-
Notifications
You must be signed in to change notification settings - Fork 12
Closed
Labels
Description
Background
For parity with R pins, a user should be able to do the following:
- upload (R code)
- accept a List[str] | str of paths to files to upload
- bundle up as a pin version of type "files"
- take largely the same arguments as pin_write: name, title, description, metdata
- create one flat list of all files and directories (included those within subdirectories, etc..)
- e.g.
['a', 'a/b', 'a/b/file.txt']
- e.g.
- delegates version uploading to pin_store
- download
- basically equivalent to pin_read, with type="files" set in data.txt.
- the R version returns a local cache path, so the python version should return the remote path, since caching happens through wrapping the remote filesystem (???).
Notes on RStudio Connect
Since RStudio Connect already uploads a bunch of files by default (e.g. data.txt, index.html, js assets), the mechanism exists to pin_store arbitrary files / directories. One issue in the R pins implementation is in the following flow:
- Internally, files are just dumped at the top-level first, then its default index.html is added.
- A user attempting to pin_upload "index.html" would have it overwritten by the one pins generates.
Research needed
- Is there a need for pin_download in python pins? Should it download locally by default (e.g. into a folder, similar to what R pins is calling the cache)?
- How does pin_read(...) when type is "files", differ from pin_download?