Store the collection information in a struct instead of a tuple

include/podio/RNTupleReader.h

@@ -5,6 +5,7 @@
 #include "podio/SchemaEvolution.h"
 #include "podio/podioVersion.h"
 #include "podio/utilities/DatamodelRegistryIOHelpers.h"
+#include "podio/utilities/RootHelpers.h"
 
 #include <string>
 #include <string_view>
@@ -161,15 +162,8 @@ class RNTupleReader {
   std::unordered_map<std::string, std::vector<unsigned>> m_readerEntries{};
   std::unordered_map<std::string, unsigned> m_totalEntries{};
 
-  struct CollectionInfo {
-    std::vector<unsigned int> id{};
-    std::vector<std::string> name{};
-    std::vector<std::string> type{};
-    std::vector<short> isSubsetCollection{};
-    std::vector<SchemaVersionT> schemaVersion{};
-  };
-
-  std::unordered_map<std::string, CollectionInfo> m_collectionInfo{};
+  /// Map each category to the collections that have been written and are available
+  std::unordered_map<std::string, std::vector<podio::root_utils::CollectionWriteInfo>> m_collectionInfo{};
 
   std::vector<std::string> m_availableCategories{};
 

include/podio/RNTupleWriter.h

@@ -105,12 +105,9 @@ class RNTupleWriter {
   struct CategoryInfo {
     std::unique_ptr<ROOT::Experimental::RNTupleWriter> writer{nullptr}; ///< The RNTupleWriter for this category
 
-    // The following are assumed to run in parallel!
-    std::vector<uint32_t> ids{};                  ///< The ids of all collections
-    std::vector<std::string> names{};             ///< The names of all collections
-    std::vector<std::string> types{};             ///< The types of all collections
-    std::vector<short> subsetCollections{};       ///< The flags identifying the subcollections
-    std::vector<SchemaVersionT> schemaVersions{}; ///< The schema versions of all collections
+    /// Collection info for this category
+    std::vector<root_utils::CollectionWriteInfo> collInfo{};
+    std::vector<std::string> names{}; ///< The names of all collections to write
 
     // Storage for the keys & values of all the parameters of this category
     // (resp. at least the current entry)

include/podio/ROOTLegacyReader.h

@@ -116,7 +116,7 @@ class ROOTLegacyReader {
 private:
   std::pair<TTree*, unsigned> getLocalTreeAndEntry(const std::string& treename);
 
-  void createCollectionBranches(const std::vector<root_utils::CollectionWriteInfoT>& collInfo);
+  void createCollectionBranches(const std::vector<root_utils::CollectionWriteInfo>& collInfo);
 
   podio::GenericParameters readEventMetaData();
 

include/podio/ROOTWriter.h

@@ -103,11 +103,10 @@ class ROOTWriter {
   /// Helper struct to group together all necessary state to write / process a
   /// given category. Created during the first writing of a category
   struct CategoryInfo {
-    TTree* tree{nullptr};                                     ///< The TTree to which this category is written
-    std::vector<root_utils::CollectionBranches> branches{};   ///< The branches for this category
-    std::vector<root_utils::CollectionWriteInfoT> collInfo{}; ///< Collection info for this category
-    podio::CollectionIDTable idTable{};                       ///< The collection id table for this category
-    std::vector<std::string> collsToWrite{};                  ///< The collections to write for this category
+    TTree* tree{nullptr};                                    ///< The TTree to which this category is written
+    std::vector<root_utils::CollectionBranches> branches{};  ///< The branches for this category
+    std::vector<root_utils::CollectionWriteInfo> collInfo{}; ///< Collection info for this category
+    std::vector<std::string> collsToWrite{};                 ///< The collections to write for this category
 
     // Storage for the keys & values of all the parameters of this category
     // (resp. at least the current entry)

include/podio/utilities/RootHelpers.h

@@ -18,7 +18,16 @@ namespace root_utils {
   // A collection of additional information that describes the collection: the
   // collectionID, the collection (data) type, whether it is a subset
   // collection, and its schema version
+  struct CollectionWriteInfo {
+    uint32_t collectionID{static_cast<uint32_t>(-1)}; ///< collection id
+    std::string dataType{};                           ///< The fully qualified data type
+    bool isSubset{false};                             ///< Whether this collection is a subset collection or not
+    unsigned int schemaVersion{0};                    ///< The schema version of the collection type
+    std::string name{};                               ///< The name of the collection
+  };
+  // The format used until version 1.2
   using CollectionWriteInfoT = std::tuple<uint32_t, std::string, bool, unsigned int>;
+
   // for backwards compatibility
   using CollectionInfoWithoutSchemaT = std::tuple<int, std::string, bool>;
 

src/RNTupleReader.cc

@@ -4,11 +4,13 @@
 #include "podio/CollectionIDTable.h"
 #include "podio/DatamodelRegistry.h"
 #include "podio/GenericParameters.h"
+#include "podio/utilities/RootHelpers.h"
 #include "rootUtils.h"
 
 #include <ROOT/RError.hxx>
 
 #include <algorithm>
+#include <cstdint>
 #include <memory>
 
 // Adjust for the move of this out of ROOT v7 in
@@ -48,27 +50,11 @@ bool RNTupleReader::initCategory(const std::string& category) {
   // Assume that the metadata is the same in all files
   auto filename = m_filenames[0];
 
-  auto& collInfo = m_collectionInfo[category];
+  auto collInfo = m_metadata_readers[filename]->GetView<std::vector<root_utils::CollectionWriteInfo>>(
+      {root_utils::collInfoName(category)});
 
-  auto id = m_metadata_readers[filename]->GetView<std::vector<unsigned int>>(root_utils::idTableName(category));
-  collInfo.id = id(0);
-
-  auto collectionName =
-      m_metadata_readers[filename]->GetView<std::vector<std::string>>(root_utils::collectionName(category));
-  collInfo.name = collectionName(0);
-
-  auto collectionType =
-      m_metadata_readers[filename]->GetView<std::vector<std::string>>(root_utils::collInfoName(category));
-  collInfo.type = collectionType(0);
-
-  auto subsetCollection =
-      m_metadata_readers[filename]->GetView<std::vector<short>>(root_utils::subsetCollection(category));
-  collInfo.isSubsetCollection = subsetCollection(0);
-
-  auto schemaVersion = m_metadata_readers[filename]->GetView<std::vector<SchemaVersionT>>("schemaVersion_" + category);
-  collInfo.schemaVersion = schemaVersion(0);
-
-  m_idTables[category] = std::make_shared<CollectionIDTable>(collInfo.id, collInfo.name);
+  m_collectionInfo[category] = collInfo(0);
+  m_idTables[category] = root_utils::makeCollIdTable(collInfo(0));
 
   return true;
 }
@@ -162,7 +148,7 @@ std::unique_ptr<ROOTFrameData> RNTupleReader::readEntry(const std::string& categ
   // Make sure to not silently ignore non-existant but requested collections
   if (!collsToRead.empty()) {
     for (const auto& name : collsToRead) {
-      if (std::ranges::find(collInfo.name, name) == collInfo.name.end()) {
+      if (std::ranges::find(collInfo, name, &root_utils::CollectionWriteInfo::name) == collInfo.end()) {
         throw std::invalid_argument(name + " is not available from Frame");
       }
     }
@@ -184,47 +170,46 @@ std::unique_ptr<ROOTFrameData> RNTupleReader::readEntry(const std::string& categ
   // we set all the fields there in any case.
   auto dentry = m_readers[category][readerIndex]->GetModel().CreateEntry();
 
-  for (size_t i = 0; i < collInfo.id.size(); ++i) {
-    if (!collsToRead.empty() && std::ranges::find(collsToRead, collInfo.name[i]) == collsToRead.end()) {
+  for (const auto& coll : collInfo) {
+    if (!collsToRead.empty() && std::ranges::find(collsToRead, coll.name) == collsToRead.end()) {
       continue;
     }
-    const auto& collType = collInfo.type[i];
+    const auto& collType = coll.dataType;
     const auto& bufferFactory = podio::CollectionBufferFactory::instance();
-    auto maybeBuffers =
-        bufferFactory.createBuffers(collType, collInfo.schemaVersion[i], collInfo.isSubsetCollection[i]);
+    auto maybeBuffers = bufferFactory.createBuffers(collType, coll.schemaVersion, coll.isSubset);
     auto collBuffers = maybeBuffers.value_or(podio::CollectionReadBuffers{});
 
     if (!maybeBuffers) {
-      std::cout << "WARNING: Buffers couldn't be created for collection " << collInfo.name[i] << " of type "
-                << collInfo.type[i] << " and schema version " << collInfo.schemaVersion[i] << std::endl;
+      std::cout << "WARNING: Buffers couldn't be created for collection " << coll.name << " of type " << coll.dataType
+                << " and schema version " << coll.schemaVersion << std::endl;
       return nullptr;
     }
 
-    if (collInfo.isSubsetCollection[i]) {
-      auto brName = root_utils::subsetBranch(collInfo.name[i]);
+    if (coll.isSubset) {
+      auto brName = root_utils::subsetBranch(coll.name);
       auto vec = new std::vector<podio::ObjectID>;
       dentry->BindRawPtr(brName, vec);
       collBuffers.references->at(0) = std::unique_ptr<std::vector<podio::ObjectID>>(vec);
     } else {
-      dentry->BindRawPtr(collInfo.name[i], collBuffers.data);
+      dentry->BindRawPtr(coll.name, collBuffers.data);
 
       const auto relVecNames = podio::DatamodelRegistry::instance().getRelationNames(collType);
       for (size_t j = 0; j < relVecNames.relations.size(); ++j) {
         const auto relName = relVecNames.relations[j];
         auto vec = new std::vector<podio::ObjectID>;
-        const auto brName = root_utils::refBranch(collInfo.name[i], relName);
+        const auto brName = root_utils::refBranch(coll.name, relName);
         dentry->BindRawPtr(brName, vec);
         collBuffers.references->at(j) = std::unique_ptr<std::vector<podio::ObjectID>>(vec);
       }
 
       for (size_t j = 0; j < relVecNames.vectorMembers.size(); ++j) {
         const auto vecName = relVecNames.vectorMembers[j];
-        const auto brName = root_utils::vecBranch(collInfo.name[i], vecName);
+        const auto brName = root_utils::vecBranch(coll.name, vecName);
         dentry->BindRawPtr(brName, collBuffers.vectorMembers->at(j).second);
       }
     }
 
-    buffers.emplace(collInfo.name[i], std::move(collBuffers));
+    buffers.emplace(coll.name, std::move(collBuffers));
   }
 
   m_readers[category][readerIndex]->LoadEntry(localEntry, *dentry);

src/RNTupleWriter.cc

@@ -2,6 +2,7 @@
 #include "podio/DatamodelRegistry.h"
 #include "podio/SchemaEvolution.h"
 #include "podio/podioVersion.h"
+#include "podio/utilities/RootHelpers.h"
 #include "rootUtils.h"
 
 #include "TFile.h"
@@ -81,14 +82,13 @@ void RNTupleWriter::writeFrame(const podio::Frame& frame, const std::string& cat
     auto model = createModels(collections);
     catInfo.writer = ROOT::Experimental::RNTupleWriter::Append(std::move(model), category, *m_file.get(), {});
 
+    catInfo.collInfo.reserve(collections.size());
     for (const auto& [name, coll] : collections) {
-      catInfo.ids.emplace_back(coll->getID());
-      catInfo.types.emplace_back(coll->getTypeName());
-      catInfo.subsetCollections.emplace_back(coll->isSubsetCollection());
-      catInfo.schemaVersions.emplace_back(coll->getSchemaVersion());
+      catInfo.collInfo.emplace_back(coll->getID(), std::string(coll->getTypeName()), coll->isSubsetCollection(),
+                                    coll->getSchemaVersion(), name);
     }
   } else {
-    if (!root_utils::checkConsistentColls(catInfo.names, collsToWrite)) {
+    if (!root_utils::checkConsistentColls(catInfo.collInfo, collsToWrite)) {
       throw std::runtime_error("Trying to write category '" + category + "' with inconsistent collection content. " +
                                root_utils::getInconsistentCollsMsg(catInfo.names, collsToWrite));
     }
@@ -251,16 +251,9 @@ void RNTupleWriter::finish() {
   }
 
   for (auto& [category, collInfo] : m_categories) {
-    auto idField = metadata->MakeField<std::vector<unsigned int>>({root_utils::idTableName(category)});
-    *idField = collInfo.ids;
-    auto collectionNameField = metadata->MakeField<std::vector<std::string>>({root_utils::collectionName(category)});
-    *collectionNameField = collInfo.names;
-    auto collectionTypeField = metadata->MakeField<std::vector<std::string>>({root_utils::collInfoName(category)});
-    *collectionTypeField = collInfo.types;
-    auto subsetCollectionField = metadata->MakeField<std::vector<short>>({root_utils::subsetCollection(category)});
-    *subsetCollectionField = collInfo.subsetCollections;
-    auto schemaVersionField = metadata->MakeField<std::vector<SchemaVersionT>>({"schemaVersion_" + category});
-    *schemaVersionField = collInfo.schemaVersions;
+    auto collInfoField =
+        metadata->MakeField<std::vector<root_utils::CollectionWriteInfo>>({root_utils::collInfoName(category)});
+    *collInfoField = collInfo.collInfo;
   }
 
   metadata->Freeze();

src/ROOTLegacyReader.cc

@@ -152,7 +152,12 @@ void ROOTLegacyReader::openFiles(const std::vector<std::string>& filenames) {
       collInfoBranch->SetAddress(&collectionInfo);
       collInfoBranch->GetEntry(0);
     }
-    createCollectionBranches(*collectionInfo);
+    std::vector<root_utils::CollectionWriteInfo> collInfo;
+    collInfo.reserve(collectionInfo->size());
+    for (auto& [id, typeName, isSubsetColl, schemaVersion] : *collectionInfo) {
+      collInfo.emplace_back(id, std::move(typeName), isSubsetColl, schemaVersion);
+    }
+    createCollectionBranches(collInfo);
     delete collectionInfo;
   } else {
     std::cout << "PODIO: Reconstructing CollectionTypeInfo branch from other sources in file: \'"
@@ -170,10 +175,10 @@ unsigned ROOTLegacyReader::getEntries(const std::string& name) const {
   return m_chain->GetEntries();
 }
 
-void ROOTLegacyReader::createCollectionBranches(const std::vector<root_utils::CollectionWriteInfoT>& collInfo) {
+void ROOTLegacyReader::createCollectionBranches(const std::vector<root_utils::CollectionWriteInfo>& collInfo) {
   size_t collectionIndex{0};
 
-  for (const auto& [collID, collType, isSubsetColl, collSchemaVersion] : collInfo) {
+  for (const auto& [collID, collType, isSubsetColl, collSchemaVersion, _] : collInfo) {
     // We only write collections that are in the collectionIDTable, so no need
     // to check here
     const auto name = m_table->name(collID).value();