com.saxonica.extra
Class MultiIndex

java.lang.Object
  extended by com.saxonica.extra.MultiIndex

public class MultiIndex
extends Object

A MultiIndex provides fast access to the items in a sequence. The sequence is referred to as the "population", and may consist of nodes and/or atomic values. The items are identified by means of a key value computed using an expression called the "use" expression, which is supplied in the form of a SequenceIterable. This may be multivalued: each item in the result of the use expression is regarded as a separate key, and a match occurs if any of the items is matched.

A MultiIndex handles a single "use" expression, for a single population.

The class is used to support a variety of indexes: indexes to support the saxon:index() and saxon:find extension functions; and indexes over the value of a variable, as determined by the Saxon-SA optimizer.

The class supports two kinds of equality matching. In ConvertUntypedToOther mode, corresponding to the XPath "=" operator, an untyped value is converted to the type of the other item, so the double 5e0 will match the untypedAtomic value "5". In ConvertUntypedToString mode, untyped atomic values are always converted to strings: this corresponds to the XPath "eq" operator (and to the rules for xsl:key).

Each MultiIndex index is in fact a set of sub-indexes, one for each primitive type encountered. In an expression such as $seq[./A = $B], one sub-index is built for each primitive type that occurs in the sequence $B. In this index, any untyped atomic values in ./A are converted to that primitive type. For an expression such as $seq[./A eq $B], a single index is built, in which untyped atomic values in ./A are converted to strings.

A particular difficulty arises where $B (in the expression $seq[./A eq $B]) is untypedAtomic. In this situation an index is built in which any untypedAtomic values in ./A are indexed as strings. While building the index, we note all the primitive types that occur in ./A, and then we convert the search term $B to each of these types in turn before doing the index lookup. In practice, of course, it is very rare for the index to be heterogeneous.

This class was previously IndexedClosure, and inherited from MemoClosure, so that the contents of the sequence were built lazily. In most cases, however, an index will be built as soon as the variable is used, so there is no benefit in delaying matters. Making the class a Value in its own right means that the class can be used in both compiled and interpreted scenarios.


Constructor Summary
MultiIndex(Value indexedValue)
          Create a MultiIndex
 
Method Summary
 SequenceIterator findItems(SequenceIterable use, SequenceIterator values, boolean convertUntypedToOther, AtomicComparer comparer, XPathContext context)
          Retrieve items that match a given index value
 
Methods inherited from class java.lang.Object
equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Constructor Detail

MultiIndex

public MultiIndex(Value indexedValue)
Create a MultiIndex

Parameters:
indexedValue - the sequence to be indexed
Method Detail

findItems

public SequenceIterator findItems(SequenceIterable use,
                                  SequenceIterator values,
                                  boolean convertUntypedToOther,
                                  AtomicComparer comparer,
                                  XPathContext context)
                           throws XPathException
Retrieve items that match a given index value

Parameters:
use - the expression whose value is to be matched
values - the set of values to be found
convertUntypedToOther - true if untypedAtomic values are to be converted to the type of the other operand
comparer - AtomicComparer used to compare values for equality (encapsulates a collation if necessary)
context - the dynamic evaluation context
Returns:
an iterator over the items that match the specified values
Throws:
XPathException


Copyright (C) Michael H. Kay. All rights reserved.