Smart Matching in Perl 5.10

Smart Matching in Perl 5.10

Tags: Perl, Perl 5, 5.10, smart match

So Perl 5.10 has been released on 18th December 2007, on the 20th birthday of Perl. There are several interesting additions to the language. In What's new in Perl 5.10? say, //, state I already wrote about a few. Now I am going to look at something called Smart Matching.

In 5.10 there is a new operator that looks like this: ~~

It is called the Smart Matching operator and it is available only if specifically requested. The easiest way to do that is to write the following in your code:

    use 5.010;

As it is a commutative operator normally you will use ~~ just like you would use "==" or "eq" between two scalar variables (but not like =~ which is not commutative).

Smart Match will check if the two values are string-equal using "eq", unless it finds some better way to match them....

so all the following will be true:

    "Foo" ~~ "Foo"
    42    ~~ 42
    42    ~~ 42.0
    42    ~~ "42.0"

and all these will be false

    "Foo"  ~~ "Bar"
    42     ~~ "42x"

And that is already a nice advantage over what we had earlier.

In every operator Perl changes the type of the value based on the operator. That is == turns both sides to Numerical values and compares them as numbers while eq turns both side to String values and compares them as strings.

When turning a string to a number Perl looks at the left side of the string and uses as many characters as it can understand as being a number and warns if there are more - non-number - characters in the string.

On the other hand ~~ fits the comparison method to the values on the two sides.
In a smart way.

This means that these are all true:

    42 == 42
    42 == 42.0
    42 == "42.0"
    42 == "42\n"

but this is false:

    42 eq "42.0"
    42 eq "42\n"

and the following are true:

    42 == "42x"
    "Foo" == "Bar"

albeit with a warning...
if you used use warnings...

This behavior while consistent is a bit hard to understand.

On the other hand the new ~~ is strange in another way. Its comparison is value driven as opposed to the other operations in Perl which are operator driven.

Let's see it in a different approach: As I wrote ~~ will compare the values as strings using eq, unless it finds some better way to compare them.

    "Foo" ~~ "Bar"

will return false but

    "Moose" ~~ "Moose"

will return true. Nothing to surprise us.

So what better ways might be there to compare two scalars?

I wrote a small function that can be used to see what smart matching does.

    sub compare {
        my ($x, $y, $description) = @_;
        my $t = ($x ~~ $y ? "True" : "False");
        $x //= '';
        $y //= '';
        printf("%-4s ~~ %-4s is %-5s   (%s)\n", $x, $y, $t, $description);
    }

This will get two values (and a description) and compare the two with smart matching printing the result. So I can now supply two values two this function and see how are the respective variables compared.

So

    "Foo" ~~ "Bar" is the same as "Foo" eq "Bar", (this is called Any ~~ Any in the documentation)

If one side is a number, it seems to be better to compare them as numbers (using ==)

    42 ~~ 42.0  are compared using == (this is called Any ~~ Num)

If one side is Number but the other one is a string (eg. "xyz" or "23xxyz") then we would be better off comparing with eq and not trying to change them to numbers. Hence:

    42 ~~ "42.0x"  are compared using eq (this is called Any ~~ Str)

But what if the string is actually including a numish thing such as "42" or "42.0" or even "42\n" to forgive those who forget to chomp()?

The following are all compared as numbers so they are true:

    42 ~~ "42"
    42 ~~ "42.0"
    42 ~~ "42\n"
    42 ~~ "42 "

This is called (Num ~~ numish)

There is a strange thing though, and I am not yet sure what is its purpose but if both are numish then they are compared as strings:

    "42" ~~ "42.0"

I have a posted a question regarding this on PerlMonks where you might find the answer.

But ~~ can do more. If one of the values in an == or eq is undef, Perl will complain about that. Smart Matching on the other hand understands that an undef is just an undef. So if one of the values is undef then ~~ checks if the other one is undef too using defined();

So these are false:

    3 ~~ undef
    "x" ~~ undef

While this is true:

    undef ~~ undef

In addition one can provide a regular expression on one side and then ~~ will apply the regex so one can write either of these without any success:

    "Perl 5.10" ~~ /Moose/
    /Moose/ ~~ "Perl 5.10"
    "Perl 5.10" ~~ qr/Moose/
    qr/Moose/ ~~ "Perl 5.10"

There are more interesting things though.

• What if one of the given values is not a real scalar?
• What if that is actually a reference to an array?

The smart match will do The Right Thing, it will check if the given scalar value is the same as one of the elements in the Array. (Str ~~ Array)

    "Moose" ~~ [qw(Foo Bar Baz)]         is false
    "Moose" ~~ [qw(Foo Bar Moose Baz)]   is true

The way the individual values are compared is based on the type of the scalar. So if the scalar is a string all the values of the array are compared to the string using "eq" while if the scalar is a number, all the comparisons will be done by "==".

(Str ~~ Array and Num ~~ Array in the documentation)

So

    42 ~~ [23, 17, 70]              false
    42 ~~ [23, 17, 42, 70]          true
    42 ~~ [23, 17, "42\n", 70]      true
    42 ~~ [23, 17, "42 ", 70]       true
    
    42 ~~ [23, 17, "42x", 70]       true with a warning Argument "42x" isn't numeric in smart match

I am not fully convinced that the last one is really good, but that's the behavior. If you would like to read more about that here is the PerlMonks post about that issue.

So that actually means we now have an operator to check if an individual scalar is represented in an array. It still slower than a hash lookup but it is faster than a grep that most people use. It is definitely easier to write.

Similarly we can have a hash reference instead of one of the values and ~~ will check if the given scalar is one of the keys in the hash: That is, using exists(); (Any ~~ Hash)

    'a' ~~ {a => 19, b => 23}        true
    42  ~~ {a => 19, b => 23}        false

But that is a bit less interesting in this context.

As a side note, instead of reference to Array or reference to Hash you can actually put there the real Array or the real Hash (but not a simple list) so this works and it is true:

    my @a = (2, 3);
    say 3 ~~ @a ? "T" : "F";

but this does not work:

    say 3 ~~ (2, 3) ? "T" : "F";

The obvious question is then, what happens when both sides are complex data structures Arrays or Hashes?

With Arrays, it will check if each element is the same (Array ~~ Array)

    ["Foo", "Bar"] ~~ ["Foo", "Bar"]          is true
    ["Foo", "Bar"] ~~ ["Foo", "Bar", "Baz"]   is false
    [1,2,3] ~~ [1,2,3]                        is true

With hashes it checks if the keys are identical (Hash ~~ Hash)

    {Foo => 19, Bar => 23} ~~ {Foo => 23, Bar => 19}     is true

and the checking is done in any deep structure:

    ["Foo", ["Bar", "Baz"]] ~~ ["Foo", ["Bar", "Baz"]], 

There are several more cases but even this was probably too much for an introduction.

To get more details search for "Smart matching in detail" after typing perldoc perlsyn on the command line - you do have 5.10 installed already, don't you, or after browsing to the perlsyn page.