Question

I am having this particular requirement where a method has to be identified by different regular expressions for different components. For example, there need to be a regex for return parameter, one for method name, one for argument type and one for argument name. I was able to come up with an expression till this step as follows –

([^,]+) ([^,]+)\((([^,]+) ([^,]+))\)

It works well for a method signature like –

ReturnType foo(Arg parameter)
The regular expression identifies ReturnType, foo, Arg and parameter separately.

Now the problem is that a method can have no/one/multiple arguments separated by commas. I am not able to get a repeating expression for this. Help will be appreciated.

Answer 1

Let’s abstract this out a bit, and say we want to match a (possibly empty) list of digits separated by commas.

(empty)
12
12,34
12,34,56

The pattern is therefore

^$|^\d+(,\d+)*$

Now you can try to replace the components to match what you want:

• Instead of \d+, whatever regex you use to match type name and identifier
• Maybe allow \s* around the comma
• Maybe you’d even add the special varargs last argument (which can also be the first and only)

Note that if you allow generic type parameters, then you definitely can’t use regex since you can nest the <...> and the language of balanced balanced parentheses of arbitrary depth is not regular.

Although you can argue that in practice, no one would ever nest type parameters deeper than, say, 3 levels, so then it becomes regular again.

That said, a proper parser is really the best tool for this. Just look for implementation of Java grammar, say, in ANTLR.

---

See also

• JLS 18.1 The Grammar of the Java Programming Language
• Java 1.6 Grammar in ANTLR