NAME

Test::Deep - Extremely flexible deep comparison

SYNOPSIS

use Test::More tests => $Num_Tests;
use Test::Deep;

cmp_deeply($actual, $expected, "equal");

cmp_deeply(\@result, set('a', 'b', {key => [1, 2]});
cmp_bag(\@result, ['a', 'b', {key => [1, 2]}]);

cmp_deeply($object, methods(name => "John", phone => "55378008");

cmp_deeply($structure, [$hash1, $hash2, ignore()]);

cmp_deeply($object, noclass({value => 5}));

DESCRIPTION

If you don't know anything about automated testing in Perl then you should probably read about Test::Simple and Test::More before preceeding. Test::Deep uses the Test::Builder framework.

Test::Deep gives you very flexible ways to check that the result you got is the result you were expecting. At it's simplest it compares two structures by going through each level, ensuring that the values match, that arrays and hashes have the same elements and that references are blessed into the correct class. It also check circular data structures.

Where it becomes more interesting is in allowing you to do something besides simple comparisons at any level of the structure comparison. So you can check that a function returned an array containing a string matching a pattern and an object with certain properties like this

cmp_deeply(
  $got,
  [
    re('^OK '),
    methods(filename => 'file.txt', handle => isa("IO::Handle"))
  ]
);

This will check that all of the following are true

$got is an array reference
$got->[0] =~ /^OK /
$got->[1]->filename eq "file.txt"
$got->[1]->handle->isa("IO::Handle")

If, for example, the wrong file had been read, causing the second condtion to fail, you get some diagnostics that look something like

Compared $data->[1]->filename()
   got : 'file.xml'
expect : 'file.txt'

TERMINOLOGY

cmp_deeply() takes 3 arguments, from now on when you $got or $expected, it means the first and second arguments resectively. As cmp_deeply() does it's work it descends into $got and $expected so that at any given time it is thinking about 2 values, one from $got and one from $exptected, I will call these $got_v and $expected_v.

COMPARISON FUNCTIONS

$ok = cmp_deeply($got, $expected, $name)

$got is the result to be checked. $expected is the structure against which $got will be check. $name is the test name.

This is the main comparison function, the others are just wrappers around this. Without any special comparisons, it will descend into $expected, following every reference and comparing $expected_v to $got_v (using eq) at the same position. If $expected_v is a special comparison in then it may do something else besides a string comparison, exactly what it does depends on the spceial comparison (see below for the spceial comparisons).

<B>Note: If $got_v is a ever a special comaparison you will get a fatal error.

$ok = cmp_bag(\@got, \@bag, $name)

Is shorthand for cmp_deeply(\@got, bag(@bag), $name)

$ok = cmp_set(\@got, \@set, $name)

Is shorthand for cmp_deeply(\@got, set(@set), $name)

$ok = cmp_methods(\@got, \@methods, $name)

Is shorthand for cmp_deeply(\@got, methods(@methods), $name)

$ok = eq_deeply($got, $expected, $name)

This is the same as cmp_deeply() but it never outputs any diagnostics, it just returns true or false.

VARIABLES

$Test::Deep::Snobby controls whether the blessed class $got_v and $expected_v should be considered when comparing them. A true value means $that class is important, false means class will be ignored.

It is probably better to control this using the useclass() and noclass() special comparisons detailed below.

WHAT ARE SPECIAL COMPARISONS?

A special comparison (SC) is simply an object that inherits from Test::Deep::Cmp. Whenever $expected_v is an SC then instead of checking $got_v eq $expected_v, we pass control over to the SC and let it do it's thing.

Test::Deep exports lots of SC constructors, to make it easy for you to use them in $expected. For example is re("hello") is just a handy way of creating a Test::Deep::Regexp object that will match any strig containing "hello". So

cmp_deeply([ 'a', 'b', 'hello world'], ['a', 'b', re("^hello")]);

will check 'a' eq 'a', 'b' eq 'b' but when it comes to comparing $got_v = 'hello world' and $got_v = re("^hello") it will see that $expected_v is an SC and so will do something like $expected_v-descend($got_v)>.

SPECIAL COMPARISONS PROVIDED

methods(%hash)

%hash is a hash of method call => expected value pairs.

This lets you call methods on an object and check the result of each call. The methods will be called in the order supplied. If you want to pass arguments to the method you should wrap the method name and arguments in an array reference.

cmp_deeply($obj, methods(name => 'John', ['favourite, 'food'] => 'taco');

is the equivalent of checking that

$obj->name eq 'John'
$obj->favourite('food') eq 'taco'

NOTE You need to be careful if the methods you call have side effects like changing the object or other objects in the structure. The order of tests is not fixed and although the methods inside a methods() SC are called in the order you supply, there is no way to know which of 2 methods() tests will be carried out first, so if $expected is

{
  manager => methods(@methods1),
  coder => methods(@methods2)
}

Then there is no way to know if manager or coder will be tested first. If the methods you are testing depend on and later global variables or if manager and coder are the same object then you may run into problems.

shallow($thing)

$thing is a ref.

This prevents cmp_deeply() from looking inside $thing. It allows you to check that $got_v and $thing are references to the same variable. So

my @a = @b = (1, 2, 3);
cmp_deeply(\@a, \@b);

will pass because @a and @b have the same elements however

cmp_deeply(\@a, shallow(\@b))

will fail because although \@a and \@b are references to different arrays.

ignore()

This makes cmp_deeply() skip tests on $got_v, matter what value $got_v has, it will be the right value. This is useful if some part of the structure you are testing is complicated and already tested elsewhere is, or is unpredictable.

cmp_deeply($got, {
  name => 'John',
  random => ignore(),
  address => ['5 A street', 'a town', 'a country'],
})

is the equivalent of checking

$got->{name} eq 'John'
exists $got->{random}
cmp_deeply($got->{address}, ['5 A street', 'a town', 'a country']);

noclass($thing)

$thing is a structure to be compared against.

This turns off comparison of classes when comparing against blessed $references inside thing. Once cmp_deeply() comes back out of $thing it will revert to it's previous setting for checking class, so if it was comparing class before entering $thing, it will start comparing class again.

Essentially it does a local $Test::Deeply::Snobby = 0.

This can be useful when you want to check that objects have been constructed correctly but you don't want to have to write lots of blesses. If $people has an array of People objects then

cmp_deeply($people, noclass([
  bless {name => 'John', phone => '555-5555'}, "Person",
  bless {name => 'Anne', phone => '444-4444'}, "Person",
]));

can be replaced with

cmp_deeply($people, noclass([
  {name => 'John', phone => '555-5555'},
  {name => 'Anne', phone => '444-4444'}
]));

if you're sure all the people were correctly blessed. However, this is testing so you should never assume anything. You could use a map to bless all those hashes or you could do a second test like

cmp_deeply($people, all_values(isa("Person"));

to make sure that all the elements of @$people have been blessed correctly.

useclass($thing)

This turns back on the class comparison inside a noclass().

cmp_deeply($got, noclass([useclass(bless {}, "Person")]))

In this example the class of $got is ignored but the class of $got->[0] is checked.

Essentially it does a local $Test::Deeply::Snobby = 1.

re($regexp)

$regexp is either a regular expression reference produced with qr/.../ or a $string which will be used to construct a regular expression.

This simply compares $got_v with the regular expression provided.

cmp_deeply($got, [ re(/ferg/i) ])

is the equivalent of

$got->[0] =~ /ferg/i

bag(@elements)

@elements is an array of elements.

This does a bag comparison, that is, it compares two arrays but ignores the order of the elements so

cmp_deeply([1, 2, 2], bag(2, 2, 1))

will be a pass.

The object returned by bag() has an add() method.

my $bag = bag(1, 2, 3);
$bag->add(2, 3, 4);

will result in a bag containing 1, 2, 2, 3, 3, 4.

NOTE If you use special comparisons within a bag or set comparison there is a danger that a test will fail when it should have passed. It can only happen if two or more special comparisons in the bag are competing to match elements. Consider this comparison

cmp_deeply(['furry', 'furball'], bag(re("^fur"), re("furb")))

There are two things that could happen, hopefully re("^fur") is paired with "furry" and re("^furb") is paired with "furb" and everything is fine but it could happen that re("^fur") is paired with "furball" and then re("^furb") cannot find a match and so the test fails. Examples of other competing comparsions are bag(1, 2, 2) vs set(1, 2) and methods(m1 = "v1", m2 => "v2")> vs methods(m1 = "v1")>

This problem is could be solved by using a slower and more complicated algorithm for set and bag matching. Something for the future...

set(@elements)

@elements is an array of elements.

This does a set comparison, that is, it compares two arrays but ignores the order of the elements and it ignores duplicate elements, so

cmp_deeply([1, 2, 2, 3], bag(3, 2, 1, 1))

will be a pass.

The object returned by set() has an add() method.

my $set = set(1, 2, 3);
$set->add(4, 5, 6);

will result in a set containing 1, 2, 3, 4, 5, 5.

NOTE See the NOTE on the bag() comparison for some dangers in using special comparisons inside set()

all(@expecteds)

@expecteds is an array of expected structures.

This allows you compare data against multiple expected results and make sure each of them matches.

cmp_deeply($got, all(isa("Person"), methods(name => 'John')))

is equivalent to

$got->isa("Person")
$got->name eq 'John'

If either test fails then the whole thing is considered a fail. This is a short-circuit test, the testing is stopped after the first failure, although in the future it may complete all tests so that diagnostics can be output for all failures. When reporting failure, the parts are counted from 1.

Thanks to the magic of overloading, you can write

all(isa("Person"), methods(name => 'John'), re("^wi"))

as

isa("Person") & methods(name => 'John') | re("^wi")

Note single | not double as || cannot be overloaded. This will only work when there is a special comparison involved. If you write

"john" | "anne" | "robert"

it's the same as

"{onort"

which is presumably not what you wanted. This is because Perl |s them together as strings before Test::Deep gets a chance to do any overload tricks.

any(@expecteds)

@expecteds is an array of expected structures.

This can be used to compare data against multiple expected results and make sure that at least one of them matches. This is a short-circuit test so if a test passes then none of the tests after that will be attempted.

You can also use overloading with & similarly to all().

isa($class)

$class is a class name.

This uses UNIVERSAL::isa() to check that $got_v is blessed into the class $class.

array_each($thing)

$thing is a structure to be compared against.

If the data we are checking is an array then each element of it will be compared to $thing. If it is a hash then all values will be compared against it. This is useful when you have an array of similar things, for example objects of a known type and you don't want to have to repeat the same test for each one.

my $common_tests = all(
   isa("MyFile"),
   methods(
     handle => isa("IO::Handle")
     filename => re("^/home/ted/tmp"),
  )
);

cmp_deeply($got, array_each($common_tests));

is something like

foreach my $got_v (@$got) {
  cmp_deeply($got_v, $common_tests)
}

so it will check that each of the objects in @$got is a MyFile and gives satisfactory results for it's methods.

You could go further, if for example there were 3 files and you knew the size of each one you could do this

cmp_deeply(
  $got,
  all(
    array_each($common_tests),
    [
      methods(size => 1000),
      methods(size => 200),
      methods(size => 20)
    ]
  )
)

which would first check that each element of @$got passes the $common_tests test and then it checks the elements one by one with $got_v->methods.

str($string)

$string is a string.

This will check $got_v eq $string, even if $got_v is a ref. It is useful for checking the stringified value of an overloaded reference.

num($number, $mode)

$number is a number.

This will check $got_v == $number, even if $got_v is a ref. It is useful for checking the numerical value of an overloaded reference.

Note in Perl "12blah" == 12 because Perl will be smart and convert "12blah" into 12. You may not want this. If you have version 1.10 or higher of Scalar::Util then you can pass in $mode as a second argument to number(). $mode can be "loose" or "strict". In strict mode, "12blah" will not be considered equal to 12.

If you cannot install the correct version of Scalar::Util then you can usually just use the string() comparison to be more strict. This will work fine for almost all situations, however it will not work when <$got_v> is an overloaded value who's string and numerical values differ.

bool($value)

$value is anything you like but it's probably best to use 0 or 1

This will check that $got_v and $value have the same truth value, that is they will give the same result when used in boolean context, like in an if() statement.

Example

Your function to be tested returns an array with three elements, the of 3 objects hashes, each hash represents a file o

You've written a module to handle people and their film interests. Say you have a function that returns an array of people from a query, each person is a hash with 2 keys: Name and Age and the array is sorted by Name. You can do

cmp_deeply(
$result,
[
  {Name => 'Anne', Age => 26},
  {Name => "Bill", Age => 47}
  {Name => 'John', Age => 25},
]
);

Soon after, your query function changes and all the results now have an ID field. Now your test is failing again because you left out ID from each of the hashes. The problem is that the IDs are generated by the database and you have no way of knowing what each person's ID is. With Test::Deep you can change your query to

cmp_deeply(
$result,
bag(
  {Name => 'John', Age => 25, ID => ignore()},
  {Name => 'Anne', Age => 26, ID => ignore()},
  {Name => "Bill", Age => 47, ID => ignore()}
)
);

Finally person gets even more complicated and includes a new field called Movies, this is a list of movies that the person has seen recently, again these movies could come back in any order so we need a bag comparions again, giving us something like

cmp_deeply(
$result,
  bag(
    {Name => 'John', Age => 25, ID => ignore(), Movies => bag(...)},
    {Name => 'Anne', Age => 26, ID => ignore(), Movies => bag(...)},
    {Name => "Bill", Age => 47, ID => ignore(), Movies => bag(...)}
  )
);

BUGS

There is a buglet in set and bag compare to do with competing SCs, it's documented in the bag() docs.

Putting bags or sets inside bags or sets is currently fragile.

SEE ALSO

Test::More

AUTHOR

Fergal Daly <fergal@esatclear.ie>, with thanks to Michael G Schwern for Test::More's is_deeply function which inspired this.

COPYRIGHT

Copyright 2003 by Fergal Daly <fergal@esatclear.ie>.

This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.

See http://www.perl.com/perl/misc/Artistic.html