I've been working through a few different approaches. What's shaping up is something more general, having a special value for directory parameters (i.e. 'attach) and auto-detection of link paths that are in the attachment directory.
The latest iterations don't move any files around, so can't actually enforce the output directory. That makes it safer overall as with my initial patch if you were to return a path to something you didn't want moved to your attachment directory you might get very surprising results.

I'll post a new patch with a different approach in a little bit.

On Mon, Oct 4, 2021 at 5:06 PM Christopher M. Miles <numbchild@gmail.com> wrote:

Ihor Radchenko <yantar92@gmail.com> writes:

Greg Minshall <minshall@umich.edu> writes:

i can imagine wanting to have input files and output files in separate directories. (for ease in "make clean", if for no other conceptual reason.) (but, probably i don't understand.)

I agree with this thought. We should separate two directories.

Makes sense. Currently, there is :dir header arg to set working directory (aka input files directory). Maybe we can introduce something like :results-dir header arg to set the output directory? It's value can be a directory path or symbol 'attach.

`:results file :results-dir 'attach` will be equivalent of `:results file attach` in the patch proposed by Ryan Scott.

WDYT?

I agree with this idea. Use :results-dir 'attach is better.

Will the patch be updated?

Best, Ihor


<#secure method=pgpmime mode=sign>
--
[ stardiviner ]
       I try to make every word tell the meaning that I want to express.

       Blog: https://stardiviner.github.io/
       IRC(freenode): stardiviner, Matrix: stardiviner
       GPG: F09F650D7D674819892591401B5DF1C95AE89AC3