pyiron / pyiron_workflow / 11096941100

Build # 11096941100

Build Type

push

github

Committed by web-flow

Commit Message

[patch] Persistent process (#476)

* Catch already-running children in Composite

* Add temporary results infrastructure

* Allow run results to be (de)serialized at run-time

By adding a new flag, having `Node.on_run` directly handle the serialization of results (making `_on_run` and `_run_args` new abstract methods that have the behaviour of the old public methods), and deserialize temporary results instead of running when a node is already running and such results exist

* Format black

* Remove redundant arg

* Save a checkpoint when running with result serialization

So that the graph gets saved with the serializer in a running state

* Refactor: extract method

* Catch the case that you re-run a node still waiting for its serialized result

* Add a helper method for resuming composite runs from broken processes

* Revert resume_from_broken_process

* Make cleaning revert to false

It's a bit messier for the filesystem, but for now let's default to keeping the data around

* Make the flag private

I want this functionality in, but I'm not at all happy with the UI, and don't totally trust edge cases (e.g. input changing under our feet), so let's put it in private for now in anticipation of changes

* Remove the checkpoint save

We'll get a recovery file when we close the parent process anyhow

* Extend doc

* Don't delete checkpoint as it's no longer written

* Refactor: slide

* Update HPC example notebook

With a real living example of `Node._serialize_result` working

---------

Co-authored-by: pyiron-runner <pyiron@mpie.de>

Run Details

3088 of 3371 relevant lines covered (91.6%)

0.92 hits per line

Source File

Press 'n' to go to next uncovered line, 'b' for previous

Source Not Available