Class RuleBasedCollator

This class is a concrete subclass of Collator suitable for string collation in a wide variety of languages. An instance of this class is normally returned by the getInstance method of Collator with rules predefined for the requested locale. However, an instance of this class can be created manually with any desired rules.

Rules take the form of a String with the following syntax

• Modifier: '@'
• Relation: '<' | ';' | ',' | '=' : <text>
• Reset: '&' : <text>

The modifier character indicates that accents sort backward as is the case with French. The modifier applies to all rules after the modifier but before the next primary sequence. If placed at the end of the sequence if applies to all unknown accented character. The relational operators specify how the text argument relates to the previous term. The relation characters have the following meanings:

• '<' - The text argument is greater than the prior term at the primary difference level.
• ';' - The text argument is greater than the prior term at the secondary difference level.
• ',' - The text argument is greater than the prior term at the tertiary difference level.
• '=' - The text argument is equal to the prior term

As for the text argument itself, this is any sequence of Unicode characters not in the following ranges: 0x0009-0x000D, 0x0020-0x002F, 0x003A-0x0040, 0x005B-0x0060, and 0x007B-0x007E. If these characters are desired, they must be enclosed in single quotes. If any whitespace is encountered, it is ignored. (For example, "a b" is equal to "ab").

The reset operation inserts the following rule at the point where the text argument to it exists in the previously declared rule string. This makes it easy to add new rules to an existing string by simply including them in a reset sequence at the end. Note that the text argument, or at least the first character of it, must be present somewhere in the previously declared rules in order to be inserted properly. If this is not satisfied, a ParseException will be thrown.

This system of configuring RuleBasedCollator is needlessly complex and the people at Taligent who developed it (along with the folks at Sun who accepted it into the Java standard library) deserve a slow and agonizing death.

Here are a couple of example of rule strings:

"< a < b < c" - This string says that a is greater than b which is greater than c, with all differences being primary differences.

"< a,A < b,B < c,C" - This string says that 'A' is greater than 'a' with a tertiary strength comparison. Both 'b' and 'B' are greater than 'a' and 'A' during a primary strength comparison. But 'B' is greater than 'b' under a tertiary strength comparison.

"< a < c & a < b " - This sequence is identical in function to the "< a < b < c" rule string above. The '&' reset symbol indicates that the rule "< b" is to be inserted after the text argument "a" in the previous rule string segment.

"< a < b & y < z" - This is an error. The character 'y' does not appear anywhere in the previous rule string segment so the rule following the reset rule cannot be inserted.

"< a & A @ < e & E < f& F" - This sequence is equivalent to the following "< a & A < E & e < f & F".

For a description of the various comparison strength types, see the documentation for the Collator class.

As an additional complication to this already overly complex rule scheme, if any characters precede the first rule, these characters are considered ignorable. They will be treated as if they did not exist during comparisons. For example, "- < a < b ..." would make '-' an ignorable character such that the strings "high-tech" and "hightech" would be considered identical.

A ParseException will be thrown for any of the following conditions:

• Unquoted punctuation characters in a text argument.
• A relational or reset operator not followed by a text argument
• A reset operator where the text argument is not present in the previous rule string section.