[main] Rewrite rootcp in C++

main/CMakeLists.txt

@@ -136,6 +136,7 @@ add_dependencies(rootcint rootcint)
 if (TARGET Gui)
   ROOT_EXECUTABLE(rootbrowse src/rootbrowse.cxx LIBRARIES RIO Core Rint Gui)
 endif()
+ROOT_EXECUTABLE(rootcp src/rootcp.cxx LIBRARIES RIO Tree Core Rint)
 ROOT_EXECUTABLE(rootls src/rootls.cxx LIBRARIES RIO Tree Core Rint ROOTNTuple)
 
 ROOT_ADD_TEST_SUBDIRECTORY(test)

main/python/cmdLineUtils.py

@@ -13,11 +13,11 @@
 # http://stackoverflow.com/questions/4675728/redirect-stdout-to-a-file-in-python/22434262#22434262
 # Thanks J.F. Sebastian !!
 
 from contextlib import contextmanager
 import os
 import sys
 from time import sleep
 from itertools import zip_longest
 
 def fileno(file_or_fd):
     """
@@ -81,10 +81,10 @@
 ROOT.PyConfig.IgnoreCommandLineOptions = True
 ROOT.gROOT.GetVersion()
 
 import argparse
 import glob
 import fnmatch
 import logging
 
 LOG_FORMAT = "%(levelname)s: %(message)s"
 logging.basicConfig(format=LOG_FORMAT)
@@ -328,12 +328,12 @@
     Open a ROOT file (like openROOTFile) with the possibility
     to change compression settings
     """
     if compress != None and os.path.isfile(fileName):
         logging.warning("can't change compression settings on existing file")
         return None
     mode = "recreate" if recreate else "update"
     theFile = openROOTFile(fileName, mode)
     if compress != None:
         theFile.SetCompressionSettings(compress)
     return theFile
 
@@ -783,52 +783,6 @@
 # End of help strings
 ##########
 
-##########
-# ROOTCP
-
-
-def _copyObjects(fileName, pathSplitList, destFile, destPathSplit, oneFile, recursive, replace):
-    retcode = 0
-    destFileName = destFile.GetName()
-    rootFile = openROOTFile(fileName) if fileName != destFileName else destFile
-    if not rootFile:
-        return 1
-    ROOT.gROOT.GetListOfFiles().Remove(rootFile)  # Fast copy necessity
-    for pathSplit in pathSplitList:
-        oneSource = oneFile and len(pathSplitList) == 1
-        retcode += copyRootObject(rootFile, pathSplit, destFile, destPathSplit, oneSource, recursive, replace)
-    if fileName != destFileName:
-        rootFile.Close()
-    return retcode
-
-
-def rootCp(sourceList, destFileName, destPathSplit, compress=None, recreate=False, recursive=False, replace=False):
-    # Check arguments
-    if sourceList == [] or destFileName == "":
-        return 1
-    if recreate and destFileName in [n[0] for n in sourceList]:
-        logging.error("cannot recreate destination file if this is also a source file")
-        return 1
-
-    # Open destination file
-    destFile = openROOTFileCompress(destFileName, compress, recreate)
-    if not destFile:
-        return 1
-    ROOT.gROOT.GetListOfFiles().Remove(destFile)  # Fast copy necessity
-
-    # Loop on the root files
-    retcode = 0
-    for fileName, pathSplitList in sourceList:
-        retcode += _copyObjects(
-            fileName, pathSplitList, destFile, destPathSplit, len(sourceList) == 1, recursive, replace
-        )
-    destFile.Close()
-    return retcode
-
-
-# End of ROOTCP
-##########
-
 ##########
 # ROOTEVENTSELECTOR
 

main/src/RootObjTree.cxx

@@ -0,0 +1,234 @@
+// \file RootObjTree.cxx
+///
+/// \author Giacomo Parolini <giacomo.parolini@cern.ch>
+/// \date 2025-10-14
+
+#include "RootObjTree.hxx"
+
+#include "wildcards.hpp"
+
+#include <TFile.h>
+
+#include <ROOT/StringUtils.hxx>
+
+#include <algorithm>
+#include <deque>
+#include <iostream>
+#include <set>
+
+static bool MatchesGlob(std::string_view haystack, std::string_view pattern)
+{
+   return wildcards::match(haystack, pattern);
+}
+
+ROOT::CmdLine::RootSource
+ROOT::CmdLine::GetMatchingPathsInFile(std::string_view fileName, std::string_view pattern, std::uint32_t flags)
+{
+   ROOT::CmdLine::RootSource source;
+   source.fFileName = fileName;
+   auto &nodeTree = source.fObjectTree;
+   nodeTree.fFile = std::unique_ptr<TFile>(TFile::Open(std::string(fileName).c_str(), "READ"));
+   if (!nodeTree.fFile) {
+      source.fErrors.push_back("Failed to open file");
+      return source;
+   }
+
+   const auto patternSplits = pattern.empty() ? std::vector<std::string>{} : ROOT::Split(pattern, "/");
+
+   // Match all objects at all nesting levels down to the deepest nesting level of `pattern` (or all nesting levels
+   // if we have the "recursive listing" flag). The nodes are visited breadth-first.
+   {
+      ROOT::CmdLine::RootObjNode rootNode = {};
+      rootNode.fName = std::string(fileName);
+      rootNode.fClassName = nodeTree.fFile->Class()->GetName();
+      rootNode.fDir = nodeTree.fFile.get();
+      nodeTree.fNodes.emplace_back(std::move(rootNode));
+   }
+   std::deque<NodeIdx_t> nodesToVisit{0};
+
+   // Keep track of the object names found at every nesting level and only add the first one.
+   std::unordered_set<std::string> namesFound;
+
+   const bool isRecursive = flags & EGetMatchingPathsFlags::kRecursive;
+   do {
+      NodeIdx_t curIdx = nodesToVisit.front();
+      nodesToVisit.pop_front();
+      ROOT::CmdLine::RootObjNode *cur = &nodeTree.fNodes[curIdx];
+      assert(cur->fDir);
+
+      // Sort the keys by name
+      std::vector<TKey *> keys;
+      keys.reserve(cur->fDir->GetListOfKeys()->GetEntries());
+      for (TKey *key : ROOT::Detail::TRangeStaticCast<TKey>(cur->fDir->GetListOfKeys()))
+         keys.push_back(key);
+
+      std::sort(keys.begin(), keys.end(),
+                [](const auto *a, const auto *b) { return strcmp(a->GetName(), b->GetName()) < 0; });
+
+      namesFound.clear();
+
+      for (TKey *key : keys) {
+         // Don't recurse lower than requested by `pattern` unless we explicitly have the `recursive listing` flag.
+         if (cur->fNesting < patternSplits.size() && !MatchesGlob(key->GetName(), patternSplits[cur->fNesting]))
+            continue;
+
+         if (namesFound.count(key->GetName()) > 0) {
+            std::cerr << "WARNING: Several versions of '" << key->GetName() << "' are present in '" << fileName
+                      << "'. Only the most recent will be considered.\n";
+            continue;
+         }
+         namesFound.insert(key->GetName());
+
+         auto &newChild = nodeTree.fNodes.emplace_back(NodeFromKey(*key));
+         // Need to get back cur since the emplace_back() may have moved it.
+         cur = &nodeTree.fNodes[curIdx];
+         newChild.fNesting = cur->fNesting + 1;
+         newChild.fParent = curIdx;
+         if (!cur->fNChildren)
+            cur->fFirstChild = nodeTree.fNodes.size() - 1;
+         cur->fNChildren++;
+
+         const auto *cl = TClass::GetClass(key->GetClassName());
+         if (cl && cl->InheritsFrom("TDirectory"))
+            newChild.fDir = cur->fDir->GetDirectory(key->GetName());
+      }
+
+      // Only recurse into subdirectories that are up to the deepest level we ask for through `pattern`.
+      if (cur->fNesting < patternSplits.size() || isRecursive) {
+         for (auto childIdx = cur->fFirstChild; childIdx < cur->fFirstChild + cur->fNChildren; ++childIdx) {
+            auto &child = nodeTree.fNodes[childIdx];
+            if (child.fDir)
+               nodesToVisit.push_back(childIdx);
+            else if (cur->fNesting < patternSplits.size())
+               nodeTree.fLeafList.push_back(childIdx);
+         }
+      }
+      if (cur->fNesting == patternSplits.size()) {
+         if (cur->fDir)
+            nodeTree.fDirList.push_back(curIdx);
+         else
+            nodeTree.fLeafList.push_back(curIdx);
+      }
+   } while (!nodesToVisit.empty());
+
+   return source;
+}
+
+ROOT::RResult<std::pair<std::string_view, std::string_view>>
+ROOT::CmdLine::SplitIntoFileNameAndPattern(std::string_view sourceRaw)
+{
+   auto prefixIdx = sourceRaw.find("://");
+   std::string_view::size_type separatorIdx = 0;
+   if (prefixIdx != std::string_view::npos) {
+      bool prefixFound = false;
+      // Handle known URI prefixes
+      static const char *const specialPrefixes[] = {"http", "https", "root", "gs", "s3"};
+      auto prefix = sourceRaw.substr(0, prefixIdx);
+      for (std::string_view knownPrefix : specialPrefixes) {
+         if (prefix == knownPrefix) {
+            prefixFound = true;
+            break;
+         }
+      }
+      if (!prefixFound) {
+         return R__FAIL("unknown file protocol");
+      }
+      separatorIdx = sourceRaw.substr(prefixIdx + 3).find_first_of(':');
+      if (separatorIdx != std::string_view::npos)
+         separatorIdx += prefixIdx + 3;
+   } else {
+      separatorIdx = sourceRaw.find_first_of(':');
+   }
+
+   if (separatorIdx != std::string_view::npos) {
+      return {{sourceRaw.substr(0, separatorIdx), sourceRaw.substr(separatorIdx + 1)}};
+   }
+   return {{sourceRaw, std::string_view{}}};
+}
+
+ROOT::CmdLine::RootSource ROOT::CmdLine::ParseRootSource(std::string_view sourceRaw, std::uint32_t flags)
+{
+   ROOT::CmdLine::RootSource source;
+
+   auto res = SplitIntoFileNameAndPattern(sourceRaw);
+   if (!res) {
+      source.fErrors.push_back(res.GetError()->GetReport());
+      return source;
+   }
+
+   auto [fileName, tokens] = res.Unwrap();
+   source = ROOT::CmdLine::GetMatchingPathsInFile(fileName, tokens, flags);
+
+   assert(source.fErrors.empty() == !!source.fObjectTree.fFile);
+   return source;
+}
+
+std::vector<ROOT::CmdLine::RootSource>
+ROOT::CmdLine::ParseRootSources(const std::vector<std::string> &sourcesRaw, std::uint32_t flags)
+{
+   std::vector<ROOT::CmdLine::RootSource> sources;
+   sources.reserve(sourcesRaw.size());
+
+   for (const auto &srcRaw : sourcesRaw) {
+      sources.push_back(ParseRootSource(srcRaw, flags));
+   }
+
+   return sources;
+}
+
+void ROOT::CmdLine::PrintObjTree(const RootObjTree &tree, std::ostream &out)
+{
+   if (tree.fNodes.empty())
+      return;
+
+   struct RevNode {
+      std::set<NodeIdx_t> fChildren;
+   };
+   std::vector<RevNode> revNodes;
+   revNodes.resize(tree.fNodes.size());
+
+   // Un-linearize the tree
+   for (int i = (int)tree.fNodes.size() - 1; i >= 0; --i) {
+      const auto *node = &tree.fNodes[i];
+      NodeIdx_t childIdx = i;
+      NodeIdx_t parentIdx = node->fParent;
+      while (childIdx != parentIdx) {
+         auto &revNodeParent = revNodes[parentIdx];
+         revNodeParent.fChildren.insert(childIdx);
+         node = &tree.fNodes[parentIdx];
+         childIdx = parentIdx;
+         parentIdx = node->fParent;
+      }
+   }
+
+   // Print out the tree.
+   // Vector of {nesting, nodeIdx}
+   std::vector<std::pair<std::uint32_t, NodeIdx_t>> nodesToVisit = {{0, 0}};
+   while (!nodesToVisit.empty()) {
+      const auto [nesting, nodeIdx] = nodesToVisit.back();
+      nodesToVisit.pop_back();
+      const auto &cur = revNodes[nodeIdx];
+      const auto &node = tree.fNodes[nodeIdx];
+      for (auto i = 0u; i < 2 * nesting; ++i)
+         out << ' ';
+      out << node.fName << " : " << node.fClassName << "\n";
+      // Add the children in reverse order to preserve alphabetical order during depth-first visit.
+      for (auto it = cur.fChildren.rbegin(); it != cur.fChildren.rend(); ++it) {
+         nodesToVisit.push_back({nesting + 1, *it});
+      }
+   }
+}
+
+std::string ROOT::CmdLine::NodeFullPath(const ROOT::CmdLine::RootObjTree &tree, ROOT::CmdLine::NodeIdx_t nodeIdx,
+                                        ROOT::CmdLine::ENodeFullPathOpt opt)
+{
+   const RootObjNode *node = &tree.fNodes[nodeIdx];
+   std::string fullPath = node->fName;
+   while (node->fParent != 0) {
+      node = &tree.fNodes[node->fParent];
+      fullPath = node->fName + (fullPath.empty() ? "" : "/") + fullPath;
+   }
+   if (opt == ENodeFullPathOpt::kIncludeFilename && nodeIdx > 0)
+      fullPath = tree.fNodes[0].fName + ":" + fullPath;
+   return fullPath;
+}