The box remote, both with and without the cache, seems to have to perform an API call for each directory level from the top. So, if I am uploading a file to /this/is/a/long/path/to/a/deep/directory/, we have to look for /this, get its ID, get /this/is/, get its ID, etc… and these lookups take longer than many of the actual transfers.
If we were to keep the folder IDs in cache, when we need /this/is/a/long/path/to/a/deep/directory/, we could identify a directory along the path (say /this/is/a/long/path/to/a/deep, use its ID to check if its parents are still “a”,“to”,“path”,“long”,…, and then proceed from there) and skip the intermediate lookups.
Is this generally useful enough to be worth doing?
The box remote does keep a directory cache, but it rebuilds it every time - you’ve seen it doing that.
As far as I’m aware there isn’t a better way of doing this as box doesn’t have a turn this path into a directory ID API. I’d love to be proved wrong though!
In general rclone tries to avoid local state as a matter of principle…
Some ideas
rclone could take an directoryID which it takes as the root directory. The drive backend does this already.
depending on what you are doing you could run a persistent rclone and use the API
That second point might be worth exploring…
What you do is run rclone rcd --rc-no-auth as the server, then run rc commands from herehttps://rclone.org/rc/ - you can copy files like that and rclone will be continuously running so no need to re-auth or fetch directory IDs.