Skip to content

[R-pkg-devel] Announcing Bazel for R packages

4 messages · Siddhartha Bagaria, Dirk Eddelbuettel

Siddhartha Bagaria
# 
Hello,

We have released some Bazel rules for R packages.
https://github.com/grailbio/rules_r

What is Bazel?

Bazel <https://bazel.build/> is an industry tested build system that can be
used for continuous integration and testing in large monorepos (multiple
software libraries as part of one repository).

When should you use Bazel for your R packages?

1. Continuous integration and testing
If your organization already uses Bazel, and you have R packages as part of
the same repo. These rules will run unit tests or R CMD check on changes to
your package, or any dependency as part of the same build system that your
organization uses. You can also deploy docker images with your R packages
installed in them as part of your continuous integration.

2. Reproducible builds
Using Bazel will force you to fix versions of your package dependencies,
and every build will be reproducible. Updates to package dependencies can
then be vetted and be done incrementally if desired. This is an alternative
to maintaining a private repository of CRAN/Bioc packages or using a CRAN
snapshot service.

If these do not concern you, then you should not consider Bazel because of
the additional complexity introduced.

I am happy to answer any questions you may have.

Thank you.

Sid
Dirk Eddelbuettel
#
On 21 September 2017 at 22:40, Siddhartha Bagaria wrote:
| We have released some Bazel rules for R packages.
| https://github.com/grailbio/rules_r
| 
| What is Bazel?
| 
| Bazel <https://bazel.build/> is an industry tested build system that can be
| used for continuous integration and testing in large monorepos (multiple
| software libraries as part of one repository).
| 
| When should you use Bazel for your R packages?
| 
| 1. Continuous integration and testing
| If your organization already uses Bazel, and you have R packages as part of
| the same repo. These rules will run unit tests or R CMD check on changes to
| your package, or any dependency as part of the same build system that your
| organization uses. You can also deploy docker images with your R packages
| installed in them as part of your continuous integration.
| 
| 2. Reproducible builds
| Using Bazel will force you to fix versions of your package dependencies,
| and every build will be reproducible. Updates to package dependencies can
| then be vetted and be done incrementally if desired. This is an alternative
| to maintaining a private repository of CRAN/Bioc packages or using a CRAN
| snapshot service.
| 
| If these do not concern you, then you should not consider Bazel because of
| the additional complexity introduced.
| 
| I am happy to answer any questions you may have.

Are you aware of any bazel users outside of Google and its spin-offs,
including Grail?

Dirk
Siddhartha Bagaria
# 
I agree the open source version has not been around for that long.

But... It is picking up adoption within many other companies this year:
https://github.com/bazelbuild/bazel/wiki/Bazel-Users

And some adoption in academia as well, e.g.
http://drake.mit.edu/bazel.html
On Thu, 21 Sep 2017 at 17:18 Dirk Eddelbuettel <edd at debian.org> wrote:

            

      
      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Dirk Eddelbuettel
    

    

      
      #
    

  


  

      
On 22 September 2017 at 00:28, Siddhartha Bagaria wrote:

        
| I agree the open source version has not been around for that long.
| 
| But... It is picking up adoption within many other companies this year:
| https://github.com/bazelbuild/bazel/wiki/Bazel-Users
| 
| And some adoption in academia as well, e.g.
| http://drake.mit.edu/bazel.html

Nice, thanks!

Dirk