Passing RDataFrame values into Python function?

ROOT
,
1

I am very new to using ROOT, so apologies if this question is unclear. I am trying to use the values stored in my RDataFrame as arguments for a Python function. Specifically, I want to define a Python function to solve a system of nonlinear equations, pass the values from each row of the RDataFrame as coefficients in the functions, and define a new column in the DF representing the solution (most likely using SciPy to find the solution). Most examples with user-defined functions seem to involve C++ functions, but is it possible to write the function in Python? Also open to other suggestions about how to approach this problem.

2

Hello @somebody_nobody,
welcome to the ROOT forum!

From what I understand, you need something like this:

import ROOT

# Create a data frame with 100 rows
rdf = ROOT.RDataFrame(100)
 
# Define a new column `x` that contains random numbers
rdf_x = rdf.Define("x", "gRandom->Rndm()")

@ROOT.Numba.Declare(['RVec<float>', 'float'], 'float') 
def pypowsum(x, y):
    s = 0
    for e in x:
        s += e**y 
    return s

pypowsum(rdf_x.Take[ROOT.Double_t]("x"), 3 )

Let me know if it helps.

Cheers,
Monica

4

Thank you for your reply. I am still looking for a solution that can define a new RDataFrame column with the result. In other words, I don’t want one result from the function, but a result corresponding to each event in the RDataFrame. My current attempt is as follows:

@ROOT.Numba.Declare([‘RVec’, ‘RVec’], ‘RVec’)
def add(x, y):
return x + y
df = df.Define(“sum”, add(“A”, “B”))

Unfortunately, I keep getting errors about the column names being “undeclared identifiers”. I have tried many modifications of this approach, but I’m not sure how to resolve it. Thanks in advance for any suggestions.

5

In this case, assuming that df is an RDF with columns A and B, what you would do is simply

df_sum = df.Define("sum", "A+B")

Is what you want to do to apply a scipy function to an RVec and then push back the result into an RDF?

6

Yes, exactly. I was only using the sum function as a test to see whether I could extract the values from the RDF. I want to use SciPy (or Sympy, or some other numerical solver) to operate on an RVec, and then export the result to a column of the RDF.

Currently the issue is that I want to use the Python Math package within the function, but this is disallowed, causing the error:
numba.core.errors.TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Untyped global name ‘Math’: Cannot determine Numba type of <class ‘Math_meta’>

7

Then the fastest solution could be to convert your RDF to awkward like this:

import awkward as awk

awk_x = awk.from_rdataframe(rdf_x, columns = ("x")) #convert to awkward
x = awk_x.x #extract column
z = np.power(x,2) #use numpy or scipy or whatever python module
awk_x["z"] = z #push back result
df_xz = awk.to_rdataframe({"x": awk_x.x,  "z":awk_x.z}) #convert back to RDF
8

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.