摘要:
A technique for updating collection-valued and other complex structured columns (see figure 6) in a nested table using a nested extension of an UPDATE statement that uses syntax and semantics to modify collection-valued columns in a way that is analogous to the syntax and semantics of the UPDATE statement that is used to modify scalar-valued columns of the table (called the outer UPDATE). Using the same syntactic and semantic constructs as the table at the outer level allows an existing implementation that processes modifications to relational tables to reuse its implementation techniques for processing outer updates to modify' collection-valued columns as well. The UPDATE extensions enable the specification of updates to nested collections embedded at arbitrary levels of depth in the object model. The new syntax is embedded inside the outer UPDATE statement in a way that parallels the structure of the data itself and thus maps more directly to the user’s conceptual model of the data. The method for implementing the UPDATE extensions uses a change descriptor, which is a data structure that aggregates substantially all changes, both scalar and collection-value that can be applied to the changed collection-valued column. The change descriptor includes hierarchical information for the cell, thereby enabling efficient application of multiple updates at various granularity levels.
摘要:
Various embodiments of the present invention are direct to the utilization of Blob Handles (BHs) which are an internal representation of a large value. BHs are immutable and stateless references to a large data object. The structure of a BH contains enough information to return an ILockBytes interface in order to provide access to the corresponding large data block, and a BH can also return information regarding its own lifetime description. A BH can be completely described using (a) a pointer to the beginning of BH data and (b) the byte-length of the BH.
摘要:
Efficient hierarchical searching is based on object type. By pre-computing additional information and storing it in a fast-lookup structure, it is possible to quickly identify objects that satisfy an object retrieval request. Furthermore, it is also possible to use this technique to avoid object hydration for operations in the store. Moreover, it is possible to leverage database statistical structures such as histograms to estimate the number of qualifying objects without having to examine each object.