Logo ROOT  
Reference Guide
InternalTreeUtils.cxx
Go to the documentation of this file.
1/*************************************************************************
2 * Copyright (C) 1995-2021, Rene Brun and Fons Rademakers. *
3 * All rights reserved. *
4 * *
5 * For the licensing terms see $ROOTSYS/LICENSE. *
6 * For the list of contributors see $ROOTSYS/README/CREDITS. *
7 *************************************************************************/
8
10#include "TTree.h"
11#include "TChain.h"
12#include "TFile.h"
13#include "TFriendElement.h"
14
15#include <utility> // std::pair
16#include <vector>
17#include <stdexcept> // std::runtime_error
18#include <string>
19
20namespace ROOT {
21namespace Internal {
22namespace TreeUtils {
23
24////////////////////////////////////////////////////////////////////////////////
25/// \fn std::vector<std::string> GetFileNamesFromTree(const TTree &tree)
26/// \ingroup tree
27/// \brief Get and store the file names associated with the input tree.
28/// \param[in] tree The tree from which friends information will be gathered.
29/// \throws std::runtime_error If no files could be associated with the input tree.
30std::vector<std::string> GetFileNamesFromTree(const TTree &tree)
31{
32 std::vector<std::string> filenames;
33
34 // If the input tree is a TChain, traverse its list of associated files.
35 if (auto chain = dynamic_cast<const TChain *>(&tree)) {
36 const auto *chainFiles = chain->GetListOfFiles();
37 if (!chainFiles) {
38 throw std::runtime_error("Could not retrieve a list of files from the input TChain.");
39 }
40 // Store this in a variable so it can be later used in `filenames.reserve`
41 // if it passes the check.
42 const auto nfiles = chainFiles->GetEntries();
43 if (nfiles == 0) {
44 throw std::runtime_error("The list of files associated with the input TChain is empty.");
45 }
46 filenames.reserve(nfiles);
47 for (const auto *f : *chainFiles)
48 filenames.emplace_back(f->GetTitle());
49 } else {
50 const TFile *f = tree.GetCurrentFile();
51 if (!f) {
52 throw std::runtime_error("The input TTree is not linked to any file, "
53 "in-memory-only trees are not supported.");
54 }
55
56 filenames.emplace_back(f->GetName());
57 }
58
59 return filenames;
60}
61
62////////////////////////////////////////////////////////////////////////////////
63/// \fn RFriendInfo GetFriendInfo(const TTree &tree)
64/// \ingroup tree
65/// \brief Get and store the names, aliases and file names of the direct friends of the tree.
66/// \param[in] tree The tree from which friends information will be gathered.
67/// \throws std::runtime_error If the input tree has a list of friends, but any
68/// of them could not be associated with any file.
69///
70/// Calls TTree::GetListOfFriends and parses its result for the names, aliases
71/// and file names, with different methodologies depending on whether the
72/// parameter is a TTree or a TChain.
73///
74/// \note This function only retrieves information about <b>direct friends</b>
75/// of the input tree. It will not recurse through friends of friends and
76/// does not take into account circular references in the list of friends
77/// of the input tree.
78///
79/// \returns An RFriendInfo struct, containing the information parsed from the
80/// list of friends. The struct will contain three vectors, which elements at
81/// position `i` represent the `i`-th friend of the input tree. If this friend
82/// is a TTree, the `i`-th element of each of the three vectors will contain
83/// respectively:
84/// - A pair with the name and alias of the tree (the alias might not be
85/// present, in which case it will be just an empty string).
86/// - A vector with a single string representing the path to current file where
87/// the tree is stored.
88/// - An empty vector.
89/// .
90/// If the `i`-th friend is a TChain instead, the `i`-th element of each of the
91/// three vectors will contain respectively:
92/// - A pair with the name and alias of the chain (if present, both might be
93/// empty strings).
94/// - A vector with all the paths to the files contained in the chain.
95/// - A vector with all the the names of the trees making up the chain,
96/// associated with the file names of the previous vector.
98{
99 std::vector<NameAlias> friendNames;
100 std::vector<std::vector<std::string>> friendFileNames;
101 std::vector<std::vector<std::string>> friendChainSubNames;
102
103 // Typically, the correct way to call GetListOfFriends would be `tree.GetTree()->GetListOfFriends()`
104 // (see e.g. the discussion at https://github.com/root-project/root/issues/6741).
105 // However, in this case, in case we are dealing with a TChain we really only care about the TChain's
106 // list of friends (which will need to be rebuilt in each processing task) while friends of the TChain's
107 // internal TTree, if any, will be automatically loaded in each task just like they would be automatically
108 // loaded here if we used tree.GetTree()->GetListOfFriends().
109 const auto *friends = tree.GetListOfFriends();
110 if (!friends)
111 return RFriendInfo();
112
113 for (auto fr : *friends) {
114 // Can't pass fr as const TObject* because TFriendElement::GetTree is not const.
115 // Also, we can't retrieve frTree as const TTree* because of TTree::GetFriendAlias(TTree *) a few lines later
116 auto frTree = static_cast<TFriendElement *>(fr)->GetTree();
117
118 // The vector of (name,alias) pairs of the current friend
119 friendFileNames.emplace_back();
120 auto &fileNames = friendFileNames.back();
121
122 // The vector of names of sub trees of the current friend, if it is a TChain.
123 // Otherwise, just an empty vector.
124 friendChainSubNames.emplace_back();
125 auto &chainSubNames = friendChainSubNames.back();
126
127 // Check if friend tree/chain has an alias
128 const auto *alias_c = tree.GetFriendAlias(frTree);
129 const std::string alias = alias_c != nullptr ? alias_c : "";
130
131 // If the current tree is a TChain
132 if (auto frChain = dynamic_cast<const TChain *>(frTree)) {
133 // Note that each TChainElement returned by TChain::GetListOfFiles has a name
134 // equal to the tree name of this TChain and a title equal to the filename.
135 // Accessing the information like this ensures that we get the correct
136 // filenames and treenames if the treename is given as part of the filename
137 // via chain.AddFile(file.root/myTree) and as well if the tree name is given
138 // in the constructor via TChain(myTree) and a file is added later by chain.AddFile(file.root).
139 // Caveat: The chain may be made of sub-trees with different names. All
140 // tree names need to be retrieved separately, see below.
141
142 // Get filelist of the current chain
143 const auto *chainFiles = frChain->GetListOfFiles();
144 if (!chainFiles || chainFiles->GetEntries() == 0) {
145 throw std::runtime_error("A TChain in the list of friends does not contain any file. "
146 "Friends with no associated files are not supported.");
147 }
148
149 // Retrieve the name of the chain and add a (name, alias) pair
150 friendNames.emplace_back(std::make_pair(frChain->GetName(), alias));
151 // Each file in the chain can contain a TTree with a different name wrt
152 // the main TChain. Retrieve the name of the file through `GetTitle`
153 // and the name of the tree through `GetName`
154 for (const auto *f : *chainFiles) {
155 chainSubNames.emplace_back(f->GetName());
156 fileNames.emplace_back(f->GetTitle());
157 }
158 } else {
159 // Get name of the tree
160 const auto realName = GetTreeFullPaths(*frTree)[0];
161 friendNames.emplace_back(std::make_pair(realName, alias));
162
163 // Get filename
164 const auto *f = frTree->GetCurrentFile();
165 if (!f)
166 throw std::runtime_error("A TTree in the list of friends is not linked to any file. "
167 "Friends with no associated files are not supported.");
168 fileNames.emplace_back(f->GetName());
169 }
170 }
171
172 return RFriendInfo{std::move(friendNames), std::move(friendFileNames), std::move(friendChainSubNames)};
173}
174
175////////////////////////////////////////////////////////////////////////////////
176/// \fn std::vector<std::string> GetTreeFullPaths(const TTree &tree)
177/// \ingroup tree
178/// \brief Retrieve the full path(s) to a TTree or the trees in a TChain.
179/// \param[in] tree The tree or chain from which the paths will be retrieved.
180/// \throws std::runtime_error If the input tree is a TChain but no files could
181/// be found associated with it.
182/// \return If the input argument is a TChain, returns a vector of strings with
183/// the name of the tree of each file in the chain. If the input
184/// argument is a TTree, returns a vector with a single element that is
185/// the full path of the tree in the file (e.g. the name of the tree
186/// itself or the path with the directories inside the file). Finally,
187/// the function returns a vector with just the name of the tree if it
188/// couldn't do any better.
189std::vector<std::string> GetTreeFullPaths(const TTree &tree)
190{
191 // Case 1: this is a TChain. For each file it contains, GetName returns the name of the tree in that file
192 if (auto chain = dynamic_cast<const TChain *>(&tree)) {
193 const auto *chainFiles = chain->GetListOfFiles();
194 if (!chainFiles || chainFiles->GetEntries() == 0) {
195 throw std::runtime_error("The input TChain does not contain any file.");
196 }
197 std::vector<std::string> treeNames;
198 for (const auto *f : *chainFiles)
199 treeNames.emplace_back(f->GetName());
200
201 return treeNames;
202 }
203
204 // Case 2: this is a TTree: we get the full path of it
205 if (const auto *treeDir = tree.GetDirectory()) {
206 // We have 2 subcases (ROOT-9948):
207 // - 1. treeDir is a TFile: return the name of the tree.
208 // - 2. treeDir is a directory: reconstruct the path to the tree in the directory.
209 // Use dynamic_cast to check whether the directory is a TFile
210 if (dynamic_cast<const TFile *>(treeDir)) {
211 return {tree.GetName()};
212 }
213 std::string fullPath = treeDir->GetPath(); // e.g. "file.root:/dir"
214 fullPath = fullPath.substr(fullPath.find(":/") + 1); // e.g. "/dir"
215 fullPath += "/";
216 fullPath += tree.GetName(); // e.g. "/dir/tree"
217 return {fullPath};
218 }
219
220 // We do our best and return the name of the tree
221 return {tree.GetName()};
222}
223
224} // namespace TreeUtils
225} // namespace Internal
226} // namespace ROOT
#define f(i)
Definition: RSha256.hxx:104
A chain is a collection of files containing TTree objects.
Definition: TChain.h:33
A ROOT file is a suite of consecutive data records (TKey instances) with a well defined format.
Definition: TFile.h:54
A TFriendElement TF describes a TTree object TF in a file.
A TTree represents a columnar dataset.
Definition: TTree.h:79
Different standalone functions to work with trees and tuples, not reqiuired to be a member of any cla...
std::vector< std::string > GetTreeFullPaths(const TTree &tree)
Retrieve the full path(s) to a TTree or the trees in a TChain.
RFriendInfo GetFriendInfo(const TTree &tree)
Get and store the names, aliases and file names of the direct friends of the tree.
std::vector< std::string > GetFileNamesFromTree(const TTree &tree)
Get and store the file names associated with the input tree.
tbb::task_arena is an alias of tbb::interface7::task_arena, which doesn't allow to forward declare tb...
Definition: tree.py:1
Information about friend trees of a certain TTree or TChain object.