On 07/23/2014 09:59 PM, Leonardo Collado Torres wrote:
On Thu, Jul 24, 2014 at 12:17 AM, Martin Morgan <mtmor...@fhcrc.org> wrote:
On 07/23/2014 06:41 PM, Leonardo Collado Torres wrote:
Finally, regarding the issue of pushing new features to devel versions
(and not release), I understand the reasons for doing so. In my case,
what I have been trying to do to minimize major differences between
the versions is to keep working on the package outside of BioC until
we are more confident on its stability. Kind of pre-devel. Pre-devel
users (just a handful) can then install it via
devtools::install_github(). I understand that not every package
workflow is like this, but well, it could be a suggestion worth
mentioning athttp://www.bioconductor.org/developers/package-guidelines/
Though of course, maybe it's better to have "pre-devel" packages be
submitted to BioC-devel and drive all the traffic through BioC. Just
some thoughts.
Frankly, this doesn't sound like a good idea to me. It confuses the user
(what, I'm supposed to be using github? I thought this was a Bioconductor
package!), makes your life as a developer difficult (three versions to
maintain), and introduces code (i.e., bugs!) where it isn't tested.
I'd just go for broke, make wild changes to your devel code, let the build
system do it's magic, get leading-edge users to knowingly try out your
changes, shake out some bugs, and have a nice clean ride into release and a
grateful audience of appreciative users. Sure work in git if that's your
flavor, but merge frequently with master and sync with bioc; never expect
your users to install 'the latest' from github.
I agree with what you said. I was just talking about the development
stage before you submit the package the first time to BioC. After it's
>>> Finally, regarding the issue of pushing new features to devel versions
>>> (and not release), I understand the reasons for doing so. In my case,
;)
in BioC then yes, don't ask users to install from GitHub. And well, as
I understand it, the Git-SVN bridge makes it easy to push to
Bioc-devel and Bioc-release (following the suggestion posted here
https://www.mail-archive.com/bioc-devel@r-project.org/msg01967.html).
Or what system do you suggest using for that first development cycle
before submitting to BioC?
There are so many ways in which code ends up as a bioc package that I don't
think I have anything useful to say here.
and...
Our way of hosting the pre-devel version on GitHub without any
publicity and iterating there until we feel the package is mature for
BioC seems to be working for us. But well, it might be slow and we
could be losing out a lot by not having it in Bioc-devel. On the other
hand, we wanted to avoid confusing users with major changes to the
package. In the end, I'm torn between the "you only make a first
impression once, so make it good" approach vs "get it out there, then
fix it" approach.
these shouldn't be mutually exclusive; aim for a well-defined problem and write
effective code for it. Things go wrong for me when I try to be too ambitious,
take short cuts hoping to make fast progress, or make public code that I was
really still just thinking through.
Martin
In the end, I'm a newbie at developing packages and I'm still learning ^^'
Martin
--
Computational Biology / Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N.
PO Box 19024 Seattle, WA 98109
Location: Arnold Building M1 B861
Phone: (206) 667-2793
--
Computational Biology / Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N.
PO Box 19024 Seattle, WA 98109
Location: Arnold Building M1 B861
Phone: (206) 667-2793
_______________________________________________
Bioc-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/bioc-devel
