note that it _is_ already possible to use `call_function` like so:
> {code:python}
> from pyspark.sql.functions import array, call_function, expr
> column = call_function("transform", array(1, 2, 3), expr("x -> x + 1"))
> {code}
> But this works on an array, not a scalar column.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
