26#include <nlohmann/json.hpp>
1482 void Exec(unsigned int slot)
1484 fPerThreadResults[slot]++;
1487 // Called at the end of the event loop.
1490 *fFinalResult = std::accumulate(fPerThreadResults.begin(), fPerThreadResults.end(), 0);
1493 // Called by RDataFrame to retrieve the name of this action.
1494 std::string GetActionName() const { return "MyCounter"; }
1498 ROOT::RDataFrame df(10);
1499 ROOT::RDF::RResultPtr<int> resultPtr = df.Book<>(MyCounter{df.GetNSlots()}, {});
1500 // The GetValue call triggers the event loop
1501 std::cout << "Number of processed entries: " << resultPtr.GetValue() << std::endl;
1505See the Book() method for more information and [this tutorial](https://root.cern/doc/master/df018__customActions_8C.html)
1506for a more complete example.
1508#### Injecting arbitrary code in the event loop with Foreach() and ForeachSlot()
1510Foreach() takes a callable (lambda expression, free function, functor...) and a list of columns and
1511executes the callable on the values of those columns for each event that passes all upstream selections.
1512It can be used to perform actions that are not already available in the interface. For example, the following snippet
1513evaluates the root mean square of column "x":
1515// Single-thread evaluation of RMS of column "x" using Foreach
1518df.Foreach([&sumSq, &n](double x) { ++n; sumSq += x*x; }, {"x"});
1519std::cout << "rms of x: " << std::sqrt(sumSq / n) << std::endl;
1521In multi-thread runs, users are responsible for the thread-safety of the expression passed to Foreach():
1522thread will execute the expression concurrently.
1523The code above would need to employ some resource protection mechanism to ensure non-concurrent writing of `rms`; but
1524this is probably too much head-scratch for such a simple operation.
1526ForeachSlot() can help in this situation. It is an alternative version of Foreach() for which the function takes an
1527additional "processing slot" parameter besides the columns it should be applied to. RDataFrame
1528guarantees that ForeachSlot() will invoke the user expression with different `slot` parameters for different concurrent
1529executions (see [Special helper columns: rdfentry_ and rdfslot_](\ref helper-cols) for more information on the slot parameter).
1530We can take advantage of ForeachSlot() to evaluate a thread-safe root mean square of column "x":
1532// Thread-safe evaluation of RMS of column "x" using ForeachSlot
1533ROOT::EnableImplicitMT();
1534const unsigned int nSlots = df.GetNSlots();
1535std::vector<double> sumSqs(nSlots, 0.);
1536std::vector<unsigned int> ns(nSlots, 0);
1538df.ForeachSlot([&sumSqs, &ns](unsigned int slot, double x) { sumSqs[slot] += x*x; ns[slot] += 1; }, {"x"});
1539double sumSq = std::accumulate(sumSqs.begin(), sumSqs.end(), 0.); // sum all squares
1540unsigned int n = std::accumulate(ns.begin(), ns.end(), 0); // sum all counts
1541std::cout << "rms of x: " << std::sqrt(sumSq / n) << std::endl;
1543Notice how we created one `double` variable for each processing slot and later merged their results via `std::accumulate`.
1547### Dataset joins with friend trees
1549Vertically concatenating multiple trees that have the same columns (creating a logical dataset with the same columns and
1550more rows) is trivial in RDataFrame: just pass the tree name and a list of file names to RDataFrame's constructor, or create a TChain
1551out of the desired trees and pass that to RDataFrame.
1553Horizontal concatenations of trees or chains (creating a logical dataset with the same number of rows and the union of the
1554columns of multiple trees) leverages TTree's "friend" mechanism.
1556Simple joins of trees that do not have the same number of rows are also possible with indexed friend trees (see below).
1558To use friend trees in RDataFrame, set up trees with the appropriate relationships and then instantiate an RDataFrame
1564main.AddFriend(&friend, "myFriend");
1567auto df2 = df.Filter("myFriend.MyCol == 42");
1570The same applies for TChains. Columns coming from the friend trees can be referred to by their full name, like in the example above,
1571or the friend tree name can be omitted in case the column name is not ambiguous (e.g. "MyCol" could be used instead of
1572"myFriend.MyCol" in the example above if there is no column "MyCol" in the main tree).
1574\note A common source of confusion is that trees that are written out from a multi-thread Snapshot() call will have their
1575 entries (block-wise) shuffled with respect to the original tree. Such trees cannot be used as friends of the original
1576 one: rows will be mismatched.
1578Indexed friend trees provide a way to perform simple joins of multiple trees over a common column.
1579When a certain entry in the main tree (or chain) is loaded, the friend trees (or chains) will then load an entry where the
1580"index" columns have a value identical to the one in the main one. For example, in Python:
1586# If a friend tree has an index on `commonColumn`, when the main tree loads
1587# a given row, it also loads the row of the friend tree that has the same
1588# value of `commonColumn`
1589aux_tree.BuildIndex("commonColumn")
1591mainTree.AddFriend(aux_tree)
1593df = ROOT.RDataFrame(mainTree)
1596RDataFrame supports indexed friend TTrees from ROOT v6.24 in single-thread mode and from v6.28/02 in multi-thread mode.
1598\anchor other-file-formats
1599### Reading data formats other than ROOT trees
1600RDataFrame can be interfaced with RDataSources. The ROOT::RDF::RDataSource interface defines an API that RDataFrame can use to read arbitrary columnar data formats.
1602RDataFrame calls into concrete RDataSource implementations to retrieve information about the data, retrieve (thread-local) readers or "cursors" for selected columns
1603and to advance the readers to the desired data entry.
1604Some predefined RDataSources are natively provided by ROOT such as the ROOT::RDF::RCsvDS which allows to read comma separated files:
1606auto tdf = ROOT::RDF::FromCSV("MuRun2010B.csv");
1607auto filteredEvents =
1608 tdf.Filter("Q1 * Q2 == -1")
1609 .Define("m", "sqrt(pow(E1 + E2, 2) - (pow(px1 + px2, 2) + pow(py1 + py2, 2) + pow(pz1 + pz2, 2)))");
1610auto h = filteredEvents.Histo1D("m");
1614See also FromNumpy (Python-only), FromRNTuple(), FromArrow(), FromSqlite().
1617### Computation graphs (storing and reusing sets of transformations)
1619As we saw, transformed dataframes can be stored as variables and reused multiple times to create modified versions of the dataset. This implicitly defines a **computation graph** in which
1620several paths of filtering/creation of columns are executed simultaneously, and finally aggregated results are produced.
1622RDataFrame detects when several actions use the same filter or the same defined column, and **only evaluates each
1623filter or defined column once per event**, regardless of how many times that result is used down the computation graph.
1624Objects read from each column are **built once and never copied**, for maximum efficiency.
1625When "upstream" filters are not passed, subsequent filters, temporary column expressions and actions are not evaluated,
1626so it might be advisable to put the strictest filters first in the graph.
1628\anchor representgraph
1629### Visualizing the computation graph
1630It is possible to print the computation graph from any node to obtain a [DOT (graphviz)](https://en.wikipedia.org/wiki/DOT_(graph_description_language)) representation either on the standard output
1633Invoking the function ROOT::RDF::SaveGraph() on any node that is not the head node, the computation graph of the branch
1634the node belongs to is printed. By using the head node, the entire computation graph is printed.
1636Following there is an example of usage:
1638// First, a sample computational graph is built
1639ROOT::RDataFrame df("tree", "f.root");
1641auto df2 = df.Define("x", []() { return 1; })
1642 .Filter("col0 % 1 == col0")
1643 .Filter([](int b1) { return b1 <2; }, {"cut1"})
1644 .Define("y", []() { return 1; });
1646auto count = df2.Count();
1648// Prints the graph to the rd1.dot file in the current directory
1649ROOT::RDF::SaveGraph(df, "./mydot.dot");
1650// Prints the graph to standard output
1651ROOT::RDF::SaveGraph(df);
1654The generated graph can be rendered using one of the graphviz filters, e.g. `dot`. For instance, the image below can be generated with the following command:
1656$ dot -Tpng computation_graph.dot -ocomputation_graph.png
1659\image html RDF_Graph2.png
1662### Activating RDataFrame execution logs
1664RDataFrame has experimental support for verbose logging of the event loop runtimes and other interesting related information. It is activated as follows:
1666#include <ROOT/RLogger.hxx>
1668// this increases RDF's verbosity level as long as the `verbosity` variable is in scope
1669auto verbosity = ROOT::RLogScopedVerbosity(ROOT::Detail::RDF::RDFLogChannel(), ROOT::ELogLevel::kInfo);
1676verbosity = ROOT.RLogScopedVerbosity(ROOT.Detail.RDF.RDFLogChannel(), ROOT.ELogLevel.kInfo)
1679More information (e.g. start and end of each multi-thread task) is printed using `ELogLevel.kDebug` and even more
1680(e.g. a full dump of the generated code that RDataFrame just-in-time-compiles) using `ELogLevel.kDebug+10`.
1682\anchor rdf-from-spec
1683### Creating an RDataFrame from a dataset specification file
1685RDataFrame can be created using a dataset specification JSON file:
1690df = ROOT.RDF.Experimental.FromSpec("spec.json")
1693The input dataset specification JSON file needs to be provided by the user and it describes all necessary samples and
1694their associated metadata information. The main required key is the "samples" (at least one sample is needed) and the
1695required sub-keys for each sample are "trees" and "files". Additionally, one can specify a metadata dictionary for each
1696sample in the "metadata" key.
1698A simple example for the formatting of the specification in the JSON file is the following:
1704 "trees": ["tree1", "tree2"],
1705 "files": ["file1.root", "file2.root"],
1709 "sample_category" = "data"
1713 "trees": ["tree3", "tree4"],
1714 "files": ["file3.root", "file4.root"],
1718 "sample_category" = "MC_background"
1725The metadata information from the specification file can be then accessed using the DefinePerSample function.
1726For example, to access luminosity information (stored as a double):
1729df.DefinePerSample("lumi", 'rdfsampleinfo_.GetD("lumi")')
1732or sample_category information (stored as a string):
1735df.DefinePerSample("sample_category", 'rdfsampleinfo_.GetS("sample_category")')
1738or directly the filename:
1741df.DefinePerSample("name", "rdfsampleinfo_.GetSampleName()")
1744An example implementation of the "FromSpec" method is available in tutorial: df106_HiggstoFourLeptons.py, which also
1745provides a corresponding exemplary JSON file for the dataset specification.
1748### Adding a progress bar
1750A progress bar showing the processed event statistics can be added to any RDataFrame program.
1751The event statistics include elapsed time, currently processed file, currently processed events, the rate of event processing
1752and an estimated remaining time (per file being processed). It is recorded and printed in the terminal every m events and every
1753n seconds (by default m = 1000 and n = 1). The ProgressBar can be also added when the multithread (MT) mode is enabled.
1755ProgressBar is added after creating the dataframe object (df):
1757ROOT::RDataFrame df("tree", "file.root");
1758ROOT::RDF::Experimental::AddProgressBar(df);
1761Alternatively, RDataFrame can be cast to an RNode first, giving the user more flexibility
1762For example, it can be called at any computational node, such as Filter or Define, not only the head node,
1763with no change to the ProgressBar function itself (please see the [Python interface](classROOT_1_1RDataFrame.html#python)
1764section for appropriate usage in Python):
1766ROOT::RDataFrame df("tree", "file.root");
1767auto df_1 = ROOT::RDF::RNode(df.Filter("x>1"));
1768ROOT::RDF::Experimental::AddProgressBar(df_1);
1770Examples of implemented progress bars can be seen by running [Higgs to Four Lepton tutorial](https://root.cern/doc/master/df106__HiggsToFourLeptons_8py_source.html) and [Dimuon tutorial](https://root.cern/doc/master/df102__NanoAODDimuonAnalysis_8C.html).
1772\anchor missing-values
1773### Working with missing values in the dataset
1775In certain situations a dataset might be missing one or more values at one or
1776more of its entries. For example:
1778- If the dataset is composed of multiple files and one or more files is
1779 missing one or more columns required by the analysis.
1780- When joining different datasets horizontally according to some index value
1781 (e.g. the event number), if the index does not find a match in one or more
1782 other datasets for a certain entry.
1784For example, suppose that column "y" does not have a value for entry 42:
1794If the RDataFrame application reads that column, for example if a Take() action
1795was requested, the default behaviour is to throw an exception indicating
1796that that column is missing an entry.
1798The following paragraphs discuss the functionalities provided by RDataFrame to
1799work with missing values in the dataset.
1801#### FilterAvailable and FilterMissing
1803FilterAvailable and FilterMissing are specialized RDataFrame Filter operations.
1804They take as input argument the name of a column of the dataset to watch for
1805missing values. Like Filter, they will either keep or discard an entire entry
1806based on whether a condition returns true or false. Specifically:
1808- FilterAvailable: the condition is whether the value of the column is present.
1809 If so, the entry is kept. Otherwise if the value is missing the entry is
1811- FilterMissing: the condition is whether the value of the column is missing. If
1812 so, the entry is kept. Otherwise if the value is present the entry is
1816df = ROOT.RDataFrame(dataset)
1818# Anytime an entry from "col" is missing, the entire entry will be filtered out
1819df_available = df.FilterAvailable("col")
1820df_available = df_available.Define("twice", "col * 2")
1822# Conversely, if we want to select the entries for which the column has missing
1823# values, we do the following
1824df_missingcol = df.FilterMissing("col")
1825# Following operations in the same branch of the computation graph clearly
1826# cannot access that same column, since there would be no value to read
1827df_missingcol = df_missingcol.Define("observable", "othercolumn * 2")
1831ROOT::RDataFrame df{dataset};
1833// Anytime an entry from "col" is missing, the entire entry will be filtered out
1834auto df_available = df.FilterAvailable("col");
1835auto df_twicecol = df_available.Define("twice", "col * 2");
1837// Conversely, if we want to select the entries for which the column has missing
1838// values, we do the following
1839auto df_missingcol = df.FilterMissing("col");
1840// Following operations in the same branch of the computation graph clearly
1841// cannot access that same column, since there would be no value to read
1842auto df_observable = df_missingcol.Define("observable", "othercolumn * 2");
1847DefaultValueFor creates a node of the computation graph which just forwards the
1848values of the columns necessary for other downstream nodes, when they are
1849available. In case a value of the input column passed to this function is not
1850available, the node will provide the default value passed to this function call
1854df = ROOT.RDataFrame(dataset)
1855# Anytime an entry from "col" is missing, the value will be the default one
1856default_value = ... # Some sensible default value here
1857df = df.DefaultValueFor("col", default_value)
1858df = df.Define("twice", "col * 2")
1862ROOT::RDataFrame df{dataset};
1863// Anytime an entry from "col" is missing, the value will be the default one
1864constexpr auto default_value = ... // Some sensible default value here
1865auto df_default = df.DefaultValueFor("col", default_value);
1866auto df_col = df_default.Define("twice", "col * 2");
1869#### Mixing different strategies to work with missing values in the same RDataFrame
1871All the operations presented above only act on the particular branch of the
1872computation graph where they are called, so that different results can be
1873obtained by mixing and matching the filtering or providing a default value
1877df = ROOT.RDataFrame(dataset)
1878# Anytime an entry from "col" is missing, the value will be the default one
1879default_value = ... # Some sensible default value here
1880df_default = df.DefaultValueFor("col", default_value).Define("twice", "col * 2")
1881df_filtered = df.FilterAvailable("col").Define("twice", "col * 2")
1883# Same number of total entries as the input dataset, with defaulted values
1884df_default.Display(["twice"]).Print()
1885# Only keep the entries where "col" has values
1886df_filtered.Display(["twice"]).Print()
1890ROOT::RDataFrame df{dataset};
1892// Anytime an entry from "col" is missing, the value will be the default one
1893constexpr auto default_value = ... // Some sensible default value here
1894auto df_default = df.DefaultValueFor("col", default_value).Define("twice", "col * 2");
1895auto df_filtered = df.FilterAvailable("col").Define("twice", "col * 2");
1897// Same number of total entries as the input dataset, with defaulted values
1898df_default.Display({"twice"})->Print();
1899// Only keep the entries where "col" has values
1900df_filtered.Display({"twice"})->Print();
1903#### Further considerations
1905Note that working with missing values is currently supported with a TTree-based
1906data source. Support of this functionality for other data sources may come in
1909\anchor special-values
1910### Dealing with NaN or Inf values in the dataset
1912RDataFrame does not treat NaNs or infinities beyond what the floating-point standards require, i.e. they will
1913propagate to the final result.
1914Non-finite numbers can be suppressed using Filter(), e.g.:
1917df.Filter("std::isfinite(x)").Mean("x")
1920\anchor rosetta-stone
1921### Translating TTree::Draw to RDataFrame
1929 <b>ROOT::RDataFrame</b>
1935// Get the tree and Draw a histogram of x for selected y values
1936auto *tree = file->Get<TTree>("myTree");
1937tree->Draw("x", "y > 2");
1942ROOT::RDataFrame df("myTree", file);
1943df.Filter("y > 2").Histo1D("x")->Draw();
1950// Draw a histogram of "jet_eta" with the desired weight
1951tree->Draw("jet_eta", "weight*(event == 1)");
1956df.Filter("event == 1").Histo1D("jet_eta", "weight")->Draw();
1963// Draw a histogram filled with values resulting from calling a method of the class of the `event` branch in the TTree.
1964tree->Draw("event.GetNtrack()");
1970df.Define("NTrack","event.GetNtrack()").Histo1D("NTrack")->Draw();
1977// Draw only every 10th event
1978tree->Draw("fNtrack","fEvtHdr.fEvtNum%10 == 0");
1983// Use the Filter operation together with the special RDF column: `rdfentry_`
1984df.Filter("rdfentry_ % 10 == 0").Histo1D("fNtrack")->Draw();
1991// object selection: for each event, fill histogram with array of selected pts
1992tree->Draw('Muon_pt', 'Muon_pt > 100');
1997// with RDF, arrays are read as ROOT::VecOps::RVec objects
1998df.Define("good_pt", "Muon_pt[Muon_pt > 100]").Histo1D("good_pt")->Draw();
2006// Draw the histogram and fill hnew with it
2007tree->Draw("sqrt(x)>>hnew","y>0");
2009// Retrieve hnew from the current directory
2010auto hnew = gDirectory->Get<TH1F>("hnew");
2015// We pass histogram constructor arguments to the Histo1D operation, to easily give the histogram a name
2016auto hist = df.Define("sqrt_x", "sqrt(x)").Filter("y>0").Histo1D({"hnew","hnew", 10, 0, 10}, "sqrt_x");
2023// Draw a 1D Profile histogram instead of TH2F
2024tree->Draw("y:x","","prof");
2026// Draw a 2D Profile histogram instead of TH3F
2027tree->Draw("z:y:x","","prof");
2033// Draw a 1D Profile histogram
2034df.Profile1D("x", "y")->Draw();
2036// Draw a 2D Profile histogram
2037df.Profile2D("x", "y", "z")->Draw();
2044// This command draws 2 entries starting with entry 5
2045tree->Draw("x", "","", 2, 5);
2050// Range function with arguments begin, end
2051df.Range(5,7).Histo1D("x")->Draw();
2058// Draw the X() component of the
2059// ROOT::Math::DisplacementVector3D in vec_list
2060tree->Draw("vec_list.X()");
2065df.Define("x", "ROOT::RVecD out; for(const auto &el: vec_list) out.push_back(el.X()); return out;").Histo1D("x")->Draw();
2072// Gather all values from a branch holding a collection per event, `pt`,
2073// and fill a histogram so that we can count the total number of values across all events
2074tree->Draw("pt>>histo");
2075auto histo = gDirectory->Get<TH1D>("histo");
2081df.Histo1D("pt")->GetEntries();
2219namespace Experimental {
2273 auto *
lm = df->GetLoopManager();
2275 throw std::runtime_error(
"Cannot print information about this RDataFrame, "
2276 "it was not properly created. It must be discarded.");
2278 auto defCols =
lm->GetDefaultColumnNames();
2280 std::ostringstream
ret;
2281 if (
auto ds = df->GetDataSource()) {
2282 ret <<
"A data frame associated to the data source \"" << cling::printValue(
ds) <<
"\"";
2284 ret <<
"An empty data frame that will create " <<
lm->GetNEmptyEntries() <<
" entries\n";
Basic types used by ROOT and required by TInterpreter.
unsigned long long ULong64_t
Portable unsigned long integer 8 bytes.
ROOT::Detail::TRangeCast< T, true > TRangeDynCast
TRangeDynCast is an adapter class that allows the typed iteration through a TCollection.
The head node of a RDF computation graph.
The dataset specification for RDataFrame.
std::shared_ptr< ROOT::Detail::RDF::RLoopManager > fLoopManager
< The RLoopManager at the root of this computation graph. Never null.
ROOT's RDataFrame offers a modern, high-level interface for analysis of data stored in TTree ,...
RDataFrame(std::string_view treeName, std::string_view filenameglob, const ColumnNames_t &defaultColumns={})
Build the dataframe.
ROOT::RDF::ColumnNames_t ColumnNames_t
Describe directory structure in memory.
A TTree represents a columnar dataset.
ROOT::RDF::Experimental::RDatasetSpec RetrieveSpecFromJson(const std::string &jsonFile)
Function to retrieve RDatasetSpec from JSON file provided.
ROOT::RDataFrame FromSpec(const std::string &jsonFile)
Factory method to create an RDataFrame from a JSON specification file.
std::vector< std::string > ColumnNames_t
std::shared_ptr< const ColumnNames_t > ColumnNamesPtr_t
