Pushing experiments to the platform is very tricky without informative warnings or errors!
The assumption is that you can pull an experiment from the platform, edit the necessary blocks (which means debugging the blocks without going through the pain of making new versions or forks for each block) and then push back to the platform.
The issue is that in each experiment the user can use toolchains, or algorithms, etc that are public from other users and when they are pulled they are organized in folders with similar usernames in the user's local prefix folder.
With current setup if users want to edit some algorithm they should first copy them to their own name space.
If you do not make new algorithms under your own name space two scenarios may happen:
- You just edited the file from original user (not a good idea).
beat pushcommand in this situation does not push anything since it already exists on the platform.
- you made new version form the same algorithm which goes under the category of the original user.
beat pushcommand will create a new algorithm with the same name under your username with version number one without any warnings and it shows on the command line that the algorithm under the original username has been created.
The problem is since there is no error or warnings in such situations and nothing mentioned in the documentation it is tricky to find out where the problem comes from. It might be very confusing for the normal user to interact with beat.