[PWGHF] Replace UPC Lc THnSparse with TTree

[Rrantu]

This PR introduces a transition from THnSparse to TTree in the UPC Lc analysis.
Main changes:

• Added UPC Lc columns and tables in utilsUpcHf.h
• Implemented production and filling of TTree in taskLc.cxx
• Removed UPC-related axes from the THnSparse

[github-actions]

O2 linter results: ❌ 0 errors, ⚠️ 0 warnings, 🔕 0 disabled

[vkucera]

@Rrantu Columns and tables are parts of the data model. They don't belong in the utility headers.

[Rrantu]

Hi @vkucera, thanks for the comment, where would you recommend moving the columns and tables to? Would it be acceptable to define them directly in the task?

[vkucera]

Hi @vkucera, thanks for the comment, where would you recommend moving the columns and tables to? Would it be acceptable to define them directly in the task?

Yes, if the tables are not used anywhere else, please put them in the task.

[Rrantu]

Hi @vkucera, thanks for the comment, where would you recommend moving the columns and tables to? Would it be acceptable to define them directly in the task?

Yes, if the tables are not used anywhere else, please put them in the task.

Thanks! I‘ve moved them to the task file

[vkucera]

Why do you duplicate columns in both tables?

[Rrantu]

One table corresponds to the BDT-applied version, while the other does not. However, both require the mass and pt information.

[vkucera]

You never fill both tables. It would be better to have one table with the common columns that you fill in any case and a second table with the extra columns for the BDT case.

[vkucera]

Why do you call this table BDT?

[Rrantu]

Because the tree includes BDT scores for the candidates.

[vkucera]

But it also includes other BDT-unrelated columns. See my comment above.

[Rrantu]

Thanks for the suggestion. Here the detector variables and the BDT output are both required together for offline selections. Given this use case, I would prefer to keep the current structure.

[vkucera]

You only have a single option to fill the trees. It seems that the attributes UPC and Lite are redundant.

[vkucera]

And the comment is wrong.

[Rrantu]

Thanks for the comment! The Lite one fills fewer candidates, including only those passing the single-gap selection.

[vkucera]

Then it should be called something like fillTreeOnlySingleGap and explained in the comment.

[Rrantu]

Thanks for the suggestion, I have renamed it.

[vkucera]

@Rrantu A conceptual question: Why can you not use the THnSparse?

[Rrantu]

@Rrantu A conceptual question: Why can you not use the THnSparse?

Thanks for the question! For large statistics, merging THnSparse objects becomes very heavy, making it difficult to obtain merged AnalysisResults.root. I also need to repeatedly adjust selections on multiple axes during the offline analysis. Projections from high-dimensional THnSparse are not very flexible.

[lubynets]

Dear @Rrantu, @vkucera, all codeowners, sorry for interfering,
Are Tasks, and not TableProducers, supposed to produce trees? https://aliceo2group.github.io/analysis-framework/docs/gettingstarted/theo2physicsrepo.html#folder-structure

[vkucera]

Dear @Rrantu, @vkucera, all codeowners, sorry for interfering, Are Tasks, and not TableProducers, supposed to produce trees? https://aliceo2group.github.io/analysis-framework/docs/gettingstarted/theo2physicsrepo.html#folder-structure

No, they are not.

[Rrantu]

Hi @lubynets, Thanks for the question. In this case, the tables are only used within this task, and the trees are produced specifically for this analysis. Also, there are other tasks that produce trees in a similar way.