It is expected that certain attributes have fixed values, ie.
It is expected that certain attributes have fixed values, ie. booleans or based on schema enums etc, this function allows such optimisations.
The qname will have been obtained via a call to either noNamespaceQName, unprefixedQName or prefixedQName, so any optimisations provided by them can be leveraged.
During parsing provides a backing array for attribute storage.
During parsing provides a backing array for attribute storage. The expected new size must be passed in, failure to give a size large enough will lead to OOB exceptions.
Implementations are advised to use the array for parsing performance reasons.
The default implementataion expands the size by 50% of the backed array. The array ownership should remain with the strategy instance
For files without large (and varied) attribute counts per repeated element it may represent significant space savings to optimise against the entire Elem itself.
For files without large (and varied) attribute counts per repeated element it may represent significant space savings to optimise against the entire Elem itself.
The Elements QName is garaunteed to be the last QName evaluated via the QName functions (i.e. Attributes are done first then the Elems QName)
Memory usage in DOMs is often dominated by repeated elements. Xerces and co use string tables to optimise memory usage, with DTM a key example.
Performing lookups is expensive so the strategies can have selective levels of lookup. Developers may therefore choose appropriate levels that best fit their trade-off between space and time.
There will, of course, be temporary garbage created for such a scheme but it should pay off for larger messages.