[Pljava-dev] CREATE FUNCTION docs inconsistent with implementation

Eric Faulhaber wrote:
> In case anyone else has been having trouble with this, hopefully this
> will save you some time...
>
> The "Function mapping" (long-form) CREATE FUNCTION sample from the Wiki
> seems to be out of sync with the backing implementation:
>
> CREATE FUNCTION getsysprop(VARCHAR)
> RETURNS VARCHAR
> AS 'java.lang.String java.lang.System.getProperty(java.lang.String)'
> LANGUAGE java;
>
> When getsysprop is run as is, the Java class name is interpreted as
> 'java.lang.Stringjava.lang.System' and the return type as 'null', as in:
>
> DEBUG: className = 'java.lang.Stringjava.lang.System', methodName =
> 'getProperty', parameters = 'java.lang.String', returnType = 'null'
> DEBUG: Loading class java.lang.Stringjava.lang.System
> ERROR: java.lang.ClassNotFoundException: java.lang.Stringjava.lang.System
>
Ouch. This is a bug. I'll have to look into that.

> The parseFunction function (line 387 in src/C/pljava/Functions.c from
> the 1.3.0 source download) looks for '=' between the return type and
> class name. Consequently, the CREATE FUNCTION statement must be:
>
> CREATE FUNCTION getsysprop(VARCHAR)
> RETURNS VARCHAR
> AS 'java.lang.String=java.lang.System.getProperty(java.lang.String)'
> LANGUAGE java;
>
> Is the '=' delimiter expected to be a permanent syntax here, or is the
> intention that the implementation eventually will match the docs?
>
The '=' is only used internally and should not be used in the CREATE FUNCTION declaration.
PL/Java will strip of every white space in the string very early on in the parse process
since they are all insignificant. Well, all but the one between the return type and the
function name. That particular white space is replaced by the '=' sign. The code at line 387
comes into play after that.

Regards,
Thomas Hallgren