Add some additional background info & links about auto-defined DBIC relationships

[catagits/Catalyst-Manual.git] / lib / Catalyst / Manual / Tutorial / 04_BasicCRUD.pod

=head1 NAME

Catalyst::Manual::Tutorial::04_BasicCRUD - Catalyst Tutorial - Chapter 4: Basic CRUD


=head1 OVERVIEW

This is B<Chapter 4 of 10> for the Catalyst tutorial.

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

=over 4

=item 1

L<Introduction|Catalyst::Manual::Tutorial::01_Intro>

=item 2

L<Catalyst Basics|Catalyst::Manual::Tutorial::02_CatalystBasics>

=item 3

L<More Catalyst Basics|Catalyst::Manual::Tutorial::03_MoreCatalystBasics>

=item 4

B<04_Basic CRUD>

=item 5

L<Authentication|Catalyst::Manual::Tutorial::05_Authentication>

=item 6

L<Authorization|Catalyst::Manual::Tutorial::06_Authorization>

=item 7

L<Debugging|Catalyst::Manual::Tutorial::07_Debugging>

=item 8

L<Testing|Catalyst::Manual::Tutorial::08_Testing>

=item 9

L<Advanced CRUD|Catalyst::Manual::Tutorial::09_AdvancedCRUD>

=item 10

L<Appendices|Catalyst::Manual::Tutorial::10_Appendices>

=back


=head1 DESCRIPTION

This chapter of the tutorial builds on the fairly primitive 
application created in Chapter 3 to add basic support for Create, 
Read, Update, and Delete (CRUD) of C<Book> objects.  Note that the 
'list' function in Chapter 2 already implements the Read portion of 
CRUD (although Read normally refers to reading a single object; you 
could implement full Read functionality using the techniques 
introduced below).  This section will focus on the Create and Delete 
aspects of CRUD.  More advanced capabilities, including full Update 
functionality, will be addressed in Chapter 9.

Although this chapter of the tutorial will show you how to build CRUD 
functionality yourself, another option is to use a "CRUD builder" type 
of tool to automate the process.  You get less control, but it's quick 
and easy.  For example, see 
L<Catalyst::Plugin::AutoCRUD|Catalyst::Plugin::AutoCRUD>, 
L<CatalystX::CRUD|CatalystX::CRUD>, and 
L<CatalystX::CRUD::YUI|CatalystX::CRUD::YUI>.

You can check out the source code for this example from the Catalyst
Subversion repository as per the instructions in
L<Catalyst::Manual::Tutorial::01_Intro|Catalyst::Manual::Tutorial::01_Intro>.


=head1 FORMLESS SUBMISSION

Our initial attempt at object creation will utilize the "URL
arguments" feature of Catalyst (we will employ the more common form-
based submission in the sections that follow).


=head2 Include a Create Action in the Books Controller

Edit C<lib/MyApp/Controller/Books.pm> and enter the following method:

    =head2 url_create
    
    Create a book with the supplied title, rating, and author
    
    =cut
    
    sub url_create : Local {
        # In addition to self & context, get the title, rating, &
        # author_id args from the URL.  Note that Catalyst automatically
        # puts extra information after the "/<controller_name>/<action_name/"
        # into @_.  The args are separated  by the '/' char on the URL.
        my ($self, $c, $title, $rating, $author_id) = @_;
    
        # Call create() on the book model object. Pass the table
        # columns/field values we want to set as hash values
        my $book = $c->model('DB::Book')->create({
                title  => $title,
                rating => $rating
            });
    
        # Add a record to the join table for this book, mapping to
        # appropriate author
        $book->add_to_book_authors({author_id => $author_id});
        # Note: Above is a shortcut for this:
        # $book->create_related('book_authors', {author_id => $author_id});
    
        # Assign the Book object to the stash for display in the view
        $c->stash->{book} = $book;
    
        # Set the TT template to use
        $c->stash->{template} = 'books/create_done.tt2';
    }

Notice that Catalyst takes "extra slash-separated information" from the
URL and passes it as arguments in C<@_>.  The C<url_create> action then
uses a simple call to the DBIC C<create> method to add the requested
information to the database (with a separate call to
C<add_to_book_authors> to update the join table).  As do virtually all
controller methods (at least the ones that directly handle user input),
it then sets the template that should handle this request.


=head2 Include a Template for the 'url_create' Action:

Edit C<root/src/books/create_done.tt2> and then enter:

    [% # Use the TT Dumper plugin to Data::Dumper variables to the browser   -%]
    [% # Not a good idea for production use, though. :-)  'Indent=1' is      -%]
    [% # optional, but prevents "massive indenting" of deeply nested objects -%]
    [% USE Dumper(Indent=1) -%]
    
    [% # Set the page title.  META can 'go back' and set values in templates -%]
    [% # that have been processed 'before' this template (here it's for      -%]
    [% # root/lib/site/html and root/lib/site/header).  Note that META only  -%]
    [% # works on simple/static strings (i.e. there is no variable           -%]
    [% # interpolation).                                                     -%]
    [% META title = 'Book Created' %]
    
    [% # Output information about the record that was added.  First title.   -%]
    <p>Added book '[% book.title %]'
    
    [% # Output the last name of the first author.                           -%]
    by '[% book.authors.first.last_name %]'
    
    [% # Output the rating for the book that was added -%]
    with a rating of [% book.rating %].</p>
    
    [% # Provide a link back to the list page                                    -%]
    [% # 'uri_for()' builds a full URI; e.g., 'http://localhost:3000/books/list' -%]
    <p><a href="[% c.uri_for('/books/list') %]">Return to list</a></p>
    
    [% # Try out the TT Dumper (for development only!) -%]
    <pre>
    Dump of the 'book' variable:
    [% Dumper.dump(book) %]
    </pre>

The TT C<USE> directive allows access to a variety of plugin modules
(TT plugins, that is, not Catalyst plugins) to add extra functionality
to the base TT capabilities.  Here, the plugin allows
L<Data::Dumper|Data::Dumper> "pretty printing" of objects and
variables.  Other than that, the rest of the code should be familiar
from the examples in Chapter 3.

Note: If you are using TT v2.15 you will need to change the code that 
outputs the "last name for the first author" above to match this:

    [% authors = book.authors %]
    by '[% authors.first.last_name IF authors.first;
           authors.list.first.value.last_name IF ! authors.first %]'

to get around an issue in TT v2.15 where blessed hash objects were not 
handled correctly.  But, if you are still using v2.15, it's probably 
time to upgrade  (v2.15 is 3.5+ years old).  If you are following 
along in Debian, then you should be on at least v2.20.  You can test 
your version of Template Toolkit with the following:

    perl -MTemplate -e 'print "$Template::VERSION\n"'


=head2 Try the 'url_create' Feature

If the application is still running from before, use C<Ctrl-C> to kill
it. Then restart the server:

    $ DBIC_TRACE=1 script/myapp_server.pl

Note that new path for C</books/url_create> appears in the startup debug
output.

B<TIP>: You can use C<script/myapp_server.pl -r> to have the development
server auto-detect changed files and reload itself (if your browser acts
odd, you should also try throwing in a C<-k>).  If you make changes to
the TT templates only, you do not need to reload the development server
(only changes to "compiled code" such as Controller and Model C<.pm>
files require a reload).

Next, use your browser to enter the following URL:

    http://localhost:3000/books/url_create/TCPIP_Illustrated_Vol-2/5/4

Your browser should display "Added book 'TCPIP_Illustrated_Vol-2' by
'Stevens' with a rating of 5." along with a dump of the new book model
object as it was returned by DBIC.  You should also see the following
DBIC debug messages displayed in the development server log messages
if you have DBIC_TRACE set:

    INSERT INTO book (rating, title) VALUES (?, ?): `5', `TCPIP_Illustrated_Vol-2'
    INSERT INTO book_author (author_id, book_id) VALUES (?, ?): `4', `6'

The C<INSERT> statements are obviously adding the book and linking it to
the existing record for Richard Stevens.  The C<SELECT> statement results
from DBIC automatically fetching the book for the C<Dumper.dump(book)>.

If you then click the "Return to list" link, you should find that 
there are now six books shown (if necessary, Shift+Reload or 
Ctrl+Reload your browser at the C</books/list> page).  You should now 
see the six DBIC debug messages similar to the following (where 
N=1-6):

    SELECT author.id, author.first_name, author.last_name 
        FROM book_author me  JOIN author author 
        ON author.id = me.author_id WHERE ( me.book_id = ? ): 'N'


=head1 CONVERT TO A CHAINED ACTION

Although the example above uses the same C<Local> action type for the
method that we saw in the previous chapter of the tutorial, there is an
alternate approach that allows us to be more specific while also
paving the way for more advanced capabilities.  Change the method
declaration for C<url_create> in C<lib/MyApp/Controller/Books.pm> you
entered above to match the following:

    sub url_create :Chained('/') :PathPart('books/url_create') :Args(3) {
        # In addition to self & context, get the title, rating, &
        # author_id args from the URL.  Note that Catalyst automatically
        # puts the first 3 arguments worth of extra information after the 
        # "/<controller_name>/<action_name/" into @_ because we specified
        # "Args(3)".  The args are separated  by the '/' char on the URL.
        my ($self, $c, $title, $rating, $author_id) = @_;
    
        ...

This converts the method to take advantage of the Chained
action/dispatch type. Chaining lets you have a single URL
automatically dispatch to several controller methods, each of which
can have precise control over the number of arguments that it will
receive.  A chain can essentially be thought of having three parts --
a beginning, a middle, and an end.  The bullets below summarize the key
points behind each of these parts of a chain:


=over 4


=item *

Beginning

=over 4

=item *

B<Use "C<:Chained('/')>" to start a chain>

=item *

Get arguments through C<CaptureArgs()>

=item *

Specify the path to match with C<PathPart()>

=back


=item *

Middle

=over 4

=item *

Link to previous part of the chain with C<:Chained('_name_')>

=item *

Get arguments through C<CaptureArgs()>

=item *

Specify the path to match with C<PathPart()>

=back


=item *

End

=over 4

=item *

Link to previous part of the chain with C<:Chained('_name_')>

=item *

B<Do NOT get arguments through "C<CaptureArgs()>," use "C<Args()>" instead to end a chain>

=item *

Specify the path to match with C<PathPart()>

=back


=back

In our C<url_create> method above, we have combined all three parts into
a single method: C<:Chained('/')> to start the chain,
C<:PathPart('books/url_create')> to specify the base URL to match, and
C<:Args(3)> to capture exactly three arguments and to end the chain.

As we will see shortly, a chain can consist of as many "links" as you
wish, with each part capturing some arguments and doing some work
along the way.  We will continue to use the Chained action type in this
chapter of the tutorial and explore slightly more advanced capabilities
with the base method and delete feature below.  But Chained dispatch
is capable of far more.  For additional information, see
L<Catalyst::Manual::Intro/Action types>,
L<Catalyst::DispatchType::Chained|Catalyst::DispatchType::Chained>,
and the 2006 Advent calendar entry on the subject:
L<http://www.catalystframework.org/calendar/2006/10>.


=head2 Try the Chained Action

If you look back at the development server startup logs from your
initial version of the C<url_create> method (the one using the
C<:Local> attribute), you will notice that it produced output similar
to the following:

    [debug] Loaded Path actions:
    .-------------------------------------+--------------------------------------.
    | Path                                | Private                              |
    +-------------------------------------+--------------------------------------+
    | /                                   | /default                             |
    | /                                   | /index                               |
    | /books                              | /books/index                         |
    | /books/list                         | /books/list                          |
    | /books/url_create                   | /books/url_create                    |
    '-------------------------------------+--------------------------------------'

Now start the development server with our basic chained method in
place and the startup debug output should change to something along
the lines of the following:

    [debug] Loaded Path actions:
    .-------------------------------------+--------------------------------------.
    | Path                                | Private                              |
    +-------------------------------------+--------------------------------------+
    | /                                   | /default                             |
    | /                                   | /index                               |
    | /books                              | /books/index                         |
    | /books/list                         | /books/list                          |
    '-------------------------------------+--------------------------------------'
    
    [debug] Loaded Chained actions:
    .-------------------------------------+--------------------------------------.
    | Path Spec                           | Private                              |
    +-------------------------------------+--------------------------------------+
    | /books/url_create/*/*/*             | /books/url_create                    |
    '-------------------------------------+--------------------------------------'

C<url_create> has disappeared form the "Loaded Path actions" section
but it now shows up under the newly created "Loaded Chained actions"
section.  And the "/*/*/*" portion clearly shows our requirement for
three arguments.

As with our non-chained version of C<url_create>, use your browser to
enter the following URL:

    http://localhost:3000/books/url_create/TCPIP_Illustrated_Vol-2/5/4

You should see the same "Added book 'TCPIP_Illustrated_Vol-2' by
'Stevens' with a rating of 5." along with a dump of the new book model
object.  Click the "Return to list" link, and you should find that there
are now seven books shown (two copies of I<TCPIP_Illustrated_Vol-2>).


=head2 Refactor to Use a 'base' Method to Start the Chains

Let's make a quick update to our initial Chained action to show a
little more of the power of chaining.  First, open
C<lib/MyApp/Controller/Books.pm> in your editor and add the following
method:

    =head2 base
    
    Can place common logic to start chained dispatch here
    
    =cut
    
    sub base :Chained('/') :PathPart('books') :CaptureArgs(0) {
        my ($self, $c) = @_;
    
        # Store the ResultSet in stash so it's available for other methods
        $c->stash->{resultset} = $c->model('DB::Book');
    
        # Print a message to the debug log
        $c->log->debug('*** INSIDE BASE METHOD ***');
    }

Here we print a log message and store the DBIC ResultSet in
C<$c-E<gt>stash-E<gt>{resultset}> so that it's automatically available
for other actions that chain off C<base>.  If your controller always
needs a book ID as its first argument, you could have the base method
capture that argument (with C<:CaptureArgs(1)>) and use it to pull the
book object with C<-E<gt>find($id)> and leave it in the stash for
later parts of your chains to then act upon. Because we have several
actions that don't need to retrieve a book (such as the C<url_create>
we are working with now), we will instead add that functionality
to a common C<object> action shortly.

As for C<url_create>, let's modify it to first dispatch to C<base>.
Open up C<lib/MyApp/Controller/Books.pm> and edit the declaration for
C<url_create> to match the following:

    sub url_create :Chained('base') :PathPart('url_create') :Args(3) {

Next, try out the refactored chain by restarting the development
server.  Notice that our "Loaded Chained actions" section has changed
slightly:

    [debug] Loaded Chained actions:
    .-------------------------------------+--------------------------------------.
    | Path Spec                           | Private                              |
    +-------------------------------------+--------------------------------------+
    | /books/url_create/*/*/*             | /books/base (0)                      |
    |                                     | => /books/url_create                 |
    '-------------------------------------+--------------------------------------'

The "Path Spec" is the same, but now it maps to two Private actions as 
we would expect.  The C<base> method is being triggered by the 
C</books> part of the URL.  However, the processing then continues to 
the C<url_create> method because this method "chained" off C<base> and 
specified C<:PathPart('url_create')> (note that we could have omitted 
the "PathPart" here because it matches the name of the method, but we 
will include it to make the logic as explicit as possible).

Once again, enter the following URL into your browser:

    http://localhost:3000/books/url_create/TCPIP_Illustrated_Vol-2/5/4

The same "Added book 'TCPIP_Illustrated_Vol-2' by 'Stevens' with a 
rating of 5." message and a dump of the new book object should appear. 
Also notice the extra "INSIDE BASE METHOD" debug message in the 
development server output from the C<base> method.  Click the "Return 
to list" link, and you should find that there are now eight books 
shown.  (You may have a larger number of books if you repeated any of 
the "create" actions more than once.  Don't worry about it as long as 
the number of books is appropriate for the number of times you added 
new books... there should be the original five books added via
C<myapp01.sql> plus one additional book for each time you ran one
of the url_create variations above.)


=head1 MANUALLY BUILDING A CREATE FORM

Although the C<url_create> action in the previous step does begin to
reveal the power and flexibility of both Catalyst and DBIC, it's
obviously not a very realistic example of how users should be expected
to enter data.  This section begins to address that concern.


=head2 Add Method to Display The Form

Edit C<lib/MyApp/Controller/Books.pm> and add the following method:

    =head2 form_create
    
    Display form to collect information for book to create
    
    =cut
    
    sub form_create :Chained('base') :PathPart('form_create') :Args(0) {
        my ($self, $c) = @_;
    
        # Set the TT template to use
        $c->stash->{template} = 'books/form_create.tt2';
    }

This action simply invokes a view containing a form to create a book.


=head2 Add a Template for the Form

Open C<root/src/books/form_create.tt2> in your editor and enter:

    [% META title = 'Manual Form Book Create' -%]
    
    <form method="post" action="[% c.uri_for('form_create_do') %]">
    <table>
      <tr><td>Title:</td><td><input type="text" name="title"></td></tr>
      <tr><td>Rating:</td><td><input type="text" name="rating"></td></tr>
      <tr><td>Author ID:</td><td><input type="text" name="author_id"></td></tr>
    </table>
    <input type="submit" name="Submit" value="Submit">
    </form>

Note that we have specified the target of the form data as
C<form_create_do>, the method created in the section that follows.


=head2 Add a Method to Process Form Values and Update Database

Edit C<lib/MyApp/Controller/Books.pm> and add the following method to
save the form information to the database:

    =head2 form_create_do
    
    Take information from form and add to database
    
    =cut
    
    sub form_create_do :Chained('base') :PathPart('form_create_do') :Args(0) {
        my ($self, $c) = @_;
    
        # Retrieve the values from the form
        my $title     = $c->request->params->{title}     || 'N/A';
        my $rating    = $c->request->params->{rating}    || 'N/A';
        my $author_id = $c->request->params->{author_id} || '1';
    
        # Create the book
        my $book = $c->model('DB::Book')->create({
                title   => $title,
                rating  => $rating,
            });
        # Handle relationship with author
        $book->add_to_book_authors({author_id => $author_id});
        # Note: Above is a shortcut for this:
        # $book->create_related('book_authors', {author_id => $author_id});
    
        # Store new model object in stash
        $c->stash->{book} = $book;
    
        # Avoid Data::Dumper issue mentioned earlier
        # You can probably omit this
        $Data::Dumper::Useperl = 1;
    
        # Set the TT template to use
        $c->stash->{template} = 'books/create_done.tt2';
    }


=head2 Test Out The Form

If the application is still running from before, use C<Ctrl-C> to kill
it.  Then restart the server:

    $ script/myapp_server.pl

Notice that the server startup log reflects the two new chained
methods that we added:

    [debug] Loaded Chained actions:
    .-------------------------------------+--------------------------------------.
    | Path Spec                           | Private                              |
    +-------------------------------------+--------------------------------------+
    | /books/form_create                  | /books/base (0)                      |
    |                                     | => /books/form_create                |
    | /books/form_create_do               | /books/base (0)                      |
    |                                     | => /books/form_create_do             |
    | /books/url_create/*/*/*             | /books/base (0)                      |
    |                                     | => /books/url_create                 |
    '-------------------------------------+--------------------------------------'

Point your browser to L<http://localhost:3000/books/form_create> and
enter "TCP/IP Illustrated, Vol 3" for the title, a rating of 5, and an
author ID of 4.  You should then see the output of the same
C<create_done.tt2> template seen in earlier examples.  Finally, click
"Return to list" to view the full list of books.

B<Note:> Having the user enter the primary key ID for the author is
obviously crude; we will address this concern with a drop-down list and
add validation to our forms in Chapter 9.


=head1 A SIMPLE DELETE FEATURE

Turning our attention to the Delete portion of CRUD, this section
illustrates some basic techniques that can be used to remove information
from the database.


=head2 Include a Delete Link in the List

Edit C<root/src/books/list.tt2> and update it to match the following (two
sections have changed: 1) the additional '<th>Links</th>' table header,
and 2) the four lines for the Delete link near the bottom):

    [% # This is a TT comment.  The '-' at the end "chomps" the newline.  You won't -%]
    [% # see this "chomping" in your browser because HTML ignores blank lines, but  -%]
    [% # it WILL eliminate a blank line if you view the HTML source.  It's purely   -%]
    [%- # optional, but both the beginning and the ending TT tags support chomping. -%]
    
    [% # Provide a title -%]
    [% META title = 'Book List' -%]
    
    <table>
    <tr><th>Title</th><th>Rating</th><th>Author(s)</th><th>Links</th></tr>
    [% # Display each book in a table row %]
    [% FOREACH book IN books -%]
      <tr>
        <td>[% book.title %]</td>
        <td>[% book.rating %]</td>
        <td>
          [% # NOTE: See "Exploring The Power of DBIC" for a better way to do this!  -%]
          [% # First initialize a TT variable to hold a list.  Then use a TT FOREACH -%]
          [% # loop in 'side effect notation' to load just the last names of the     -%]
          [% # authors into the list. Note that the 'push' TT vmethod doesn't return -%]
          [% # a value, so nothing will be printed here.  But, if you have something -%]
          [% # in TT that does return a value and you don't want it printed, you can -%]
          [% # 1) assign it to a bogus value, or                                     -%]
          [% # 2) use the CALL keyword to call it and discard the return value.      -%]
          [% tt_authors = [ ];
             tt_authors.push(author.last_name) FOREACH author = book.authors %]
          [% # Now use a TT 'virtual method' to display the author count in parens   -%]
          [% # Note the use of the TT filter "| html" to escape dangerous characters -%]
          ([% tt_authors.size | html %])
          [% # Use another TT vmethod to join & print the names & comma separators   -%]
          [% tt_authors.join(', ') | html %]
        </td>
        <td>
          [% # Add a link to delete a book %]
          <a href="[% c.uri_for(c.controller.action_for('delete'), [book.id]) %]">Delete</a>
        </td>
      </tr>
    [% END -%]
    </table>

The additional code is obviously designed to add a new column to the
right side of the table with a C<Delete> "button" (for simplicity, links
will be used instead of full HTML buttons; in practice, anything that
modifies data should be handled with a form sending a POST request).

Also notice that we are using a more advanced form of C<uri_for> than
we have seen before.  Here we use
C<$c-E<gt>controller-E<gt>action_for> to automatically generate a URI
appropriate for that action based on the method we want to link to
while inserting the C<book.id> value into the appropriate place.  Now,
if you ever change C<:PathPart('delete')> in your controller method to
C<:PathPart('kill')>, then your links will automatically update
without any changes to your .tt2 template file.  As long as the name
of your method does not change (here, "delete"), then your links will
still be correct.  There are a few shortcuts and options when using
C<action_for()>:

=over 4

=item *

If you are referring to a method in the current controller, you can
use C<$self-E<gt>action_for('_method_name_')>.

=item *

If you are referring to a method in a different controller, you need
to include that controller's name as an argument to C<controller()>, as in
C<$c-E<gt>controller('_controller_name_')-E<gt>action_for('_method_name_')>.

=back

B<Note:> In practice you should B<never> use a GET request to delete a
record -- always use POST for actions that will modify data.  We are
doing it here for illustrative and simplicity purposes only.


=head2 Add a Common Method to Retrieve a Book for the Chain

As mentioned earlier, since we have a mixture of actions that operate
on a single book ID and others that do not, we should not have C<base>
capture the book ID, find the corresponding book in the database and
save it in the stash for later links in the chain.  However, just
because that logic does not belong in C<base> doesn't mean that we
can't create another location to centralize the book lookup code.  In
our case, we will create a method called C<object> that will store the
specific book in the stash.  Chains that always operate on a single
existing book can chain off this method, but methods such as
C<url_create> that don't operate on an existing book can chain
directly off base.

To add the C<object> method, edit C<lib/MyApp/Controller/Books.pm>
and add the following code:

    =head2 object
    
    Fetch the specified book object based on the book ID and store
    it in the stash
    
    =cut
    
    sub object :Chained('base') :PathPart('id') :CaptureArgs(1) {
        # $id = primary key of book to delete
        my ($self, $c, $id) = @_;
    
        # Find the book object and store it in the stash
        $c->stash(object => $c->stash->{resultset}->find($id));
    
        # Make sure the lookup was successful.  You would probably
        # want to do something like this in a real app:
        #   $c->detach('/error_404') if !$c->stash->{object};
        die "Book $id not found!" if !$c->stash->{object};
    
        # Print a message to the debug log
        $c->log->debug("*** INSIDE OBJECT METHOD for obj id=$id ***");
    }

Now, any other method that chains off C<object> will automatically
have the appropriate book waiting for it in
C<$c-E<gt>stash-E<gt>{object}>.

Also note that we are using a different technique for setting
C<$c-E<gt>stash>.  The advantage of this style is that it lets you set
multiple stash variables at a time.  For example:

    $c->stash(object => $c->stash->{resultset}->find($id),
              another_thing => 1);

or as a hashref:

    $c->stash({object => $c->stash->{resultset}->find($id),
              another_thing => 1});

Either format works, but the C<$c-E<gt>stash(name =E<gt> value);>
style is growing in popularity -- you may wish to use it all
the time (even when you are only setting a single value).


=head2 Add a Delete Action to the Controller

Open C<lib/MyApp/Controller/Books.pm> in your editor and add the
following method:

    =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_author' entries
        $c->stash->{object}->delete;
    
        # Set a status message to be displayed at the top of the view
        $c->stash->{status_msg} = "Book deleted.";
    
        # Forward to the list action/method in this controller
        $c->forward('list');
    }

This method first deletes the book object saved by the C<object> method.
However, it also removes the corresponding entry from the
C<book_author> table with a cascading delete.

Then, rather than forwarding to a "delete done" page as we did with the
earlier create example, it simply sets the C<status_msg> to display a
notification to the user as the normal list view is rendered.

The C<delete> action uses the context C<forward> method to return the
user to the book list.  The C<detach> method could have also been used.
Whereas C<forward> I<returns> to the original action once it is
completed, C<detach> does I<not> return.  Other than that, the two are
equivalent.


=head2 Try the Delete Feature

If the application is still running from before, use C<Ctrl-C> to kill
it.  Then restart the server:

    $ DBIC_TRACE=1 script/myapp_server.pl

The C<delete> method now appears in the "Loaded Chained actions" section
of the startup debug output:

    [debug] Loaded Chained actions:
    .-------------------------------------+--------------------------------------.
    | Path Spec                           | Private                              |
    +-------------------------------------+--------------------------------------+
    | /books/id/*/delete                  | /books/base (0)                      |
    |                                     | -> /books/object (1)                 |
    |                                     | => /books/delete                     |
    | /books/form_create                  | /books/base (0)                      |
    |                                     | => /books/form_create                |
    | /books/form_create_do               | /books/base (0)                      |
    |                                     | => /books/form_create_do             |
    | /books/url_create/*/*/*             | /books/base (0)                      |
    |                                     | => /books/url_create                 |
    '-------------------------------------+--------------------------------------'

Then point your browser to L<http://localhost:3000/books/list> and click
the "Delete" link next to the first "TCPIP_Illustrated_Vol-2".  A green
"Book deleted" status message should display at the top of the page,
along with a list of the eight remaining books.  You will also see the
cascading delete operation via the DBIC_TRACE output:

    SELECT me.id, me.title, me.rating FROM book me WHERE ( ( me.id = ? ) ): '6'
    DELETE FROM book WHERE ( id = ? ): '6'
    SELECT me.book_id, me.author_id FROM book_author me WHERE ( me.book_id = ? ): '6'
    DELETE FROM book_author WHERE ( author_id = ? AND book_id = ? ): '4', '6'


=head2 Fixing a Dangerous URL

Note the URL in your browser once you have performed the deletion in the
prior step -- it is still referencing the delete action:

    http://localhost:3000/books/id/6/delete

What if the user were to press reload with this URL still active?  In
this case the redundant delete is harmless (although it does generate
an exception screen, it doesn't perform any undesirable actions on the
application or database), but in other cases this could clearly be
extremely dangerous.

We can improve the logic by converting to a redirect.  Unlike
C<$c-E<gt>forward('list'))> or C<$c-E<gt>detach('list'))> that perform
a server-side alteration in the flow of processing, a redirect is a
client-side mechanism that causes the browser to issue an entirely
new request.  As a result, the URL in the browser is updated to match
the destination of the redirection URL.

To convert the forward used in the previous section to a redirect,
open C<lib/MyApp/Controller/Books.pm> and edit the existing
C<sub delete> method to match:

    =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_author' entries
        $c->stash->{object}->delete;
    
        # Set a status message to be displayed at the top of the view
        $c->stash->{status_msg} = "Book deleted.";
    
        # Redirect the user back to the list page.  Note the use
        # of $self->action_for as earlier in this section (BasicCRUD)
        $c->response->redirect($c->uri_for($self->action_for('list')));
    }


=head2 Try the Delete and Redirect Logic

Restart the development server and point your browser to
L<http://localhost:3000/books/list> (don't just hit "Refresh" in your
browser since we left the URL in an invalid state in the previous
section!) and delete the first copy of the remaining two
"TCPIP_Illustrated_Vol-2" books.  The URL in your browser should return
to the L<http://localhost:3000/books/list> URL, so that is an
improvement, but notice that I<no green "Book deleted" status message is
displayed>.  Because the stash is reset on every request (and a redirect
involves a second request), the C<status_msg> is cleared before it can
be displayed.


=head2 Using 'uri_for' to Pass Query Parameters

There are several ways to pass information across a redirect. One 
option is to use the C<flash> technique that we will see in Chapter 5 
of this tutorial; however, here we will pass the information via query 
parameters on the redirect itself.  Open 
C<lib/MyApp/Controller/Books.pm> and update the existing C<sub delete> 
method to match the following:

    =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_author' entries
        $c->stash->{object}->delete;
    
        # Redirect the user back to the list page with status msg as an arg
        $c->response->redirect($c->uri_for($self->action_for('list'),
            {status_msg => "Book deleted."}));
    }

This modification simply leverages the ability of C<uri_for> to include
an arbitrary number of name/value pairs in a hash reference.  Next, we
need to update C<root/src/wrapper.tt2> to handle C<status_msg> as a
query parameter:

    ...
    <div id="content">
        [%# Status and error messages %]
        <span class="message">[% status_msg || c.request.params.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 the Delete and Redirect With Query Param Logic

Restart the development server and point your browser to
L<http://localhost:3000/books/list> (you should now be able to safely
hit "refresh" in your browser).  Then delete the remaining copy of
"TCPIP_Illustrated_Vol-2".  The green "Book deleted" status message
should return.

B<NOTE:> Another popular method for maintaining server-side
information across a redirect is to use the C<flash> technique we
discuss in the next chapter of the tutorial,
L<Authentication|Catalyst::Manual::Tutorial::05_Authentication>. While
C<flash> is a "slicker" mechanism in that it's all handled by the
server and doesn't "pollute" your URLs, B<it is important to note that
C<flash> can lead to situations where the wrong information shows up
in the wrong browser window if the user has multiple windows or
browser tabs open>.  For example, Window A causes something to be
placed in the stash, but before that window performs a redirect,
Window B makes a request to the server and gets the status information
that should really go to Window A.  For this reason, you may wish
to use the "query param" technique shown here in your applications.


=head1 EXPLORING THE POWER OF DBIC

In this section we will explore some additional capabilities offered
by DBIx::Class.  Although these features have relatively little to do
with Catalyst per se, you will almost certainly want to take advantage
of them in your applications.


=head2 Add Datetime Columns to Our Existing Books Table

Let's add two columns to our existing C<books> table to track when
each book was added and when each book is updated:

    $ sqlite3 myapp.db
    sqlite> ALTER TABLE book ADD created INTEGER;
    sqlite> ALTER TABLE book ADD updated INTEGER;
    sqlite> UPDATE book SET created = DATETIME('NOW'), updated = DATETIME('NOW');
    sqlite> SELECT * FROM book;
    1|CCSP SNRS Exam Certification Guide|5|2009-03-08 16:26:35|2009-03-08 16:26:35
    2|TCP/IP Illustrated, Volume 1|5|2009-03-08 16:26:35|2009-03-08 16:26:35
    3|Internetworking with TCP/IP Vol.1|4|2009-03-08 16:26:35|2009-03-08 16:26:35
    4|Perl Cookbook|5|2009-03-08 16:26:35|2009-03-08 16:26:35
    5|Designing with Web Standards|5|2009-03-08 16:26:35|2009-03-08 16:26:35
    9|TCP/IP Illustrated, Vol 3|5|2009-03-08 16:26:35|2009-03-08 16:26:35
    sqlite> .quit
    $

This will modify the C<books> table to include the two new fields
and populate those fields with the current time.


=head2 Update DBIx::Class to Automatically Handle the Datetime Columns

Next, we should re-run the DBIC helper to update the Result Classes
with the new fields:

    $ script/myapp_create.pl model DB DBIC::Schema MyApp::Schema \
        create=static components=TimeStamp dbi:SQLite:myapp.db \
        on_connect_do="PRAGMA foreign_keys = ON"
     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"

Notice that we modified our use of the helper slightly: we told
it to include the L<DBIx::Class::TimeStamp|DBIx::Class::TimeStamp>
in the C<load_components> line of the Result Classes.

If you open C<lib/MyApp/Schema/Result/Book.pm> in your editor you
should see that the C<created> and C<updated> fields are now included
in the call to C<add_columns()>, but our relationship information below
the "C<# DO NOT MODIFY...>" line was automatically preserved.

While we have this file open, let's update it with some additional
information to have DBIC automatically handle the updating of these
two fields for us.  Insert the following code at the bottom of the
file (it B<must> be B<below> the "C<# DO NOT MODIFY...>" line and
B<above> the C<1;> on the last line):

    #
    # Enable automatic date handling
    #
    __PACKAGE__->add_columns(
        "created",
        { data_type => 'datetime', set_on_create => 1 },
        "updated",
        { data_type => 'datetime', set_on_create => 1, set_on_update => 1 },
    );

This will override the definition for these fields that Schema::Loader 
placed at the top of the file.  The C<set_on_create> and 
C<set_on_update> options will cause DBIx::Class to automatically 
update the timestamps in these columns whenever a row is created or 
modified.

To test this out, restart the development server using the
C<DBIC_TRACE=1> option:

    DBIC_TRACE=1 script/myapp_server.pl

Then enter the following URL into your web browser:

    http://localhost:3000/books/url_create/TCPIP_Illustrated_Vol-2/5/4

You should get the same "Book Created" screen we saw above.  However,
if you now use the sqlite3 command-line tool to dump the C<books> table,
you will see that the new book we added has an appropriate date and
time entered for it (see the last line in the listing below):

    $ sqlite3 myapp.db "select * from book"
    1|CCSP SNRS Exam Certification Guide|5|2009-03-08 16:26:35|2009-03-08 16:26:35
    2|TCP/IP Illustrated, Volume 1|5|2009-03-08 16:26:35|2009-03-08 16:26:35
    3|Internetworking with TCP/IP Vol.1|4|2009-03-08 16:26:35|2009-03-08 16:26:35
    4|Perl Cookbook|5|2009-03-08 16:26:35|2009-03-08 16:26:35
    5|Designing with Web Standards|5|2009-03-08 16:26:35|2009-03-08 16:26:35
    9|TCP/IP Illustrated, Vol 3|5|2009-03-08 16:26:35|2009-03-08 16:26:35
    10|TCPIP_Illustrated_Vol-2|5|2009-03-08 16:29:08|2009-03-08 16:29:08

Notice in the debug log that the SQL DBIC generated has changed to
incorporate the datetime logic:

    INSERT INTO book ( created, rating, title, updated ) VALUES ( ?, ?, ?, ? ): 
    '2009-05-25 20:39:41', '5', 'TCPIP_Illustrated_Vol-2', '2009-05-25 20:39:41'
    INSERT INTO book_author ( author_id, book_id ) VALUES ( ?, ? ): '4', '10'


=head2 Create a ResultSet Class

An often overlooked but extremely powerful features of DBIC is that it
allows you to supply your own subclasses of C<DBIx::Class::ResultSet>.
It allows you to pull complex and unsightly "query code" out of your
controllers and encapsulate it in a method of your ResultSet Class.
These "canned queries" in your ResultSet Class can then be invoked
via a single call, resulting in much cleaner and easier to read
controller code.

To illustrate the concept with a fairly simple example, let's create a
method that returns books added in the last 10 minutes.  Start by
making a directory where DBIx::Class will look for our ResultSet Class:

    $ mkdir lib/MyApp/Schema/ResultSet

Then open C<lib/MyApp/Schema/ResultSet/Book.pm> and enter the following:

    package MyApp::Schema::ResultSet::Book;
    
    use strict;
    use warnings;
    use base 'DBIx::Class::ResultSet';
    
    =head2 created_after
    
    A predefined search for recently added books
    
    =cut
    
    sub created_after {
        my ($self, $datetime) = @_;
    
        my $date_str = $self->result_source->schema->storage
                              ->datetime_parser->format_datetime($datetime);
    
        return $self->search({
            created => { '>' => $date_str }
        });
    }
    
    1;

Then add the following method to the C<lib/MyApp/Controller/Books.pm>:

    =head2 list_recent
    
    List recently created books
    
    =cut
    
    sub list_recent :Chained('base') :PathPart('list_recent') :Args(1) {
        my ($self, $c, $mins) = @_;
    
        # Retrieve all of the book records as book model objects and store in the
        # stash where they can be accessed by the TT template, but only
        # retrieve books created within the last $min number of minutes
        $c->stash->{books} = [$c->model('DB::Book')
                                ->created_after(DateTime->now->subtract(minutes => $mins))];
    
        # Set the TT template to use.  You will almost always want to do this
        # in your action methods (action methods respond to user input in
        # your controllers).
        $c->stash->{template} = 'books/list.tt2';
    }

Now start the development server with C<DBIC_TRACE=1> and try
different values for the minutes argument (the final number value) for
the URL C<http://localhost:3000/books/list_recent/10>.  For example,
this would list all books added in the last fifteen minutes:

    http://localhost:3000/books/list_recent/15

Depending on how recently you added books, you might want to
try a higher or lower value.


=head2 Chaining ResultSets

One of the most helpful and powerful features in DBIx::Class is that 
it allows you to "chain together" a series of queries (note that this 
has nothing to do with the "Chained Dispatch" for Catalyst that we 
were discussing above).  Because each ResultSet returns another 
ResultSet, you can take an initial query and immediately feed that 
into a second query (and so on for as many queries you need). Note 
that no matter how many ResultSets you chain together, the database 
itself will not be hit until you use a method that attempts to access 
the data. And, because this technique carries over to the ResultSet 
Class feature we implemented in the previous section for our "canned 
search", we can combine the two capabilities.  For example, let's add 
an action to our C<Books> controller that lists books that are both 
recent I<and> have "TCP" in the title.  Open up 
C<lib/MyApp/Controller/Books.pm> and add the following method:

    =head2 list_recent_tcp
    
    List recently created books
    
    =cut
    
    sub list_recent_tcp :Chained('base') :PathPart('list_recent_tcp') :Args(1) {
        my ($self, $c, $mins) = @_;
    
        # Retrieve all of the book records as book model objects and store in the
        # stash where they can be accessed by the TT template, but only
        # retrieve books created within the last $min number of minutes
        # AND that have 'TCP' in the title
        $c->stash->{books} = [$c->model('DB::Book')
                                ->created_after(DateTime->now->subtract(minutes => $mins))
                                ->search({title => {'like', '%TCP%'}})
                             ];
    
        # Set the TT template to use.  You will almost always want to do this
        # in your action methods (action methods respond to user input in
        # your controllers).
        $c->stash->{template} = 'books/list.tt2';
    }

To try this out, restart the development server with:

    DBIC_TRACE=1 script/myapp_server.pl

And enter the following URL into your browser:

    http://localhost:3000/books/list_recent_tcp/100

And you should get a list of books added in the last 100 minutes that
contain the string "TCP" in the title.  However, if you look at all
books within the last 100 minutes, you should get a longer list
(again, you might have to adjust the number of minutes depending on
how recently you added books to your database):

    http://localhost:3000/books/list_recent/100

Take a look at the DBIC_TRACE output in the development server log for
the first URL and you should see something similar to the following:

    SELECT me.id, me.title, me.rating, me.created, me.updated FROM book me 
    WHERE ( ( title LIKE ? AND created > ? ) ): '%TCP%', '2009-05-25 19:09:13'

However, let's not pollute our controller code with this raw "TCP"
query -- it would be cleaner to encapsulate that code in a method on
our ResultSet Class.  To do this, open
C<lib/MyApp/Schema/ResultSet/Book.pm> and add the following method:

    =head2 title_like
    
    A predefined search for books with a 'LIKE' search in the string
    
    =cut
    
    sub title_like {
        my ($self, $title_str) = @_;
    
        return $self->search({
            title => { 'like' => "%$title_str%" }
        });
    }

We defined the search string as C<$title_str> to make the method more
flexible.  Now update the C<list_recent_tcp> method in
C<lib/MyApp/Controller/Books.pm> to match the following (we have
replaced the C<-E<gt>search> line with the C<-E<gt>title_like> line
shown here -- the rest of the method should be the same):

    =head2 list_recent_tcp
    
    List recently created books
    
    =cut
    
    sub list_recent_tcp :Chained('base') :PathPart('list_recent_tcp') :Args(1) {
        my ($self, $c, $mins) = @_;
    
        # Retrieve all of the book records as book model objects and store in the
        # stash where they can be accessed by the TT template, but only
        # retrieve books created within the last $min number of minutes
        # AND that have 'TCP' in the title
        $c->stash->{books} = [$c->model('DB::Book')
                                ->created_after(DateTime->now->subtract(minutes => $mins))
                                ->title_like('TCP')
                             ];
    
        # Set the TT template to use.  You will almost always want to do this
        # in your action methods (action methods respond to user input in
        # your controllers).
        $c->stash->{template} = 'books/list.tt2';
    }

Then restart the development server and try out the C<list_recent_tcp>
and C<list_recent> URL as we did above.  It should work just the same,
but our code is obviously cleaner and more modular, while also being
more flexible at the same time.


=head2 Adding Methods to Result Classes

In the previous two sections we saw a good example of how we could use 
DBIx::Class ResultSet Classes to clean up our code for an entire query 
(for example, our "canned searches" that filtered the entire query). 
We can do a similar improvement when working with individual rows as 
well.  Whereas the ResultSet construct is used in DBIC to correspond 
to an entire query, the Result Class construct is used to represent a 
row. Therefore, we can add row-specific "helper methods" to our Result 
Classes stored in C<lib/MyApp/Schema/Result/>. For example, open 
C<lib/MyApp/Schema/Result/Author.pm> and add the following method (as 
always, it must be above the closing "C<1;>"):

    #
    # Helper methods
    #
    sub full_name {
        my ($self) = @_;
    
        return $self->first_name . ' ' . $self->last_name;
    }

This will allow us to conveniently retrieve both the first and last
name for an author in one shot.  Now open C<root/src/books/list.tt2>
and change the definition of C<tt_authors> from this:

    ...
      [% tt_authors = [ ];
         tt_authors.push(author.last_name) FOREACH author = book.authors %]
    ...

to:

    ...
      [% tt_authors = [ ];
         tt_authors.push(author.full_name) FOREACH author = book.authors %]
    ...

(Only C<author.last_name> was changed to C<author.full_name> -- the
rest of the file should remain the same.)

Now restart the development server and go to the standard book list
URL:

    http://localhost:3000/books/list

The "Author(s)" column will now contain both the first and last name.
And, because the concatenation logic was encapsulated inside our
Result Class, it keeps the code inside our TT template nice and clean
(remember, we want the templates to be as close to pure HTML markup as
possible). Obviously, this capability becomes even more useful as you
use to to remove even more complicated row-specific logic from your
templates!


=head2 Moving Complicated View Code to the Model

The previous section illustrated how we could use a Result Class 
method to print the full names of the authors without adding any extra 
code to our view, but it still left us with a fairly ugly mess (see
C<root/src/books/list.tt2>):

    ...
    <td>
      [% # NOTE: See Chapter 4 for a better way to do this!                      -%]
      [% # First initialize a TT variable to hold a list.  Then use a TT FOREACH -%]
      [% # loop in 'side effect notation' to load just the last names of the     -%]
      [% # authors into the list. Note that the 'push' TT vmethod does not print -%]
      [% # a value, so nothing will be printed here.  But, if you have something -%]
      [% # in TT that does return a method and you don't want it printed, you    -%]
      [% # can: 1) assign it to a bogus value, or 2) use the CALL keyword to     -%]
      [% # call it and discard the return value.                                 -%]
      [% tt_authors = [ ];
         tt_authors.push(author.full_name) FOREACH author = book.authors %]
      [% # Now use a TT 'virtual method' to display the author count in parens   -%]
      [% # Note the use of the TT filter "| html" to escape dangerous characters -%]
      ([% tt_authors.size | html %])
      [% # Use another TT vmethod to join & print the names & comma separators   -%]
      [% tt_authors.join(', ') | html %]
    </td>
    ...

Let's combine some of the techniques used earlier in this section to 
clean this up.  First, let's add a method to our Book Result Class to 
return the number of authors for a book.  Open 
C<lib/MyApp/Schema/Result/Book.pm> and add the following method:

    =head2 author_count
    
    Return the number of authors for the current book
    
    =cut
    
    sub author_count {
        my ($self) = @_;
    
        # Use the 'many_to_many' relationship to fetch all of the authors for the current
        # and the 'count' method in DBIx::Class::ResultSet to get a SQL COUNT
        return $self->authors->count;
    }

Next, let's add a method to return a list of authors for a book to the
same C<lib/MyApp/Schema/Result/Book.pm> file:

    =head2 author_list
    
    Return a comma-separated list of authors for the current book
    
    =cut
    
    sub author_list {
        my ($self) = @_;
    
        # Loop through all authors for the current book, calling all the 'full_name' 
        # Result Class method for each
        my @names;
        foreach my $author ($self->authors) {
            push(@names, $author->full_name);
        }
    
        return join(', ', @names);
    }

This method loops through each author, using the C<full_name> Result 
Class method we added to C<lib/MyApp/Schema/Result/Author.pm> in the 
prior section.

Using these two methods, we can simplify our TT code.  Open
C<root/src/books/list.tt2> and update the "Author(s)" table cell to
match the following:

    ...
    <td>
      [% # Print count and author list using Result Class methods -%]
      ([% book.author_count | html %]) [% book.author_list | html %]
    </td>
    ...

Although most of the code we removed comprised comments, the overall 
effect is dramatic... because our view code is so simple, we don't 
huge comments to clue people in to the gist of our code.  The view 
code is now self-documenting and readable enough that you could 
probably get by with no comments at all.  All of the "complex" work is 
being done in our Result Class methods (and, because we have broken 
the code into nice, modular chucks, the Result Class code is hardly 
something you would call complex).  

As we saw in this section, always strive to keep your view AND 
controller code as simple as possible by pulling code out into your 
model objects.  Because DBIx::Class can be easily extended in so many 
ways, it's an excellent to way accomplish this objective.  It will 
make your code cleaner, easier to write, less error-prone, and easier 
to debug and maintain.

Before you conclude this section, fire up the development server and
hit Refresh in your browser... the output should be the same even
though the backend code has been trimmed down.


=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/5.80/trunk/lib/Catalyst/Manual/Tutorial/>.

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