Files saved with double extension about arc_ecto HOT 4 OPEN

This happens to me too. I lost the custom filename as a quick fix, but obvs that's not ideal.

from arc_ecto.

@joshuaswilcox
Try:

def filename(version, {file, _scope}), do: "#{version}_#{Path.basename(file.file_name, Path.extname(file.file_name))}"

See:
Arc Local Tests

from arc_ecto.

Ah - Thanks @davidkuhta.

As I currently have it, the "filename" should not include the extension, because different versions of a file may have different extensions. (i.e. you may want the original jpg file stored, but a thumbnail png).

If you have any suggestions on how to make that more clear I'm all 👂

from arc_ecto.

Hi @stavro
If I trace the code right I think this is actually more of an Arc consideration.
I believe the confusion stems from some examples in the Arc docs, that don't necessarily allude when you override the filename/2 function in storage.ex, you lose the default extension stripping:
def filename(_, {file, _}), do: Path.basename(file.file_name, Path.extname(file.file_name))

Particularly Local Configuration and File Names that show:
def filename(version, {file, scope}), do: "#{version}-#{file.file_name}"

So my thought would be to either:

1. Provide a comment in the Arc docs that communicates file_name != basename (and includes Path.basename(file.file_name, Path.extname(file.file_name)))
2. (My preference) Expose a helper function (say filename/1 in Arc.Definition.Storage) that strips the extension (ultimately what the default filename/2 was doing).
   So in storage.ex you'd now have:

def filename(_, {file, _}), do: filename(file)
def filename(file), do: Path.basename(file.file_name, Path.extname(file.file_name))

Option 2. would mean you could do things like:
def filename(version, {file, _scope}), do: "#{version}_#{filename(file)}"
It also wouldn't break anything because the filename/2 still provides the same implementation as it did before.

Note: I chose filename because basename/1 is used by Path but there's nothing to say it couldn't just be that. (Which IMHO is what I would've assumed Path.basename/1 did...) (stemname/1 was another thought)

Anyhow, let me know if either fits your vision for arc/arc_ecto and I'll be happy to send over a PR.

from arc_ecto.