DEV Community

Cover image for Regular Expressions
Elizabeth Mattijsen
Elizabeth Mattijsen

Posted on • Edited on

Regular Expressions

Regular expressions have been completely re-imagined from the original Perl regular expressions. Even to the point that it was decided they were no longer regular in Raku. Hence the use of the word "regex".

The noticeable differences between Perl and Raku become bigger when the regex becomes more complicated.

Operator Changes

In Perl, the =~ (true on match) and !~ (true if no match) operators

# Perl
say "matched" if "foo" =~ /o/;         # matched
say “did not match" if "foo" !~ /x/;   # did not match
Enter fullscreen mode Exit fullscreen mode

have been replaced by ~~ and !~~ in Raku:

# Raku
say "matched" if "foo" ~~ /o/;         # matched
say "did not match" if "foo" !~~ /x/;  # did not match
Enter fullscreen mode Exit fullscreen mode

Note that the ! is a generic Raku feature that you can use to negate the result of an infix operator, even custom defined infix operators.

Raku also has some methods that may be better readable if you're just interested in whether there is an occurrence:

Some example usage:

# Raku
say "contains 'o'"    if "foo".contains("o");     # contains 'o'
say "has letters"     if "foo".contains(/ \w+ /); # has letters
say "starts with 'f'" if "foo".starts-with("f");  # starts with 'f'
say "ends-with 'o'"   if "foo".ends-with("o");    # ends with 'o'
say "from 2nd char" if "foo".substr-eq("oo",1);   # from 2nd char
Enter fullscreen mode Exit fullscreen mode

The same operator changes apply for substitution. Code in Perl that would look like:

# Perl
my $string = "foo";
$string =~ s/o/x/;
say $string;                  # fxo
Enter fullscreen mode Exit fullscreen mode

looks like this in Raku (just the operator is different):

# Raku
my $string = “foo”;
$string ~~ s/o/x/;
say $string;                  # fxo
Enter fullscreen mode Exit fullscreen mode

Newer versions of Perl also allow the r modifier, to just return the substitution rather than attempt to modify the source string:

# Perl
say "foo" =~ s/o/x/r;         # fxo
Enter fullscreen mode Exit fullscreen mode

Raku does not have that modifier. But it does have other syntax for achieving the same result, using the subst method:

# Raku
say "foo".subst( /o/, “x” );  # fxo
Enter fullscreen mode Exit fullscreen mode

Note that the subst method does not actually require a regular expression, but can be used with bare strings only also:

# Raku
say "foo".subst( “o”, “x” );  # fxo
Enter fullscreen mode Exit fullscreen mode

Which is actually quite a bit more performant as well.

Whitespace is not significant

In Perl’s regular expressions, whitespace is significant:

# Perl
say "matched" if "foo" =~ / o /;       # no output
Enter fullscreen mode Exit fullscreen mode

One can add the x modifier to make whitespace not significant in Perl:

# Perl
say "matched" if "foo" =~ / o /x;      # matched
Enter fullscreen mode Exit fullscreen mode

One could consider the x modifier to be always specified in Raku:

# Raku
say "matched" if "foo" ~~ / o /;       # matched
Enter fullscreen mode Exit fullscreen mode

If you do want to match a string with possible whitespace, you must quote it in Raku:

# Raku
say “matched” if "f o o" ~~ / " o " /; # matched
Enter fullscreen mode Exit fullscreen mode

Note that in this case, the whitespace outside of the quoted string, is ignored again.

Positional Captures

Positional captures in Perl start at number 1:

# Perl
say $1 if "foo" =~ /(.)/;              # f
Enter fullscreen mode Exit fullscreen mode

In Raku, they start at 0, like any other array index:

# Raku
say $0 if "foo" ~~ / (.) /;            # 「f」
Enter fullscreen mode Exit fullscreen mode

This is because $0 is short-hand for $/[0], in which $/ is the latest Match object (the result of the smartmatch with ~~).

Named Captures

Perl has support for named captures since version 5.22. Raku has this from the start.

In Perl, this is achieved with the (?<name>regex) syntax, and obtaining the matched string is by accessing the %+ hash:

# Perl
"The colour is blue" =~ /is (?<colour>\w+)/;
say "Found colour '$+{colour}';  # Found colour 'blue'
Enter fullscreen mode Exit fullscreen mode

In Raku, the syntax is a little different: $<name>=(regex), and obtaining the matched string uses the same syntax for indicating the name $<name>:

# Raku
"The colour is blue" ~~ /is \s $<colour>=(\w+)/;
say "Found colour '$<colour>'";  # Found colour 'blue'
Enter fullscreen mode Exit fullscreen mode

Note that you need to be specific about the whitespace between "is" and the colour!

Character Classes

Specification of character classes in Raku is slightly different from Perl, and a little more flexible.

Basically, the square brackets have been repurposed as non-capturing grouping syntax (as that in most cases, will be used more often than character classes). Also indicating ranges of characters uses the standard Range syntax in Raku.

Some examples in Perl:

# Perl
"foo" =~ /[aeiuo]/;           # match a char of a e i o u
"bar" =~ /[^aeiou]/;          # match a char that is NOT a e i o u
"baz" =~ /[a-z]/;             # match a char between a and z inclusive
"AMS" =~ /[[:upper:]]/;       # match a char that is uppercase
"CMI" =~ /[aeiou[:upper:]]/;  # combined character classes
Enter fullscreen mode Exit fullscreen mode

The equivalents in Raku:

# Raku
"foo" ~~ /<[aeiuo]>/;         # match a char of a e i o u
"bar" ~~ /<-[aeiou]>/;        # match a char that is NOT a e i o u
"baz" ~~ /<[a..z]>/;          # match a char between a and z inclusive
"AMS" ~~ /<:Upper>/;          # match a char that is uppercase
"CMI" ~~ /<[aeiou]+:upper>/;  # combined character classes
Enter fullscreen mode Exit fullscreen mode

And because whitespace is not significant inside of the <[ ]> either in Raku, you can write them in a much more readable way:

# Raku
"foo" ~~ / <[ a e i u o ]>/;
"bar" ~~ / <-[ a e i o u ]>/;
"baz" ~~ / <[ a .. z ]> /;
"AMS" ~~ / <:upper> /;
"CMI" ~~ / <[ a e i o u ] + :upper> /;
Enter fullscreen mode Exit fullscreen mode

Note that in Raku you could consider specifying a character class as a grouping of characters, so using [ ] inside of the < > should make sense mnemonically.

Under the hood

Under the hood, the regex engines of Perl and Raku couldn't be more different. In Perl, the regex engine is basically a state machine with extensions that allow for code execution. In Raku, the regex engine is just executable code.

So when you "run a regex" in Raku, it's really just Raku code that is being executed under the hood. Not a state machine that is written in another language.

A named Regex in Raku is really just a piece of executable code, much like a method. And a Grammar could be considered a special type of class with regexes as methods.

This has implications for composability. A whole book could probably be written about regexes and grammars of the Raku Programming Language. And Moritz Lenz actually did this in "Parsing with Regexes and Grammars - A Recursive Descent into Parsing". Well recommended!

Summary

This covers the most obvious user visible changes between regexes in Perl and Raku.

There's definitely a lot more to be said about the differences, but it feels the difference in approach to regexes is so large, that doing further comparisons would not be treating either version well.

If you're interested in pursuing Raku regexes further, then these tutorials are recommended reading:

Latest comments (2)

Collapse
 
scimon profile image
Simon Proctor
"The colour is blue" =~ /is $<colour>=(\w+)/;
Enter fullscreen mode Exit fullscreen mode

Is wrong no?
Using =~ and expecting the whitepace to be significant?

"The colour is blue" ~~ /is " " $<colour>=(\w+)/;
Enter fullscreen mode Exit fullscreen mode

Is closer yes?

Collapse
 
lizmat profile image
Elizabeth Mattijsen

Good catch! Fixed by using \s for better visibility.