apex-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Yogi Devendra <yogideven...@apache.org>
Subject Re: Default value for tupleSeperator
Date Sat, 13 Aug 2016 13:41:04 GMT
Since, there is difference of opinion and no clear win I would keep this
unchanged.
We can revisit it in the future if required.

~ Yogi

On 13 August 2016 at 11:04, Venkatesh Kottapalli <venkatesh@datatorrent.com>
wrote:

> Yes Yogi.
>
> What you pointed is one such example, another example for similar case
> would be CSVFormatter and AbstractFileOutputOperator combination. If we go
> with default values two new lines get inserted here as well.
>
> What I meant before is to have the line delimiter as it is because these
> two operators might not be combined all the time. When such case arise, the
> user can always configure empty string as delimiter in the operator he
> would prefer to.
>
> -Venkatesh.
>
>
> > On Aug 12, 2016, at 10:08 PM, Yogi Devendra <
> devendra.vyavahare@gmail.com> wrote:
> >
> > @Venkatesh
> >
> > Yes. Property is configurable by the end user.
> > But, we are saying that better to have meaningful defaults where ever
> > possible.
> >
> > ~ Yogi
> >
> > On 13 August 2016 at 08:39, Vlad Rozov <v.rozov@datatorrent.com> wrote:
> >
> >> StringFileOutput should default to empty. It should be possible to send
> >> multiple strings to a file without every new string being on a new line.
> >>
> >> Possibly a better option will be to have a separate operator that
> inserts
> >> necessary delimiters as csvformatter may be connected to jdbc output
> >> operator and should not insert new lines after each tuple. That brings
> >> overall design question: is it better to compose dag from multiple
> >> operators or overload each operator functionality with extra features.
> IMO,
> >> composition is better.
> >>
> >> Thank you,
> >> Vlad
> >>
> >> Отправлено с iPhone
> >>
> >>> On Aug 12, 2016, at 19:09, Venkatesh Kottapalli <
> >> venkatesh@datatorrent.com> wrote:
> >>>
> >>>
> >>> IMO, since we have an option to configure the line delimiter, leaving
> it
> >> to the user is a good idea as there might be cases where either one of
> them
> >> is used and not both.
> >>>
> >>> -Venkatesh.
> >>>
> >>>
> >>>> On Aug 12, 2016, at 6:29 PM, Yogi Devendra <yogidevendra@apache.org>
> >> wrote:
> >>>>
> >>>> Hi,
> >>>>
> >>>> Currently, for the usecases involving writing POJO to file output we
> use
> >>>> DAG simillar to following:
> >>>> POJOEmittingOperator->CSVFormatter->StringFileOutputOperator.
> >>>>
> >>>> Both CSVFormatter as well as StringFileOutputOperator have
> >>>> configurable tupleSeperator defaulting to newline. This introduces two
> >>>> newlines for every tuple. We should make one of the defaults to empty
> >>>> string and the other to be newline.
> >>>>
> >>>> Any opinions on which default should be empty string?
> >>>>
> >>>> ~ Yogi
> >>>
> >>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message