Brett Viren <bv@bnl.gov> writes:

> I want what may be two conflicting things:
>
> - Produce a monolithic HTML export of an main org file that #+include:'s
>   other org files.
>
> - Have links in the other org files that are valid both when the
>   monolith is exported to HTML and when visiting the individual org
>   files in Emacs.

I can see the rationale, but it is tricky.

> === a.org ===
> * A
>
> This is A.
> === b.org ===
> * B
>
> This is B.  Links:
> - [[file:a.org::*A][A by file with headline]].
> - [[A][A by headline only]].
>   
> === main.org ===
> ...
> * Includes
>
> #+include: a.org
> #+include: b.org

In this scenario, it indeed makes sense to replace links to a.org and
b.org with internal links.

However, someone who is exporting multiple files, including a.org/b.org
may actually want to keep links to a.org/b.org. In particular, when only
portions of a.org/b.org are included (like a single src block or certain
lines).

> * Links
>
> - A :: [[file:a.org]]
>
> - B :: [[file:b.org]]

And this is actually a scenario where it also makes more sense to keep
the links to a.org/b.org. If we follow your suggestion, these would be

> - A :: [[file:main.org]]
>
> - B :: [[file:main.org]]

which would look confusing.

For reference, I am attaching a patch that blindly converts links to
included files into links to the top-level includer. However, the patch
will fail for partially included files where the portions the links are
referencing are not actually included. It will also replace A:: and B::
links in the above example. So, it is not suitable for inclusion in Org
- we need some better idea how to handle other possible scenarios.

Suggestions welcome!
(For now, I do not see how we can handle the described scenarios
cleanly)