Timeline for RDataFrame to be made compatible with RNTuple

[jcatmore]

---

Dear ROOT team,

my understanding is that at present, RDataFrame can’t read RNTuple as an input format, at least in the production versions of ROOT. My question is, do you have an approximate timeline for when this will work? Is this likely to be months away or not until during or after Run 3?

I should add that this is for my own curiosity rather than anything urgent.

Thanks for your help!

Best wishes,

James Catmore

ROOT Version: Not Provided
Platform: Not Provided
Compiler: Not Provided

---

[eguiraud]

Hi @jcatmore ,
there is an experimental ROOT::Experimental::MakeNTupleDataFrame in master that constructs a dataframe that uses an RNTuple as input. It is currently in development and missing some features and performance optimizations, but work in progress!

We have some tests and tutorials, see e.g. https://github.com/root-project/root/search?q=MakeNTupleDataFrame&type=

The ETA is the same for RNTuple itself, they grow together.
Cheers,
Enrico

P.S.
MakeNTupleDataFrame is also available in v6.24, but I recommend using ROOT’s nightly builds to try it out to get the latest, as-stable-as-it-gets version.

[jcatmore]

Hi @eguiraud ,

thanks a lot for this! We tested it and it works very nicely (even via the python bindings).

Thanks again,

James.

[eguiraud]

Great! cc: @jblomer

The performance overhead w.r.t. direct RNTuple usage is still too large for our taste and we know of some things that need improving/fixing, but anyways please let us know if you have any feedback.

Cheers,
Enrico