26#include <nlohmann/json.hpp>
1433 void Exec(unsigned int slot)
1435 fPerThreadResults[slot]++;
1438 // Called at the end of the event loop.
1441 *fFinalResult = std::accumulate(fPerThreadResults.begin(), fPerThreadResults.end(), 0);
1444 // Called by RDataFrame to retrieve the name of this action.
1445 std::string GetActionName() const { return "MyCounter"; }
1449 ROOT::RDataFrame df(10);
1450 ROOT::RDF::RResultPtr<int> resultPtr = df.Book<>(MyCounter{df.GetNSlots()}, {});
1451 // The GetValue call triggers the event loop
1452 std::cout << "Number of processed entries: " << resultPtr.GetValue() << std::endl;
1456See the Book() method for more information and [this tutorial](https://root.cern/doc/master/df018__customActions_8C.html)
1457for a more complete example.
1459#### Injecting arbitrary code in the event loop with Foreach() and ForeachSlot()
1461Foreach() takes a callable (lambda expression, free function, functor...) and a list of columns and
1462executes the callable on the values of those columns for each event that passes all upstream selections.
1463It can be used to perform actions that are not already available in the interface. For example, the following snippet
1464evaluates the root mean square of column "x":
1466// Single-thread evaluation of RMS of column "x" using Foreach
1469df.Foreach([&sumSq, &n](double x) { ++n; sumSq += x*x; }, {"x"});
1470std::cout << "rms of x: " << std::sqrt(sumSq / n) << std::endl;
1472In multi-thread runs, users are responsible for the thread-safety of the expression passed to Foreach():
1473thread will execute the expression concurrently.
1474The code above would need to employ some resource protection mechanism to ensure non-concurrent writing of `rms`; but
1475this is probably too much head-scratch for such a simple operation.
1477ForeachSlot() can help in this situation. It is an alternative version of Foreach() for which the function takes an
1478additional "processing slot" parameter besides the columns it should be applied to. RDataFrame
1479guarantees that ForeachSlot() will invoke the user expression with different `slot` parameters for different concurrent
1480executions (see [Special helper columns: rdfentry_ and rdfslot_](\ref helper-cols) for more information on the slot parameter).
1481We can take advantage of ForeachSlot() to evaluate a thread-safe root mean square of column "x":
1483// Thread-safe evaluation of RMS of column "x" using ForeachSlot
1484ROOT::EnableImplicitMT();
1485const unsigned int nSlots = df.GetNSlots();
1486std::vector<double> sumSqs(nSlots, 0.);
1487std::vector<unsigned int> ns(nSlots, 0);
1489df.ForeachSlot([&sumSqs, &ns](unsigned int slot, double x) { sumSqs[slot] += x*x; ns[slot] += 1; }, {"x"});
1490double sumSq = std::accumulate(sumSqs.begin(), sumSqs.end(), 0.); // sum all squares
1491unsigned int n = std::accumulate(ns.begin(), ns.end(), 0); // sum all counts
1492std::cout << "rms of x: " << std::sqrt(sumSq / n) << std::endl;
1494Notice how we created one `double` variable for each processing slot and later merged their results via `std::accumulate`.
1498### Dataset joins with friend trees
1500Vertically concatenating multiple trees that have the same columns (creating a logical dataset with the same columns and
1501more rows) is trivial in RDataFrame: just pass the tree name and a list of file names to RDataFrame's constructor, or create a TChain
1502out of the desired trees and pass that to RDataFrame.
1504Horizontal concatenations of trees or chains (creating a logical dataset with the same number of rows and the union of the
1505columns of multiple trees) leverages TTree's "friend" mechanism.
1507Simple joins of trees that do not have the same number of rows are also possible with indexed friend trees (see below).
1509To use friend trees in RDataFrame, set up trees with the appropriate relationships and then instantiate an RDataFrame
1515main.AddFriend(&friend, "myFriend");
1518auto df2 = df.Filter("myFriend.MyCol == 42");
1521The same applies for TChains. Columns coming from the friend trees can be referred to by their full name, like in the example above,
1522or the friend tree name can be omitted in case the column name is not ambiguous (e.g. "MyCol" could be used instead of
1523"myFriend.MyCol" in the example above if there is no column "MyCol" in the main tree).
1525\note A common source of confusion is that trees that are written out from a multi-thread Snapshot() call will have their
1526 entries (block-wise) shuffled with respect to the original tree. Such trees cannot be used as friends of the original
1527 one: rows will be mismatched.
1529Indexed friend trees provide a way to perform simple joins of multiple trees over a common column.
1530When a certain entry in the main tree (or chain) is loaded, the friend trees (or chains) will then load an entry where the
1531"index" columns have a value identical to the one in the main one. For example, in Python:
1537# If a friend tree has an index on `commonColumn`, when the main tree loads
1538# a given row, it also loads the row of the friend tree that has the same
1539# value of `commonColumn`
1540aux_tree.BuildIndex("commonColumn")
1542mainTree.AddFriend(aux_tree)
1544df = ROOT.RDataFrame(mainTree)
1547RDataFrame supports indexed friend TTrees from ROOT v6.24 in single-thread mode and from v6.28/02 in multi-thread mode.
1549\anchor other-file-formats
1550### Reading data formats other than ROOT trees
1551RDataFrame can be interfaced with RDataSources. The ROOT::RDF::RDataSource interface defines an API that RDataFrame can use to read arbitrary columnar data formats.
1553RDataFrame calls into concrete RDataSource implementations to retrieve information about the data, retrieve (thread-local) readers or "cursors" for selected columns
1554and to advance the readers to the desired data entry.
1555Some predefined RDataSources are natively provided by ROOT such as the ROOT::RDF::RCsvDS which allows to read comma separated files:
1557auto tdf = ROOT::RDF::FromCSV("MuRun2010B.csv");
1558auto filteredEvents =
1559 tdf.Filter("Q1 * Q2 == -1")
1560 .Define("m", "sqrt(pow(E1 + E2, 2) - (pow(px1 + px2, 2) + pow(py1 + py2, 2) + pow(pz1 + pz2, 2)))");
1561auto h = filteredEvents.Histo1D("m");
1565See also FromNumpy (Python-only), FromRNTuple(), FromArrow(), FromSqlite().
1568### Computation graphs (storing and reusing sets of transformations)
1570As 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
1571several paths of filtering/creation of columns are executed simultaneously, and finally aggregated results are produced.
1573RDataFrame detects when several actions use the same filter or the same defined column, and **only evaluates each
1574filter or defined column once per event**, regardless of how many times that result is used down the computation graph.
1575Objects read from each column are **built once and never copied**, for maximum efficiency.
1576When "upstream" filters are not passed, subsequent filters, temporary column expressions and actions are not evaluated,
1577so it might be advisable to put the strictest filters first in the graph.
1579\anchor representgraph
1580### Visualizing the computation graph
1581It 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
1584Invoking the function ROOT::RDF::SaveGraph() on any node that is not the head node, the computation graph of the branch
1585the node belongs to is printed. By using the head node, the entire computation graph is printed.
1587Following there is an example of usage:
1589// First, a sample computational graph is built
1590ROOT::RDataFrame df("tree", "f.root");
1592auto df2 = df.Define("x", []() { return 1; })
1593 .Filter("col0 % 1 == col0")
1594 .Filter([](int b1) { return b1 <2; }, {"cut1"})
1595 .Define("y", []() { return 1; });
1597auto count = df2.Count();
1599// Prints the graph to the rd1.dot file in the current directory
1600ROOT::RDF::SaveGraph(df, "./mydot.dot");
1601// Prints the graph to standard output
1602ROOT::RDF::SaveGraph(df);
1605The 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:
1607$ dot -Tpng computation_graph.dot -ocomputation_graph.png
1610\image html RDF_Graph2.png
1613### Activating RDataFrame execution logs
1615RDataFrame has experimental support for verbose logging of the event loop runtimes and other interesting related information. It is activated as follows:
1617#include <ROOT/RLogger.hxx>
1619// this increases RDF's verbosity level as long as the `verbosity` variable is in scope
1620auto verbosity = ROOT::RLogScopedVerbosity(ROOT::Detail::RDF::RDFLogChannel(), ROOT::ELogLevel::kInfo);
1627verbosity = ROOT.RLogScopedVerbosity(ROOT.Detail.RDF.RDFLogChannel(), ROOT.ELogLevel.kInfo)
1630More information (e.g. start and end of each multi-thread task) is printed using `ELogLevel.kDebug` and even more
1631(e.g. a full dump of the generated code that RDataFrame just-in-time-compiles) using `ELogLevel.kDebug+10`.
1633\anchor rdf-from-spec
1634### Creating an RDataFrame from a dataset specification file
1636RDataFrame can be created using a dataset specification JSON file:
1641df = ROOT.RDF.Experimental.FromSpec("spec.json")
1644The input dataset specification JSON file needs to be provided by the user and it describes all necessary samples and
1645their associated metadata information. The main required key is the "samples" (at least one sample is needed) and the
1646required sub-keys for each sample are "trees" and "files". Additionally, one can specify a metadata dictionary for each
1647sample in the "metadata" key.
1649A simple example for the formatting of the specification in the JSON file is the following:
1655 "trees": ["tree1", "tree2"],
1656 "files": ["file1.root", "file2.root"],
1660 "sample_category" = "data"
1664 "trees": ["tree3", "tree4"],
1665 "files": ["file3.root", "file4.root"],
1669 "sample_category" = "MC_background"
1676The metadata information from the specification file can be then accessed using the DefinePerSample function.
1677For example, to access luminosity information (stored as a double):
1680df.DefinePerSample("lumi", 'rdfsampleinfo_.GetD("lumi")')
1683or sample_category information (stored as a string):
1686df.DefinePerSample("sample_category", 'rdfsampleinfo_.GetS("sample_category")')
1689or directly the filename:
1692df.DefinePerSample("name", "rdfsampleinfo_.GetSampleName()")
1695An example implementation of the "FromSpec" method is available in tutorial: df106_HiggstoFourLeptons.py, which also
1696provides a corresponding exemplary JSON file for the dataset specification.
1699### Adding a progress bar
1701A progress bar showing the processed event statistics can be added to any RDataFrame program.
1702The event statistics include elapsed time, currently processed file, currently processed events, the rate of event processing
1703and an estimated remaining time (per file being processed). It is recorded and printed in the terminal every m events and every
1704n seconds (by default m = 1000 and n = 1). The ProgressBar can be also added when the multithread (MT) mode is enabled.
1706ProgressBar is added after creating the dataframe object (df):
1708ROOT::RDataFrame df("tree", "file.root");
1709ROOT::RDF::Experimental::AddProgressBar(df);
1712Alternatively, RDataFrame can be cast to an RNode first, giving the user more flexibility
1713For example, it can be called at any computational node, such as Filter or Define, not only the head node,
1714with no change to the ProgressBar function itself (please see the [Python interface](classROOT_1_1RDataFrame.html#python)
1715section for appropriate usage in Python):
1717ROOT::RDataFrame df("tree", "file.root");
1718auto df_1 = ROOT::RDF::RNode(df.Filter("x>1"));
1719ROOT::RDF::Experimental::AddProgressBar(df_1);
1721Examples 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).
1723\anchor missing-values
1724### Working with missing values in the dataset
1726In certain situations a dataset might be missing one or more values at one or
1727more of its entries. For example:
1729- If the dataset is composed of multiple files and one or more files is
1730 missing one or more columns required by the analysis.
1731- When joining different datasets horizontally according to some index value
1732 (e.g. the event number), if the index does not find a match in one or more
1733 other datasets for a certain entry.
1735For example, suppose that column "y" does not have a value for entry 42:
1745If the RDataFrame application reads that column, for example if a Take() action
1746was requested, the default behaviour is to throw an exception indicating
1747that that column is missing an entry.
1749The following paragraphs discuss the functionalities provided by RDataFrame to
1750work with missing values in the dataset.
1752#### FilterAvailable and FilterMissing
1754FilterAvailable and FilterMissing are specialized RDataFrame Filter operations.
1755They take as input argument the name of a column of the dataset to watch for
1756missing values. Like Filter, they will either keep or discard an entire entry
1757based on whether a condition returns true or false. Specifically:
1759- FilterAvailable: the condition is whether the value of the column is present.
1760 If so, the entry is kept. Otherwise if the value is missing the entry is
1762- FilterMissing: the condition is whether the value of the column is missing. If
1763 so, the entry is kept. Otherwise if the value is present the entry is
1767df = ROOT.RDataFrame(dataset)
1769# Anytime an entry from "col" is missing, the entire entry will be filtered out
1770df_available = df.FilterAvailable("col")
1771df_available = df_available.Define("twice", "col * 2")
1773# Conversely, if we want to select the entries for which the column has missing
1774# values, we do the following
1775df_missingcol = df.FilterMissing("col")
1776# Following operations in the same branch of the computation graph clearly
1777# cannot access that same column, since there would be no value to read
1778df_missingcol = df_missingcol.Define("observable", "othercolumn * 2")
1782ROOT::RDataFrame df{dataset};
1784// Anytime an entry from "col" is missing, the entire entry will be filtered out
1785auto df_available = df.FilterAvailable("col");
1786auto df_twicecol = df_available.Define("twice", "col * 2");
1788// Conversely, if we want to select the entries for which the column has missing
1789// values, we do the following
1790auto df_missingcol = df.FilterMissing("col");
1791// Following operations in the same branch of the computation graph clearly
1792// cannot access that same column, since there would be no value to read
1793auto df_observable = df_missingcol.Define("observable", "othercolumn * 2");
1798DefaultValueFor creates a node of the computation graph which just forwards the
1799values of the columns necessary for other downstream nodes, when they are
1800available. In case a value of the input column passed to this function is not
1801available, the node will provide the default value passed to this function call
1805df = ROOT.RDataFrame(dataset)
1806# Anytime an entry from "col" is missing, the value will be the default one
1807default_value = ... # Some sensible default value here
1808df = df.DefaultValueFor("col", default_value)
1809df = df.Define("twice", "col * 2")
1813ROOT::RDataFrame df{dataset};
1814// Anytime an entry from "col" is missing, the value will be the default one
1815constexpr auto default_value = ... // Some sensible default value here
1816auto df_default = df.DefaultValueFor("col", default_value);
1817auto df_col = df_default.Define("twice", "col * 2");
1820#### Mixing different strategies to work with missing values in the same RDataFrame
1822All the operations presented above only act on the particular branch of the
1823computation graph where they are called, so that different results can be
1824obtained by mixing and matching the filtering or providing a default value
1828df = ROOT.RDataFrame(dataset)
1829# Anytime an entry from "col" is missing, the value will be the default one
1830default_value = ... # Some sensible default value here
1831df_default = df.DefaultValueFor("col", default_value).Define("twice", "col * 2")
1832df_filtered = df.FilterAvailable("col").Define("twice", "col * 2")
1834# Same number of total entries as the input dataset, with defaulted values
1835df_default.Display(["twice"]).Print()
1836# Only keep the entries where "col" has values
1837df_filtered.Display(["twice"]).Print()
1841ROOT::RDataFrame df{dataset};
1843// Anytime an entry from "col" is missing, the value will be the default one
1844constexpr auto default_value = ... // Some sensible default value here
1845auto df_default = df.DefaultValueFor("col", default_value).Define("twice", "col * 2");
1846auto df_filtered = df.FilterAvailable("col").Define("twice", "col * 2");
1848// Same number of total entries as the input dataset, with defaulted values
1849df_default.Display({"twice"})->Print();
1850// Only keep the entries where "col" has values
1851df_filtered.Display({"twice"})->Print();
1854#### Further considerations
1856Note that working with missing values is currently supported with a TTree-based
1857data source. Support of this functionality for other data sources may come in
1860\anchor special-values
1861### Dealing with NaN or Inf values in the dataset
1863RDataFrame does not treat NaNs or infinities beyond what the floating-point standards require, i.e. they will
1864propagate to the final result.
1865Non-finite numbers can be suppressed using Filter(), e.g.:
1868df.Filter("std::isfinite(x)").Mean("x")
2003namespace Experimental {
2057 auto *
lm = df->GetLoopManager();
2059 throw std::runtime_error(
"Cannot print information about this RDataFrame, "
2060 "it was not properly created. It must be discarded.");
2062 auto defCols =
lm->GetDefaultColumnNames();
2064 std::ostringstream
ret;
2065 if (
auto ds = df->GetDataSource()) {
2066 ret <<
"A data frame associated to the data source \"" << cling::printValue(
ds) <<
"\"";
2068 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
