Using ROOT.Numba: confusing imports are confusing

[swertz]

Hello,

I was trying to use Numba to jit python function (relying on numpy), to call from an RDataFrame graph. It works (amazing!), but I’m a bit confused about how to import pyROOT to use Numba as ROOT.Numba.

I thought that from cppyy import gbl as ROOT and import ROOT should be equivalent in practice, but that does not seem to be the case:

In [1]: from cppyy import gbl

In [2]: gbl.Numba
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-2-6f9dc22788fc> in <module>
----> 1 gbl.Numba

AttributeError: <namespace cppyy.gbl at 0x516b4d0> has no attribute 'Numba'. Full details:
  type object '' has no attribute 'Numba'
  'Numba' is not a known C++ class
  'Numba' is not a known C++ template
  'Numba' is not a known C++ enum

In [3]: gbl.ROOT.Numba
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-3-0909b72b901b> in <module>
----> 1 gbl.ROOT.Numba

AttributeError: <namespace cppyy.gbl.ROOT at 0x6d0f060> has no attribute 'Numba'. Full details:                                                                                             
  type object 'ROOT' has no attribute 'Numba'
  'ROOT::Numba' is not a known C++ class
  'Numba' is not a known C++ template
  'Numba' is not a known C++ enum

In [4]: import ROOT

In [5]: ROOT.Numba
Out[5]: <namespace cppyy.gbl.Numba at 0x72ce970>

Even stranger, once import ROOT has been called, this now works:

In [6]: gbl.Numba
Out[6]: <namespace cppyy.gbl.Numba at 0x72ce970>

Could someone help to clarify in which way the different imports are not equivalent, and why using Numba seems to require to import ROOT?

Cheers,
Sébastien

---

ROOT Version: 6.24/06
Platform: CC7
Compiler: GCC11

---

[etejedor]

Hello,

If you want to use ROOT.Numba you can only do it via the ROOT module, it’s a feature that is not provided by cppyy alone (ROOT does the injection of that feature).

[swertz]

I see, thanks. In general, is it therefore recommended to always use import ROOT? Is there any advantange of doing from cppyy import gbl as ROOT in some situations?

[etejedor]

cppyy.gbl and ROOT are actually not exactly the same. cppyy.gbl is the global namespace. ROOT is a facade module that redirects lookups to cppyy.gbl, but also injects some extra functionality as it is the case of ROOT.Numba, and registers a number of pythonizations (extra behaviour in Python) for ROOT classes. Therefore if you are working with ROOT I would recommend to use import ROOT.

[swertz]

Perfect, thanks for clarifying!