Re: Output to directory patch

[Al Haji-Ali]

Hi Ikumi,

Thank you for taking the time to review my patch.

On 23/02/2021, Ikumi Keita wrote:
> 1. When a subfile is located in subdirectory, region typeset causes
> error, even when TeX-output-dir is nil.
>
Indeed, this is a bug that I can probably fix but see below.

> 2. Suppose that TeX-output-dir is non-nil in the subfile while it is
> nil in the main file. This inconsistency can lead to endless loop in
>
Agreed. The value of TeX-output-dir in master should be the correct one (the 
value in the subfile should be ignored). The infinite loop was unexpected but 
should be easy to fix.

> I don't think it's a good idea to modify `TeX-master-file' and
> `TeX-region-file'. Their usage and implementation in AUCTeX have a lot
> of subtleness, so modifying them is fragile in general. I can't
> guarantee that shortcomings other than the above two don't come out
> later.
>
>
> Hence I'd recommend to make the basic design more similar to Mosè's
> approach:
> (1) Keep `TeX-master-file' and `TeX-region-file' as-is.
> (2) Make a function dedicated to build a path (including the file name)
>     for generated files such as .pdf, .dvi, .aux etc.
> (3) Use that dedicated function for "%o" and alikes in
>     `TeX-expand-list(-builtin)' and everywhere AUCTeX calls
>     `TeX-check-path'.
> (4) Adjust `TeX-clean' and `TeX-view' to accomodate those changes.


There are two things my patch is trying to do:
- Implement `--output-directory` options in tex tools. This requires a minimal 
change to TeX-master-file which is ostensibly an automatic version of what you 
suggested above. If 'extension' is an output extension like "pdf", "dvi" or etc 
and TeX-output-dir is not nil then the function TeX-output-master-file is 
called which constructs the paths as you suggest. This should mean that the 
return value of TeX-master-file is different only for those extensions and only 
when `TeX-output-dir` is non-nil. This was a decision that I made early on 
since the full implementation that you suggested will require replacing 
(manually) any calls to TeX-master-file with calls to TeX-output-master-file 
based on the extension.

- Allow `TeX-region` to have a sub-directory. I thought this was an easy fix to 
TeX-region-file but it turned out to cause many headaches (including the first 
bug above and some of the previous bugs related to preview-latex).

While the precise implementation is being discussed here, I will remove the 
changes related to allowing TeX-region to be a sub-directory to increase 
reliability and fix the infinite loop bug above.

>
> I suppose that (1)-(4) would be almost inevitable to implement reliably
> the output-diretory feature. In addtion, they would gain another
> advantage of being free of complication discussed below.
>
Hmm.. I don't think the complications I talked about in TeX-command and 
TeX-save-document are a result of the patch implementation but rather are a 
result of these functions accepting file and file-function arguments instead of 
calling the relevant functions (like TeX-active-master) but not passing the 
file-function down to other functions.

Even a full implementation would have to change the interface of these commands 
to allow passing of two functions and/or file paths (one for the normal path 
and one for the output one).

Best regards,
Al