MichaelChirico commented on pull request #28386:
URL: https://github.com/apache/spark/pull/28386#issuecomment-725187311
@felixcheung I think this PR was ready to merge, could you help to reopen?
This is an automated message f
MichaelChirico commented on pull request #28386:
URL: https://github.com/apache/spark/pull/28386#issuecomment-727694593
@HyukjinKwon unfortunately I've since switched companies so I'm going from
memory. But the basic idea AIRI is that `mutate` tries to auto-infer column
names when not supp
MichaelChirico commented on pull request #28386:
URL: https://github.com/apache/spark/pull/28386#issuecomment-727768929
@HyukjinKwon test added, please have a look
This is an automated message from the Apache Git Service.
To
MichaelChirico commented on pull request #28386:
URL: https://github.com/apache/spark/pull/28386#issuecomment-728044077
collapse is an aggregation method. input is length>1 & collapse guarantees
length-1 output.
what's happening is separate has variable return length -- usually it
MichaelChirico commented on pull request #28386:
URL: https://github.com/apache/spark/pull/28386#issuecomment-728182679
then what happens when users send an expression with width>500?
not likely but it's safer to just collapse the result
On Mon, Nov 16, 2020, 11:40 AM S Daniel
MichaelChirico commented on pull request #28386:
URL: https://github.com/apache/spark/pull/28386#issuecomment-728186184
e.g. in [`SPARK-31517`](https://issues.apache.org/jira/browse/SPARK-31517),
the supplied width was 98:
```
over(rank(), orderBy(windowPartitionBy(column("cyl")),
MichaelChirico commented on pull request #28386:
URL: https://github.com/apache/spark/pull/28386#issuecomment-728724666
> Ideally, we'd like both to behave the same way (possibly resolving the
problem with mutate), that's however a breaking change.
A reasonable request (though defini
MichaelChirico commented on pull request #28386:
URL: https://github.com/apache/spark/pull/28386#issuecomment-728726963
> Adding magrittr as a dependency
Best saved for a different thread, but TL;DR: yes, I think that's a good
idea, `magrittr` will help writing `SparkR` code that's m
MichaelChirico commented on pull request #28386:
URL: https://github.com/apache/spark/pull/28386#issuecomment-623224648
Thanks @felixcheung. both backports are directly from `base` R:
https://github.com/wch/r-source/blob/08ebf253e44e10bfb445f27b53b2a43bc7e6740d/src/library/base/R/uti