Pandas dataframe to TTree. Problem with types 'uint8' and 'int8'

[jruizvid]

When trying to convert a python array into a TTree, the method ROOT.RDF.MakeNumpyDataFrame(array) does not work if the columns have dtypes ‘int8’ or ‘uint8’.

I managed to make my script work by changing them to int32 in the initial pandas’ data frame.

# df is a pandas dataframe
i8_columns = df.select_dtypes(['uint8']).columns
df[i8_columns] = df[i8_columns].apply(lambda x: x.astype('int32'))

# Convert data to a dictionary with arrays
arraydata = {key: df[key].values for key in df.keys()}

# Write in a TTree
import ROOT
rdf = ROOT.RDF.MakeNumpyDataFrame(arraydata)
rdf.Snapshot('mytree', 'myfile.root')

Just pointing it out since I could not find this issue anywhere.

(ROOT 6.24)

[etejedor]

Hello,

Thank you for the report, indeed looking at the code it seems that int8 and uint8 are not supported data types in MakeNumpyDataFrame.