Ensure source method search regex handles complex sigs about jitwatch HOT 4 CLOSED

Turns out regex is the wrong tool for the job once you go past simple method signatures and get into heavy generics.
Looks like a full parser like ANTLR is needed.

http://stackoverflow.com/questions/3007139/regex-to-match-a-java-method-signature

http://stackoverflow.com/questions/5744549/java-regex-for-a-method-signature-with-a-collectiongenerictype-return-type

Will investigate further to decide correct fix.

from jitwatch.

If this can help, parboiled provides a Java parser in its examples section: https://github.com/sirthias/parboiled/tree/master/examples-java/src/main/java/org/parboiled/examples/java/

from jitwatch.

Thanks :)

Thinking about this more, JITWatch knows the declared method on the Class e.g.

public static java.lang.Object[] java.util.Arrays.copyOf(java.lang.Object[],int,java.lang.Class)

and is trying to create a regex that will match this in the Java source file (a big String) by adding wildcards for things it cannot know (variable names, if types are fully qualified, etc.)

Currently it comes up with:

^(.*)public static ([0-9a-zA-Z_\.]*)Object\[\] copyOf( )*\(( )*([0-9a-zA-Z_\.]*)Object\[\]( )+([0-9a-zA-Z_]+),( )*int( )+([0-9a-zA-Z_]+),( )*([0-9a-zA-Z_\.]*)Class( )+([0-9a-zA-Z_]+)( )*\)(.*)$

Which will match with the type-erased signature but not with the generics syntactic sugar.

Given that generics can substitute the type names with E,T,U, ... I think the best solution might be to try a first pass search with the regex and then attempt a "best fit" search if that fails.

from jitwatch.

Added best match fallback algorithm where regex fails to match source code method signatures with generics.

from jitwatch.