| Extension of
RulesBase for complex schema.
This is an extension of the basic pattern matching scheme
intended to improve support for mapping complex xml-schema.
It is intended to be a minimal extension of the standard rules
big enough to support complex schema but without the full generality
offered by more exotic matching pattern rules.
When should you use this rather than the original?
This pattern-matching engine is complex and slower than the basic
default RulesBase class, but offers more functionality:
- Universal patterns allow patterns to be specified which will match
regardless of whether there are "better matching" patterns available.
- Parent-match patterns (eg "a/b/?") allow matching for all direct
children of a specified element.
- Ancestor-match patterns (eg "a/b/*") allow matching all elements
nested within a specified element to any nesting depth.
- Completely-wild patterns ("*" or "!*") allow matching all elements.
Universal Match Patterns
The default RulesBase pattern-matching engine always attempts to find
the "best matching pattern", and will ignore rules associated with other
patterns that match but are not "as good". As an example, if the pattern
"a/b/c" is associated with rules 1 and 2, and "*/c" is associated with
rules 3 and 4 then element "a/b/c" will cause only rules 1 and 2 to execute.
Rules 3 and 4 do have matching patterns, but because the patterns are shorter
and include wildcard characters they are regarded as being "not as good" as
a direct match. In general, exact patterns are better than wildcard patterns,
and among multiple patterns with wildcards, the longest is preferred.
See the RulesBase class for more information.
This feature of preferring "better" patterns can be a powerful tool.
However it also means that patterns can interact in unexpected ways.
When using the ExtendedBaseRules, any pattern prefixed with '!' bypasses
the "best match" feature. Even if there is an exact match or a longer
wildcard match, patterns prefixed by '!' will still be tested to see if
they match, and if so their associated Rule objects will be included in
the set of rules to be executed in the normal manner.
- Pattern
"!*/a/b" matches whenever an 'b' element
is inside an 'a'.
- Pattern
"!a/b/?" matches any child of a parent
matching "a/b" (see "Parent Match Patterns").
- Pattern
"!*/a/b/?" matches any child of a parent
matching "!*/a/b" (see "Parent Match Patterns").
- Pattern
"!a/b/*" matches any element whose path
starts with "a" then "b" (see "Ancestor Match Patterns").
- Pattern
"!*/a/b/*" matches any elements whose path
contains 'a/b' (see "Ancestor Match Patterns").
Parent Match Patterns
These will match direct child elements of a particular parent element.
-
"a/b/c/?" matches any child whose parent matches
"a/b/c". Exact parent rules take precedence over Ancestor
Match patterns.
-
"*/a/b/c/?" matches any child whose parent matches
"*/a/b/c". The longest matching still applies to parent
matches but the length excludes the '?', which effectively means
that standard wildcard matches with the same level of depth are
chosen in preference.
Ancestor Match Patterns
These will match elements whose parentage includes a particular sequence
of elements.
-
"a/b/*" matches any element whose path starts with
'a' then 'b'. Exact parent and parent match rules take precedence.
The longest ancestor match will take precedence.
-
"*/a/b/*" matches any elements whose path contains
an element 'a' followed by an element 'b'. The longest matching still
applies but the length excludes the '*' at the end.
Completely Wild Patterns
Pattern "*" matches every pattern that isn't matched by
any other basic rule.
Pattern "!*" matches every pattern.
Using The Extended Rules
By default, a Digester instance uses a
RulesBase instance as
its pattern matching engine. To use an ExtendedBaseRules instance, call
the Digester.setRules method before adding any Rule objects to the digester
instance:
Digester digester = new Digester();
digester.setRules( new ExtendedBaseRules() );
The most important thing to remember when using the extended rules is
that universal and non-universal patterns are completely independent.
Universal patterns are never affected by the addition of new patterns
or the removal of existing ones. Non-universal patterns are never affected
by the addition of new universal patterns or the removal of
existing universal patterns. As in the basic matching rules,
non-universal (basic) patterns can be affected by the
addition of new non-universal patterns or the removal of existing
non-universal patterns, because only rules associated with the
"best matching" pattern for each xml element are executed.
This means that you can use universal patterns to build up the simple
parts of your structure - for example defining universal creation and
property setting rules. More sophisticated and complex mapping will require
non-universal patterns and this might mean that some of the universal rules
will need to be replaced by a series of special cases using non-universal
rules. But by using universal rules as your backbone, these additions
should not break your existing rules.
| 