The on-storage metadata of an RNTuple.
Represents the on-disk (on storage) information about an RNTuple. The metadata consists of a header, a footer, and potentially multiple page lists. The header carries the RNTuple schema, i.e. the fields and the associated columns and their relationships. The footer carries information about one or several cluster groups and links to their page lists. For every cluster group, a page list envelope stores cluster summaries and page locations. For every cluster, it stores for every column the range of element indexes as well as a list of pages and page locations.
The descriptor provides machine-independent (de-)serialization of headers and footers, and it provides lookup routines for RNTuple objects (pages, clusters, ...). It is supposed to be usable by all RPageStorage implementations.
The serialization does not use standard ROOT streamers in order to not let it depend on libCore. The serialization uses the concept of envelopes and frames: header, footer, and page list envelopes have a preamble with a type ID and length. Substructures are serialized in frames and have a size and number of items (for list frames). This allows for forward and backward compatibility when the metadata evolves.
Definition at line 633 of file RNTupleDescriptor.hxx.
Classes | |
class | RClusterDescriptorIterable |
Used to loop over all the clusters of an RNTuple (in unspecified order) More... | |
class | RClusterGroupDescriptorIterable |
Used to loop over all the cluster groups of an RNTuple (in unspecified order) More... | |
class | RColumnDescriptorIterable |
Used to loop over a field's associated columns. More... | |
struct | RCreateModelOptions |
Modifiers passed to CreateModel() More... | |
class | RExtraTypeInfoDescriptorIterable |
Used to loop over all the extra type info record of an RNTuple (in unspecified order) More... | |
class | RFieldDescriptorIterable |
Used to loop over a field's child fields. More... | |
class | RHeaderExtension |
Summarizes information about fields and the corresponding columns that were added after the header has been serialized. More... | |
Static Public Attributes | |
static constexpr unsigned int | kFeatureFlagTest = 137 |
Private Member Functions | |
RNTupleDescriptor | CloneSchema () const |
Creates a descriptor containing only the schema information about this RNTuple, i.e. | |
ROOT::DescriptorId_t | FindClusterId (ROOT::NTupleSize_t entryIdx) const |
Private Attributes | |
std::unordered_map< ROOT::DescriptorId_t, RClusterDescriptor > | fClusterDescriptors |
Potentially a subset of all the available clusters. | |
std::unordered_map< ROOT::DescriptorId_t, RClusterGroupDescriptor > | fClusterGroupDescriptors |
std::unordered_map< ROOT::DescriptorId_t, RColumnDescriptor > | fColumnDescriptors |
std::string | fDescription |
Free text from the user. | |
std::vector< RExtraTypeInfoDescriptor > | fExtraTypeInfoDescriptors |
std::set< unsigned int > | fFeatureFlags |
std::unordered_map< ROOT::DescriptorId_t, RFieldDescriptor > | fFieldDescriptors |
ROOT::DescriptorId_t | fFieldZeroId = ROOT::kInvalidDescriptorId |
Set by the descriptor builder. | |
std::uint64_t | fGeneration = 0 |
The generation of the descriptor. | |
std::unique_ptr< RHeaderExtension > | fHeaderExtension |
std::string | fName |
The RNTuple name needs to be unique in a given storage location (file) | |
std::uint64_t | fNClusters = 0 |
Updated by the descriptor builder when the cluster groups are added. | |
std::uint64_t | fNEntries = 0 |
Updated by the descriptor builder when the cluster groups are added. | |
std::uint64_t | fNPhysicalColumns = 0 |
Updated by the descriptor builder when columns are added. | |
std::uint64_t | fOnDiskFooterSize = 0 |
Like fOnDiskHeaderSize, contains both cluster summaries and page locations. | |
std::uint64_t | fOnDiskHeaderSize = 0 |
Set by the descriptor builder when deserialized. | |
std::uint64_t | fOnDiskHeaderXxHash3 = 0 |
Set by the descriptor builder when deserialized. | |
std::vector< ROOT::DescriptorId_t > | fSortedClusterGroupIds |
References cluster groups sorted by entry range and thus allows for binary search. | |
Friends | |
RNTupleDescriptor | Internal::CloneDescriptorSchema (const RNTupleDescriptor &desc) |
class | Internal::RNTupleDescriptorBuilder |
#include <ROOT/RNTupleDescriptor.hxx>
|
default |
|
delete |
|
default |
ROOT::RResult< void > ROOT::RNTupleDescriptor::AddClusterGroupDetails | ( | ROOT::DescriptorId_t | clusterGroupId, |
std::vector< RClusterDescriptor > & | clusterDescs ) |
Methods to load and drop cluster group details (cluster IDs and page locations)
Definition at line 598 of file RNTupleDescriptor.cxx.
ROOT::RNTupleDescriptor ROOT::RNTupleDescriptor::Clone | ( | ) | const |
Definition at line 691 of file RNTupleDescriptor.cxx.
|
private |
Creates a descriptor containing only the schema information about this RNTuple, i.e.
all the information needed to create a new RNTuple with the same schema as this one but not necessarily the same clustering. This is used when merging two RNTuples.
Definition at line 666 of file RNTupleDescriptor.cxx.
std::unique_ptr< ROOT::RNTupleModel > ROOT::RNTupleDescriptor::CreateModel | ( | const RCreateModelOptions & | options = RCreateModelOptions() | ) | const |
Re-create the C++ model from the stored metadata.
Definition at line 640 of file RNTupleDescriptor.cxx.
ROOT::RResult< void > ROOT::RNTupleDescriptor::DropClusterGroupDetails | ( | ROOT::DescriptorId_t | clusterGroupId | ) |
Definition at line 626 of file RNTupleDescriptor.cxx.
ROOT::DescriptorId_t ROOT::RNTupleDescriptor::FindClusterId | ( | ROOT::DescriptorId_t | physicalColumnId, |
ROOT::NTupleSize_t | index ) const |
Definition at line 403 of file RNTupleDescriptor.cxx.
|
private |
Definition at line 465 of file RNTupleDescriptor.cxx.
ROOT::DescriptorId_t ROOT::RNTupleDescriptor::FindFieldId | ( | std::string_view | fieldName | ) | const |
Searches for a top-level field.
Definition at line 372 of file RNTupleDescriptor.cxx.
ROOT::DescriptorId_t ROOT::RNTupleDescriptor::FindFieldId | ( | std::string_view | fieldName, |
ROOT::DescriptorId_t | parentId ) const |
Definition at line 341 of file RNTupleDescriptor.cxx.
ROOT::DescriptorId_t ROOT::RNTupleDescriptor::FindLogicalColumnId | ( | ROOT::DescriptorId_t | fieldId, |
std::uint32_t | columnIndex, | ||
std::uint16_t | representationIndex ) const |
Definition at line 377 of file RNTupleDescriptor.cxx.
ROOT::DescriptorId_t ROOT::RNTupleDescriptor::FindNextClusterId | ( | ROOT::DescriptorId_t | clusterId | ) | const |
Definition at line 517 of file RNTupleDescriptor.cxx.
ROOT::DescriptorId_t ROOT::RNTupleDescriptor::FindPhysicalColumnId | ( | ROOT::DescriptorId_t | fieldId, |
std::uint32_t | columnIndex, | ||
std::uint16_t | representationIndex ) const |
Definition at line 392 of file RNTupleDescriptor.cxx.
ROOT::DescriptorId_t ROOT::RNTupleDescriptor::FindPrevClusterId | ( | ROOT::DescriptorId_t | clusterId | ) | const |
Definition at line 528 of file RNTupleDescriptor.cxx.
|
inline |
Definition at line 759 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 755 of file RNTupleDescriptor.hxx.
ROOT::RNTupleDescriptor::RClusterGroupDescriptorIterable ROOT::RNTupleDescriptor::GetClusterGroupIterable | ( | ) | const |
Definition at line 1376 of file RNTupleDescriptor.cxx.
ROOT::RNTupleDescriptor::RClusterDescriptorIterable ROOT::RNTupleDescriptor::GetClusterIterable | ( | ) | const |
Definition at line 1381 of file RNTupleDescriptor.cxx.
|
inline |
Definition at line 751 of file RNTupleDescriptor.hxx.
ROOT::RNTupleDescriptor::RColumnDescriptorIterable ROOT::RNTupleDescriptor::GetColumnIterable | ( | ) | const |
Definition at line 1359 of file RNTupleDescriptor.cxx.
ROOT::RNTupleDescriptor::RColumnDescriptorIterable ROOT::RNTupleDescriptor::GetColumnIterable | ( | const RFieldDescriptor & | fieldDesc | ) | const |
Definition at line 1365 of file RNTupleDescriptor.cxx.
ROOT::RNTupleDescriptor::RColumnDescriptorIterable ROOT::RNTupleDescriptor::GetColumnIterable | ( | ROOT::DescriptorId_t | fieldId | ) | const |
Definition at line 1371 of file RNTupleDescriptor.cxx.
|
inline |
Definition at line 788 of file RNTupleDescriptor.hxx.
ROOT::RNTupleDescriptor::RExtraTypeInfoDescriptorIterable ROOT::RNTupleDescriptor::GetExtraTypeInfoIterable | ( | ) | const |
Definition at line 1386 of file RNTupleDescriptor.cxx.
std::vector< std::uint64_t > ROOT::RNTupleDescriptor::GetFeatureFlags | ( | ) | const |
Definition at line 578 of file RNTupleDescriptor.cxx.
|
inline |
Definition at line 747 of file RNTupleDescriptor.hxx.
ROOT::RNTupleDescriptor::RFieldDescriptorIterable ROOT::RNTupleDescriptor::GetFieldIterable | ( | const RFieldDescriptor & | fieldDesc | ) | const |
Definition at line 1323 of file RNTupleDescriptor.cxx.
ROOT::RNTupleDescriptor::RFieldDescriptorIterable ROOT::RNTupleDescriptor::GetFieldIterable | ( | const RFieldDescriptor & | fieldDesc, |
const std::function< bool(ROOT::DescriptorId_t, ROOT::DescriptorId_t)> & | comparator ) const |
Definition at line 1328 of file RNTupleDescriptor.cxx.
ROOT::RNTupleDescriptor::RFieldDescriptorIterable ROOT::RNTupleDescriptor::GetFieldIterable | ( | ROOT::DescriptorId_t | fieldId | ) | const |
Definition at line 1336 of file RNTupleDescriptor.cxx.
ROOT::RNTupleDescriptor::RFieldDescriptorIterable ROOT::RNTupleDescriptor::GetFieldIterable | ( | ROOT::DescriptorId_t | fieldId, |
const std::function< bool(ROOT::DescriptorId_t, ROOT::DescriptorId_t)> & | comparator ) const |
Definition at line 1341 of file RNTupleDescriptor.cxx.
|
inline |
Definition at line 804 of file RNTupleDescriptor.hxx.
|
inline |
Returns the logical parent of all top-level RNTuple data fields.
Definition at line 803 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 831 of file RNTupleDescriptor.hxx.
|
inline |
Return header extension information; if the descriptor does not have a header extension, return nullptr
Definition at line 824 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 795 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 787 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 793 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 794 of file RNTupleDescriptor.hxx.
ROOT::NTupleSize_t ROOT::RNTupleDescriptor::GetNElements | ( | ROOT::DescriptorId_t | physicalColumnId | ) | const |
Definition at line 328 of file RNTupleDescriptor.cxx.
|
inline |
We know the number of entries from adding the cluster summaries.
Definition at line 799 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 796 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 790 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 791 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 792 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 745 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 744 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 743 of file RNTupleDescriptor.hxx.
std::string ROOT::RNTupleDescriptor::GetQualifiedFieldName | ( | ROOT::DescriptorId_t | fieldId | ) | const |
Walks up the parents of the field ID and returns a field name of the form a.b.c.d In case of invalid field ID, an empty string is returned.
Definition at line 360 of file RNTupleDescriptor.cxx.
ROOT::RNTupleDescriptor::RFieldDescriptorIterable ROOT::RNTupleDescriptor::GetTopLevelFields | ( | ) | const |
Definition at line 1348 of file RNTupleDescriptor.cxx.
ROOT::RNTupleDescriptor::RFieldDescriptorIterable ROOT::RNTupleDescriptor::GetTopLevelFields | ( | const std::function< bool(ROOT::DescriptorId_t, ROOT::DescriptorId_t)> & | comparator | ) | const |
Definition at line 1353 of file RNTupleDescriptor.cxx.
Definition at line 820 of file RNTupleDescriptor.hxx.
|
inline |
Definition at line 832 of file RNTupleDescriptor.hxx.
|
delete |
|
default |
bool ROOT::RNTupleDescriptor::operator== | ( | const RNTupleDescriptor & | other | ) | const |
Definition at line 313 of file RNTupleDescriptor.cxx.
void ROOT::RNTupleDescriptor::PrintInfo | ( | std::ostream & | output | ) | const |
Definition at line 79 of file RNTupleDescriptorFmt.cxx.
|
friend |
|
friend |
Definition at line 634 of file RNTupleDescriptor.hxx.
|
private |
Potentially a subset of all the available clusters.
Definition at line 682 of file RNTupleDescriptor.hxx.
|
private |
Definition at line 676 of file RNTupleDescriptor.hxx.
|
private |
Definition at line 652 of file RNTupleDescriptor.hxx.
|
private |
Free text from the user.
Definition at line 644 of file RNTupleDescriptor.hxx.
|
private |
Definition at line 654 of file RNTupleDescriptor.hxx.
|
private |
Definition at line 650 of file RNTupleDescriptor.hxx.
|
private |
Definition at line 651 of file RNTupleDescriptor.hxx.
|
private |
Set by the descriptor builder.
Definition at line 646 of file RNTupleDescriptor.hxx.
|
private |
The generation of the descriptor.
Once constructed by an RNTupleDescriptorBuilder, the descriptor is mostly immutable except for the set of active page locations. During the lifetime of the descriptor, page location information for clusters can be added or removed. When this happens, the generation should be increased, so that users of the descriptor know that the information changed. The generation is increased, e.g., by the page source's exclusive lock guard around the descriptor. It is used, e.g., by the descriptor cache in RNTupleReader.
Definition at line 674 of file RNTupleDescriptor.hxx.
|
private |
Definition at line 655 of file RNTupleDescriptor.hxx.
|
private |
The RNTuple name needs to be unique in a given storage location (file)
Definition at line 642 of file RNTupleDescriptor.hxx.
|
private |
Updated by the descriptor builder when the cluster groups are added.
Definition at line 665 of file RNTupleDescriptor.hxx.
|
private |
Updated by the descriptor builder when the cluster groups are added.
Definition at line 664 of file RNTupleDescriptor.hxx.
|
private |
Updated by the descriptor builder when columns are added.
Definition at line 648 of file RNTupleDescriptor.hxx.
|
private |
Like fOnDiskHeaderSize, contains both cluster summaries and page locations.
Definition at line 662 of file RNTupleDescriptor.hxx.
|
private |
Set by the descriptor builder when deserialized.
Definition at line 660 of file RNTupleDescriptor.hxx.
|
private |
Set by the descriptor builder when deserialized.
Definition at line 661 of file RNTupleDescriptor.hxx.
|
private |
References cluster groups sorted by entry range and thus allows for binary search.
Note that this list is empty during the descriptor building process and will only be created when the final descriptor is extracted from the builder.
Definition at line 680 of file RNTupleDescriptor.hxx.
Definition at line 693 of file RNTupleDescriptor.hxx.