feat: add tree to virtual array conversion

[pfackeldey]

This waits for scikit-hep/awkward#3364 (and a corresponding awkward release).

I may likely not have attempted the most optimal solution here. Happy for feedback & input.

[ianna]

@pfackeldey - what is the plan for this PR? thanks!

[ikrommyd]

@pfackeldey - what is the plan for this PR? thanks!

I think we were discussing with Peter to add some caching support. Uproot will deserialize each electron branch for example separately with its own offsets. All those will have the same count_branch however so it's probably best to not deserialize the same offsets dozens of times. It's probably best to cache count_branch deserialization result (length) and use it for the other branches that have the same count_branch.

[ariostas]

I think we were discussing with Peter to add some caching support.

Isn't there already some caching being done in Uproot? When trying to read the count_branch multiple times it should already be hitting the cache

[ikrommyd]

I think we were discussing with Peter to add some caching support.

Isn't there already some caching being done in Uproot? When trying to read the count_branch multiple times it should already be hitting the cache

Yeah I need to try if it's hitting it, haven't done that yet. Will do today. Do you the best way to log that (the number of deserializations per branch)?

[ariostas]

Do you the best way to log that (the number of deserializations per branch)?

I'm not sure. I've just skimmed the code since at some point I'll have to do that for RNTuples

[pfackeldey]

@pfackeldey - what is the plan for this PR? thanks!

I'm not sure. I'm not a big fan of this implementation, but I also don't know how it can be done in a better way. I was hoping for some input here.