[
https://issues.apache.org/jira/browse/BEAM-524?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Frank Yellin updated BEAM-524:
------------------------------
Description:
Two problems with documentation for "type" argument.
Trivial: Remove "by default". This phrase implies that there are other
alternatives besides what is listed. There aren't.
Non trivial. The documentation says "types appropriate to the combine_fn" are
accepted. I tried
Accumulator("foo", max, datetime.datetime)
This failed even though "datetime.datetime" is a perfectly reasonable type to
want to take the max of. (I wanted to know precisely when the last job
finished.)
Either the documentation needs to be changed to specify that max/min only apply
to numeric types, or the code needs to be changed to allow other uses of min
and max.
I suspect that the other operators also suffer similar limitations. I doubt
that I'd be allowed to sum numpy arrays, even though they support +.
was:
Two problems with documentation for "type" argument.
Trivial: Remove "by default". This phrase implies that there are other
alternatives besides what is listed. There aren't.
Non trivial. The documentation says "types appropriate to the combine_fn" are
accepted. I tried
Accumulator("foo", max, datetime.datetime)
This failed even though "datetime.datetime" is a perfectly reasonable type to
want to take the max of. (I wanted to know precisely when the last job
finished.)
Either the documentation needs to be changed to specify that max/min only apply
to numeric types, or the code needs to be changed to allow other uses of min
and max.
> Description of "type" argument in Aggregator is incorrect
> ---------------------------------------------------------
>
> Key: BEAM-524
> URL: https://issues.apache.org/jira/browse/BEAM-524
> Project: Beam
> Issue Type: Bug
> Components: sdk-py
> Reporter: Frank Yellin
> Priority: Minor
>
> Two problems with documentation for "type" argument.
> Trivial: Remove "by default". This phrase implies that there are other
> alternatives besides what is listed. There aren't.
> Non trivial. The documentation says "types appropriate to the combine_fn"
> are accepted. I tried
> Accumulator("foo", max, datetime.datetime)
> This failed even though "datetime.datetime" is a perfectly reasonable type to
> want to take the max of. (I wanted to know precisely when the last job
> finished.)
> Either the documentation needs to be changed to specify that max/min only
> apply to numeric types, or the code needs to be changed to allow other uses
> of min and max.
> I suspect that the other operators also suffer similar limitations. I doubt
> that I'd be allowed to sum numpy arrays, even though they support +.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
