Can't use formula as branch

[alexpearce]

root_numpy supports using formula as branch names. For example:

>>> root_numpy.root2array('f.root', 'tree', branches=['sqrt(mass)'], stop=3)
array([(739.8708446958261,), (1225.6622271912993,), (1096.497904427192,)],
      dtype=[('sqrt(mass)', '<f8')])

root_pandas doesn't support this, because it pattern matches branch names, failing if a name in the branches argument doesn't match a TBranch in the TTree.

[ibab]

Would you suggest that we silently ignore patterns that don't match and let root_numpy fail if it doesn't find something?

It seems like we can't have both branch name matching and formulas in branch names, as the * in a * b would apply to both things at once.

[alexpearce]

It seems like we can't have both branch name matching and formulas in branch names, as the * in a * b would apply to both things at once.

Indeed. I really like the globbing and brace expansion, so I'd be sad to see that go.

Maybe if formulas had to be prefixed with some magic string? Like formula:sqrt(mass). Then anything with the prefix would be ignored by the expansion logic, and the prefix is stripped before being passed to root_numpy.

It's not ideal though, and would require explicit documentation (although the support for formula is isn't mentioned in the root_numpy.root2array docs…).

[ibab]

A noexpand: prefix would make a lot of sense, especially for cases where your variable names contain braces and *.

Or there could just be a expand=False option in read_root that does it for all columns.
Then the user might have to call read_root several times (which shouldn't be too bad in terms of efficiency).

[alexpearce]

👍 for noexpand:.

It's pretty obvious what noexpand: should do and why, whereas understanding an expand kwarg requires you understanding the internals a bit more.