df019_Cache¶

Cache a processed RDataFrame in memory for further usage.

This tutorial shows how the content of a data frame can be cached in memory in form of a dataframe. The content of the columns is stored in memory in contiguous slabs of memory and is "ready to use", i.e. no ROOT IO operation is performed.

Creating a cached data frame storing all of its content deserialised and uncompressed in memory is particularly useful when dealing with datasets of a moderate size (small enough to fit the RAM) over which several explorative loops need to be performed as fast as possible. In addition, caching can be useful when no file on disk needs to be created as a side effect of checkpointing part of the analysis.

All steps in the caching are lazy, i.e. the cached data frame is actually filled only when the event loop is triggered on it.

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 09:46 AM.

In [1]:

import ROOT
import os

Welcome to JupyROOT 6.29/01

We create a data frame on top of the hsimple example.

In [2]:

hsimplePath = os.path.join(str(ROOT.gROOT.GetTutorialDir().Data()), "hsimple.root")
df = ROOT.RDataFrame("ntuple", hsimplePath)

We apply a simple cut and define a new column.

In [3]:

df_cut = df.Filter("py > 0.f")\
           .Define("px_plus_py", "px + py")

We cache the content of the dataset. Nothing has happened yet: the work to accomplish has been described.

In [4]:

df_cached = df_cut.Cache()

h = df_cached.Histo1D("px_plus_py")

Now the event loop on the cached dataset is triggered by accessing the histogram. This event triggers the loop on the df data frame lazily.

In [5]:

c = ROOT.TCanvas()
h.Draw()
c.SaveAs("df019_Cache.png")

print("Saved figure to df019_Cache.png")

Saved figure to df019_Cache.png

Info in <TCanvas::Print>: png file df019_Cache.png has been created

Draw all canvases

In [6]:

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