Re: rmarkdown-like production of multiple plots in org

[John Hendy]

On Tue, Mar 31, 2020 at 2:22 PM Matt Price <address@hidden> wrote:
>
> I'm completely new to R.
>
> I've started working with a project that creates plots using the ggplot 
> package -- so by default it creates grid objects, rather than writing to 
> files.
>
> In rmarkdown/rstudio, I can write something like this in a SOMEFILE.Rmd :
>
> ```
> install_github('eeholmes/CoV19')
> library(CoV19)
> getdata();
> plot4(world, 'Ontario Canada')
> plot2(world, 'Italy')
> plot4(states, "WA")
> ```
>

Interesting. I hadn't really thought that approach through. For
exploratory analysis, this sounds awesome and I don't think I've ever
tried multiple plots in a single chunk in RStudio. Only after seeing
Thomas' reply just now did I realize this isn't just plot(), though...
where are those functions from? There might be hidden conveniences
that don't apply to pure ggplot() calls... dunno.

I will add that as soon as you want to start tailoring sizes or how
these appear in the resultant file, I think you'd have to split these
into separate chunks in order to set the options, no?

> I sort of love how the rmarkdown package will just create all 3 of those 
> plots, save them to auto-named files, and render to HTML.  In RStudio, 
> running just that block will also create all three blocks and display them in 
> the editor.
>
> By contrast, creating a series of many plots in org is fairly tedious.  I 
> have to name the plot individually & put each function call in its own src 
> block. Is there any way to mimic the behaviour of rmarkdown instead? I odn't 
> understand babel or R enough to really even see how something like that could 
> be implemented, but I'd appreciate some pointers.  Thank you!

Perhaps an alternative is running all your plots in one block, but
using ggsave() (or similar) to save out the files directly (vs. using
:output/:file to do it). Then you could have file links in the org
file instead of the typical 1:1 match-up of a single block to a single
result. I did this once during an effort to optimize a neural network.
I had a big loop iterating through parameters, and would
programmatically save out one residual plot per combination, e.g.
model_var1-value_var2-value_etc.png, I also generated an org-mode
section and exported headers for each combination, embedding the
corresponding [[./plots/foo.png]] image link within that heading.
Exporting to pdf let me page through all my residual plots handily to
compare them.

Anyway, maybe something helpful in there?

John