single file download

[joehand]

Having single file download would be cool. A couple options:

1. dat download <key>/file.txt, dat <key>/file.txt - download file with in memory metadata
2. dat clone <key>/file.txt - download file with sparse state

The first would be easier and can be implemented now. Second requires some sparse state tracker. This is because if you have a .dat folder and run dat sync it will try to download the whole dat by default, even if you only cloned a single file. We need some way to track:

1. If a dat was initialized in sparse mode
2. What files have been requested for download

This should probably happen in dat-node, since Desktop would have the same issue of sparse downloads ending up with the whole dat.

[okdistribute]

sparse ness on the commandline seems like it might benefit from some sort of 'include' config (file?) that says which files should be included in the sync

[okdistribute]

Or maybe just use the .datignore for that :P

[joehand]

sparse ness on the commandline seems like it might benefit from some sort of 'include' config (file?) that says which files should be included in the sync

Ya, something like that would be necessary.

Or maybe just use the .datignore for that :P

This would be deleted and/or changed by hyperdrive if the source dat had a .datignore file in it.

[mafintosh]

.datdownload cause it's the opposite of .datignore and we'd add that file to the default ignore list

[mafintosh]

just noting that this should work with folders and not just files as well :)

[mitar]

And adding that this should work also on subdirectories. So:

dat download dat://key/directory/subdirectory

[dcposch]

Tools like dat ls / dat download would def make it easier to work with large dat repositories.

I have a medium-size dat repository (dat://dcpos.ch, ~10GB). Parts of it aren't loading correctly in Beaker. To narrow down whether this is a Beaker or Dat issue, I tried downloading those files via command-line dat... and found there's no way short of dat clone-ing the whole thing.

http:// has wget and curl, IMO dat:// needs an equivalent.

[millette]

@dcposch meanwhile, there's https://github.com/millette/dat-shell/ to help you navigate a dat (cd, ls, cat) without downloading the whole thing.

[pfrazee]

Neither dat ls or dat download should be terribly hard to implement.