Automated Testing with

WWW::Mechanize
Copyright 2004, Andy Lester
Automated testing
philosophy
 Why test?
Humans make mistakes
Bugs cost time & money
Bugs are demoralizing
Testing shows that your code works
At least in the cases that you're testing
 Why automated
testing?
Humans hate testing. Computers don't.
Instantly verify changes haven't broken code
Regularly verify all code
More tests don’t make the computer grumble
 Automated testing
philosophy
You are human
Testing is an investment
Test for now
Test for the future
Test constantly
 You are human
You're good, but not that good
Your coworkers are human, too
Machines work, humans think
Machines are good at repetitive tasks.
A cron job will never blow off its tasks because
they're boring.
 Test for the future
Testing will detect breakage in the future
(at least if you actually run the tests)
99.99% of the time, the test runs fine
This is not a waste
Proves that your code still works as you wrote it.
Eliminate future stress and frustration!
 How Perl handles
testing
• Test::Harness runs your
.t files

• The .t files run tests, and

reports results to
standard output.

• T::H analyzes the output

and gives the thumbs up
or thumbs down.
 What's a .t file?
A .t file is just a Perl file
Test::Harness looks for the .t extension by
convention
Each one runs one or more tests
Results of each test are reported to standard
output for parsing
The test results format
Simple text-based output
One line per test says the test passes or fails.
The following shows four tests run, but the third
one failed.
1..4
ok 1
ok 2
not ok 3
ok 4
 Running prove
prove is a wrapper around Test::Harness
Made to make testing simple and fast
Running prove
t/check_title....ok
All tests successful.
Files=1, Tests=3, 1 wallclock secs
(0.39 cusr + 0.07 csys = 0.46 CPU)
Starting with
Test::More
 Starting with
Test::More
Create a .t file
Use the Test::Simple module
Exercise part of your code
Pass a true/false result to ok()
ok() reports the result
ok() keeps track of numbering
 Test::More
Helper functions for common tests
All tests can be done with ok(), but why?
Everything is a wrapper around ok()
Helper functions make doing the right thing trivial,
so you're less likely to take shortcuts.
Use function messages
Most helper functions take a message
Provide meaningful messages
Messages document your intent
Messages help you find failed tests
 diag()
Adds comments to your output
Printed with a “#” in front
Ignored by Test::Harness and prove
Sprinkle them liberally -- They'll make detective
work easier
diag( "Checking for foo" );

# Checking for foo

Checking for correct
values
Many tests are "did I get this value"
my $stooges = stooge_count();
ok( $stooges == 3, "The holy trinity" );

Bummer if it fails:
not ok 1 - The holy trinity
# Failed test (stooges.t at line 8)

What was it if not 3?

 is() to the rescue
The is() function tells you what's wrong
my $stooges = stooge_count();
is( $stooges, 3, "Check stooge count" );

not ok 1 - Check stooge count

# Failed test (stooges.t at line 8)
# got: '4'
# expected: '3'

Aha! We probably counted Shemp!

 isnt()
Sometimes you want to avoid a certain value...
my $fave = get_favorite_stooge();
isnt( $fave, "Larry", "Nobody liked Larry" );

not ok 4 - Nobody likes Larry

# Failed test (stooges.t at line 18)
# 'Larry'
# ne
# 'Larry'
 like() is like is()
like() is just is() with regexes
my $str = "I always liked Shemp";
like( $str, qr/Larry|Moe|Curly/,
"Need at least one real stooge" );
not ok 1 - Need at least one real stooge
# Failed test (stooges.t at line 14)
# 'I always liked Shemp'
# doesn't match '(?-xism:Larry|Moe|Curly)'

unlike() is the negation of like()

 cmp_ok()
When ne and eq won't do
Specifies any binary operator

my @stooges = get_stooges();
cmp_ok( scalar @stooges, ">=", 3,
"We'll allow Shemp" );
Return values of ok()
ok() et al return booleans
Use to know at runtime a test's status
die if a critical failure means all others will fail, oo

my $bot = new WWW::Mechanize();

isa_ok( $bot, "Got an object OK" ) or
die "Couldn't create a bot";

$bot->dostuff(); # continue testing

 isa_ok()
Checks that your object is the correct type, and
isn't undef

use_ok( 'HTML::Lint' );
my $linter = new HTML::Lint;
isa_ok( $linter, 'HTML::Lint' );

ok 1 - use HTML::Lint;
ok 2 - The object isa HTML::Lint
 TODO blocks
When you don't have the code done yet, you can
still have tests.
T::H doesn't count these as failures
You'll be told if they unexpectedly succeed

TODO: {
local $TODO = "read_minds() not done";

ok( read_minds(), "Reading mind" );

}
Writing more and
more tests
 Get extravagant
Resources you have in abundance
Machine cycles
Disk space
Resources you don't
Time
Hair
 Make many test files
Disk space is cheap
More granular test files make it easier to write
more accurate tests
The importance of the
plan
Note the plan
use Test::More tests=>2;

You could specify

use Test::More 'no_plan';

Specify the tests if you can

What if the .t stops unexpectedly?
Having the correct number of tests is a test in itself
 Expect failure
Document what you expect
Explain why you're testing it
Make it easy to grep for failure

is( $nrows, 100, "Should have 99 rows, plus the one we added" );
Make autonomous test
files
Don't make test files rely on each other
Tests should clean up after themselves
You can't rely on the order that test files get run.*

*But if you have to, use Test::Manifest

 Test unrealistically
Many errors occur at the extremes
quantities of 0, 1, negatives
empty strings, and of maximum length
Don't think "that can never happen"
It will.
You won't like it.
Prove that your code can handle it.
 Test::HTML::Lint
Checks structural integrity of HTML
html_ok( $generated_html );

# Errors:
# (184:42) <IMG> tag has no HEIGHT and WIDTH attributes.
 WWW::Mechanize
Subclass of LWP::UserAgent
Acts like a web browser
Has a back() function
Handles forms and fields for you
Many handy convenience functions
 WWW::Mechanize
Wraps up web and form pages
Handles proxies, agent strings, etc
my $mech = new WWW::Mechanize;
$mech->agent( "TestBot/1.0" );
$mech->get( "http:://foo.com" );
if ( $mech->success ) {
print $mech->content;
} else {
print $mech->response->error_as_HTML;
}
 Get a page
Get a page with Mech
use WWW::Mechanize;
my $m = WWW::Mechanize->new();
$m->get( "http://plainblack.com/" );
print $m->title, "\n";

$ ./fetch.pl
Plain Black, makers of WebGUI - Home
 Accessing links
links() returns array of all links
Each link is an object containing:
The text of the link
The raw URL
The absolute URL
Contents of the NAME attribute
Links also include <FRAME> tags.
 Check title and links
Check title & links
my $m = WWW::Mechanize->new();
$m->get( "http://www.plainblack.com/" );
print "Title: ", $m->title, "\n";
print $_->text, " --> ", $_->url, "\n"
for $m->links;

$ ./links.pl
Title: Plain Black, makers of WebGUI - Home
Plain Black --> http://www.plainblack.com/home
Store --> http://www.plainblack.com/store
WebGUI Users --> http://www.plainblack.com/wg
Write a test file
 title.t
Make sure your title is correct
use Test::More tests => 2;
use WWW::Mechanize;

my $m = WWW::Mechanize->new() or die;
$m->get( "http://plainblack.com/" );
ok( $m->success, 'Fetched OK' );
like( $m->title, qr/Plain Black/ );
 Running title.t
Running title.t
$ prove title.t
title....ok
All tests successful.
Files=1, Tests=2, 2 wallclock secs
( 0.70 cusr + 0.08 csys = 0.78 CPU)

$ prove -v title.t
title....1..2
ok 1 - Fetched OK
ok 2 - Right title
All tests successful.
Files=1, Tests=2, 2 wallclock secs
( 0.60 cusr + 0.12 csys = 0.72 CPU)
 Follow a link
The follow_link() method takes a string to match,
or a number of the link to follow.

$a->follow_link( text => 'JT' );

finds first link that matches "JT" exactly.
Mech then fetches the target page for you.
Current page stored in history.
Follow a link -- other
methods
follow_link() parameters
text => "string"
text_regex => qr/regex/
url => "string"
url_regex => qr/regex/
tag => "string"
n => number
# and many many more

Can combine n=>number with text or url

# 3rd download link
follow_link( text=>"download", n=>3 );
Follow a link -- Testing
Testing a followed link
$a->get( "http://plainblack.com/" );
ok( $a->success, "Got front page" );

$a->follow_link( text_regex => qr/Contact/ );

ok( $a->success, "Followed Contact page" );
is( $a->title, qr/Contact Us/ );
# Make sure we got the right page
 Using back()
The back() method goes back in your history, just
like a back button in a browser.
Doesn't actually refetch the page.
$a->follow_link( text=>"link" );
# do some stuff
$a->back()
 Checking links on a
page
Fetch the page
Get the list of links on the page
Call get() on each link target
get() takes care of the relative URLs!
Remember each link in links() gives:
The link
The text of the link
Contents of the NAME attribute
Link checking in action
Check all your links
$m->get( "http://www.plainblack.com/" );
ok( $m->success, 'Fetched OK' );

for my $link ( $m->links ) {

$m->get( $link->url );
ok( $m->success, $link->url . " fetched" );

$a->back();
ok( $m->success, "Went back" );
}
 Handling forms
Parsed them automatically on get()
Each form represented by a HTML::Form object
Forms specified by number or name
Fields specified by name
 Fill out a form
# Get http://plainblack.com/search, then...
$a->form_number(1);
$a->field( "as_q" => "Perl" );
$a->submit;

like( $a->content,
qr/Results \d+ - \d+ of about \d+/,
"Got search results back" );
 Submit the form
Two ways to submit a form
$a->submit(); # No "click"
$a->click("name"); # "click" the "name" button

Use $a->submit() unless you need to specify the

button.
Use $a->click("name",$x,$y) to simulate clicking an
imagemap.
 HTML::Lint
Finds errors in HTML
Types of errors to check can be specified
Structure
Helpers
Fluff
Comes with weblint, standalone utility
HTML::Lint error types
Structure
Ex: Unclosed <TABLE> tags, incorrect values for
attributes, and repeated attributes.
Helper
Ex: Missing HEIGHT and WIDTH in an IMG tag.
Fluff
Unnecessary items that don't hurt your page.
Usually an unknown attribute on a tag.
 weblint
Wrapper around HTML::Lint
Takes any combination of URLs and files

$ weblint http://www.plainblack.com
http://www.plainblack.com (184:42) <IMG> tag has no HEIGHT and
WIDTH attributes.
 Using HTML::Lint
Create an HTML::Lint object
Set your parms
Feed it text or a filename
Handle the errors programmatically
 Test::HTML::Lint
Wraps HTML::Lint in Test::* framework
Creates the HTML::Lint object for you
Can take an object you create
Allows you customizations
 Test::HTML::Lint in
action
Validate HTML on the site
use Test::More tests=>2;
use Test::HTML::Lint;
use WWW::Mechanize;

my $m = WWW::Mechanize->new;
$m->get( "http://www.plainblack.com/" );
ok( $m-success, "Fetched OK" );
html_ok( $m->content, "Home page valid HTML" );
 Test::HTML::Lint in
action
Running the HTML checker
ok 1 - Fetched OK
not ok 2 - Home page valid HTML
# Failed test (foo.pl at line 8)
# Errors: Home page valid HTML
# (30:1) Unknown attribute "rightmargin" for tag <body>
# (30:1) Unknown attribute "bottommargin" for tag <body>
# Looks like you failed 1 tests of 2.

Note that html_ok() dumps your errors for you.

 Custom HTML::Lint
testing
Make my checks less stringent
my $m = WWW::Mechanize->new;
$m->get( "http://www.plainblack.com/" );

my $lint = HTML::Lint->new;
$lint->only_types( HTML::Lint::STRUCTURE );
html_ok( $lint, $m->content );

HTML passes clean now!

Test::HTML::Content
Checks for content in the HTML
link_ok($HTML, "http://www.perl.com",
"We link to Perl");

no_link($HTML, "http://www.pearl.com",
"We have no embarassing typos");

link_ok($HTML, qr"http://[a-z]+\.perl.com",
"We have a link to perl.com");
Testing more than just
your website
Testing more than your
code
As long as you can call ok(), you can test anything
using Perl.
Project-wide rules
Database validity
Web pages
Adequate disk space
Anything else at all
 Where to Get More
Information
• http://qa.perl.org

• Test::Tutorial

• Test::WWW::Mechanize

• Perl & LWP