Remove old hack for Data::Dumper that should no longer be required

[catagits/Catalyst-Manual.git] / lib / Catalyst / Manual / Tutorial / Authentication.pod

=head1 NAME

Catalyst::Manual::Tutorial::Authentication - Catalyst Tutorial - Chapter 5: Authentication


=head1 OVERVIEW

This is B<Chapter 5 of 10> for the Catalyst tutorial.

L<Tutorial Overview|Catalyst::Manual::Tutorial>

=over 4

=item 1

L<Introduction|Catalyst::Manual::Tutorial::Intro>

=item 2

L<Catalyst Basics|Catalyst::Manual::Tutorial::CatalystBasics>

=item 3

L<More Catalyst Basics|Catalyst::Manual::Tutorial::MoreCatalystBasics>

=item 4

L<Basic CRUD|Catalyst::Manual::Tutorial::BasicCRUD>

=item 5

B<Authentication>

=item 6

L<Authorization|Catalyst::Manual::Tutorial::Authorization>

=item 7

L<Debugging|Catalyst::Manual::Tutorial::Debugging>

=item 8

L<Testing|Catalyst::Manual::Tutorial::Testing>

=item 9

L<Advanced CRUD|Catalyst::Manual::Tutorial::AdvancedCRUD>

=item 10

L<Appendices|Catalyst::Manual::Tutorial::Appendices>

=back


=head1 DESCRIPTION

Now that we finally have a simple yet functional application, we can
focus on providing authentication (with authorization coming next in
Chapter 6).

This chapter of the tutorial is divided into two main sections: 1) basic,
cleartext authentication and 2) hash-based authentication.

You can checkout the source code for this example from the catalyst
subversion repository as per the instructions in
L<Catalyst::Manual::Tutorial::Intro|Catalyst::Manual::Tutorial::Intro>.


=head1 BASIC AUTHENTICATION

This section explores how to add authentication logic to a Catalyst
application.


=head2 Add Users and Roles to the Database

First, we add both user and role information to the database (we will
add the role information here although it will not be used until the
authorization section, Chapter 6).  Create a new SQL script file by opening
C<myapp02.sql> in your editor and insert:

    --
    -- Add users and roles tables, along with a many-to-many join table
    --
    CREATE TABLE users (
            id            INTEGER PRIMARY KEY,
            username      TEXT,
            password      TEXT,
            email_address TEXT,
            first_name    TEXT,
            last_name     TEXT,
            active        INTEGER
    );
    CREATE TABLE roles (
            id   INTEGER PRIMARY KEY,
            role TEXT
    );
    CREATE TABLE user_roles (
            user_id INTEGER,
            role_id INTEGER,
            PRIMARY KEY (user_id, role_id)
    );
    --
    -- Load up some initial test data
    --
    INSERT INTO users VALUES (1, 'test01', 'mypass', 't01@na.com', 'Joe',  'Blow', 1);
    INSERT INTO users VALUES (2, 'test02', 'mypass', 't02@na.com', 'Jane', 'Doe',  1);
    INSERT INTO users VALUES (3, 'test03', 'mypass', 't03@na.com', 'No',   'Go',   0);
    INSERT INTO roles VALUES (1, 'user');
    INSERT INTO roles VALUES (2, 'admin');
    INSERT INTO user_roles VALUES (1, 1);
    INSERT INTO user_roles VALUES (1, 2);
    INSERT INTO user_roles VALUES (2, 1);
    INSERT INTO user_roles VALUES (3, 1);

Then load this into the C<myapp.db> database with the following command:

    $ sqlite3 myapp.db < myapp02.sql


=head2 Add User and Role Information to DBIC Schema

Although we could manually edit the DBIC schema information to include
the new tables added in the previous step, let's use the C<create=static>
option on the DBIC model helper to do most of the work for us:

    $ script/myapp_create.pl model DB DBIC::Schema MyApp::Schema \
        create=static components=TimeStamp dbi:SQLite:myapp.db
     exists "/root/dev/MyApp/script/../lib/MyApp/Model"
     exists "/root/dev/MyApp/script/../t"
    Dumping manual schema for MyApp::Schema to directory /root/dev/MyApp/script/../lib ...
    Schema dump completed.
     exists "/root/dev/MyApp/script/../lib/MyApp/Model/DB.pm"
    $
    $ ls lib/MyApp/Schema/Result
    Authors.pm  BookAuthors.pm  Books.pm  Roles.pm  UserRoles.pm  Users.pm

Notice how the helper has added three new table-specific result source
files to the C<lib/MyApp/Schema/Result> directory.  And, more
importantly, even if there were changes to the existing result source
files, those changes would have only been written above the C<# DO NOT
MODIFY THIS OR ANYTHING ABOVE!> comment and your hand-edited
enhancements would have been preserved.

Speaking of "hand-edit ted enhancements," we should now add
relationship information to the three new result source files.  Edit
each of these files and add the following information between the C<#
DO NOT MODIFY THIS OR ANYTHING ABOVE!> comment and the closing C<1;>:

C<lib/MyApp/Schema/Result/Users.pm>:

    #
    # Set relationships:
    #
    
    # has_many():
    #   args:
    #     1) Name of relationship, DBIC will create accessor with this name
    #     2) Name of the model class referenced by this relationship
    #     3) Column name in *foreign* table (aka, foreign key in peer table)
    __PACKAGE__->has_many(map_user_role => 'MyApp::Schema::Result::UserRoles', 'user_id');
    
    # many_to_many():
    #   args:
    #     1) Name of relationship, DBIC will create accessor with this name
    #     2) Name of has_many() relationship this many_to_many() is shortcut for
    #     3) Name of belongs_to() relationship in model class of has_many() above
    #   You must already have the has_many() defined to use a many_to_many().
    __PACKAGE__->many_to_many(roles => 'map_user_role', 'role');


C<lib/MyApp/Schema/Result/Roles.pm>:

    #
    # Set relationships:
    #
    
    # has_many():
    #   args:
    #     1) Name of relationship, DBIC will create accessor with this name
    #     2) Name of the model class referenced by this relationship
    #     3) Column name in *foreign* table (aka, foreign key in peer table)
    __PACKAGE__->has_many(map_user_role => 'MyApp::Schema::Result::UserRoles', 'role_id');


C<lib/MyApp/Schema/Result/UserRoles.pm>:

    #
    # Set relationships:
    #
    
    # belongs_to():
    #   args:
    #     1) Name of relationship, DBIC will create accessor with this name
    #     2) Name of the model class referenced by this relationship
    #     3) Column name in *this* table
    __PACKAGE__->belongs_to(user => 'MyApp::Schema::Result::Users', 'user_id');
    
    # belongs_to():
    #   args:
    #     1) Name of relationship, DBIC will create accessor with this name
    #     2) Name of the model class referenced by this relationship
    #     3) Column name in *this* table
    __PACKAGE__->belongs_to(role => 'MyApp::Schema::Result::Roles', 'role_id');


The code for these three sets of updates is obviously very similar to
the edits we made to the C<Books>, C<Authors>, and C<BookAuthors>
classes created in Chapter 3.

Note that we do not need to make any change to the
C<lib/MyApp/Schema.pm> schema file.  It simply tells DBIC to load all
of the Result Class and ResultSet Class files it finds in below the
C<lib/MyApp/Schema> directory, so it will automatically pick up our
new table information.


=head2 Sanity-Check Reload of Development Server

We aren't ready to try out the authentication just yet; we only want
to do a quick check to be sure our model loads correctly.  Press
C<Ctrl-C> to kill the previous server instance (if it's still running)
and restart it:

    $ script/myapp_server.pl

Look for the three new model objects in the startup debug output:

    ...
     .-------------------------------------------------------------------+----------.
    | Class                                                             | Type     |
    +-------------------------------------------------------------------+----------+
    | MyApp::Controller::Books                                          | instance |
    | MyApp::Controller::Root                                           | instance |
    | MyApp::Model::DB                                                  | instance |
    | MyApp::Model::DB::Author                                          | class    |
    | MyApp::Model::DB::Books                                           | class    |
    | MyApp::Model::DB::BookAuthors                                     | class    |
    | MyApp::Model::DB::Roles                                           | class    |
    | MyApp::Model::DB::Users                                           | class    |
    | MyApp::Model::DB::UserRoles                                       | class    |
    | MyApp::View::TT                                                   | instance |
    '-------------------------------------------------------------------+----------'
    ...

Again, notice that your "Result Class" classes have been "re-loaded"
by Catalyst under C<MyApp::Model>.


=head2 Include Authentication and Session Plugins

Edit C<lib/MyApp.pm> and update it as follows (everything below
C<StackTrace> is new):

    # Load plugins
    use Catalyst qw/-Debug
                ConfigLoader
                Static::Simple
    
                StackTrace
    
                Authentication
    
                Session
                Session::Store::FastMmap
                Session::State::Cookie
                /;

B<Note:> As discussed in MoreCatalystBasics, different versions of
C<Catalyst::Devel> have used a variety of methods to load the plugins.
You can put the plugins in the C<use Catalyst> statement if you prefer.

The C<Authentication> plugin supports Authentication while the
C<Session> plugins are required to maintain state across multiple HTTP
requests.

Note that the only required Authentication class is the main one. This
is a change that occurred in version 0.09999_01 of the
C<Authentication> plugin. You B<do not need> to specify a particular
Authentication::Store or Authentication::Credential plugin. Instead,
indicate the Store and Credential you want to use in your application
configuration (see below).

Note that there are several options for
L<Session::Store|Catalyst::Plugin::Session::Store>
(L<Session::Store::FastMmap|Catalyst::Plugin::Session::Store::FastMmap>
is generally a good choice if you are on Unix; try
L<Session::Store::File|Catalyst::Plugin::Session::Store::File> if you
are on Win32) -- consult
L<Session::Store|Catalyst::Plugin::Session::Store> and its subclasses
for additional information and options (for example to use a database-
backed session store).


=head2 Configure Authentication

There are a variety of way to provide configuration information to
L<Catalyst::Plugin::Authentication|Catalyst::Plugin::Authentication>.
Here we will use 
L<Catalyst::Authentication::Realm::SimpleDB|Catalyst::Authentication::Realm::SimpleDB>
because it automatically sets a reasonable set of defaults for us. Open 
C<lib/MyApp.pm> and place the following text above the call to
C<__PACKAGE__-E<gt>setup();>:

    # Configure SimpleDB Authentication
    __PACKAGE__->config->{'Plugin::Authentication'} = {
            default => {
                class           => 'SimpleDB',
                user_model      => 'DB::Users',
                password_type   => 'clear',
            },
        };

We could have placed this configuration in C<myapp.conf>, but placing 
it in C<lib/MyApp.pm> is probably a better place since it's not likely 
something that users of your application will want to change during 
deployment.


=head2 Add Login and Logout Controllers

Use the Catalyst create script to create two stub controller files:

    $ script/myapp_create.pl controller Login
    $ script/myapp_create.pl controller Logout

You could easily use a single controller here.  For example, you could
have a C<User> controller with both C<login> and C<logout> actions.
Remember, Catalyst is designed to be very flexible, and leaves such
matters up to you, the designer and programmer.

Then open C<lib/MyApp/Controller/Login.pm>, locate the
C<sub index :Path :Args(0)> method (or C<sub index : Private> if you
are using an older version of Catalyst) that was automatically
inserted by the helpers when we created the Login controller above,
and update the definition of C<sub index> to match:

    =head2 index
    
    Login logic
    
    =cut
    
    sub index :Path :Args(0) {
        my ($self, $c) = @_;
    
        # Get the username and password from form
        my $username = $c->request->params->{username} || "";
        my $password = $c->request->params->{password} || "";
    
        # If the username and password values were found in form
        if ($username && $password) {
            # Attempt to log the user in
            if ($c->authenticate({ username => $username,
                                   password => $password  } )) {
                # If successful, then let them use the application
                $c->response->redirect($c->uri_for(
                    $c->controller('Books')->action_for('list')));
                return;
            } else {
                # Set an error message
                $c->stash->{error_msg} = "Bad username or password.";
            }
        }
    
        # If either of above don't work out, send to the login page
        $c->stash->{template} = 'login.tt2';
    }

This controller fetches the C<username> and C<password> values from the
login form and attempts to authenticate the user.  If successful, it
redirects the user to the book list page.  If the login fails, the user
will stay at the login page and receive an error message.  If the
C<username> and C<password> values are not present in the form, the
user will be taken to the empty login form.

Note that we could have used something like "C<sub default :Path>",
however, it is generally recommended (partly for historical reasons,
and partly for code clarity) only to use C<default> in
C<MyApp::Controller::Root>, and then mainly to generate the 404 not
found page for the application.

Instead, we are using "C<sub somename :Path :Args(0) {...}>" here to
specifically match the URL C</login>. C<Path> actions (aka, "literal
actions") create URI matches relative to the namespace of the
controller where they are defined.  Although C<Path> supports
arguments that allow relative and absolute paths to be defined, here
we use an empty C<Path> definition to match on just the name of the
controller itself.  The method name, C<index>, is arbitrary. We make
the match even more specific with the C<:Args(0)> action modifier --
this forces the match on I<only> C</login>, not
C</login/somethingelse>.

Next, update the corresponding method in
C<lib/MyApp/Controller/Logout.pm> to match:

    =head2 index
    
    Logout logic
    
    =cut
    
    sub index :Path :Args(0) {
        my ($self, $c) = @_;
    
        # Clear the user's state
        $c->logout;
    
        # Send the user to the starting point
        $c->response->redirect($c->uri_for('/'));
    }

As with the login controller, be sure to delete the
C<$c-E<gt>response-E<gt>body('Matched MyApp::Controller::Logout in Logout.');>
line of the C<sub index>.


=head2 Add a Login Form TT Template Page

Create a login form by opening C<root/src/login.tt2> and inserting:

    [% META title = 'Login' %]
    
    <!-- Login form -->
    <form method="post" action="[% c.uri_for('/login') %]">
      <table>
        <tr>
          <td>Username:</td>
          <td><input type="text" name="username" size="40" /></td>
        </tr>
        <tr>
          <td>Password:</td>
          <td><input type="password" name="password" size="40" /></td>
        </tr>
        <tr>
          <td colspan="2"><input type="submit" name="submit" value="Submit" /></td>
        </tr>
      </table>
    </form>


=head2 Add Valid User Check

We need something that provides enforcement for the authentication
mechanism -- a I<global> mechanism that prevents users who have not
passed authentication from reaching any pages except the login page.
This is generally done via an C<auto> action/method (prior to Catalyst
v5.66, this sort of thing would go in C<MyApp.pm>, but starting in
v5.66, the preferred location is C<lib/MyApp/Controller/Root.pm>).

Edit the existing C<lib/MyApp/Controller/Root.pm> class file and insert
the following method:

    =head2 auto
    
    Check if there is a user and, if not, forward to login page
    
    =cut
    
    # Note that 'auto' runs after 'begin' but before your actions and that
    # 'auto's "chain" (all from application path to most specific class are run)
    # See the 'Actions' section of 'Catalyst::Manual::Intro' for more info.
    sub auto : Private {
        my ($self, $c) = @_;
    
        # Allow unauthenticated users to reach the login page.  This
        # allows unauthenticated users to reach any action in the Login
        # controller.  To lock it down to a single action, we could use:
        #   if ($c->action eq $c->controller('Login')->action_for('index'))
        # to only allow unauthenticated access to the 'index' action we
        # added above.
        if ($c->controller eq $c->controller('Login')) {
            return 1;
        }
    
        # If a user doesn't exist, force login
        if (!$c->user_exists) {
            # Dump a log message to the development server debug output
            $c->log->debug('***Root::auto User not found, forwarding to /login');
            # Redirect the user to the login page
            $c->response->redirect($c->uri_for('/login'));
            # Return 0 to cancel 'post-auto' processing and prevent use of application
            return 0;
        }
    
        # User found, so return 1 to continue with processing after this 'auto'
        return 1;
    }

As discussed in
L<Catalyst::Manual::Tutorial::MoreCatalystBasics/CREATE A CATALYST CONTROLLER>,
every C<auto> method from the application/root controller down to the
most specific controller will be called.  By placing the
authentication enforcement code inside the C<auto> method of
C<lib/MyApp/Controller/Root.pm> (or C<lib/MyApp.pm>), it will be
called for I<every> request that is received by the entire
application.


=head2 Displaying Content Only to Authenticated Users

Let's say you want to provide some information on the login page that
changes depending on whether the user has authenticated yet.  To do
this, open C<root/src/login.tt2> in your editor and add the following
lines to the bottom of the file:

    ...
    <p>
    [%
       # This code illustrates how certain parts of the TT
       # template will only be shown to users who have logged in
    %]
    [% IF c.user_exists %]
        Please Note: You are already logged in as '[% c.user.username %]'.
        You can <a href="[% c.uri_for('/logout') %]">logout</a> here.
    [% ELSE %]
        You need to log in to use this application.
    [% END %]
    [%#
       Note that this whole block is a comment because the "#" appears
       immediate after the "[%" (with no spaces in between).  Although it
       can be a handy way to temporarily "comment out" a whole block of
       TT code, it's probably a little too subtle for use in "normal"
       comments.
    %]
    </p>

Although most of the code is comments, the middle few lines provide a
"you are already logged in" reminder if the user returns to the login
page after they have already authenticated.  For users who have not yet
authenticated, a "You need to log in..." message is displayed (note the
use of an IF-THEN-ELSE construct in TT).


=head2 Try Out Authentication

Press C<Ctrl-C> to kill the previous server instance (if it's still
running) and restart it:

    $ script/myapp_server.pl

B<IMPORTANT NOTE:> If you are having issues with authentication on
Internet Explorer, be sure to check the system clocks on both your
server and client machines.  Internet Explorer is very picky about
timestamps for cookies.  You can quickly sync a Debian system by
installing the "ntpdate" package:

    sudo aptitude -y install ntpdate

And then run the following command:

    sudo ntpdate-debian

Or, depending on your firewall configuration:

    sudo ntpdate-debian -u

Note: NTP can be a little more finicky about firewalls because it uses
UDP vs. the more common TCP that you see with most Internet protocols.
Worse case, you might have to manually set the time on your development
box instead of using NTP.

Now trying going to L<http://localhost:3000/books/list> and you should
be redirected to the login page, hitting Shift+Reload or Ctrl+Reload
if necessary (the "You are already logged in" message should I<not>
appear -- if it does, click the C<logout> button and try again). Note
the C<***Root::auto User not found...> debug message in the
development server output.  Enter username C<test01> and password
C<mypass>, and you should be taken to the Book List page.

Open C<root/src/books/list.tt2> and add the following lines to the
bottom (below the closing </table> tag):

    <p>
      <a href="[% c.uri_for('/login') %]">Login</a>
      <a href="[% c.uri_for(c.controller.action_for('form_create')) %]">Create</a>
    </p>

Reload your browser and you should now see a "Login" and "Create" links
at the bottom of the page (as mentioned earlier, you can update template
files without reloading the development server).  Click the first link
to return to the login page.  This time you I<should> see the "You are
already logged in" message.

Finally, click the C<You can logout here> link on the C</login> page.
You should stay at the login page, but the message should change to "You
need to log in to use this application."


=head1 USING PASSWORD HASHES

In this section we increase the security of our system by converting 
from cleartext passwords to SHA-1 password hashes that include a 
random "salt" value to make them extremely difficult to crack with
dictionary and "rainbow table" attacks.

B<Note:> This section is optional.  You can skip it and the rest of the
tutorial will function normally.

Be aware that even with the techniques shown in this section, the browser
still transmits the passwords in cleartext to your application.  We are
just avoiding the I<storage> of cleartext passwords in the database by
using a salted SHA-1 hash. If you are concerned about cleartext passwords
between the browser and your application, consider using SSL/TLS, made
easy with the Catalyst plugin Catalyst::Plugin:RequireSSL.


=head2 Install DBIx::Class::EncodedColumn

L<DBIx::Class::EncodedColumn|DBIx::Class::EncodedColumn> provides features
that can greatly simplify the maintenance of passwords.  It's currently 
not available as a .deb package in the normal Debian repositories, so let's
install it directly from CPAN:

    $ sudo cpan DBIx::Class::EncodedColumn


=head2 Re-Run the DBIC::Schema Model Helper to Include DBIx::Class::EncodedColumn

Next, we can re-run the model helper to have it include 
L<DBIx::Class::EncodedColumn|DBIx::Class::EncodedColumn> in all of the 
Result Classes it generates for us.  Simply use the same command we 
saw in Chapters 3 and 4, but add C<,EncodedColumn> to the C<components>
argument:

    $ script/myapp_create.pl model DB DBIC::Schema MyApp::Schema \
        create=static components=TimeStamp,EncodedColumn dbi:SQLite:myapp.db

If you then open one of the Result Classes, you will see that it 
includes EncodedColumn in the C<load_components> line.  Take a look at 
C<lib/MyApp/Schema/Result/Users.pm> since that's the main class where we
want to use hashed and salted passwords:

    __PACKAGE__->load_components("InflateColumn::DateTime", "TimeStamp", "EncodedColumn", "Core");


=head2 Modify the "password" Column to Use EncodedColumn

Open the file C<lib/MyApp/Schema/Result/Users.pm> and enter the following
text below the "# DO NOT MODIFY THIS OR ANYTHING ABOVE!" line but above
the closing "1;":

    # Have the 'password' column use a SHA-1 hash and 10-character salt
    # with hex encoding; Generate the 'check_password" method
    __PACKAGE__->add_columns(
        'password' => {
            data_type           => "TEXT",
            size                => undef,
            encode_column       => 1,
            encode_class        => 'Digest',
            encode_args         => {salt_length => 10},
            encode_check_method => 'check_password',
        },
    );

This redefines the automatically generated definition for the password 
fields at the top of the Result Class file to now use EncodedColumn 
logic (C<encoded_column> is set to 1).  C<encode_class> can be set to 
either C<Digest> to use 
L<DBIx::Class::EncodedColumn::Digest|DBIx::Class::EncodedColumn::Digest>, 
or C<Crypt::Eksblowfish::Bcrypt> for 
L<DBIx::Class::EncodedColumn::Crypt::Eksblowfish::Bcrypt|DBIx::Class::EncodedColumn::Crypt::Eksblowfish::Bcrypt>.
C<encode_args> is then used to customize the type of Digest you 
selected. Here we only specified the size of the salt to use, but
we could have also modified the hashing algorithm ('SHA-256' is 
the default) and the format to use ('base64' is the default, but
'hex' and 'binary' are other options).  To use these, you could 
change the C<encode_args> to something like:

            encode_args         => {algorithm => 'SHA-1', 
                                    format => 'hex', 
                                    salt_length => 10},


=head2 Load Hashed Passwords in the Database

Next, let's create a quick script to load some hashed and salted passwords
into the C<password> column of our C<users> table.  Open the file
C<set_hashed_passwords.pl> in your editor and enter the following text:

    #!/usr/bin/perl
    
    use strict;
    use warnings;
    
    use MyApp::Schema;
    
    my $schema = MyApp::Schema->connect('dbi:SQLite:myapp.db');
    
    my @users = $schema->resultset('Users')->all;
    
    foreach my $user (@users) {
        $user->password('mypass');
        $user->update;
    }

EncodedColumn lets us simple call C<$user->check_password($password)> 
to see if the user has supplied the correct password, or, as we show 
above, call C<$user->update($new_password)> to update the hashed 
password stored for this user.

Then run the following command:

    $ perl -Ilib set_hashed_passwords.pl

We had to use the C<-Ilib> arguement to tell perl to look under the 
C<lib> directory for our C<MyApp::Schema> model.

Then dump the users table to verify that it worked:

    $ sqlite3 myapp.db "select * from users"
    1|test01|38d3974fa9e9263099f7bc2574284b2f55473a9bM=fwpX2NR8|t01@na.com|Joe|Blow|1
    2|test02|6ed8586587e53e0d7509b1cfed5df08feadc68cbMJlnPyPt0I|t02@na.com|Jane|Doe|1
    3|test03|af929a151340c6aed4d54d7e2651795d1ad2e2f7UW8dHoGv9z|t03@na.com|No|Go|0

As you can see, the passwords are much harder to steal from the 
database.  Also note that this demonstrates how to use a DBIx::Class 
model outside of your web application -- a very useful feature in many 
situations.


=head2 Enable Hashed and Salted Passwords

Edit C<lib/MyApp.pm> and update it to match the following text (the only change
is to the C<password_type> field):

    # Configure SimpleDB Authentication
    __PACKAGE__->config->{'Plugin::Authentication'} = {
            default => {
                class           => 'SimpleDB',
                user_model      => 'DB::Users',
                password_type   => 'self_check',
            },
        };

The use of C<self_check> will cause 
Catalyst::Plugin::Authentication::Store::DBIC to call the 
C<check_password> method we enabled on our C<password> columns.


=head2 Try Out the Hashed Passwords

Press C<Ctrl-C> to kill the previous server instance (if it's still
running) and restart it:

    $ script/myapp_server.pl

You should now be able to go to L<http://localhost:3000/books/list> and
login as before.  When done, click the "logout" link on the login page
(or point your browser at L<http://localhost:3000/logout>).


=head1 USING THE SESSION FOR FLASH

As discussed in the previous chapter of the tutorial, C<flash> allows 
you to set variables in a way that is very similar to C<stash>, but it 
will remain set across multiple requests.  Once the value is read, it 
is cleared (unless reset).  Although C<flash> has nothing to do with 
authentication, it does leverage the same session plugins.  Now that 
those plugins are enabled, let's go back and update the "delete and 
redirect with query parameters" code seen at the end of the L<Basic 
CRUD|Catalyst::Manual::Tutorial::BasicCRUD> chapter of the tutorial to 
take advantage of C<flash>.

First, open C<lib/MyApp/Controller/Books.pm> and modify C<sub delete>
to match the following (everything after the model search line of code
has changed):

    =head2 delete
    
    Delete a book
    
    =cut
    
    sub delete :Chained('object') :PathPart('delete') :Args(0) {
        my ($self, $c) = @_;
    
        # Use the book object saved by 'object' and delete it along
        # with related 'book_authors' entries
        $c->stash->{object}->delete;
    
        # Use 'flash' to save information across requests until it's read
        $c->flash->{status_msg} = "Book deleted";
    
        # Redirect the user back to the list page
        $c->response->redirect($c->uri_for($self->action_for('list')));
    }

Next, open C<root/src/wrapper.tt2> and update the TT code to pull from
flash vs. the C<status_msg> query parameter:

    ...
    <div id="content">
        [%# Status and error messages %]
        <span class="message">[% status_msg || c.flash.status_msg %]</span>
        <span class="error">[% error_msg %]</span>
        [%# This is where TT will stick all of your template's contents. -%]
        [% content %]
    </div><!-- end content -->
    ...

Although the sample above only shows the C<content> div, leave the
rest of the file intact -- the only change we made to the C<wrapper.tt2>
was to add "C<|| c.request.params.status_msg>" to the
C<E<lt>span class="message"E<gt>> line.


=head2 Try Out Flash

Restart the development server, log in, and then point your browser to
L<http://localhost:3000/books/url_create/Test/1/4> to create an extra
several books.  Click the "Return to list" link and delete one of the
"Test" books you just added.  The C<flash> mechanism should retain our
"Book deleted" status message across the redirect.

B<NOTE:> While C<flash> will save information across multiple requests,
I<it does get cleared the first time it is read>.  In general, this is
exactly what you want -- the C<flash> message will get displayed on
the next screen where it's appropriate, but it won't "keep showing up"
after that first time (unless you reset it).  Please refer to
L<Catalyst::Plugin::Session|Catalyst::Plugin::Session> for additional
information.


=head2 Switch To Flash-To-Stash

Although the a use of flash above works well, the
C<status_msg || c.flash.status_msg> statement is a little ugly. A nice
alternative is to use the C<flash_to_stash> feature that automatically
copies the content of flash to stash.  This makes your controller
and template code work regardless of where it was directly access, a
forward, or a redirect.  To enable C<flash_to_stash>, you can either
set the value in C<lib/MyApp.pm> by changing the default
C<__PACKAGE__-E<gt>config> setting to something like:

    __PACKAGE__->config(
            name    => 'MyApp',
            session => {flash_to_stash => 1}
        );

B<or> add the following to C<myapp.conf>:

    <session>
        flash_to_stash   1
    </session>

The C<__PACKAGE__-E<gt>config> option is probably preferable here
since it's not something you will want to change at runtime without it
possibly breaking some of your code.

Then edit C<root/src/wrapper.tt2> and change the C<status_msg> line
to match the following:

    <span class="message">[% status_msg %]</span>

Restart the development server and go to
L<http://localhost:3000/books/list> in your browser.  Delete another
of the "Test" books you added in the previous step.  Flash should still
maintain the status message across the redirect even though you are no
longer explicitly accessing C<c.flash>.


=head1 AUTHOR

Kennedy Clark, C<hkclark@gmail.com>

Please report any errors, issues or suggestions to the author.  The
most recent version of the Catalyst Tutorial can be found at
L<http://dev.catalyst.perl.org/repos/Catalyst/Catalyst-Manual/5.70/trunk/lib/Catalyst/Manual/Tutorial/>.

Copyright 2006-2008, Kennedy Clark, under Creative Commons License
(L<http://creativecommons.org/licenses/by-sa/3.0/us/>).