search

opensafely / research-template

The template for new research projects that use the OpenSAFELY framework.
MIT License
16 stars 13 forks source link

Deal with conflicting input CSVs #4

Closed alexwalkerepi closed 4 years ago

alexwalkerepi commented 4 years ago

After making a change to the study definition in a PR, the input.csv then conflicts with the one in master. This makes it difficult to merge. I guess we always want to keep the more recently generated version. image

sebbacon commented 4 years ago

This is tricky. @inglesp's idea seems the best: have branches generate a differently-named file (temp_branch_input.csv?) which is deleted on merge; and have the "real" one generated in master on merge.

sebbacon commented 4 years ago

Hmm. Having started this, it feels a bit messy.

I think perhaps making the CSVs release assets might be the best thing.

We can make latest always be current master so it can be linked to from the README.

We can also a per-branch release which is deleted when the branch is merged.

It's ugly but we should be encouraging most users to use the command line anyway.

It means we don't end up polluting the git history with meaningless diffs too.

sebbacon commented 4 years ago
alexwalkerepi commented 4 years ago

Yes that does sound better, and would mean there's less restriction on the size of dummy data population.