26#include <nlohmann/json.hpp>
1426 void Exec(unsigned int slot)
1428 fPerThreadResults[slot]++;
1431 // Called at the end of the event loop.
1434 *fFinalResult = std::accumulate(fPerThreadResults.begin(), fPerThreadResults.end(), 0);
1437 // Called by RDataFrame to retrieve the name of this action.
1438 std::string GetActionName() const { return "MyCounter"; }
1442 ROOT::RDataFrame df(10);
1443 ROOT::RDF::RResultPtr<int> resultPtr = df.Book<>(MyCounter{df.GetNSlots()}, {});
1444 // The GetValue call triggers the event loop
1445 std::cout << "Number of processed entries: " << resultPtr.GetValue() << std::endl;
1449See the Book() method for more information and [this tutorial](https://root.cern/doc/master/df018__customActions_8C.html)
1450for a more complete example.
1452#### Injecting arbitrary code in the event loop with Foreach() and ForeachSlot()
1454Foreach() takes a callable (lambda expression, free function, functor...) and a list of columns and
1455executes the callable on the values of those columns for each event that passes all upstream selections.
1456It can be used to perform actions that are not already available in the interface. For example, the following snippet
1457evaluates the root mean square of column "x":
1459// Single-thread evaluation of RMS of column "x" using Foreach
1462df.Foreach([&sumSq, &n](double x) { ++n; sumSq += x*x; }, {"x"});
1463std::cout << "rms of x: " << std::sqrt(sumSq / n) << std::endl;
1465In multi-thread runs, users are responsible for the thread-safety of the expression passed to Foreach():
1466thread will execute the expression concurrently.
1467The code above would need to employ some resource protection mechanism to ensure non-concurrent writing of `rms`; but
1468this is probably too much head-scratch for such a simple operation.
1470ForeachSlot() can help in this situation. It is an alternative version of Foreach() for which the function takes an
1471additional "processing slot" parameter besides the columns it should be applied to. RDataFrame
1472guarantees that ForeachSlot() will invoke the user expression with different `slot` parameters for different concurrent
1473executions (see [Special helper columns: rdfentry_ and rdfslot_](\ref helper-cols) for more information on the slot parameter).
1474We can take advantage of ForeachSlot() to evaluate a thread-safe root mean square of column "x":
1476// Thread-safe evaluation of RMS of column "x" using ForeachSlot
1477ROOT::EnableImplicitMT();
1478const unsigned int nSlots = df.GetNSlots();
1479std::vector<double> sumSqs(nSlots, 0.);
1480std::vector<unsigned int> ns(nSlots, 0);
1482df.ForeachSlot([&sumSqs, &ns](unsigned int slot, double x) { sumSqs[slot] += x*x; ns[slot] += 1; }, {"x"});
1483double sumSq = std::accumulate(sumSqs.begin(), sumSqs.end(), 0.); // sum all squares
1484unsigned int n = std::accumulate(ns.begin(), ns.end(), 0); // sum all counts
1485std::cout << "rms of x: " << std::sqrt(sumSq / n) << std::endl;
1487Notice how we created one `double` variable for each processing slot and later merged their results via `std::accumulate`.
1491### Dataset joins with friend trees
1493Vertically concatenating multiple trees that have the same columns (creating a logical dataset with the same columns and
1494more rows) is trivial in RDataFrame: just pass the tree name and a list of file names to RDataFrame's constructor, or create a TChain
1495out of the desired trees and pass that to RDataFrame.
1497Horizontal concatenations of trees or chains (creating a logical dataset with the same number of rows and the union of the
1498columns of multiple trees) leverages TTree's "friend" mechanism.
1500Simple joins of trees that do not have the same number of rows are also possible with indexed friend trees (see below).
1502To use friend trees in RDataFrame, set up trees with the appropriate relationships and then instantiate an RDataFrame
1508main.AddFriend(&friend, "myFriend");
1511auto df2 = df.Filter("myFriend.MyCol == 42");
1514The same applies for TChains. Columns coming from the friend trees can be referred to by their full name, like in the example above,
1515or the friend tree name can be omitted in case the column name is not ambiguous (e.g. "MyCol" could be used instead of
1516"myFriend.MyCol" in the example above if there is no column "MyCol" in the main tree).
1518\note A common source of confusion is that trees that are written out from a multi-thread Snapshot() call will have their
1519 entries (block-wise) shuffled with respect to the original tree. Such trees cannot be used as friends of the original
1520 one: rows will be mismatched.
1522Indexed friend trees provide a way to perform simple joins of multiple trees over a common column.
1523When a certain entry in the main tree (or chain) is loaded, the friend trees (or chains) will then load an entry where the
1524"index" columns have a value identical to the one in the main one. For example, in Python:
1530# If a friend tree has an index on `commonColumn`, when the main tree loads
1531# a given row, it also loads the row of the friend tree that has the same
1532# value of `commonColumn`
1533aux_tree.BuildIndex("commonColumn")
1535mainTree.AddFriend(aux_tree)
1537df = ROOT.RDataFrame(mainTree)
1540RDataFrame supports indexed friend TTrees from ROOT v6.24 in single-thread mode and from v6.28/02 in multi-thread mode.
1542\anchor other-file-formats
1543### Reading data formats other than ROOT trees
1544RDataFrame can be interfaced with RDataSources. The ROOT::RDF::RDataSource interface defines an API that RDataFrame can use to read arbitrary columnar data formats.
1546RDataFrame calls into concrete RDataSource implementations to retrieve information about the data, retrieve (thread-local) readers or "cursors" for selected columns
1547and to advance the readers to the desired data entry.
1548Some predefined RDataSources are natively provided by ROOT such as the ROOT::RDF::RCsvDS which allows to read comma separated files:
1550auto tdf = ROOT::RDF::FromCSV("MuRun2010B.csv");
1551auto filteredEvents =
1552 tdf.Filter("Q1 * Q2 == -1")
1553 .Define("m", "sqrt(pow(E1 + E2, 2) - (pow(px1 + px2, 2) + pow(py1 + py2, 2) + pow(pz1 + pz2, 2)))");
1554auto h = filteredEvents.Histo1D("m");
1558See also FromNumpy (Python-only), FromRNTuple(), FromArrow(), FromSqlite().
1561### Computation graphs (storing and reusing sets of transformations)
1563As 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
1564several paths of filtering/creation of columns are executed simultaneously, and finally aggregated results are produced.
1566RDataFrame detects when several actions use the same filter or the same defined column, and **only evaluates each
1567filter or defined column once per event**, regardless of how many times that result is used down the computation graph.
1568Objects read from each column are **built once and never copied**, for maximum efficiency.
1569When "upstream" filters are not passed, subsequent filters, temporary column expressions and actions are not evaluated,
1570so it might be advisable to put the strictest filters first in the graph.
1572\anchor representgraph
1573### Visualizing the computation graph
1574It 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
1577Invoking the function ROOT::RDF::SaveGraph() on any node that is not the head node, the computation graph of the branch
1578the node belongs to is printed. By using the head node, the entire computation graph is printed.
1580Following there is an example of usage:
1582// First, a sample computational graph is built
1583ROOT::RDataFrame df("tree", "f.root");
1585auto df2 = df.Define("x", []() { return 1; })
1586 .Filter("col0 % 1 == col0")
1587 .Filter([](int b1) { return b1 <2; }, {"cut1"})
1588 .Define("y", []() { return 1; });
1590auto count = df2.Count();
1592// Prints the graph to the rd1.dot file in the current directory
1593ROOT::RDF::SaveGraph(df, "./mydot.dot");
1594// Prints the graph to standard output
1595ROOT::RDF::SaveGraph(df);
1598The 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:
1600$ dot -Tpng computation_graph.dot -ocomputation_graph.png
1603\image html RDF_Graph2.png
1606### Activating RDataFrame execution logs
1608RDataFrame has experimental support for verbose logging of the event loop runtimes and other interesting related information. It is activated as follows:
1610#include <ROOT/RLogger.hxx>
1612// this increases RDF's verbosity level as long as the `verbosity` variable is in scope
1613auto verbosity = ROOT::Experimental::RLogScopedVerbosity(ROOT::Detail::RDF::RDFLogChannel(), ROOT::Experimental::ELogLevel::kInfo);
1620verbosity = ROOT.Experimental.RLogScopedVerbosity(ROOT.Detail.RDF.RDFLogChannel(), ROOT.Experimental.ELogLevel.kInfo)
1623More information (e.g. start and end of each multi-thread task) is printed using `ELogLevel.kDebug` and even more
1624(e.g. a full dump of the generated code that RDataFrame just-in-time-compiles) using `ELogLevel.kDebug+10`.
1626\anchor rdf-from-spec
1627### Creating an RDataFrame from a dataset specification file
1629RDataFrame can be created using a dataset specification JSON file:
1634df = ROOT.RDF.Experimental.FromSpec("spec.json")
1637The input dataset specification JSON file needs to be provided by the user and it describes all necessary samples and
1638their associated metadata information. The main required key is the "samples" (at least one sample is needed) and the
1639required sub-keys for each sample are "trees" and "files". Additionally, one can specify a metadata dictionary for each
1640sample in the "metadata" key.
1642A simple example for the formatting of the specification in the JSON file is the following:
1648 "trees": ["tree1", "tree2"],
1649 "files": ["file1.root", "file2.root"],
1653 "sample_category" = "data"
1657 "trees": ["tree3", "tree4"],
1658 "files": ["file3.root", "file4.root"],
1662 "sample_category" = "MC_background"
1669The metadata information from the specification file can be then accessed using the DefinePerSample function.
1670For example, to access luminosity information (stored as a double):
1673df.DefinePerSample("lumi", 'rdfsampleinfo_.GetD("lumi")')
1676or sample_category information (stored as a string):
1679df.DefinePerSample("sample_category", 'rdfsampleinfo_.GetS("sample_category")')
1682or directly the filename:
1685df.DefinePerSample("name", "rdfsampleinfo_.GetSampleName()")
1688An example implementation of the "FromSpec" method is available in tutorial: df106_HiggstoFourLeptons.py, which also
1689provides a corresponding exemplary JSON file for the dataset specification.
1692### Adding a progress bar
1694A progress bar showing the processed event statistics can be added to any RDataFrame program.
1695The event statistics include elapsed time, currently processed file, currently processed events, the rate of event processing
1696and an estimated remaining time (per file being processed). It is recorded and printed in the terminal every m events and every
1697n seconds (by default m = 1000 and n = 1). The ProgressBar can be also added when the multithread (MT) mode is enabled.
1699ProgressBar is added after creating the dataframe object (df):
1701ROOT::RDataFrame df("tree", "file.root");
1702ROOT::RDF::Experimental::AddProgressBar(df);
1705Alternatively, RDataFrame can be cast to an RNode first, giving the user more flexibility
1706For example, it can be called at any computational node, such as Filter or Define, not only the head node,
1707with no change to the ProgressBar function itself (please see the [Python interface](classROOT_1_1RDataFrame.html#python)
1708section for appropriate usage in Python):
1710ROOT::RDataFrame df("tree", "file.root");
1711auto df_1 = ROOT::RDF::RNode(df.Filter("x>1"));
1712ROOT::RDF::Experimental::AddProgressBar(df_1);
1714Examples 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).
1716\anchor missing-values
1717### Working with missing values in the dataset
1719In certain situations a dataset might be missing one or more values at one or
1720more of its entries. For example:
1722- If the dataset is composed of multiple files and one or more files is
1723 missing one or more columns required by the analysis.
1724- When joining different datasets horizontally according to some index value
1725 (e.g. the event number), if the index does not find a match in one or more
1726 other datasets for a certain entry.
1728For example, suppose that column "y" does not have a value for entry 42:
1738If the RDataFrame application reads that column, for example if a Take() action
1739was requested, the default behaviour is to throw an exception indicating
1740that that column is missing an entry.
1742The following paragraphs discuss the functionalities provided by RDataFrame to
1743work with missing values in the dataset.
1745#### FilterAvailable and FilterMissing
1747FilterAvailable and FilterMissing are specialized RDataFrame Filter operations.
1748They take as input argument the name of a column of the dataset to watch for
1749missing values. Like Filter, they will either keep or discard an entire entry
1750based on whether a condition returns true or false. Specifically:
1752- FilterAvailable: the condition is whether the value of the column is present.
1753 If so, the entry is kept. Otherwise if the value is missing the entry is
1755- FilterMissing: the condition is whether the value of the column is missing. If
1756 so, the entry is kept. Otherwise if the value is present the entry is
1760df = ROOT.RDataFrame(dataset)
1762# Anytime an entry from "col" is missing, the entire entry will be filtered out
1763df_available = df.FilterAvailable("col")
1764df_available = df_available.Define("twice", "col * 2")
1766# Conversely, if we want to select the entries for which the column has missing
1767# values, we do the following
1768df_missingcol = df.FilterMissing("col")
1769# Following operations in the same branch of the computation graph clearly
1770# cannot access that same column, since there would be no value to read
1771df_missingcol = df_missingcol.Define("observable", "othercolumn * 2")
1775ROOT::RDataFrame df{dataset};
1777// Anytime an entry from "col" is missing, the entire entry will be filtered out
1778auto df_available = df.FilterAvailable("col");
1779auto df_twicecol = df_available.Define("twice", "col * 2");
1781// Conversely, if we want to select the entries for which the column has missing
1782// values, we do the following
1783auto df_missingcol = df.FilterMissing("col");
1784// Following operations in the same branch of the computation graph clearly
1785// cannot access that same column, since there would be no value to read
1786auto df_observable = df_missingcol.Define("observable", "othercolumn * 2");
1791DefaultValueFor creates a node of the computation graph which just forwards the
1792values of the columns necessary for other downstream nodes, when they are
1793available. In case a value of the input column passed to this function is not
1794available, the node will provide the default value passed to this function call
1798df = ROOT.RDataFrame(dataset)
1799# Anytime an entry from "col" is missing, the value will be the default one
1800default_value = ... # Some sensible default value here
1801df = df.DefaultValueFor("col", default_value)
1802df = df.Define("twice", "col * 2")
1806ROOT::RDataFrame df{dataset};
1807// Anytime an entry from "col" is missing, the value will be the default one
1808constexpr auto default_value = ... // Some sensible default value here
1809auto df_default = df.DefaultValueFor("col", default_value);
1810auto df_col = df_default.Define("twice", "col * 2");
1813#### Mixing different strategies to work with missing values in the same RDataFrame
1815All the operations presented above only act on the particular branch of the
1816computation graph where they are called, so that different results can be
1817obtained by mixing and matching the filtering or providing a default value
1821df = ROOT.RDataFrame(dataset)
1822# Anytime an entry from "col" is missing, the value will be the default one
1823default_value = ... # Some sensible default value here
1824df_default = df.DefaultValueFor("col", default_value).Define("twice", "col * 2")
1825df_filtered = df.FilterAvailable("col").Define("twice", "col * 2")
1827# Same number of total entries as the input dataset, with defaulted values
1828df_default.Display(["twice"]).Print()
1829# Only keep the entries where "col" has values
1830df_filtered.Display(["twice"]).Print()
1834ROOT::RDataFrame df{dataset};
1836// Anytime an entry from "col" is missing, the value will be the default one
1837constexpr auto default_value = ... // Some sensible default value here
1838auto df_default = df.DefaultValueFor("col", default_value).Define("twice", "col * 2");
1839auto df_filtered = df.FilterAvailable("col").Define("twice", "col * 2");
1841// Same number of total entries as the input dataset, with defaulted values
1842df_default.Display({"twice"})->Print();
1843// Only keep the entries where "col" has values
1844df_filtered.Display({"twice"})->Print();
1847#### Further considerations
1849Note that working with missing values is currently supported with a TTree-based
1850data source. Support of this functionality for other data sources may come in
1985namespace Experimental {
2039 auto *
lm = df->GetLoopManager();
2041 throw std::runtime_error(
"Cannot print information about this RDataFrame, "
2042 "it was not properly created. It must be discarded.");
2044 auto defCols =
lm->GetDefaultColumnNames();
2046 std::ostringstream
ret;
2047 if (
auto ds = df->GetDataSource()) {
2048 ret <<
"A data frame associated to the data source \"" << cling::printValue(
ds) <<
"\"";
2050 ret <<
"An empty data frame that will create " <<
lm->GetNEmptyEntries() <<
" entries\n";
unsigned long long ULong64_t
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
tbb::task_arena is an alias of tbb::interface7::task_arena, which doesn't allow to forward declare tb...
std::shared_ptr< const ColumnNames_t > ColumnNamesPtr_t