On Thu, Dec 9, 2021 at 9:44 AM David Mertz, Ph.D. firstname.lastname@example.org wrote:
On Wed, Dec 8, 2021, 2:58 PM Rob Cliffe via Python-ideas
On 08/12/2021 19:27, Paul Moore wrote:
The reason deferred objects keep coming up is because they *do* have a much more compelling benefit - they help in a much broader range of cases.
Can anyone provide some realistic use cases? I've read the whole thread and I can only recall at most one, viz. the default value is expensive to compute and may not be needed.
This library is widely used, extremely powerful, and expressive. Basically, a built-in capability would have every win of Dask Delayed, but decrease the impedance mismatch (which isn't terrible as-is) and avoid the need for external tooling.
Most of that is a *massive* YAGNI as regards function default arguments. We do not need parallel execution just to figure out the length of a list passed as a previous parameter. So you've just added weight to my argument that a generic "delayed" feature is a completely separate proposal, nothing whatsoever to do with PEP 671.