Proposed C API for Regular Expressions

A proposal for a C API to ICU regular expressions is here:

http://oss.software.ibm.com/cvs/icu/icu/source/i18n/unicode/uregex.h?rev=1.1&content-type=text/x-cvsweb-markup

This is brand new; It's certain to have things wrong, and things that 
could be done better, or more simply or in ways that are easier to 
understand or use.

The functions pretty much follow those from the C++ API.  The main areas 
of difference are these:

1)  The C++ API has two main classes, RegexPattern, representing a 
compiled regexp pattern, and RegexMatcher, representing the state of a 
specific match.  Patterns are read-only and shareable by multiple 
matchers.  This structure was copied from the Java JDK regular 
expression package.

For C, I have proposed combining these two concepts into one.  Doing so 
makes user code simpler, and doesn't appear to have any real 
disadvantages.  Internally, the implementation still has both types, and 
compiled patterns can still be shared when appropriate.  (They are 
shared when a regexp object is cloned.)

In our C++ API, without Java's garbage collection, managing the 
dependency between a pattern and the Matcher objects that use it is a 
manual and potentially error prone process.  With the C API, it will all 
be automatic.

2)  A reduction in the number of overloaded function variants. 
Overloads are awkward to translate to C because the function name must 
change too.  The upshot is longer parameter lists to some functions, as 
the simplified overloads and/or defaulted parameters from C++ are gone.

3)  String types change from class UnicodeString to (UChar *) everywhere.

So please, everyone, let me know what you think.

   Thanks,


   -- Andy Heninger
      [email protected]