To use content elements for Ranking, Sorting, and Filtering and Refinements and Binning, you must tell
the indexer to treat them specially. This is done by fast-indexing those content
elements. Fast-indexing is the general term for indexing methods that enable content to be
accessed quickly, and which is flexible enough to be used to filter and sort query results.
Watson Explorer Engine supports two different fast-indexing methods:
Note: When fast-indexing is specified, all instances of that content element are
stored in RAM. Therefore, to optimize search application performance, careful consideration
should be given to which content elements are fast-indexed.
- Fast-Indexing allows you to define a list of name|type entries in the
fast-index option on a search collection's Indexing tab, and
stores this fast-index data in memory for quick access. This list identifies each content
element that you want to fast-index and its datatype. Alternatively, you can add the
fast-index="type" attribute to the content elements in
your data that you want to be able to use in fast-index comparisons.
- Indexed Fast-Indexing is the same as fast-indexing in terms of being able to do
fast-index comparisons, but permanently stores additional data in the index for a search
collection. This can improve run-time/compare-time performance, but increases indexing time
and index storage requirements. To use indexed fast-indexing, you must specify
indexed-fast-index="type" attribute on content elements
in your data that you want to be able to use in fast-index comparisons. (See Activating Indexed Fast-Indexing for more detailed information about defining indexed fast-indexed
contents.)
You should only choose one fast-indexing method for any name of a content in the collection.
If both the fast-index and indexed fast-index attributes are set for the same
content name, the benefit of indexed fast-indexing will be lost.
Deciding when to use these fast-indexing methods largely depends upon the characteristics of
your system, please contact IBM product support for further information.
Note: When you use any form of fast-indexing, the content elements that you are
fast-indexing must be defined as fields in the syntax for your project to be able to use them
in queries. If the element(s) that you are fast-indexing are not a part of the default syntax,
you must add them to the syntax for your project.
Valid types that can be used with fast-index and indexed-fast-index are as
follows:
- date - a special data type that automatically invokes the IBM viv:parse-date
function on a date string to produce an integer that can be fast-indexed. Allowable dates
are 4 bytes and between the dates 1901-12-14 and 2038-01-18.
- float - single-precision floating point numbers with decimal points. 4 bytes.
- int - integers. variable # of bytes.
- double or number - double-precision floating point numbers with decimal
points. 8 bytes.
- set - a set of values, such as a character string. This is the default type.