bionic
bionic copied to clipboard
square
A Python framework for data science.
Parallel evaluation of a flow doesn't seem to work when `core__persist_by_default = False`, even if all entities in the flow (in particular, the ones on the parallel branches of the...
This PR refactors refactors our task structure to make saving and loading into separate tasks. There should be no behavior changes (except for some slight changes in logging). There are...
Sometimes users pass a single argument to `@bn.outputs`, expecting it to work like `@bn.output`: builder = bn.FlowBuilder('test') @builder @bn.outputs('value') def _(): return 7 flow = builder.build() flow.get('value') This fails because...
Currently, when an entity has both persistence and memoization disabled, we store its value in memory for the duration the `Flow.get()` call (as documented [here](https://bionic.readthedocs.io/en/stable/concepts.html#disabling-in-memory-caching)). I think we should change...
If a flow span multiple files (e.g. a flow merges another flow from a different module), reloading may not work correctly. Possible issues - Reload ordering: If a flow X...
The `FlowBuilder.merge` method excepts a flow, not a builder, which seems to be a common source of surprise. What's worse, if you pass it a builder you get a cryptic...
In general, almost any exception while reading a cached file will currently be reported as an `InvalidCacheStateError`, which means the user will be told the cache is corrupted and needs...
Could you please document in Warnings and Pitfalls what the user should be aware of, if they have mutiple processes or threads open, which are using the same flow?
Sometimes GCS file uploads (and presumably downloads) can time out (stack trace attached below). For most of these operations we use the GCS Python API rather than gsutil, so it's...