Out_of_Date targets when downloaded from Cluster

[Skourtis]

Hi,

I'm running my _targets.R on an SGE cluster and then copy everything locally to work locally on it. But everything is out of date. I also tried going through git by pulling the targets/meta and targets/objects but still everything is out of date on my local machine. Is the an issue only for me or a known issue? Locally I'm working on Windows 10. Could this be a cluster-linux, local-windows issue?

How do I use targets if I want to have collaborators on git who need to know that the targets they see are up to date?

Thank you?
Savvas

[wlandau]

It is hard to say. Does the pipeline have any external input files that are system-dependent or missing when you just copy over _targets/? Does tar_visnetwork() point to anything specific upstream that might be invalidating the rest of the targets? It would help if you would share a small reproducible example.

[Skourtis]

Hi, it is as you predicted. After some digging I realised that the path structure are different in my local drive and cluster. I'm using here::here to create relative paths in tar_target which are identical for both, but when this is saved as a variable using the format = "file", it creates and saves the absolute path. So when I pull from git, it seems that I have changed the input files (which I haven't). I think this is the problem but not sure how to confirm or solve it. Let's say I am using this to collaborate through git, I cannot expect them to have the same absolute path as me (which is why I was using here::here).

[wlandau]

here::here() creates an absolute path. For targets, it is better to use a relative path. I recommend something like fs::path_rel(here::here("...")) instead.

[Skourtis]

This has solved the issue! thank you!