You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It is my understanding that the editor can keep a cached state of the workflow that has not yet been serialized to disk. In cases where asserting the state of the software is important, e.g. asserting if the repository is dirty at start-up, this feature makes it difficult to know what is the state that run the experiment as the cached state is not serialized to disk. One way to go around this may be to "force" the workflow to save itself each time the program starts.
The text was updated successfully, but these errors were encountered:
Temp files instead of saving on run? Would this help?
I don't think this is especially helpful for Bruno's original stated use-case when this issue was made. IIRC he was embedding the hash of the workflow in logs so there was traceability as to which workflow was used for the experiment.
It might make sense for Bonsai to do this in the background as a method for auto-recovery though.
Have a toggle button for "save on run"?
If this is being envisioned as a button on the toolbar this sounds like extreme clutter to me. The behavior Bruno is describing is exactly what Visual Studio and VSCode do.
IIRC Visual Studio used to prompt you before saving (with a checkbox to remember the choice) but these days it just saves automatically.
Not opposed to the behavior being optional, but Bonsai really needs a proper settings menu for this sort of thing.
It is possible to develop "buggy" custom extensions which will build, save, but not deserialize correctly from XML after save, is this a problem?
IMO it's a bad thing that it's even possible to run a workflow that won't be able to be deserialized. It honestly might not even be a bad thing for compilation to be done on a copy of the workflow that has gone through a serialize→deserialize trip. (This is probably a separate issue to consider though.)
It is my understanding that the editor can keep a cached state of the workflow that has not yet been serialized to disk. In cases where asserting the state of the software is important, e.g. asserting if the repository is dirty at start-up, this feature makes it difficult to know what is the state that run the experiment as the cached state is not serialized to disk. One way to go around this may be to "force" the workflow to save itself each time the program starts.
The text was updated successfully, but these errors were encountered: