[Bioc-devel] new package IGV needs web browsers on the build machines to pass build & check

[Paul Shannon]

Many thanks, Vince! I will fix the printf and markdown errors.  
With release date now not far off, I propose (Bioc policymakers please weigh in if this is inadequate) to

- remove web browser communication from vignette, examples and unit tests
- evaluate stubs and/or "mocks" for possible use in a future release
- run full web browser tests on my own, separate from the bioc build

Any objections to this approach?

 - Paul


> On Apr 9, 2018, at 7:15 PM, Vincent Carey <stvjc at channing.harvard.edu> wrote:
> 
> Thanks Paul.  Package works.  Browser visualizations look really nice.
> 
> I tried to run your unit tests and they
> use a function called "printf".  Setting printf = sprintf allowed the tests
> to all run to completion.
> 
> The following vignette chunk is problematic ... maybe you meant eval=FALSE, echo=TRUE to show what the
> user needs to do?
> 
> ```{r createLoad, echo=TRUE, echo=FALSE, results='hide'}
> igv <- IGV()
> setBrowserWindowTitle(igv, "MEF2C")
> setGenome(igv, "hg19")
> ```
> 
> It does seem to me that you would need to avoid running
> the browser in tests and vignette and examples.  Using
> if (interactive()) can help with vignette and examples.
> The unit testing of such a package should I think focus
> on verifying that the essential data manipulations succeed,
> without requiring browser communication.  But others in the
> group will have more authoritative remarks on this.
> 
> I don't know whether the methods sketched in
> 
> https://stackoverflow.com/questions/15509231/unit-testing-node-js-and-websockets-socket-io
> 
> could be used in this task, to go beyond checking the basic
> R manipulations.
> 
> 
>> On Mon, Apr 9, 2018 at 8:15 PM, Paul Shannon <pshannon at systemsbiology.org> wrote:
>> Hi Vince,
>> 
>> My dumb mistake, sorry.  Now fixed, version 0.99.9, https://github.com/paul-shannon/IGV
>> 
>> I’m looking forward to hearing your suggestions.
>> 
>>  - Paul
>> 
>> 
>> 
>> > On Apr 9, 2018, at 4:49 PM, Vincent Carey <stvjc at channing.harvard.edu> wrote:
>> >
>> > Hi Paul -- I am trying to build your vignette but it has
>> >
>> > load("~/s/work/priceLab/AD/tbl.gwas.level_1.RData")
>> >
>> > I think it should be possible to get your package through
>> > check, but I would like to get the vignette built and
>> > check out the tests before commenting further.
>> >
>> >
>> > On Mon, Apr 9, 2018 at 1:52 PM, Paul Shannon <paul.thurmond.shannon at gmail.com> wrote:
>> > > "Once your package builds and checks without errors or (avoidable) warnings, a Bioconductor team member will provide a technical review of your package. Other Bioconductor developers and users with domain expertise are encouraged to provide additional community commentary. Reviewers will add comments to the issue you created.”
>> >
>> > I don’t believe my submission, IGV, quite fits the mold.  As best I can tell, no web browser can be summoned on any of the architectures by IGV during build or check.  So the package build always times out.  And so, according to the guideline quoted above, it will never be reviewed!
>> >
>> > IGV is a BrowserViz version >=2 subclass; BrowserViz was recently uprev’d in devel.  Like RCyjs, IGV's raison d’être is the provision of R access to interactive graphics in your web browser, in this case to Jim Robinson’s igv.js.
>> >
>> > My analysis may be incorrect.  Can this matter receive a little attention, despite the bioc submission guideline quoted above, in hopes that the package can move forward?
>> >
>> > Thank you,
>> >
>> >  - Paul
>> > _______________________________________________
>> > Bioc-devel at r-project.org mailing list
>> > https://stat.ethz.ch/mailman/listinfo/bioc-devel
>> >
>> 
> 

	[[alternative HTML version deleted]]