Strange behaviour regular expressions with binding

[enieuwen]

When using parantheses around regular expressions in bindings, the binding result sometimes seems incorrect:

// recognize 'something' between X-es
rascal>[ <comment,rest> | /<comment:(X.*X)><rest:.*>/ := "before XcommentX after"]
lrel[str,str]: [<"XcommentX","XcommentX">]

The variables comment and rest should never match the same part of the string.

rascal>[ rest | /<comment:(X.*X)><rest:.*>/ := "before XcommentX after"]
list[str]: ["XcommentX"]

The result above is consistent with the first test but if you remove the first (unused) binding of comment, the result is different:

rascal>[ rest | /(X.*X)<rest:.*>/ := "before XcommentX after"]
list[str]: [" after"]

[jurgenvinju]

Thanks for the report. We analyzed it and the bug is that the round brackets should not be interpreted as groups (angular brackets are groups in Rascal).

[jurgenvinju]

@PaulKlint could you have a look? and what about the compiler code?

[edwardmp]

We have encountered this bug too today. These two regular expressions are the same, except for the added parentheses to capture a group (see where it binds to group variable). The behavior is different but it shouldn't make any difference.

/\A<group:.*>\*\/\s*<code:(.*)>\z/
/\A<group:(.*)>\*\/\s*<code:(.*)>\z/

[jurgenvinju]

I still think we should not support Java regex syntax anymore but just our own regular expression notation from the syntax definitions.

[DavyLandman]

👍

[DavyLandman]

although for now we could remove the group parens for now? (since we also generate our own groups)

[jurgenvinju]

This is still open. @lmove just encountered it.