Closed mdwint closed 4 years ago
Hi,
Sorry it's not working for you. I've never really thought about supporting dask.persist()
before. I will see about writing some tests and seeing what I can do.
My guess is that:
Future
instances.persist_with_trace
which would implement the same code structure as eliot.dask.compute_with_trace
does—pre-optimize with logging injected, then call dask.persist()
on that.Please try the PR above and let me know if it works.
I'm trying to add logging to a list of Dask collections, some of which may contain futures (if
dask.persist()
was previously called on them).Replacing
dask.compute(*args)
witheliot.dask.compute_with_trace(*args)
results in this error:It makes sense that this fails, since the calculation is already running in the background. Surely, logging would need to be added before the call to
dask.persist()
.Does anyone have any advice on how to achieve this?