How to use colType in RDF::FromCSV?

Newbie
,
1

Hi,

I’m trying to read a text file into an RDataFrame, and need to specify that all columns should be of type Double_t. Here is a toy example to illustrate:

{
   std::ofstream outfile;
   outfile.open("./toydata.csv");
   outfile << "5,3\n7,8.1\n4.1,4.67\n-1.3,5.01";
   outfile.close();

   auto df1 = ROOT::RDF::FromCSV("./toydata.csv", false);
   auto df2 = ROOT::RDF::FromCSV(
      "./toydata.csv", false, ',', -1, {{"Col0", 'D'}});
   auto df3 = ROOT::RDF::FromCSV(
      "./toydata.csv", false, ',', -1, {{"Col1", 'D'}});
   auto df4 = ROOT::RDF::FromCSV(
      "./toydata.csv", false, ',', -1, {{"Col0", 'D'},{"Col1", 'D'}});

   std::cout << "colTypes = {}" << std::endl;
   df1.Display()->Print();
   std::cout << "colTypes = {{\"Col0\", \'D\'}}" << std::endl;
   df2.Display()->Print();
   std::cout << "colTypes = {{\"Col1\", \'D\'}}" << std::endl;
   df3.Display()->Print();
   std::cout << "colTypes = {{\"Col0\", \'D\'},{\"Col1\", \'D\'}}" << std::endl;
   df4.Display()->Print();
}

This prints

colTypes = {}
+-----+------+------+
| Row | Col0 | Col1 |
+-----+------+------+
| 0   | 5    | 3    |
+-----+------+------+
| 1   | 7    | 8    |
+-----+------+------+
| 2   | 4    | 4    |
+-----+------+------+
| 3   | -1   | 5    |
+-----+------+------+
colTypes = {{"Col0", 'D'}}
+-----+-----------+------+
| Row | Col0      | Col1 |
+-----+-----------+------+
| 0   | 0.0000000 | 0    |
+-----+-----------+------+
| 1   | 0.0000000 | 0    |
+-----+-----------+------+
| 2   | 0.0000000 | 0    |
+-----+-----------+------+
| 3   | 0.0000000 | 0    |
+-----+-----------+------+
colTypes = {{"Col1", 'D'}}
+-----+------+-----------+
| Row | Col0 | Col1      |
+-----+------+-----------+
| 0   | 5    | 0.0000000 |
+-----+------+-----------+
| 1   | 7    | 0.0000000 |
+-----+------+-----------+
| 2   | 4    | 0.0000000 |
+-----+------+-----------+
| 3   | -1   | 0.0000000 |
+-----+------+-----------+
colTypes = {{"Col0", 'D'},{"Col1", 'D'}}
+-----+-----------+-----------+
| Row | Col0      | Col1      |
+-----+-----------+-----------+
| 0   | 0.0000000 | 0.0000000 |
+-----+-----------+-----------+
| 1   | 0.0000000 | 0.0000000 |
+-----+-----------+-----------+
| 2   | 0.0000000 | 0.0000000 |
+-----+-----------+-----------+
| 3   | 0.0000000 | 0.0000000 |
+-----+-----------+-----------+

What is happening here? And what is the correct way to specify column type?

ROOT Version: 6.28/00 from conda-forge
Platform: macosx 10.15.7
Compiler: Not Provided

2

Hi @JSHeines ,

thank you for the report! I can reproduce the problem, this is a bug but I don’t know where, exactly, things are going wrong yet.

The only workaround I can suggest at this time is to avoid colTypes and add a . to the numbers in the first line of the CSV so they are auto-detected as doubles, which works.

Can you please open an issue at Issues · root-project/root · GitHub ? I will take a look as soon as possible.

Sorry for the trouble!
Enrico

3

Hi @eguiraud,

Thanks for your reply, I have opened an issue.
It’s no problem to workaround, reading the file into a TTree also works fine. I was mostly wondering what went wrong.

Thank you for your help
Johannes

4

HI,

thank you for your report, this is now fixed in the master and v6-28-00-patches branches (future 6.28.04).

Cheers,
Enrico

1 Like