Contributing to Mir libraries
jmh530
john.michael.hall at gmail.com
Tue Aug 18 15:11:58 UTC 2026
On Tuesday, 18 August 2026 at 13:18:17 UTC, Serg Gini wrote:
> On Tuesday, 18 August 2026 at 12:57:51 UTC, Iskaban10 wrote:
>> Hi everyone,
>>
>> I’m planning to submit a proposal for SAOC this year and
>> wanted some feedback on my idea.
>>
>> I am interested in mathematics and optimization, and am
>> thinking of expanding the Mir libraries. Looking at the
>> repositories, it seems like these libraries aren't actively
>> maintained at the moment.
>
> I think jmh530 is still maintaining it =)
>
I would distinguish between maintaining and actively adding new
functionality. I don't have a lot of free time right now to add
new functionality*, but I've been checking on things and pay
attention to new issues or PRs.
mir-algorithm was still compiling for me back in February, so I
see no reason why it wouldn't work fine now. There are some
failures in the CI, but they were dependent on getting some
things fixed in LDC, which I believe subsequently happened. So
the test suite should pass more of the options now.
With respect to optimization, my recommendation would be to just
use importC and bring in something like nlopt (or something
else). Why reinvent the wheel? At one point I had a wrapper for
nlpot, but I wrote it a number of years ago and haven't used it
much since then). I can't recall using mir-optim. For that
matter, I think there is scope in D for bringing in more
numerical libraries with importC.
With respect to the path forward generally on mir, Ilya (@9il)
was the driving force behind mir-algorithm. He continues to
provide support, but he is not not adding new features and does
not want any breaking changes. I think that the folks at Symmetry
still make use of it, so you might want to reach out to someone
who works with them to get a better sense of what they might
need/want most.
* The outstanding work I never got around to finishing for
mir-stat was support for histograms and adding in blas support so
that I could do multiple regression
More information about the Digitalmars-d
mailing list