uima-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Richard Eckart de Castilho <...@apache.org>
Subject Re: CasIOUtils class - some meta-questions
Date Thu, 04 Aug 2016 09:16:17 GMT
+1 for URL.

-- Richard

> On 04.08.2016, at 09:14, Peter Kl├╝gl <peter.kluegl@averbis.com> wrote:
> 
> We need at least one of these because it is not really feasible to
> distinguish xcas and xmi only based on the content (first few bytes). I
> would prefer URL to Path and File because you can point to a file within
> a jar.
> 
> 
> Am 03.08.2016 um 22:49 schrieb Marshall Schor:
>> The "mitigating" factor would be if we could easily imagine a significant
>> sub-community of UIMA users appreciating these variants.  In this particular
>> case, I'm leaning toward agreeing with Richard, but am fine with having some
>> variants if needed by Peter. 
>> 
>> -Marshall
>> 
>> On 8/3/2016 4:32 PM, Richard Eckart de Castilho wrote:
>>> Under that policy, should we really introduce all kinds of variants using 
>>> File, Path, and URL not rather stick to InputStream (maybe to URL which
>>> incurs additional overhead opening/closing streams)?
>>> 
>>> -- Richard
>>> 
>>>> On 03.08.2016, at 22:29, Marshall Schor <msa@schor.com> wrote:
>>>> 
>>>> My general feeling: anytime we make something "public" it becomes "set in
>>>> stone".  So, it is best not to make things public ahead of clear
>>>> needs/use-cases.  That way, if at some point in the future, we find we now
have
>>>> a clear use-case for the loadBinary kind of thing, we can implement it then,
>>>> without any worries about backwards compatibility :-).
>>>> 
>>>> Kind of a "lazy" API creation, I guess.
>>>> 
>>>> -Marshall


Mime
View raw message