Hi René & Taylor,

On Wed, 26 Jan 2022, Taylor Blau wrote:

> On Wed, Jan 26, 2022 at 10:34:04AM +0100, René Scharfe wrote:
> > Am 26.01.22 um 09:41 schrieb Johannes Schindelin via GitGitGadget:
> > > Note: originally, Scalar was implemented in C# using the .NET API, where
> > > we had the luxury of a comprehensive standard library that includes
> > > basic functionality such as writing a `.zip` file. In the C version, we
> > > lack such a commodity. Rather than introducing a dependency on, say,
> > > libzip, we slightly abuse Git's `archive` command: Instead of writing
> > > the `.zip` file directly, we stage the file contents in a Git index of a
> > > temporary, bare repository, only to let `git archive` have at it, and
> > > finally removing the temporary repository.
> >
> > git archive allows you to include untracked files in an archive with its
> > option --add-file.  You can see an example in Git's Makefile; search for
> > GIT_ARCHIVE_EXTRA_FILES.  It still requires a tree argument, but the
> > empty tree object should suffice if you don't want to include any
> > tracked files.  It doesn't currently support streaming, though, i.e.
> > files are fully read into memory, so it's impractical for huge ones.

That's a good point.

I did not want to invent any `fast-import`-like streaming protocol just
for the sake of supporting the "funny" use case of `scalar diagnose`, so I
invented a new option `--add-file-with-content=<path>:<content>` (with the
obvious limitation that the `<path>` cannot contain any colon, if that is
desired, users will still need to write out untracked files).

> Using `--add-file` would likely be preferable to setting up a temporary
> repository just to invoke `git archive` in it. Johannes would be the
> expert to ask whether or not big files are going to be a problem here
> (based on a cursory scan of the new functions in scalar.c, I don't
> expect this to be the case).

Indeed, it is unlikely that any large files are included.

> The new stage_directory() function _could_ add `--add-file` arguments in
> a loop around readdir(), but it might also be nice to add a new
> `--add-directory` function to `git archive` which would do the "heavy"
> lifting for us.

I went one step further and used `write_archive()` to do the
heavy-lifting. That way, we truly avoid spawning any separate process let
alone creating any throw-away repository.

Ciao,
Dscho