On 17 January 2014 00:57, Gary Gregory <garydgreg...@gmail.com> wrote:
> On Thu, Jan 16, 2014 at 7:24 PM, sebb <seb...@gmail.com> wrote:
>
>> On 16 January 2014 20:58, Gary Gregory <garydgreg...@gmail.com> wrote:
>> > On Thu, Jan 16, 2014 at 7:49 AM, Emmanuel Bourg <ebo...@apache.org>
>> wrote:
>> >
>> >> Le 15/01/2014 14:04, Gary Gregory a écrit :
>> >>
>> >> > Uh, I want to go the other way around. See my use case.
>> >>
>> >> Why isn't this suitable to your use case? You could write:
>> >>
>> >> factory.create(record.toMap());
>> >>
>> >> and then remove the create(CSVRecord) method with the duplicated
>> >> implementation.
>> >>
>> >> What am I missing?
>> >>
>> >
>> > That would work. What is still not clean or OO is that toMap() means
>> > nothing when no headers are defined.
>> >
>> > I've split the record into a mapped record subclass here:
>> >
>> > https://issues.apache.org/jira/browse/CSV-104
>> >
>> > Thoughts?
>> >
>> > We can still have toMap() implemented on top of the code now, but it
>> would
>> > have to return null if there are no headers.
>>
>> Why not return an empty Map?
>>
>
> An empty map is, well, empty but the record is not empty, so that's kinda
> confusing,
Depends on the method Javadoc whether it is confusing or not.
If toMap() is defined to return the empty Map if there are no column
names, then that is clear to me.
AFAICT an empty map cannot be generated otherwise.
Or is there a situation where an empty map means something usefully
distinct from no header names?
> whereas null is clearly something that does exist. If toMap() is
> only in the subclass, then there is no problem.
>
> The alternative is to create a real map with columns name "Column1",
> "Column2", and so on. But then that would be confusing because
> record.get("Column1) would give you null whereas
> record.toMap().get("Column1") would give you a value. Then we get to talk
> about having "Column1", "Column2"... as the default column names if none
> are provided... kinda too clever.
Yuk.
> All of this to say that over time it is likely as we need to provide more
> APIs, and it seems easier to say null than an empty structure. First
> toMap(), then toProperties() and so on.
I don't think that follows.
> The KISS solution is toMap() -> null for unmapped records IMO.
IMO the empty map is better.
> Gary
>
>
>>
>> > Gary
>> >
>> >
>> >>
>> >> Emmanuel Bourg
>> >>
>> >>
>> >> ---------------------------------------------------------------------
>> >> To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
>> >> For additional commands, e-mail: dev-h...@commons.apache.org
>> >>
>> >>
>> >
>> >
>> > --
>> > E-Mail: garydgreg...@gmail.com | ggreg...@apache.org
>> > Java Persistence with Hibernate, Second Edition<
>> http://www.manning.com/bauer3/>
>> > JUnit in Action, Second Edition <http://www.manning.com/tahchiev/>
>> > Spring Batch in Action <http://www.manning.com/templier/>
>> > Blog: http://garygregory.wordpress.com
>> > Home: http://garygregory.com/
>> > Tweet! http://twitter.com/GaryGregory
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
>> For additional commands, e-mail: dev-h...@commons.apache.org
>>
>>
>
>
> --
> E-Mail: garydgreg...@gmail.com | ggreg...@apache.org
> Java Persistence with Hibernate, Second
> Edition<http://www.manning.com/bauer3/>
> JUnit in Action, Second Edition <http://www.manning.com/tahchiev/>
> Spring Batch in Action <http://www.manning.com/templier/>
> Blog: http://garygregory.wordpress.com
> Home: http://garygregory.com/
> Tweet! http://twitter.com/GaryGregory
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
For additional commands, e-mail: dev-h...@commons.apache.org
