df007_snapshot¶

Write ROOT data with RDataFrame.

This tutorial shows how to write out datasets in ROOT format using RDataFrame.

Author: Danilo Piparo (CERN)
This notebook tutorial was automatically generated with ROOTBOOK-izer from the macro found in the ROOT repository on Monday, March 27, 2023 at 09:45 AM.

In [1]:

import ROOT

Welcome to JupyROOT 6.29/01

A simple helper function to fill a test tree: this makes the example stand-alone.

In [2]:

def fill_tree(treeName, fileName):
    df = ROOT.RDataFrame(10000)
    df.Define("b1", "(int) rdfentry_")\
      .Define("b2", "(float) rdfentry_ * rdfentry_").Snapshot(treeName, fileName)

We prepare an input tree to run on

In [3]:

fileName = "df007_snapshot_py.root"
outFileName = "df007_snapshot_output_py.root"
outFileNameAllColumns = "df007_snapshot_output_allColumns_py.root"
treeName = "myTree"
fill_tree(treeName, fileName)

We read the tree from the file and create a RDataFrame

In [4]:

d = ROOT.RDataFrame(treeName, fileName)

Select entries¶

We now select some entries in the dataset

In [5]:

d_cut = d.Filter("b1 % 2 == 0")

Enrich the dataset¶

Build some temporary columns: we'll write them out

In [6]:

getVector_code ='''
std::vector<float> getVector (float b2)
{
   std::vector<float> v;
   for (int i = 0; i < 3; i++) v.push_back(b2*i);
   return v;
}
'''
ROOT.gInterpreter.Declare(getVector_code)

d2 = d_cut.Define("b1_square", "b1 * b1") \
          .Define("b2_vector", "getVector( b2 )")

Write it to disk in ROOT format¶

We now write to disk a new dataset with one of the variables originally present in the tree and the new variables. The user can explicitly specify the types of the columns as template arguments of the Snapshot method, otherwise they will be automatically inferred.

In [7]:

d2.Snapshot(treeName, outFileName, \
    ["b1", "b1_square", "b2_vector"])

Out[7]:

<cppyy.gbl.ROOT.RDF.RResultPtr<ROOT::RDF::RInterface<ROOT::Detail::RDF::RLoopManager,void> > object at 0x12e1e890>

Open the new file and list the columns of the tree

In [8]:

f1 = ROOT.TFile(outFileName)
t = f1.myTree
print("These are the columns b1, b1_square and b2_vector:")
for branch in t.GetListOfBranches():
    print("Branch: %s" %branch.GetName())

f1.Close()

These are the columns b1, b1_square and b2_vector:
Branch: b1
Branch: b1_square
Branch: b2_vector

We are not forced to write the full set of column names. We can also specify a regular expression for that. In case nothing is specified, all columns are persistified.

In [9]:

d2.Snapshot(treeName, outFileNameAllColumns)

Out[9]:

<cppyy.gbl.ROOT.RDF.RResultPtr<ROOT::RDF::RInterface<ROOT::Detail::RDF::RLoopManager,void> > object at 0x14b58690>

Open the new file and list the columns of the tree

In [10]:

f2 = ROOT.TFile(outFileNameAllColumns)
t = f2.myTree
print("These are all the columns available to this dataframe:")
for branch in t.GetListOfBranches():
    print("Branch: %s" %branch.GetName())

f2.Close()

These are all the columns available to this dataframe:
Branch: b1_square
Branch: b2_vector
Branch: b1
Branch: b2

We can also get a fresh RDataFrame out of the snapshot and restart the analysis chain from it.

In [11]:

snapshot_df = d2.Snapshot(treeName, outFileName, ["b1_square"]);
h = snapshot_df.Histo1D("b1_square")

c = ROOT.TCanvas()
h.Draw()
c.SaveAs("df007_snapshot.png")

print("Saved figure to df007_snapshot.png")

Saved figure to df007_snapshot.png

Info in <TCanvas::Print>: png file df007_snapshot.png has been created

Draw all canvases

In [12]:

from ROOT import gROOT 
gROOT.GetListOfCanvases().Draw()