[Boostheaders-devel] Building RStan with BH

[Ben Goodrich]

Hi Dirk,

A few weeks ago, we have managed to build the rstan package using LinkingTo
for BH (and RcppEigen) but had to include the result of bcp for
circular_buffer.hpp ourselves. I don't even know what the circular_buffer
sublibrary does, but it is only about 5 files, so if you include it the
next time you update BH, then we can drop it from rstan.

For its next major release, Stan is likely to merge pull requests that
involve the multiarray and odeint sublibraries of Boost. These are less
small but presumably useful to other R projects besides rstan. We can
include those in rstan like we do for circular_buffer if you really don't
want them in BH for some reason.

Finally, if you are doing a poll, we would be in favor of bumping BH to
Boost 1.56 sometime this year if no major bugs emerge (fingers crossed)
from its reoganization / transition to git.

No rush on any of this. Thanks!

Ben


On Wed, Oct 23, 2013 at 10:15 AM, Dirk Eddelbuettel <edd at debian.org> wrote:

>
> Hi Ben,
>
> On 23 October 2013 at 09:10, Ben Goodrich wrote:
> |     I can do that and R-Forge will have a traball for you, I can also
> make a
> |     prelim tarball available (this eve).  Could you then test?
> |
> |
> | Yes, I can test any time. But there is no real rush since it would take
> us a
>
> Perfect. Let's put this on the TODO list. I want to rejig a few things
> anyway
> (ie no longer start bcp with a look at bigmemory a la Jay and Mike but make
> it run over an explicit boost libs list) so there will be testing.
>
> | while to get CRAN-compliant anyway. We would need to split the existing
> RStan
> | into three packages in order for each to compile within a few minutes.
>
> AFAIK there is no compile-time limit. There is one for tests.  (My 10-years
> old RQuantLib package is a beast, also due to heavy Boost and C++ use in
> QuantLib. Try (re-) building that for a comparison -- on Ubuntu/Debian you
> just need libquantlib-dev install. Might be useful for you in discussing /
> deciding what to do with RStan...)
>
> Dirk
>
> --
> Dirk Eddelbuettel | edd at debian.org | http://dirk.eddelbuettel.com
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/boostheaders-devel/attachments/20140807/a99946ff/attachment.html>