Learn Perl in about 2 hours 30 minutes
Perl is a dynamic, dynamically-typed, high-level, scripting (interpreted) language most comparable with PHP and Python. Perl's syntax owes a lot to ancient shell scripting tools, and it is famed for its overuse of confusing symbols, the majority of which are impossible to Google for. Perl's shell scripting heritage makes it great for writingglue code: scripts which link together other scripts and programs. Perl is ideally suited for processing text data and producing more text data. Perl is widespread, popular, highly portable and well-supported. Perl was designed with the philosophy "There's More Than One Way To Do It" (TMTOWTDI) (contrast with Python, where "there should be one - and preferably only one - obvious way to do it").
Perl has horrors, but it also has some great redeeming features. In this respect it is like every other programming language ever created.
This document is intended to be informative, not evangelical. It is aimed at people who, like me:
- dislike the official Perl documentation at http://perl.org/ for being intensely technical and giving far too much space to very unusual edge cases
- learn new programming languages most quickly by "axiom and example"
- wish Larry Wall would get to the point
- already know how to program in general terms
- don't care about Perl beyond what's necessary to get the job done.
This document is intended to be as short as possible, but no shorter.
Preliminary notes
- The following can be said of almost every declarative statement in this document: "that's not, strictly speaking, true; the situation is actually a lot more complicated". If you see a serious lie, point it out, but I reserve the right to preserve certain critical lies-to-children.
- Throughout this document I'm using example
print statements to output data but not explicitly appending line breaks. This is done to prevent me from going crazy and to give greater attention to the actual string being printed in each case, which is invariably more important. In many examples, this results in alotofwordsallsmusheduptogetherononeline if the code is run in reality. Try to ignore this.
Hello world
A Perl script is a text file with the extension .pl.
Here's the full text of helloworld.pl:
use strict;
use warnings;
print "Hello world";
Perl scripts are interpreted by the Perl interpreter, perl or perl.exe:
perl helloworld.pl [arg0 [arg1 [arg2 ...]]]
A few immediate notes. Perl's syntax is highly permissive and it will allow you to do things which result in ambiguous-looking statements with unpredictable behaviour. There's no point in me explaining what these behaviours are, because you want to avoid them. The way to avoid them is to put use strict; use warnings; at the very top of every Perl script or module that you create. Statements of the form use foo; are pragmas. A pragma is a signal to perl.exe, which takes effect when initial syntactic validation is being performed, before the program starts running. These lines have no effect when the interpreter encounters them at run time.
The symbol # begins a comment. A comment lasts until the end of the line. Perl has no block comment syntax.
Variables
Perl variables come in three types: scalars, arrays and hashes. Each type has its own sigil: $, @ and % respectively. Variables are declared using my, and remain in scope until the end of the enclosing block or file.
Scalar variables
A scalar variable can contain:
undef (corresponds to None in Python, null in PHP)- a number (Perl does not distinguish between an integer and a float)
- a string
- a reference to any other variable.
my $undef = undef;
print $undef; # prints the empty string "" and raises a warning
# implicit undef:
my $undef2;
print $undef2; # prints "" and raises exactly the same warning
my $num = 4040.5;
print $num; # "4040.5"
my $string = "world";
print $string; # "world"
(References are coming up shortly.)
String concatenation using the . operator (same as PHP):
print "Hello ".$string; # "Hello world"
"Booleans"
Perl has no boolean data type. A scalar in an if statement evaluates to boolean "false" if and only if it is one of the following:
undef- number
0
- string
""
- string
"0".
The Perl documentation repeatedly claims that functions return "true" or "false" values in certain situations. In practice, when a function is claimed to return "true" it usually returns 1, and when it is claimed to return false it usually returns the empty string, "".
Weak typing
It is impossible to determine whether a scalar contains a "number" or a "string". More precisely, it should never be necessary to do this. Whether a scalar behaves like a number or a string depends on the operator with which it is used. When used as a string, a scalar will behave like a string. When used as a number, a scalar will behave like a number (raising a warning if this isn't possible):
my $str1 = "4G";
my $str2 = "4H";
print $str1 . $str2; # "4G4H"
print $str1 + $str2; # "8" with two warnings
print $str1 eq $str2; # "" (empty string, i.e. false)
print $str1 == $str2; # "1" with two warnings
# The classic error
print "yes" == "no"; # "1" with two warnings; both values evaluate to 0 when used as numbers
The lesson is to always using the correct operator in the correct situation. There are separate operators for comparing scalars as numbers and comparing scalars as strings:
# Numerical operators: <, >, <=, >=, ==, !=, <=>, +, *
# String operators: lt, gt, le, ge, eq, ne, cmp, ., x
Array variables
An array variable is a list of scalars indexed by integers beginning at 0. In Python this is known as a list, and in PHP this is known as an array. An array is declared using a parenthesised list of scalars:
my @array = (
"print",
"these",
"strings",
"out",
"for",
"me", # trailing comma is okay
);
You have to use a dollar sign to access a value from an array, because the value being retrieved is not an array but a scalar:
print $array[0]; # "print"
print $array[1]; # "these"
print $array[2]; # "strings"
print $array[3]; # "out"
print $array[4]; # "for"
print $array[5]; # "me"
print $array[6]; # returns undef, prints "" and raises a warning
You can use negative indices to retrieve entries starting from the end and working backwards:
print $array[-1]; # "me"
print $array[-2]; # "for"
print $array[-3]; # "out"
print $array[-4]; # "strings"
print $array[-5]; # "these"
print $array[-6]; # "print"
print $array[-7]; # returns undef, prints "" and raises a warning
There is no collision between a scalar $var and an array @var containing a scalar entry $var[0]. There may, however, be reader confusion, so avoid this.
To get an array's length:
print "This array has ".(scalar @array)."elements"; # "This array has 6 elements"
print "The last populated index is ".$#array; # "The last populated index is 5"
Variables can be interpolated into strings:
print "Hello $string"; # "Hello world"
print "@array"; # "print these strings out for me"
Caution. One day you will put somebody's email address inside a string, "jeff@gmail.com". This will cause Perl to look for an array variable called @gmail to interpolate into the string, and not find it, resulting in a runtime error. Interpolation can be prevented in two ways: by backslash-escaping the sigil, or by using single quotes instead of double quotes.
print "Hello \$string"; # "Hello $string"
print 'Hello $string'; # "Hello $string"
print "\@array"; # "@array"
print '@array'; # "@array"
Hash variables
A hash variable is a list of scalars indexed by strings. In Python this is known as a dictionary, and in PHP it is known as an array.
my %scientists = (
"Newton" => "Isaac",
"Einstein" => "Albert",
"Darwin" => "Charles",
);
Notice how similar this declaration is to an array declaration. In fact, the double arrow symbol => is called a "fat comma", because it is just a synonym for the comma separator. A hash is declared using a list with an even number of elements, where the even-numbered elements (0, 2, ...) are all taken as strings.
Once again, you have to use a dollar sign to access a value from a hash, because the value being retrieved is not a hash but a scalar:
print $scientists{"Newton"}; # "Isaac"
print $scientists{"Einstein"}; # "Albert"
print $scientists{"Darwin"}; # "Charles"
print $scientists{"Dyson"}; # returns undef, prints "" and raises a warning
Note the braces used here. Again, there is no collision between a scalar $var and a hash %var containing a scalar entry $var{"foo"}.
You can convert a hash straight to an array with twice as many entries, alternating between key and value (and the reverse is equally easy):
my @scientists = %scientists;
However, unlike an array, the keys of a hash have no underlying order. They will be returned in whatever order is more efficient. So, notice the rearranged order but preserved pairs in the resulting array:
print "@scientists"; # something like "Einstein Albert Darwin Charles Newton Isaac"
To recap, you have to use square brackets to retrieve a value from an array, but you have to use braces to retrieve a value from a hash. The square brackets are effectively a numerical operator and the braces are effectively a string operator. The fact that the index supplied is a number or a string is of absolutely no significance:
my $data = "orange";
my @data = ("purple");
my %data = ( "0" => "blue");
print $data; # "orange"
print $data[0]; # "purple"
print $data["0"]; # "purple"
print $data{0}; # "blue"
print $data{"0"}; # "blue"
Lists
A list in Perl is a different thing again from either an array or a hash. You've just seen several lists:
(
"print",
"these",
"strings",
"out",
"for",
"me",
)
(
"Newton" => "Isaac",
"Einstein" => "Albert",
"Darwin" => "Charles",
)
A list is not a variable. A list is an ephemeral value which can be assigned to an array or a hash variable. This is why the syntax for declaring array and hash variables is identical. There are many situations where the terms "list" and "array" can be used interchangeably, but there are equally many where lists and arrays display subtly different and extremely confusing behaviour.
Okay. Remember that => is just , in disguise and then look at this example:
("one", 1, "three", 3, "five", 5)
("one" => 1, "three" => 3, "five" => 5)
The use of => hints that one of these lists is an array declaration and the other is a hash declaration. But on their own, neither of them are declarations of anything. They are just lists. Identical lists. Also:
()
There aren't even hints here. This list could be used to declare an empty array or an empty hash and the perl interpreter clearly has no way of telling either way. Once you understand this odd aspect of Perl, you will also understand why the following fact must be true: List values cannot be nested. Try it:
my @array = (
"apples",
"bananas",
(
"inner",
"list",
"several",
"entries",
),
"cherries",
);
Perl has no way of knowing whether ("inner", "list", "several", "entries") is supposed to be an inner array or an inner hash. Therefore, Perl assumes that it is neither and flattens the list out into a single long list:
print $array[0]; # "apples"
print $array[1]; # "bananas"
print $array[2]; # "inner"
print $array[3]; # "list"
print $array[4]; # "several"
print $array[5]; # "entries"
print $array[6]; # "cherries"
The same is true whether the fat comma is used or not:
my %hash = (
"beer" => "good",
"bananas" => (
"green" => "wait",
"yellow" => "eat",
),
);
# The above raises a warning because the hash was declared using a 7-element list
print $hash{"beer"}; # "good"
print $hash{"bananas"}; # "green"
print $hash{"wait"}; # "yellow";
print $hash{"eat"}; # undef, so prints "" and raises a warning
Of course, this does make it easy to concatenate multiple arrays together:
my @bones = ("humerus", ("jaw", "skull"), "tibia");
my @fingers = ("thumb", "index", "middle", "ring", "little");
my @parts = (@bones, @fingers, ("foot", "toes"), "eyeball", "knuckle");
print @parts;
More on this shortly.
Context
Perl's most distinctive feature is that its code is
context-sensitive.
Every expression in Perl is evaluated either in scalar context or list context, depending on whether it is expected to produce a scalar or a list. Many Perl expressions and
built-in functions display radically different behaviour depending on the context in which they are evaluated.
A scalar assignment such as $scalar = evaluates its expression in scalar context. In this case, the expression is "Mendeleev" and the returned value is the same scalar value "Mendeleev":
my $scalar = "Mendeleev";
An array or hash assignment such as @array = or %hash = evaluates its expression in list context. A list value evaluated in list context returns the list, which then gets fed in to populate the array or hash:
my @array = ("Alpha", "Beta", "Gamma", "Pie");
my %hash = ("Alpha" => "Beta", "Gamma" => "Pie");
No surprises so far.
A scalar expression evaluated in list context turns into a single-element list:
my @array = "Mendeleev"; # same as 'my @array = ("Mendeleev");'
A list expression evaluated in scalar context returns the final scalar in the list:
my $scalar = ("Alpha", "Beta", "Gamma", "Pie"); # Value of $scalar is now "Pie"
An array expression (an array is different from a list, remember?) evaluated in scalar context returns the length of the array:
my @array = ("Alpha", "Beta", "Gamma", "Pie");
my $scalar = @array; # Value of $scalar is now 4
The
print built-in function evaluates all of its arguments in list context. In fact,
print accepts an unlimited list of arguments and prints each one after the other, which means it can be used to print arrays directly:
my @array = ("Alpha", "Beta", "Goo");
my $scalar = "-X-";
print @array; # "AlphaBetaGoo";
print $scalar, @array, 98; # "-X-AlphaBetaGoo98";
You can force any expression to be evaluated in scalar context using the
scalar built-in function. In fact, this is why we use
scalar to retrieve the length of an array.
You are not bound by law or syntax to return a scalar value when a subroutine is evaluated in scalar context, nor to return a list value in list context. As seen above, Perl is perfectly capable of fudging the result for you.
References and nested data structures
In the same way that lists cannot contain lists as elements, arrays and hashes cannot contain other arrays and hashes as elements. They can only contain scalars. Watch what happens when we try:
my @outer = ("Sun", "Mercury", "Venus", undef, "Mars");
my @inner = ("Earth", "Moon");
$outer[3] = @inner;
print $outer[3]; # "2"
$outer[3] is a scalar, so it demands a scalar value. When you try to assign an array value like @inner to it, @inner is evaluated in scalar context. This is the same as assigning scalar @inner, which is the length of array @inner, which is 2.
However, a scalar variable may contain a reference to any variable, including an array variable or a hash variable. This is how more complicated data structures are created in Perl.
A reference is created using a backslash.
my $colour = "Indigo";
my $scalarRef = \$colour;
Any time you would use the name of a variable, you can instead just put some braces in, and, within the braces, put a reference to a variable instead.
print $colour; # "Indigo"
print $scalarRef; # e.g. "SCALAR(0x182c180)"
print ${ $scalarRef }; # "Indigo"
As long as the result is not ambiguous, you can omit the braces too:
print $$scalarRef; # "Indigo"
If your reference is a reference to an array or hash variable, you can get data out of it using braces or using the more popular arrow operator, ->:
my @colours = ("Red", "Orange", "Yellow", "Green", "Blue");
my $arrayRef = \@colours;
print $colours[0]; # direct array access
print ${ $arrayRef }[0]; # use the reference to get to the array
print $arrayRef->[0]; # exactly the same thing
my %atomicWeights = ("Hydrogen" => 1.008, "Helium" => 4.003, "Manganese" => 54.94);
my $hashRef = \%atomicWeights;
print $atomicWeights{"Helium"}; # direct hash access
print ${ $hashRef }{"Helium"}; # use a reference to get to the hash
print $hashRef->{"Helium"}; # exactly the same thing - this is very common
Declaring a data structure
Here are four examples, but in practice the last one is the most useful.
my %owner1 = (
"name" => "Santa Claus",
"DOB" => "1882-12-25",
);
my $owner1Ref = \%owner1;
my %owner2 = (
"name" => "Mickey Mouse",
"DOB" => "1928-11-18",
);
my $owner2Ref = \%owner2;
my @owners = ( $owner1Ref, $owner2Ref );
my $ownersRef = \@owners;
my %account = (
"number" => "12345678",
"opened" => "2000-01-01",
"owners" => $ownersRef,
);
That's obviously unnecessarily laborious, because you can shorten it to:
my %owner1 = (
"name" => "Santa Claus",
"DOB" => "1882-12-25",
);
my %owner2 = (
"name" => "Mickey Mouse",
"DOB" => "1928-11-18",
);
my @owners = ( \%owner1, \%owner2 );
my %account = (
"number" => "12345678",
"opened" => "2000-01-01",
"owners" => \@owners,
);
It is also possible to declare anonymous arrays and hashes using different symbols. Use square brackets for an anonymous array and braces for an anonymous hash. The value returned in each case is a reference to the anonymous data structure in question. Watch carefully, this results in exactly the same %account as above:
# Braces denote an anonymous hash
my $owner1Ref = {
"name" => "Santa Claus",
"DOB" => "1882-12-25",
};
my $owner2Ref = {
"name" => "Mickey Mouse",
"DOB" => "1928-11-18",
};
# Square brackets denote an anonymous array
my $ownersRef = [ $owner1Ref, $owner2Ref ];
my %account = (
"number" => "12345678",
"opened" => "2000-01-01",
"owners" => $ownersRef,
);
Or, for short (and this is the form you should actually use when declaring complex data structures in-line):
my %account = (
"number" => "31415926",
"opened" => "3000-01-01",
"owners" => [
{
"name" => "Philip Fry",
"DOB" => "1974-08-06",
},
{
"name" => "Hubert Farnsworth",
"DOB" => "2841-04-09",
},
],
);
Getting information out of a data structure
Now, let's assume that you still have %account kicking around but everything else (if there was anything else) has fallen out of scope. You can print the information out by reversing the same procedure in each case. Again, here are four examples, of which the last is the most useful:
my $ownersRef = $account{"owners"};
my @owners = @{ $ownersRef };
my $owner1Ref = $owners[0];
my %owner1 = %{ $owner1Ref };
my $owner2Ref = $owners[1];
my %owner2 = %{ $owner2Ref };
print "Account #", $account{"number"}, "\n";
print "Opened on ", $account{"opened"}, "\n";
print "Joint owners:\n";
print "\t", $owner1{"name"}, " (born ", $owner1{"DOB"}, ")\n";
print "\t", $owner2{"name"}, " (born ", $owner2{"DOB"}, ")\n";
Or, for short:
my @owners = @{ $account{"owners"} };
my %owner1 = %{ $owners[0] };
my %owner2 = %{ $owners[1] };
print "Account #", $account{"number"}, "\n";
print "Opened on ", $account{"opened"}, "\n";
print "Joint owners:\n";
print "\t", $owner1{"name"}, " (born ", $owner1{"DOB"}, ")\n";
print "\t", $owner2{"name"}, " (born ", $owner2{"DOB"}, ")\n";
Or using references and the -> operator:
my $ownersRef = $account{"owners"};
my $owner1Ref = $ownersRef->[0];
my $owner2Ref = $ownersRef->[1];
print "Account #", $account{"number"}, "\n";
print "Opened on ", $account{"opened"}, "\n";
print "Joint owners:\n";
print "\t", $owner1Ref->{"name"}, " (born ", $owner1Ref->{"DOB"}, ")\n";
print "\t", $owner2Ref->{"name"}, " (born ", $owner2Ref->{"DOB"}, ")\n";
And if we completely skip all the intermediate values:
print "Account #", $account{"number"}, "\n";
print "Opened on ", $account{"opened"}, "\n";
print "Joint owners:\n";
print "\t", $account{"owners"}->[0]->{"name"}, " (born ", $account{"owners"}->[0]->{"DOB"}, ")\n";
print "\t", $account{"owners"}->[1]->{"name"}, " (born ", $account{"owners"}->[1]->{"DOB"}, ")\n";
How to shoot yourself in the foot with array references
This array has five elements:
my @array1 = (1, 2, 3, 4, 5);
print @array1; # "12345"
This array, however, has ONE element (which happens to be a reference to an anonymous, five-element array):
my @array2 = [1, 2, 3, 4, 5];
print @array2; # e.g. "ARRAY(0x182c180)"
This scalar is a reference to an anonymous, five-element array:
my $array3Ref = [1, 2, 3, 4, 5];
print $array3Ref; # e.g. "ARRAY(0x22710c0)"
print @{ $array3Ref }; # "12345"
print @$array3Ref; # "12345"
Conditionals
if ... elsif ... else ...
No surprises here, other than the spelling of elsif:
my $word = "antidisestablishmentarianism";
my $strlen = length $word;
if($strlen >= 15) {
print "'", $word, "' is a very long word";
} elsif(10 <= $strlen && $strlen < 15) {
print "'", $word, "' is a medium-length word";
} else {
print "'", $word, "' is a a short word";
}
Perl provides a shorter "statement if condition" syntax which is highly recommended for short statements:
print "'", $word, "' is actually enormous" if $strlen >= 20;
unless ... else ...
my $temperature = 20;
unless($temperature > 30) {
print $temperature, " degrees Celsius is not very hot";
} else {
print $temperature, " degrees Celsius is actually pretty hot";
}
unless blocks are generally best avoided like the plague because they are very confusing. An "unless [... else]" block can be trivially refactored into an "if [... else]" block by negating the condition [or by keeping the condition and swapping the blocks]. Mercifully, there is no elsunless keyword.
This, by comparison, is highly recommended because it is so easy to read:
print "Oh no it's too cold" unless $temperature > 15;
Ternary operator
The ternary operator ?: allows simple if statements to be embedded in a statement. The canonical use for this is singular/plural forms:
my $gain = 48;
print "You gained ", $gain, " ", ($gain == 1 ? "experience point" : "experience points"), "!";
Aside: singulars and plurals are best spelled out in full in both cases. Don't do something clever like the following, because anybody searching the codebase to replace the words "tooth" or "teeth" will never find this line:
my $lost = 1;
print "You lost ", $lost, " t", ($lost == 1 ? "oo" : "ee"), "th!";
Ternary operators may be nested:
my $eggs = 5;
print "You have ", $eggs == 0 ? "no eggs" :
$eggs == 1 ? "an egg" :
"some eggs";
if statements evaluate their conditions in scalar context. For example, if(@array) returns true if and only if @array has 1 or more elements. It doesn't matter what those elements are - they may contain undef or other false values for all we care.
Loops
There's More Than One Way To Do It.
Perl has a conventional while loop:
my $i = 0;
while($i < scalar @array) {
print $i, ": ", $array[$i];
$i++;
}
Perl also offers the until keyword:
my $i = 0;
until($i >= scalar @array) {
print $i, ": ", $array[$i];
$i++;
}
These do loops are almost equivalent to the above (a warning would be raised if @array were empty):
my $i = 0;
do {
print $i, ": ", $array[$i];
$i++;
} while ($i < scalar @array);
and
my $i = 0;
do {
print $i, ": ", $array[$i];
$i++;
} until ($i >= scalar @array);
Basic C-style for loops are available too. Notice how we put a my inside the for statement, declaring $i only for the scope of the loop:
for(my $i = 0; $i < scalar @array; $i++) {
print $i, ": ", $array[$i];
}
# $i has ceased to exist here, which is much tidier.
This kind of for loop is considered old-fashioned and should be avoided where possible. Native iteration over a list is much nicer. Note: unlike PHP, the for and foreachkeywords are synonyms. Just use whatever looks most readable:
foreach my $string ( @array ) {
print $string;
}
If you do need the indices, the
range operator .. creates an anonymous list of integers:
foreach my $i ( 0 .. $#array ) {
print $i, ": ", $array[$i];
}
You can't iterate over a hash. However, you can iterate over its keys. Use the keys built-in function to retrieve an array containing all the keys of a hash. Then use theforeach approach that we used for arrays:
foreach my $key (keys %scientists) {
print $key, ": ", $scientists{$key};
}
Since a hash has no underlying order, the keys may be returned in any order. Use the sort built-in function to sort the array of keys alphabetically beforehand:
foreach my $key (sort keys %scientists) {
print $key, ": ", $scientists{$key};
}
If you don't provide an explicit iterator, Perl uses a default iterator,
$_.
$_ is the first and friendliest of the
built-in variables:
foreach ( @array ) {
print $_;
}
If using the default iterator, and you only wish to put a single statement inside your loop, you can use the super-short loop syntax:
print $_ foreach @array;
Loop control
next and last can be used to control the progress of a loop. In most programming languages these are known as continue and break respectively. We can also optionally provide a label for any loop. By convention, labels are written in ALLCAPITALS. Having labelled the loop, next and last may target that label. This example finds primes below 100:
CANDIDATE: for my $candidate ( 2 .. 100 ) {
for my $divisor ( 2 .. sqrt $candidate ) {
next CANDIDATE if $candidate % $divisor == 0;
}
print $candidate." is prime\n";
}
Array functions
In-place array modification
We'll use @stack to demonstrate these:
my @stack = ("Fred", "Eileen", "Denise", "Charlie");
print @stack; # "FredEileenDeniseCharlie"
pop extracts and returns the final element of the array. This can be thought of as the top of the stack:
print pop @stack; # "Charlie"
print @stack; # "FredEileenDenise"
push appends extra elements to the end of the array:
push @stack, "Bob", "Alice";
print @stack; # "FredEileenDeniseBobAlice"
shift extracts and returns the first element of the array:
print shift @stack; # "Fred"
print @stack; # "EileenDeniseBobAlice"
unshift inserts new elements at the beginning of the array:
unshift @stack, "Hank", "Grace";
print @stack; # "HankGraceEileenDeniseBobAlice"
pop,
push,
shift and
unshift are all special cases of
splice.
splice removes and returns an array slice, replacing it with a different array slice:
print splice(@stack, 1, 4, "<<<", ">>>"); # "GraceEileenDeniseBob"
print @stack; # "Hank<<<>>>Alice"
Creating new arrays from old
Perl provides the following functions which act on arrays to create other arrays.
The
join function concatenates many strings into one:
my @elements = ("Antimony", "Arsenic", "Aluminum", "Selenium");
print @elements; # "AntimonyArsenicAluminumSelenium"
print "@elements"; # "Antimony Arsenic Aluminum Selenium"
print join(", ", @elements); # "Antimony, Arsenic, Aluminum, Selenium"
In list context, the
reverse function returns a list in reverse order. In scalar context,
reverse concatenates the whole list together and then reverses it as a single word.
print reverse("Hello", "World"); # "WorldHello"
print reverse("HelloWorld"); # "HelloWorld"
print scalar reverse("HelloWorld"); # "dlroWolleH"
print scalar reverse("Hello", "World"); # "dlroWolleH"
The
map function takes an array as input and applies an operation to every scalar
$_ in this array. It then constructs a new array out of the results. The operation to perform is provided in the form of a single expression inside braces:
my @capitals = ("Baton Rouge", "Indianapolis", "Columbus", "Montgomery", "Helena", "Denver", "Boise");
print join ", ", map { uc $_ } @capitals;
# "BATON ROUGE, INDIANAPOLIS, COLUMBUS, MONTGOMERY, HELENA, DENVER, BOISE"
The
grep function takes an array as input and returns a filtered array as output. The syntax is similar to
map. This time, the second argument is evaluated for each scalar
$_ in the input array. If a boolean true value is returned, the scalar is put into the output array, otherwise not.
print join ", ", grep { length $_ == 6 } @capitals;
# "Helena, Denver"
Obviously, the length of the resulting array is the number of successful matches, which means you can use grep to quickly check whether an array contains an element:
print scalar grep { $_ eq "Columbus" } @capitals; # "1"
grep and
map may be combined to form
list comprehensions, an exceptionally powerful feature conspicuously absent from many other programming languages.
By default, the
sort function returns the input array, sorted into lexical (alphabetical) order:
my @elevations = (19, 1, 2, 100, 3, 98, 100, 1056);
print join ", ", sort @elevations;
# "1, 100, 100, 1056, 19, 2, 3, 98"
However, similar to grep and map, you may supply some code of your own. Sorting is always performed using a series of comparisons between two elements. Your block receives $a and $b as inputs and should return -1 if $a is "less than" $b, 0 if they are "equal" or 1 if $a is "greater than" $b.
The cmp operator does exactly this for strings:
print join ", ", sort { $a cmp $b } @elevations;
# "1, 100, 100, 1056, 19, 2, 3, 98"
The "spaceship operator", <=>, does the same for numbers:
print join ", ", sort { $a <=> $b } @elevations;
# "1, 2, 3, 19, 98, 100, 100, 1056"
$a and $b are always scalars, but they can be references to quite complex objects which are difficult to compare. If you need more space for the comparison, you can create a separate subroutine and provide its name instead:
sub comparator {
# lots of code...
# return -1, 0 or 1
}
print join ", ", sort comparator @elevations;
You can't do this for grep or map operations.
Notice how the subroutine and block are never explicitly provided with $a and $b. Like $_, $a and $b are, in fact, global variables which are populated with a pair of values to be compared each time.
Built-in functions
By now you have seen at least a dozen built-in functions: print, sort, map, grep, keys, scalar and so on. Built-in functions are one of Perl's greatest strengths. They
- are numerous
- are very useful
- are extensively documented
- vary greatly in syntax, so check the documentation
- sometimes accept regular expressions as arguments
- sometimes accept entire blocks of code as arguments
- sometimes don't require commas between arguments
- sometimes will consume an arbitrary number of comma-separated arguments and sometimes will not
- sometimes will fill in their own arguments if too few are supplied
- generally don't require brackets around their arguments except in ambiguous circumstances
The best advice regarding built-in functions is to know that they exist. Skim the documentation for future reference. If you are carrying out a task which feels like it's low-level and common enough that it's been done many times before, the chances are that it has.
User-defined subroutines
Subroutines are declared using the
sub keyword. In contrast with built-in functions, user-defined subroutines always accept the same input: a list of scalars. That list may of course have a single element, or be empty. A single scalar is taken as a list with a single element. A hash with
N elements is taken as a list with 2
N elements.
Although the brackets are optional, subroutines should always be invoked using brackets, even when called with no arguments. This makes it clear that a subroutine call is happening.
sub hyphenate {
# Extract the first argument from the array, ignore everything else
my $word = shift @_;
# An overly clever list comprehension
$word = join "-", map { substr $word, $_, 1 } (0 .. (length $word) - 1);
return $word;
}
print hyphenate("exterminate"); # "e-x-t-e-r-m-i-n-a-t-e"
Unpacking arguments
There's More Than One Way To unpack @_, but some are superior to others.
The example subroutine left_pad below pads a string out to the required length using the supplied pad character. (The x function concatenates multiple copies of the same string in a row.) (Note: for brevity, these subroutines all lack some elementary error checking, i.e. ensuring the pad character is only 1 character, checking that the width is greater than or equal to the length of existing string, checking that all needed arguments were passed at all.)
left_pad is typically invoked as follows:
print left_pad("hello", 10, "+"); # "+++++hello"
- Some people don't unpack the arguments at all and use
@_ "live". This is ugly and discouraged:
sub left_pad {
my $newString = ($_[2] x ($_[1] - length $_[0])) . $_[0];
return $newString;
}
- Unpacking
@_ is only slightly less strongly discouraged:
sub left_pad {
my $oldString = $_[0];
my $width = $_[1];
my $padChar = $_[2];
my $newString = ($padChar x ($width - length $oldString)) . $oldString;
return $newString;
}
- Unpacking
@_ by removing data from it using shift is recommended for up to 4 arguments:
sub left_pad {
my $oldString = shift @_;
my $width = shift @_;
my $padChar = shift @_;
my $newString = ($padChar x ($width - length $oldString)) . $oldString;
return $newString;
}
If no array is provided to the shift function, then it operates on @_ implicitly. This approach is seen very commonly:
sub left_pad {
my $oldString = shift;
my $width = shift;
my $padChar = shift;
my $newString = ($padChar x ($width - length $oldString)) . $oldString;
return $newString;
}
Beyond 4 arguments it becomes hard to keep track of what is being assigned where.
- You can unpack
@_ all in one go using multiple simultaneous scalar assignment. Again, this is okay for up to 4 arguments:
sub left_pad {
my ($oldString, $width, $padChar) = @_;
my $newString = ($padChar x ($width - length $oldString)) . $oldString;
return $newString;
}
- For subroutines with large numbers of arguments or where some arguments are optional or cannot be used in combination with others, best practice is to require the user to provide a hash of arguments when calling the subroutine, and then unpack
@_ back into that hash of arguments. For this approach, our subroutine call would look a little different:
print left_pad("oldString" => "pod", "width" => 10, "padChar" => "+");
And the subroutine itself looks like this:
sub left_pad {
my %args = @_;
my $newString = ($args{"padChar"} x ($args{"width"} - length $args{"oldString"})) . $args{"oldString"};
return $newString;
}
Returning values
Like other Perl expressions, subroutine calls may display contextual behaviour. You can use the
wantarray function (which should be called
wantlist but never mind) to detect what context the subroutine is being evaluated in, and return a result appropriate to that context:
sub contextualSubroutine {
# Caller wants a list. Return a list
return ("Everest", "K2", "Etna") if wantarray;
# Caller wants a scalar. Return a scalar
return 3;
}
my @array = contextualSubroutine();
print @array; # "EverestK2Etna"
my $scalar = contextualSubroutine();
print $scalar; # "3"
System calls
Apologies if you already know the following non-Perl-related facts. Every time a process finishes on a Windows or Linux system (and, I assume, on most other systems), it concludes with a 16-bit status word. The highest 8 bits constitute a return code between 0 and 255 inclusive, with 0 conventionally representing unqualified success, and other values representing various degrees of failure. The other 8 bits are less frequently examined - they "reflect mode of failure, like signal death and core dump information".
You can exit from a Perl script with the return code of your choice (from 0 to 255) using
exit.
Perl provides More Than One Way To - in a single call - spawn a child process, pause the current script until the child process has finished, and then resume interpretation of the current script. Whichever method is used, you will find that immediately afterwards, the
built-in scalar variable $? has been populated with the status word that was returned from that child process's termination. You can get the return code by taking just the highest 8 of those 16 bits:
$? >> 8.
The
system function can be used to invoke another program with the arguments listed. The value returned by
system is the same value with which
$? is populated:
my $rc = system "perl", "anotherscript.pl", "foo", "bar", "baz";
$rc >>= 8;
print $rc; # "37"
Alternatively, you can use backticks `` to run an actual command at the command line and capture the standard output from that command. In scalar context the entire output is returned as a single string. In list context, the entire output is returned as an array of strings, each one representing a line of output.
my $text = `perl anotherscript.pl foo bar baz`;
print $text; # "foobarbaz"
This is the behaviour which would be seen if anotherscript.pl contained, for example:
use strict;
use warnings;
print @ARGV;
exit 37;
Files and file handles
A scalar variable may contain a file handle instead of a number/string/reference or undef. A file handle is essentially a reference to a specific location inside a specific file.
Use
open to turn a scalar variable into a file handle.
open must be supplied with a
mode. The mode
< indicates that we wish to open the file to read from it:
my $f = "text.txt";
my $result = open my $fh, "<", $f;
if(!$result) {
die "Couldn't open '".$f."' for reading because: ".$!;
}
If successful, open returns a true value. Otherwise, it returns false and an error message is stuffed into the built-in variable $!. As seen above, you should always check that the open operation completed successfully. This checking being rather tedious, a common idiom is:
open(my $fh, "<", $f) || die "Couldn't open '".$f."' for reading because: ".$!;
Note the need for parentheses around the open call's arguments.
To read a line of text from a filehandle, use the
readline built-in function.
readline returns a full line of text, with a line break intact at the end of it (except possibly for the final line of the file), or
undef if you've reached the end of the file.
while(1) {
my $line = readline $fh;
last unless defined $line;
# process the line...
}
To truncate that possible trailing line break, use
chomp:
chomp $line;
Note that chomp acts on $line in place. $line = chomp $line is probably not what you want.
You can also use
eof to detect that the end of the file has been reached:
while(!eof $fh) {
my $line = readline $fh;
# process $line...
}
But beware of just using while(my $line = readline $fh), because if $line turns out to be "0", the loop will terminate early. If you want to write something like that, Perl provides the <> operator which wraps up readline in a fractionally safer way. This is very commonly-seen and perfectly safe:
while(my $line = <$fh>) {
# process $line...
}
And even:
while(<$fh>) {
# process $_...
}
Writing to a file involves first opening it in a different mode. The mode > indicates that we wish to open the file to write to it. (> will clobber the content of the target file if it already exists and has content. To merely append to an existing file, use mode >>.) Then, simply provide the filehandle as a zeroth argument for the print function.
open(my $fh2, ">", $f) || die "Couldn't open '".$f."' for writing because: ".$!;
print $fh2 "The eagles have left the nest";
Notice the absence of a comma between $fh2 and the next argument.
File handles are actually closed automatically when they drop out of scope, but otherwise:
close $fh2;
close $fh;
Three filehandles exist as global constants: STDIN, STDOUT and STDERR. These are open automatically when the script starts. To read a single line of user input:
my $line = ;
To just wait for the user to hit Enter:
;
Calling <> with no filehandle reads data from STDIN, or from any files named in arguments when the Perl script was called.
As you may have gathered, print prints to STDOUT by default if no filehandle is named.
File tests
The function -e is a built-in function which tests whether the named file exists.
print "what" unless -e "/usr/bin/perl";
The function -d is a built-in function which tests whether the named file is a directory.
The function -f is a built-in function which tests whether the named file is a plain file.
These are just three of
a large class of functions of the form
-X where
X is some lower- or upper-case letter. These functions are called
file tests. Note the leading minus sign. In a Google query, the minus sign indicates to exclude results containing this search term. This makes file tests hard to Google for! Just search for "perl file test" instead.
Regular expressions
Regular expressions appear in many languages and tools other than Perl. Perl's core regular expression syntax is basically the same as everywhere else, but Perl's fullregular expression capabilities are terrifyingly complex and difficult to understand. The best advice I can give you is to avoid this complexity wherever possible.
Match operations are performed using =~ m//. In scalar context, =~ m// returns true on success, false on failure.
my $string = "Hello world";
if($string =~ m/(\w+)\s+(\w+)/) {
print "success";
}
Parentheses perform sub-matches. After a successful match operation is performed, the sub-matches get stuffed into the built-in variables $1, $2, $3, ...:
print $1; # "Hello"
print $2; # "world"
In list context, =~ m// returns $1, $2, ... as a list.
my $string = "colourless green ideas sleep furiously";
my @matches = $string =~ m/(\w+)\s+((\w+)\s+(\w+))\s+(\w+)\s+(\w+)/;
print join ", ", map { "'".$_."'" } @matches;
# prints "'colourless', 'green ideas', 'green', 'ideas', 'sleep', 'furiously'"
Substitution operations are performed using =~ s///.
my $string = "Good morning world";
$string =~ s/world/Vietnam/;
print $string; # "Good morning Vietnam"
Notice how the contents of $string have changed. You have to pass a scalar variable on the left-hand side of an =~ s/// operation. If you pass a literal string, you'll get an error.
The /g flag indicates "group match".
In scalar context, each =~ m//g call finds another match after the previous one, returning true on success, false on failure. You can access $1 and so on afterwards in the usual way. For example:
my $string = "a tonne of feathers or a tonne of bricks";
while($string =~ m/(\w+)/g) {
print "'".$1."'\n";
}
In list context, an =~ m//g call returns all of the matches at once.
my @matches = $string =~ m/(\w+)/g;
print join ", ", map { "'".$_."'" } @matches;
An =~ s///g call performs a global search/replace and returns the number of matches. Here, we replace all vowels with the letter "r".
# Try once without /g.
$string =~ s/[aeiou]/r/;
print $string; # "r tonne of feathers or a tonne of bricks"
# Once more.
$string =~ s/[aeiou]/r/;
print $string; # "r trnne of feathers or a tonne of bricks"
# And do all the rest using /g
$string =~ s/[aeiou]/r/g;
print $string, "\n"; # "r trnnr rf frrthrrs rr r trnnr rf brrcks"
The /i flag makes matches and substitutions case-insensitive.
The /x flag allows your regular expression to contain whitespace (e.g., line breaks) and comments.
"Hello world" =~ m/
(\w+) # one or more word characters
[ ] # single literal space, stored inside a character class
world # literal "world"
/x;
# returns true
Modules and packages
In Perl, modules and packages are different things.
Modules
A module is a .pm file that you can include in another Perl file (script or module). A module is a text file with exactly the same syntax as a .pl Perl script. An example module might be located at C:\foo\bar\baz\Demo\StringUtils.pm or /foo/bar/baz/Demo/StringUtils.pm, and read as follows:
use strict;
use warnings;
sub zombify {
my $word = shift @_;
$word =~ s/[aeiou]/r/g;
return $word;
}
return 1;
Because a module is executed from top to bottom when it is loaded, you need to return a true value at the end to show that it was loaded successfully.
So that the Perl interpreter can find them, directories containing Perl modules should be listed in your environment variable PERL5LIB before calling perl. List the root directory containing the modules, don't list the module directories or the modules themselves:
set PERL5LIB=C:\foo\bar\baz;%PERL5LIB%
or
export PERL5LIB=/foo/bar/baz:$PERL5LIB
Once the Perl module is created and
perl knows where to look for it, you can use the
require built-in function to search for and execute it during a Perl script. For example, calling
require Demo::StringUtils causes the Perl interpreter to search each directory listed in
PERL5LIB in turn, looking for a file called
Demo/StringUtils.pm. After the module has been executed, the subroutines that were defined there suddenly become available to the main script. Our example script might be called
main.pland read as follows:
use strict;
use warnings;
require Demo::StringUtils;
print zombify("i want brains"); # "r wrnt brrrns"
Note the use of the double colon :: as a directory separator.
Now a problem surfaces: if main.pl contains many require calls, and each of the modules so loaded contains more require calls, then it can become difficult to track down the original declaration of the zombify() subroutine. The solution to this problem is to use packages.
Packages
A
package is a namespace in which subroutines can be declared. Any subroutine you declare is implicitly declared within the current package. At the beginning of execution, you are in the
main package, but you can switch package using the
package built-in function:
use strict;
use warnings;
sub subroutine {
print "universe";
}
package Food::Potatoes;
# no collision:
sub subroutine {
print "kingedward";
}
Note the use of the double colon :: as a namespace separator.
Any time you call a subroutine, you implicitly call a subroutine which is inside the current package. Alternatively, you can explicitly provide a package. See what happens if we continue the above script:
subroutine(); # "kingedward"
main::subroutine(); # "universe"
Food::Potatoes::subroutine(); # "kingedward"
So the logical solution to the problem described above is to modify C:\foo\bar\baz\Demo\StringUtils.pm or /foo/bar/baz/Demo/StringUtils.pm to read:
use strict;
use warnings;
package Demo::StringUtils;
sub zombify {
my $word = shift @_;
$word =~ s/[aeiou]/r/g;
return $word;
}
return 1;
And modify main.pl to read:
use strict;
use warnings;
require Demo::StringUtils;
print Demo::StringUtils::zombify("i want brains"); # "r wrnt brrrns"
Now read this next bit carefully.
Packages and modules are two completely separate and distinct features of the Perl programming language. The fact that they both use the same double colon delimiter is a huge red herring. It is possible to switch packages multiple times over the course of a script or module, and it is possible to use the same package declaration in multiple locations in multiple files. Calling require Foo::Bar does not look for and load a file with a package Foo::Bar declaration somewhere inside it, nor does it necessarily load subroutines in the Foo::Bar namespace. Calling require Foo::Bar merely loads a file called Foo/Bar.pm, which need not have any kind of package declaration inside it at all, and in fact might declare package Baz::Qux and other nonsense inside it for all you know.
Likewise, a subroutine call Baz::Qux::processThis() need not necessarily have been declared inside a file named Baz/Qux.pm. It could have been declared literally anywhere.
Separating these two concepts is one of the stupidest features of Perl, and treating them as separate concepts invariably results in chaotic, maddening code. Fortunately for us, the majority of Perl programmers obey the following two laws:
- A Perl script (
.pl file) must always contain exactly zero package declarations.
- A Perl module (
.pm file) must always contain exactly one package declaration, corresponding exactly to its name and location. E.g. moduleDemo/StringUtils.pm must begin with package Demo::StringUtils.
Because of this, in practice you will find that most "packages" and "modules" produced by reliable third parties can be regarded and referred to interchangeably. However, it is important that you do not take this for granted, because one day you will meet code produced by a madman.
Object-oriented Perl
Perl is not a great language for OO programming. Perl's OO capabilities were grafted on after the fact, and this shows.
- An object is simply a reference (i.e. a scalar variable) which happens to know which class its referent belongs to. To tell a reference that its referent belongs to a class, use
bless. To find out what class a reference's referent belongs to (if any), use ref.
- A method is simply a subroutine that expects an object (or, in the case of class methods, a package name) as its first argument. Object methods are invoked using
$obj->method(); class methods are invoked using Package::Name->method().
- A class is simply a package that happens to contain methods.
A quick example makes this clearer. An example module Animal.pm containing a class Animal reads like this:
use strict;
use warnings;
package Animal;
sub eat {
# First argument is always the object to act upon.
my $self = shift @_;
foreach my $food ( @_ ) {
if($self->can_eat($food)) {
print "Eating ", $food;
} else {
print "Can't eat ", $food;
}
}
}
# For the sake of argument, assume an Animal can eat anything.
sub can_eat {
return 1;
}
return 1;
And we might make use of this class like so:
require Animal;
my $animal = {
"legs" => 4,
"colour" => "brown",
}; # $animal is an ordinary hash reference
print ref $animal; # "HASH"
bless $animal, "Animal"; # now it is an object of class "Animal"
print ref $animal; # "Animal"
Note: literally any reference can be blessed into any class. It's up to you to ensure that (1) the referent can actually be used as an instance of this class and (2) that the class in question exists and has been loaded.
You can still work with the original hash in the usual way:
print "Animal has ", $animal->{"legs"}, " leg(s)";
But you can now also call methods on the object using the same -> operator, like so:
$animal->eat("insects", "curry", "eucalyptus");
This final call is equivalent to Animal::eat($animal, "insects", "curry", "eucalyptus").
Constructors
A constructor is a class method which returns a new object. If you want one, just declare one. You can use any name you like. For class methods, the first argument passed is not an object but a class name. In this case, "Animal":
use strict;
use warnings;
package Animal;
sub new {
my $class = shift @_;
return bless { "legs" => 4, "colour" => "brown" }, $class;
}
# ...etc.
And then use it like so:
my $animal = Animal->new();
Inheritance
To create a class inheriting from a parent class, use use parent. Let's suppose we subclassed Animal with Koala, located at Koala.pm:
use strict;
use warnings;
package Koala;
# Inherit from Animal
use parent ("Animal");
# Override one method
sub can_eat {
my $self = shift @_; # Not used. You could just put "shift @_;" here
my $food = shift @_;
return $food eq "eucalyptus";
}
return 1;
And some sample code:
use strict;
use warnings;
require Koala;
my $koala = Koala->new();
$koala->eat("insects", "curry", "eucalyptus"); # eat only the eucalyptus
This final method call tries to invoke Koala::eat($koala, "insects", "curry", "eucalyptus"), but a subroutine eat() isn't defined in the Koala package. However, because Koala has a parent class Animal, the Perl interpreter tries calling Animal::eat($koala, "insects", "curry", "eucalyptus") instead, which works. Note how the class Animal was loaded automatically by Koala.pm.
Since use parent accepts a list of parent class names, Perl supports multiple inheritance, with all the benefits and horrors this entails.
BEGIN blocks
A
BEGIN block is executed as soon as
perl has finished parsing that block, even before it parses the rest of the file. It is ignored at execution time:
use strict;
use warnings;
print "This gets printed second";
BEGIN {
print "This gets printed first";
}
print "This gets printed third";
A BEGIN block is always executed first. If you create multiple BEGIN blocks (don't), they are executed in order from top to bottom as the compiler encounters them. ABEGIN block always executes first even if it is placed halfway through a script (don't do this) or at the end (or this). Do not mess with the natural order of code. PutBEGIN blocks at the beginning!
A BEGIN block is executed as soon as the block has been parsed. Once this is done, parsing resumes at the end of the BEGIN block. Only once the whole script or module has been parsed is any of the code outside of BEGIN blocks executed.
use strict;
use warnings;
print "This 'print' statement gets parsed successfully but never executed";
BEGIN {
print "This gets printed first";
}
print "This, also, is parsed successfully but never executed";
...because e4h8v3oitv8h4o8gch3o84c3 there is a huge parsing error down here.
Because they are executed at compilation time, a BEGIN block placed inside a conditional block will still be executed first, even if the conditional evaluates to false and despite the fact that the conditional has not been evaluated at all yet and in fact may never be evaluated.
if(0) {
BEGIN {
print "This will definitely get printed";
}
print "Even though this won't";
}
Do not put BEGIN blocks in conditionals! If you want to do something conditionally at compile time, you need to put the conditional inside the BEGIN block:
BEGIN {
if($condition) {
# etc.
}
}
use
Okay. Now that you understand the obtuse behaviour and semantics of packages, modules, class methods and
BEGIN blocks, I can explain the exceedingly commonly-seen
use function.
The following three statements:
use Caterpillar ("crawl", "pupate");
use Caterpillar ();
use Caterpillar;
are respectively equivalent to:
BEGIN {
require Caterpillar;
Caterpillar->import("crawl", "pupate");
}
BEGIN {
require Caterpillar;
}
BEGIN {
require Caterpillar;
Caterpillar->import();
}
- No, the three examples are not in the wrong order. It is just that Perl is dumb.
- A
use call is a disguised BEGIN block. The same warnings apply. use statements must always be placed at the top of the file, and never inside conditionals.
import() is not a built-in Perl function. It is a user-defined class method. The burden is on the programmer of the Caterpillar package to define or inheritimport(), and the method could theoretically accept anything as arguments and do anything with those arguments. use Caterpillar; could do anything. Consult the documentation of Caterpillar.pm to find out exactly what will happen.- Notice how
require Caterpillar loads a module named Caterpillar.pm, whereas Caterpillar->import() calls the import() subroutine that was defined inside the Caterpillar package. Let's hope the module and the package coincide!
Exporter
The most common way to define an
import() method is to inherit it from the
Exporter module. Exporter is a core module, and a
de facto core feature of the Perl programming language. In Exporter's implementation of
import(), the list of arguments that you pass in is interpreted as a list of subroutine names. When a subroutine is
import()ed, it becomes available in the current package as well as in its own original package.
This concept is easiest to grasp using an example. Here's what Caterpillar.pm looks like:
use strict;
use warnings;
package Caterpillar;
# Inherit from Exporter
use parent ("Exporter");
sub crawl { print "inch inch"; }
sub eat { print "chomp chomp"; }
sub pupate { print "bloop bloop"; }
our @EXPORT_OK = ("crawl", "eat");
return 1;
The package variable @EXPORT_OK should contain a list of subroutine names.
Another piece of code may then import() these subroutines by name, typically using a use statement:
use strict;
use warnings;
use Caterpillar ("crawl");
crawl(); # "inch inch"
In this case, the current package is main, so the crawl() call is actually a call to main::crawl(), which (because it was imported) maps to Caterpillar::crawl().
Note: regardless of the content of @EXPORT_OK, every method can always be called "longhand":
use strict;
use warnings;
use Caterpillar (); # no subroutines named, no import() call made
# and yet...
Caterpillar::crawl(); # "inch inch"
Caterpillar::eat(); # "chomp chomp"
Caterpillar::pupate(); # "bloop bloop"
Perl has no private methods. Customarily, a method intended for private use is named with a leading underscore or two.
@EXPORT
The Exporter module also defines a package variable called @EXPORT, which can also be populated with a list of subroutine names.
use strict;
use warnings;
package Caterpillar;
# Inherit from Exporter
use parent ("Exporter");
sub crawl { print "inch inch"; }
sub eat { print "chomp chomp"; }
sub pupate { print "bloop bloop"; }
our @EXPORT = ("crawl", "eat", "pupate");
return 1;
The subroutines named in @EXPORT are exported if import() is called with no arguments at all, which is what happens here:
use strict;
use warnings;
use Caterpillar; # calls import() with no arguments
crawl(); # "inch inch"
eat(); # "chomp chomp"
pupate(); # "bloop bloop"
But notice how we are back in a situation where, without other clues, it might not be easy to tell where crawl() was originally defined. The moral of this story is twofold:
- When creating a module which makes use of Exporter, never use
@EXPORT to export subroutines by default. Always make the user call subroutines "longhand" orimport() them explicitly (using e.g. use Caterpillar ("crawl"), which is a strong clue to look in Caterpillar.pm for the definition of crawl()).
- When
useing a module which makes use of Exporter, always explicitly name the subroutines you want to import(). If you don't want to import() any subroutines and wish to refer to them longhand, you must supply an explicit empty list: use Caterpillar ().
Miscellaneous notes
The end.
I’m highlighting this group of companies first, not because I think they’re the best (although that might well be), but because I’m truly excited about the panel they’ll be featured on at our conference next month. The panel is moderated by Facebook engineering manager Ravi Murthy– a guy who knows his way around a database — so they’ll have to answer some tough questions from one of the most-advanced and most-aggressive Hadoop and analytics tools users out there:
Platfora: Technically not a SQL product, Platfora is red-hot right now and is trying to re-imagine the world of business intelligence for a big data world. Essentially an HTML5 canvas laid atop Hadoop and an in-memory, massively parallel processing engine, the company’s software, which it unveiled in October, is designed to make analyzing data stored in Hadoop a fast and visually intuitive process.
Shark: Shark isn’t technically Hadoop, but it’s cut from the same cloth. Shark, in this case, stands for “Hive on Spark,” with Hive meaning the same thing it does to Hadoop, but with Spark being an in-memory platform designed to run parallel-processing jobs 100 times faster than MapReduce (a speed improve over traditional Hive that Shark also claims). Shark also includes APIs for turning query results into a type of data format amenable to machine learning algorithms. Both Shark and Spark are developed by the University of California, Berkeley’s AMPLab.
Stinger Initiative: Launched on Wednesday (along with a security gateway called Knox and a faster, simpler processing framework called Tez), the Stinger Initiative is a Hortonworks-led effort to make Hive faster — up too 100x — and more functional. Stinger adds more SQL analytics capabilities to Hive, but the most-important aspects are infrastructural: an optimized execution engine, a columnar file format and the ability to avoid MapReduce bottlenecks by running atop Tez.
Splice Machine: Database startup Splice Machine is also trying to get into the operational space by building its Splice SQL Engine atop the naturally distributed HBase database. Splice Machine focuses its message on transactional integrity, which is really where it separates itself from scalable NoSQL databases and analytics-focused SQL-on-Hadoop efforts. It relies on HBase’s aut0-sharding feature in order to making scaling an easy process.
