Creating and manipulating groups of objects inside an HDF5 file
- H5G_CLOSE — Closes the specified group
- H5G_CREATE — Creates a new empty group and links it to a location in the file
- H5G_CREATE_ANON — Creates a new empty group without linking it into the file structure
- H5G_FLUSH — Flushes all buffers associated with a group to disk.
- H5G_GET_COMMENT — Retrieves comment for specified object.
- H5G_GET_CREATE_PLIST — Gets a group creation property list identifier.
- H5G_GET_INFO — Retrieves information about a group
- H5G_GET_INFO_BY_IDX — Retrieves information about a group, according to the group’s position within an index
- H5G_GET_INFO_BY_NAME — Retrieves information about a group
- H5G_GET_LINKVAL — Returns the name of the object that the symbolic link points to.
- H5G_GET_NUM_OBJS — Returns number of objects in the group specified by its identifier
- H5G_GET_OBJINFO — Returns information about an object.
- H5G_GET_OBJNAME_BY_IDX — Returns a name of an object specified by an index.
- H5G_GET_OBJTYPE_BY_IDX — Returns the type of an object specified by an index.
- H5G_ITERATE — Iterates an operation over the entries of a group.
- H5G_LINK — Creates a link of the specified type from new_name to current_name.
- H5G_LINK2 — Creates a link of the specified type from current_name to new_name.
- H5G_MOVE — Renames an object within an HDF5 file.
- H5G_MOVE2 — Renames an object within an HDF5 file.
- H5G_OPEN — Opens an existing group in a file
- H5G_REFRESH — Refreshes all buffers associated with a group.
- H5G_SET_COMMENT — Sets comment for specified object.
- H5G_UNLINK — Removes the link to an object from a group.
A group associates names with objects and provides a mechanism for mapping a name to an object. Since all objects appear in at least one group (with the possible exception of the root object) and since objects can have names in more than one group, the set of all objects in an HDF5 file is a directed graph. The internal nodes (nodes with out-degree greater than zero) must be groups while the leaf nodes (nodes with out-degree zero) are either empty groups or objects of some other type. Exactly one object in every non-empty file is the root object. The root object always has a positive in-degree because it is pointed to by the file super block.
An object name consists of one or more components separated from one another by slashes. An absolute name begins with a slash and the object is located by looking for the first component in the root object, then looking for the second component in the first object, etc., until the entire name is traversed. A relative name does not begin with a slash and the traversal begins at the location specified by the create or access function.
The original HDF5 group implementation provided a single indexed structure for link storage. A new group implementation, in HDF5 Release 1.8.0, enables more efficient compact storage for very small groups, improved link indexing for large groups, and other advanced features.
- The original indexed format remains the default. Links are stored in a B-tree in the group’s local heap.
- Groups created in the new compact-or-indexed format, the implementation introduced with Release 1.8.0, can be tuned for performance, switching between the compact and indexed formats at thresholds set in the user application.
- The compact format will conserve file space and processing overhead when working with small groups and is particularly valuable when a group contains no links. Links are stored as a list of messages in the group’s header.
- The indexed format will yield improved performance when working with large groups, e.g., groups containing thousands to millions of members. Links are stored in a fractal heap and indexed with an improved B-tree.
- The new implementation also enables the use of link names consisting of non-ASCII character sets (see
H5P_SET_CHAR_ENCODING) and is required for all link types other than hard or soft links, e.g., external and user-defined links (see the H5L APIs).
The original group structure and the newer structures are not directly interoperable. By default, a group will be created in the original indexed format. An existing group can be changed to a compact-or-indexed format if the need arises; there is no capability to change back. As stated above, once in the compact-or-indexed format, a group can switch between compact and indexed as needed.
Groups will be initially created in the compact-or-indexed format only when one or more of the following conditions is met:
- The low version bound value of the library version bounds property has been set to Release 1.8.0 or later in the file access property list (see
H5P_SET_LIBVER_BOUNDS). Currently, that would require an
H5Pset_libver_boundscall with the low parameter set to
When this property is set for an HDF5 file, all objects in the file will be created using the latest available format; no effort will be made to create a file that can be read by older libraries.
- The creation order tracking property,
H5P_CRT_ORDER_TRACKED, has been set in the group creation property list (see
An existing group, currently in the original indexed format, will be converted to the compact-or-indexed format upon the occurrence of any of the following events:
- An external or user-defined link is inserted into the group.
- A link named with a string composed of non-ASCII characters is inserted into the group.
The compact-or-indexed format offers performance improvements that will be most notable at the extremes, i.e., in groups with zero members and in groups with tens of thousands of members. But measurable differences may sometimes appear at a threshold as low as eight group members. Since these performance thresholds and criteria differ from application to application, tunable settings are provided to govern the switch between the compact and indexed formats (see
H5P_SET_LINK_PHASE_CHANGE). Optimal thresholds will depend on the application and the operating environment.
Future versions of HDF5 will retain the ability to create, read, write, and manipulate all groups stored in either the original indexed format or the compact-or-indexed format.