Skip to content

[Bioc-devel] jupyter notebooks as vignettes?

3 messages · Martin Morgan, Gabriel Becker, Michael Lawrence

#
I'm forwarding a couple of messages from the support site.

I'd appreciate some discussion.

It seems like jupyter notebooks are becoming increasingly plausible as a 
possible vignette format, and I was wondering whether there is any real-world 
experience with these, including their pros and cons as vignettes. For instance...

Developer / build system

- Presumably these notebooks need to be run as part of the package check -- no 
sense distributed non-functional notebooks. Are there facilities for this, along 
the lines of a traditional vignette that builds and potentially generates errors 
during R and / or document processing?

- Are notebooks actually mature enough that, e.g., their format will not change 
dramatically?

End-user

- I guess most end users will get static, evaluated versions of the notebooks, 
unless they install python & friends. Is that right?

- Can the user easily obtain the equivalent of a Stangle'd vignette, i.e., a 
runnable R script?

- If a typical user only has access to a static version of the notebook, then 
what have we gained relative to traditional vignettes?

-------- Forwarded Message --------
Subject: 	[bioc] A: BiocCheck: REQUIRED: vignette sources in vignettes/
directory (for custom vignet
Date: 	Wed, 20 May 2015 16:32:24 -0000
From: 	Dan Tenenbaum [bioc] <noreply at bioconductor.org>
Reply-To: 	reply+9665d4a8+code at bioconductor.org
To: 	mtmorgan at fhcrc.org



Activity on a post you are following on support.bioconductor.org
<https://support.bioconductor.org>

User Dan Tenenbaum <https://support.bioconductor.org/u/4256/> wrote Answer:
BiocCheck: REQUIRED: vignette sources in vignettes/ directory (for custom
vignettebuilder) <https://support.bioconductor.org/p/67884/#67885>:

BTW, questions about package development and submission should go to the
bioc-devel mailing list <http://bioconductor.org/help/support/#bioc-devel>.

We don't yet have a policy about .ipynb vignettes and we'd have to make sure the
build system is capable of building them. However, I would ignore the error for
now and you can discuss it with your reviewer when you submit the package.

--------------------------------------------------------------------------------

-------- Forwarded Message --------
Subject: 	[bioc] BiocCheck: REQUIRED: vignette sources in vignettes/ directory
(for custom vignettebuilder)
Date: 	Wed, 20 May 2015 16:29:49 -0000
From: 	flying-sheep [bioc] <noreply at bioconductor.org>
Reply-To: 	reply+c01911bc+code at bioconductor.org
To: 	mtmorgan at fhcrc.org



Activity on a post you are following on support.bioconductor.org
<https://support.bioconductor.org>

User flying-sheep <https://support.bioconductor.org/u/7779/> wrote Question:
BiocCheck: REQUIRED: vignette sources in vignettes/ directory (for custom
vignettebuilder) <https://support.bioconductor.org/p/67884/>:

I?m using a .ipynb notebook as source for my Vignette, so I created a Vignette
engine inside my package to build it.

That one registers a the necessary extensions, and R CMD build obediently
creates a vignette from the .ipynb file.

Yet R CMD BiocCheck still says `REQUIRED: vignette sources in vignettes/
directory`. should I ignore that when submitting my package, or will the
bioconductor people have to fix that check beforehand?

--------------------------------------------------------------------------------

You may reply via email or visit https://support.bioconductor.org/p/67884/
#
Martin,

I have a package on Github (DynDocModel) that was part of my thesis which
will happily process jupyter notebooks (they were called IPython Notebooks
at the time).  It (IIRC) supports tangling, weaving, restructuring, etc.

I can look at productionizing it if that is something Bioc thinks is
important. Right now it is tied in pretty closely with my caching mechanism
(another part of my thesis), and supports non-linear notebooks (the third
part), but those features are probably "lower priority" for what you want.

~G

On Wed, May 20, 2015 at 10:15 AM, Martin Morgan <mtmorgan at fredhutch.org>
wrote:

  
    
#
I would guess a potential major gain for the users would be the ability to
interact with the vignette on the Bioc site, where notebooks would run
inside Docker containers or something. Beyond software demonstration, this
would be very useful for reproducible research.

On Wed, May 20, 2015 at 10:15 AM, Martin Morgan <mtmorgan at fredhutch.org>
wrote: