Dear all,
the mailing list analysis part of Codeface relies substantially
on package snatm. This package was written and published by Angela
Bohn, and I've provided numerous changes. So far, no one else seems
to have contributed to the package.
Mitchell Joblin recently created a patch for snatm to get it
working with the current revisions of package TM. This raises
the question how to handle development of snatm in the future.
To me, it seems that Codeface is the only major user of snatm these
days, so we should think about how to better integrate the two
approaches.
While I do have commit access to snatm, I don't think maintaining
snatm separately has any real benefit. Besides, snatm does not have a
mailing list where patches can be discussed, so we would have to do
this on the Codeface list anyway.
I can think of two ways how to proceed:
1.) Integrate the snatm source base into Codeface, turning it from
an external package into native code.
This would require some development effort for the transision, and
would make a stand-alone deployment of snatm impossible (I don't
assume the latter exists, anyway)
2.) Move a copy of the snatm repository to github. Discuss development
on the codeface list, and require a github installation of snatm
for codeface. To provision snatm, I can think of two ways
a) Use devtools::github_install() for snatm when provisioning
codeface
b) Integrate snatm as a submodule into codeface, and require the
package to be installed from this local source when provisioning.
I tend towards 2.a), but would like to hear if there are any arguments
against this solution. The idea is that changes to snatm can be handled
using the same tools and processes as for Codeface, which should
benefit everyone. Mitchell, since you'll be likely working with snatm
a lot for current research, your opinion on this is particularly
requested.
Thanks & best regards, Wolfgang Mauerer
