Draft: Switch CP algs to ConfigBlock configuration
This depends on atlas/athena!65122 (merged).
For now it is expected to fail in CI.
The lepton blocks are easy to do as standalone pieces, and can be reintegrated into the ComponentAccumulator
configuration straightforwardly, by merging.
For any sequences that depend on upstream collections, such as MET, overlap removal and jets (muons for b-jet calibration), we need to join up the ConfigAccumulator
instances, as the provenance is traced backwards. So cpalgs_config.py
will need a rewrite to pass and aggregate the various components into one ConfigAccumulator
. This seems like it must also have to extend downstream to overlap removal, but while some things (e.g. muons) depend on pileup reweighting, they didn't seem to need this incorporated.
IMO we should not yet commit to using ConfigAccumulator
to handle the TTree output until and unless we decide that the entire configuration including analysis-specific components should be in ConfigAccumulator
. That's not to say this is undesirable, but it would be a major undertaking that requires one additional layer (that might still be evolving) to be understood in development.
TODO:
-
Move early operations (trigger, PRW, DQ filter) to ConfigAccumulator
-
Move jets to ConfigAccumulator
-
Define a ConfigAccumulator
for b-jet calibration
-
-
Define a ConfigAccumulator
module for the systematics linker (!255 (closed), athena:!65087) -
Aggregate ConfigAccumulator
instances incpalgs_config.py
-
Move MET to ConfigAccumulator
(easy but needs the previous step) -
Move OR to ConfigAccumulator
Moved to !262 (merged).
FOLLOW-UPS