Re: [eigen] Numerical Integration module

Pavel,

I noticed you have some quadrature code in C at http://www.holoborodko.com/pavel/numerical-methods/numerical-integration/. Do you think the function gauss_legendre_tbl would function if it were adapted (converted to C++ and templated) to use mpreal? If so, would you be willing to impart a copy, or portion of gauss_legendre.c under the MPL2 license to Sree and I for adapting to include in the Eigen Quadrature module? Of course, I am happy to include any attribution notes, so long as they are compatible with MPL2. Feel free to contact me at my personal email if you prefer.

Best,

Jeff

On Tue, Aug 26, 2014 at 1:44 AM, Jeff J <complexzeros@xxxxxxxxx> wrote:

I submitted a pull request with some quadrature code. It uses hard-coded abscissae and weights, and it doesn't allow for custom scalars yet. There are a few details to work out, but it seems to function well.

Sree, I think we can start working on combining our code.

On Mon, Aug 18, 2014 at 10:01 AM, Jeff J <complexzeros@xxxxxxxxx> wrote:

Hi Sree,

I'm working to get my code in decent shape by the end of this week. It will use hard-coded doubles, but we should still be able to combine our code. I can work on custom scalars after that. Feel free to contact me at my personal email directly.

On Fri, Aug 15, 2014 at 6:10 AM, Sreekumar Thaithara Balan <tbs1980@xxxxxxxxx> wrote:

I have a working version of a code implementing the Laurie 1997 algorithm. You can find it at https://github.com/tbs1980/GaussKronrod .

Jeff, may we should have a chat with each other on how to combine the codes. As Jitse suggested below, I am now reading the documentation on contributing to Eigen!

Sree

On 6 Aug 2014, at 08:34, Jitse Niesen <jitseniesen@xxxxxxxxx> wrote:

On Tue, 5 Aug 2014, Jeff J wrote:

Thanks Sree! We can fork Eigen when we're ready to start combining our work.
I'm not sure how to go about submitting a new unsupported module. I guess
just follow the existing format and submit a pull request?

Yes, a pull request would probably be best.

I think you do not need to wait until the module has all the bells and whistles before you submit a pull request. I'd argue that it is better to work incrementally: implement a simple case (e.g. no custom scalars, fixed order, fixed and hard-coded abscissae), include some documentation and test cases, and submit a pull request for that. That should hopefully give you some feedback; at least with a pull request you can prod me and the other core developers to do something. Then you can extend to include all the nice functionality. All IMHO of course.

Jitse