Ideas / Examples of "quick start" and other useful code snippets for coffea 2023.

[lgray]

Using this thread to collect code snippets from folks that can be included into coffea 2023 documentation.

[btovar]

Here is a draft for taskvine. I'm using pip install --pre coffea so that the examples run. (Making a note to remember to update on release.)
tvqs.md

[btovar]

From old coffea:

runner = processor.Runner(exec_instance, schema=NanoAODSchema, chunksize=chunksize, maxchunks=nchunks)

schema should be set in NanoEventsFactory.
chunksize and maxchunks do not exists anymore.
chunksize closest equivalent in NanoEventsFactory is steps_per_file, but coffea warns that it should only be used in small tests.

[lgray]

Not quite the case!

max_chunks is now handled with a function, and it operates on dataset definitions rather than being a parameter of the analysis itself.

https://github.com/CoffeaTeam/coffea/blob/master/src/coffea/dataset_tools/manipulations.py#L12

You can see it being used in the following test:
https://github.com/CoffeaTeam/coffea/blob/master/tests/test_dataset_tools.py#L189

There's also the more general slice_chunks function now.

[lgray]

So you specify the chunks in the steps section of the dataset definition, and then operate upon that.
You could also create by hand a test dataset that that only specifies some single, small step to process.

[lgray]

And you specify chunksize now through the separate preprocess function that spits out a completely specified preprocessed dataset.

You can see its use here:
https://github.com/CoffeaTeam/coffea/blob/master/tests/test_dataset_tools.py#L204