You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@uima.apache.org by jo...@refinitiv.com on 2018/12/11 17:41:24 UTC
What's the future for C++ UIMA?
Dear UIMA committers,
I've been following this list for a number of years, and I have benefitted
from UIMA by using it in industry research as well as in academic teaching
(thank you, IBM, for releasing it under Apache 2 license!).
One of the things that would need a bit of work, in my view, is getting the
C++ implementation on par with the Java implementation, and I'm aware this
is a tough call because C++ changed so much, and C++ compilers compare many
different languages whereas Java compilers all appear to be more 'aligned'.
But the current state of the C++ implementation is "less well attended and harder
to get to compile", which limits the potential impact UIMA can have (in
principle, it permits Java, C++ and also mixed projects). What I would love to see
is more activity on the C++ side, including;
- feature parity with the Java version;
- quality parity (in terms of ongoing testing efforts and the ability to get
the system built). For example, the uima-dev Debian package could be a
wrapper around uima-dev-cpp and uima-dev-java packages; I would like to
see an integrated, simplified build approach (e.g. batch script that
triggers C++ (bjam/gmake/CMake) and Java (Maven) build systems).
- parity of documentation (hard to maintain two parallel documents; could
they be integrated with Java/C++ code snippets? There are great tools for
that like Slate - https://github.com/lord/slate).
There appears to be a broader opportunity to re-factor the C++ version so
that it cam make use of the latest C++11 idioms.
I believe the result would re-vitalize the adoption, in particular there has
been a lot of work gone into Python toolkits, and that world is presently
locked out of UIMA; a refresh of the C++ implementation permits wrapping
UIMA s a Python package, so here's another ask:
- provide a Python wrapper for UIMA (Python module 'uima’). I can’t wait
to type ‘import uima’ in a Jupyter/IPython notebook!
Are there folks out there who also think the above suggestions are good
ideas, and if so, have time to implement some of that?
Best
Jochen
PS: Thanks to all developers who have made UIMA what it is (not just core
devs, but also component contributors)
Jochen L. Leidner, Ph.D.
Director of Research, Research & Development
Refinitiv Labs
The Financial and Risk business of Thomson Reuters is now Refinitiv<http://www.refinitiv.com/>
[cid:A7913272-7D78-455F-AB40-5FA7D61C79CF]<http://www.refinitiv.com/>
________________________________
This e-mail is for the sole use of the intended recipient and contains information that may be privileged and/or confidential. If you are not an intended recipient, please notify the sender by return e-mail and delete this e-mail and any attachments. Certain required legal entity disclosures can be accessed on our website.<https://www.refinitiv.com/>
Re: What's the future for C++ UIMA?
Posted by Marshall Schor <ms...@schor.com>.
Thanks, Dan, and welcome!
Please feel free to reach out for any (non- C++) questions on getting things
done here :-)
-Marshall
On 12/11/2018 12:59 PM, Daniel Gruhl wrote:
> Good morning!
>
> I am actually digging through the uimacpp code and trying to get it to at
> least compile with a modern C++ compiler. I have the core code done and am
> working through the test suite. Hopefully earlyish next year I'll have
> something to commit (it's a bit of a side project for me).
>
> Once that is done I agree it could use a refactor and clean up for sure!
>
> -= Dan
>
>
>
> On Tue, Dec 11, 2018 at 9:47 AM <jo...@refinitiv.com> wrote:
>
>> Dear UIMA committers,
>>
>> I've been following this list for a number of years, and I have benefitted
>> from UIMA by using it in industry research as well as in academic teaching
>> (thank you, IBM, for releasing it under Apache 2 license!).
>>
>> One of the things that would need a bit of work, in my view, is getting the
>> C++ implementation on par with the Java implementation, and I'm aware this
>> is a tough call because C++ changed so much, and C++ compilers compare many
>> different languages whereas Java compilers all appear to be more 'aligned'.
>>
>> But the current state of the C++ implementation is "less well attended and
>> harder
>> to get to compile", which limits the potential impact UIMA can have (in
>> principle, it permits Java, C++ and also mixed projects). What I would
>> love to see
>> is more activity on the C++ side, including;
>> - feature parity with the Java version;
>> - quality parity (in terms of ongoing testing efforts and the ability to
>> get
>> the system built). For example, the uima-dev Debian package could be a
>> wrapper around uima-dev-cpp and uima-dev-java packages; I would like to
>> see an integrated, simplified build approach (e.g. batch script that
>> triggers C++ (bjam/gmake/CMake) and Java (Maven) build systems).
>> - parity of documentation (hard to maintain two parallel documents; could
>> they be integrated with Java/C++ code snippets? There are great tools for
>> that like Slate - https://github.com/lord/slate).
>> There appears to be a broader opportunity to re-factor the C++ version so
>> that it cam make use of the latest C++11 idioms.
>>
>> I believe the result would re-vitalize the adoption, in particular there
>> has
>> been a lot of work gone into Python toolkits, and that world is presently
>> locked out of UIMA; a refresh of the C++ implementation permits wrapping
>> UIMA s a Python package, so here's another ask:
>> - provide a Python wrapper for UIMA (Python module 'uima’). I can’t wait
>> to type ‘import uima’ in a Jupyter/IPython notebook!
>>
>> Are there folks out there who also think the above suggestions are good
>> ideas, and if so, have time to implement some of that?
>>
>> Best
>> Jochen
>>
>> PS: Thanks to all developers who have made UIMA what it is (not just core
>> devs, but also component contributors)
>>
>>
>> *Jochen L. Leidner, Ph.D.*
>>
>> Director of Research, Research & Development
>>
>> Refinitiv Labs
>>
>>
>>
>> The Financial and Risk business of Thomson Reuters is now *Refinitiv
>> <http://www.refinitiv.com/>*
>>
>>
>>
>> <http://www.refinitiv.com/>
>>
>>
>>
>>
>>
>> ------------------------------
>>
>> This e-mail is for the sole use of the intended recipient and contains
>> information that may be privileged and/or confidential. If you are not an
>> intended recipient, please notify the sender by return e-mail and delete
>> this e-mail and any attachments. Certain required legal entity disclosures
>> can be accessed on our website. <https://www.refinitiv.com/>
>>
Re: What's the future for C++ UIMA?
Posted by José Tomás Atria <jt...@gmail.com>.
Heh, funny coincidence: I was just thinking about this the other day and
playing with the idea of doing what Dan mentioned myself. Good to see
others interested in this!
I'd be particularly interested in uimacpp to bring uima to the R world; I'm
currently using rJava to configure and run uima pipelines from R code, but
it would be great to have native support built over uimacpp. I'd be very
happy to collaborate on an eventual collective effort to bring uimacpp up
to speed with uimaj.
I'll take a look at the current uimacpp to familiarize myself with the code
and stay tuned on your efforts.
Best,
jta
On Tue, Dec 11, 2018 at 10:20 AM <jo...@refinitiv.com> wrote:
> Thanks, Dan, this is great news!
> -Jochen
>
>
> PS: I’m familiar with your PhD thesis work and also WebFountain, which I
> saw
> an early demo of.
>
>
> On 11/12/2018, 17:59, "Daniel Gruhl" <da...@gmail.com> wrote:
>
> >Good morning!
> >
> >I am actually digging through the uimacpp code and trying to get it to at
> >least compile with a modern C++ compiler. I have the core code done and am
> >working through the test suite. Hopefully earlyish next year I'll have
> >something to commit (it's a bit of a side project for me).
> >
> >Once that is done I agree it could use a refactor and clean up for sure!
> >
> > -= Dan
>
>
--
entia non sunt multiplicanda praeter necessitatem
Re: What's the future for C++ UIMA?
Posted by jo...@refinitiv.com.
Thanks, Dan, this is great news!
-Jochen
PS: I’m familiar with your PhD thesis work and also WebFountain, which I
saw
an early demo of.
On 11/12/2018, 17:59, "Daniel Gruhl" <da...@gmail.com> wrote:
>Good morning!
>
>I am actually digging through the uimacpp code and trying to get it to at
>least compile with a modern C++ compiler. I have the core code done and am
>working through the test suite. Hopefully earlyish next year I'll have
>something to commit (it's a bit of a side project for me).
>
>Once that is done I agree it could use a refactor and clean up for sure!
>
> -= Dan
Re: What's the future for C++ UIMA?
Posted by Daniel Gruhl <da...@gmail.com>.
Good morning!
I am actually digging through the uimacpp code and trying to get it to at
least compile with a modern C++ compiler. I have the core code done and am
working through the test suite. Hopefully earlyish next year I'll have
something to commit (it's a bit of a side project for me).
Once that is done I agree it could use a refactor and clean up for sure!
-= Dan
On Tue, Dec 11, 2018 at 9:47 AM <jo...@refinitiv.com> wrote:
> Dear UIMA committers,
>
> I've been following this list for a number of years, and I have benefitted
> from UIMA by using it in industry research as well as in academic teaching
> (thank you, IBM, for releasing it under Apache 2 license!).
>
> One of the things that would need a bit of work, in my view, is getting the
> C++ implementation on par with the Java implementation, and I'm aware this
> is a tough call because C++ changed so much, and C++ compilers compare many
> different languages whereas Java compilers all appear to be more 'aligned'.
>
> But the current state of the C++ implementation is "less well attended and
> harder
> to get to compile", which limits the potential impact UIMA can have (in
> principle, it permits Java, C++ and also mixed projects). What I would
> love to see
> is more activity on the C++ side, including;
> - feature parity with the Java version;
> - quality parity (in terms of ongoing testing efforts and the ability to
> get
> the system built). For example, the uima-dev Debian package could be a
> wrapper around uima-dev-cpp and uima-dev-java packages; I would like to
> see an integrated, simplified build approach (e.g. batch script that
> triggers C++ (bjam/gmake/CMake) and Java (Maven) build systems).
> - parity of documentation (hard to maintain two parallel documents; could
> they be integrated with Java/C++ code snippets? There are great tools for
> that like Slate - https://github.com/lord/slate).
> There appears to be a broader opportunity to re-factor the C++ version so
> that it cam make use of the latest C++11 idioms.
>
> I believe the result would re-vitalize the adoption, in particular there
> has
> been a lot of work gone into Python toolkits, and that world is presently
> locked out of UIMA; a refresh of the C++ implementation permits wrapping
> UIMA s a Python package, so here's another ask:
> - provide a Python wrapper for UIMA (Python module 'uima’). I can’t wait
> to type ‘import uima’ in a Jupyter/IPython notebook!
>
> Are there folks out there who also think the above suggestions are good
> ideas, and if so, have time to implement some of that?
>
> Best
> Jochen
>
> PS: Thanks to all developers who have made UIMA what it is (not just core
> devs, but also component contributors)
>
>
> *Jochen L. Leidner, Ph.D.*
>
> Director of Research, Research & Development
>
> Refinitiv Labs
>
>
>
> The Financial and Risk business of Thomson Reuters is now *Refinitiv
> <http://www.refinitiv.com/>*
>
>
>
> <http://www.refinitiv.com/>
>
>
>
>
>
> ------------------------------
>
> This e-mail is for the sole use of the intended recipient and contains
> information that may be privileged and/or confidential. If you are not an
> intended recipient, please notify the sender by return e-mail and delete
> this e-mail and any attachments. Certain required legal entity disclosures
> can be accessed on our website. <https://www.refinitiv.com/>
>