BC selection task and alias accounting added#5757
Conversation
|
Does everyone need always both tasks (event selection and bc selection), or shall we make that configurable / in split files? |
Most of our analyses rely on collision-based event selection so they will always need the bc-based selection in the pipe. On the other hand, bc-based analyses (like muon UPC) do not need collision-based event selection. I am not sure if we need excepions for this rare case. I prefer to avoid configurables now because event-selection qa and multiplicity tasks rely on the presence of the evsel table. |
|
Thanks for your reply. I agree. Let me test it also with my tasks, before we merge. |
|
Hi @ekryshen |
| { | ||
| // TODO bool arrays are not supported? Storing in int32 for the moment | ||
| DECLARE_SOA_COLUMN(Alias, alias, int32_t[kNaliases]); | ||
| DECLARE_SOA_COLUMN(BBT0A, bbT0A, bool); // beam-beam time in T0A |
There was a problem hiding this comment.
There is no need to define the columns again. I propose you put them in the same name space, and then you can reuse them for the second definition
Thanks for the confirmation. I suspect those extra events are due to a bug in trigger aliases stored in CCDB (essentially counting some events from triggerMaskNext50). I found and fixed it in the morning. If you ran your code yesterday, that can explain extra counts. I will check other comments to the code now. |
I ran my code this morning (few minutes before my comment), both: with and without your changes, so it should have picked up the same CCDB |
| kNaliases | ||
| }; | ||
|
|
||
| static const std::string aliasLabels[kNaliases] = { |
There was a problem hiding this comment.
Isn't const char* enough?
jgrosseo
left a comment
There was a problem hiding this comment.
I still had a comment but I am fine with merging this, you can change that eventually in the next PR
|
Hi @jgrosseo, ok, thanks. Just for the record: the difference in the old and new event selection was traced back to the issue in the Run2MatchedSparse matcher that was used in the old approach. Anton is looking into it. |
* BC selection task and alias accounting added * clang-format fixes * bcsel namespace removed. alias labels added. +bug fixes * clang-format fixes
Switching to bc-based selection: