We can use group numbers to reuse the same piece of code while inside regex string, like that:

StringCases["x = y", RegularExpression["([\\w\\s]+)=(?1)"]]

{"x = y"}

To make it more readable and robust (like avoiding the situation when the groups can be renumbered either internally or through a redesign of the regex code), we can name patterns (and yes, (?<n>) / (?&n) syntax works as well as (?P<n>) / (?P>n)):

StringCases["x = y", RegularExpression["(?<var>[\\w\\s]+)=(?&var)"]]
StringCases["x = y", RegularExpression["(?P<var>[\\w\\s]+)=(?P>var)"]]

{"x = y"}

{"x = y"}

Perfect so far. However, what if I want to use the groups in the replacement rule? This works:

StringCases["x = y", RegularExpression["([\\w\\s]+)=((?1))"] :> {"$1", "$2"}]

{{"x ", " y"}}

But this doesn't:

StringCases["x = y", RegularExpression["(?<var>[\\w\\s]+)=(?<rhs>(?&var))"] :> {"${var}", "${rhs}"}]

{{"${var}", "${rhs}"}}

Of course, I can still use the numbers to refer to the groups:

StringCases["x = y", RegularExpression["(?<var>[\\w\\s]+)=(?<rhs>(?&var))"] :> {"$1", "$2"}]

{{"x ", " y"}}

But this is exactly what I want to avoid. I want to use the names, as they won't be affected by the code redesign.

I also tried $+{name}, \g<name>, but they all don't work.

So, the question is: can we use names of the groups in the replacement rule?

  • $\begingroup$ Is rewriting your regular expression as Mathematica StringExpression and option for you? $\endgroup$ – halirutan Apr 12 '17 at 12:57
  • $\begingroup$ @halirutan Potentially yes, but the regex is rather big, so the StringExpression would be huge and there is no way to automate this conversion. I also know that sometimes StringExpression actually might have an advantage when extracting groups (25677), so it is an option, but I thought Mathematica would support named groups (why shouldn't it?). $\endgroup$ – Stitch Apr 12 '17 at 13:09
  • $\begingroup$ Come and join the Mathematica Chat for a moment. $\endgroup$ – halirutan Apr 12 '17 at 13:39

Unfortunately, this is not an answer to the problem but rather some information on the subject that is hopefully of interest. As I said in my comment, one way to solve this issue is to transfer the regex into a StringExpression.

StringCases["x = y", 
 x : ((WhitespaceCharacter | LetterCharacter) ..) ~~ "=" ~~ y__ :> {x,y}

The above code should be equivalent to your regex. What might be of interest for you is how Mathematica handles the named patterns. For this, we can use the undocumented function StringPattern`PatternConvert (undocumented, but you can access its full implementation which is very enlightening).

 x : ((WhitespaceCharacter | LetterCharacter) ..) ~~ "=" ~~ y__]

(* {"(?ms)((?:[[:space:][:alpha:]])+)=(.+)", {{Hold[x], 1}, {Hold[y], 2}}, {}, Hold[None]} *)

Interestingly, you see that Mathematica converts it into a regex with normal grouping and stores the mapping between the groups and the names you gave. Therefore, it knows that x belongs to the first group.

How this helps you? It probably doesn't. As far as I can see there is no function to automatically turn a regex into a string expression.

  • $\begingroup$ Actually StringPattern`PatternConvert is documented under the "Implementation Details" section of the tutorial "Working with String Patterns". $\endgroup$ – Alexey Popkov Apr 14 '17 at 8:36

Not the answer you're looking for? Browse other questions tagged or ask your own question.