Alternate approach for issue #17

[uhhhh2]

I've created a possible approach for addressing Issue #17 .

First, I added the abstract method declaration in JsonGenerator. Next, I added implementations of this method in two of the JsonGenerator implementations. Iin one implementation of the JsonGenerator (WriterBasedJsonGenerator), I created a separate buffer field in the class (using new char[] to avoid calling the same IOContext.allocX method twice which apparently results in an exception). In another implementation of the JsonGenerator (UTF8JsonGenerator), I tried to reuse the (apparently pre-existing) _charBuffer. I am not sure if these are the best approaches for getting a char[] for use as a temporary buffer for reading from a Reader. Then, I added an implementation in the JsonGeneratorDelegate that calls the appropriate function of the delegate instance.

As for a unit test, I copied one of the pre-existing unit test files (StringGenerationTest) and created a new version (StringGenerationReaderTest to avoid deleting the original tests) In StringGenerationReaderTest, the boolean parameters to the methods are removed (as it only tests using a Reader). Also, each function calls

StringReader reader = new StringReader(text); gen.writeString(reader, -1)

where it would previously call

gen.writeString(text); 

where text is the name of the String variable with the text to write.

Let me know if there is anything I can do to further test and/or improve this approach.

[uhhhh2]

I've edited the code so that it throws on a null reader or after reading too little. I've added some more unit tests (adapted from GeneratorFailTest and saved as GeneratorFailFromReaderTest) to test these types of failures.
Also, added a default implementation in GeneratorBase that simply reports the operation as "unsupported".

[cowtowncoder]

Looks good, and since I got CLA will merge for 2.9.