Small cover
Embedding Perl in HTML with Mason
Dave Rolsky
Ken Williams

Table of Contents | Foreword | Preface
Chapters: 1 2 3 4 5 6 7 8 9 10 11 12
Appendices: A B C D
Glossary | Colophon | Copyright


Chapter 2: Components

As mentioned in Chapter 1, the basic building block of Mason is called a component. A component consists of text of any sort as well as Mason-specific markup syntax. This chapter briefly introduces some core Mason concepts and then goes into the nitty-gritty of component syntax.

In this chapter we'll introduce you to the syntax of Mason components, but we won't spend much time on semantics. In most of the sections, we refer to other parts of the book where you can find out more about each concept.

Mason from 10,000 Feet

In order to put Mason into perspective, a basic understanding of how Mason processes a request is helpful. Each request is defined by an initial component path and a set of arguments to be passed to that component.

Requests are handled by the Interpreter object. You can use it directly or its API can be called by the ApacheHandler or CGIHandler modules provided with Mason.

The Interpreter asks the Resolver to fetch the requested component from the filesystem. Then the Interpreter asks the Compiler to create a "compiled" representation of the component. Mason's compilation process consists of turning Mason source code into Perl code, which is then executed in order to create an object representing the component. Mason stores this generated Perl code on disk, so that it doesn't need to go through the parsing and compilation process for every request, and stores the compiled code in an LRU (least recently used) cache in memory.

Once Mason has an object representing the initial component, it creates a request object and tells it to execute that component. The initial component might call several other components during the request. Any output a component generates is sent to STDOUT, which is a reasonable default for most environments in which Mason might be used. Of course, it is possible to change this default and send output elsewhere.

Several parameters can change how elements of this process happen, and you can replace the core Mason classes with your own customized subclasses for specialized behavior. When using the ApacheHandler module, all of these parameters can be specified in the web server's configuration file.

If a fatal error occurs during any part of this process, Mason throws an exception via Perl's built-in die() function. In a mod_perl or CGI environment, Mason will make sure that this exception is handled in a reasonable way, by showing the error in the browser and/or recording the error in the server's error log. You can also catch exceptions in your own code and handle them as you please.

Core Concepts

Before diving into component syntax, it is important to understand a few basic Mason concepts, with the key concepts highlighted in italics.

First there is the component . A component is a combination of text and Mason-specific markup. The markup sections may contain Perl code or special Mason directives. A component can correspond to a single web page, but more often a page is built up from several components. However, a component always corresponds to a single file.

A component is usually expected to generate output of some sort, whether HTML, an email message, or an image file. Components are closely analogous to Perl subroutines.

The component root is a directory or list of directories on the filesystem under which Mason expects to find all of your components. This is important in determining how component calls are resolved. If you ask Mason to execute the component /view/books.comp, Mason needs to know where to find such a thing. If your component root is /var/www/mason, Mason will look for a file called /var/www/mason/view/books.comp.

The process of resolving a component path to a component can actually be a bit more complex than that, because you may actually specify multiple directories in which to search for components or use another storage mechanism altogether. We'll leave those complexities aside for now.1 When running under Apache, either via mod_perl or CGI, Mason will default to using the

web server's document root as the component root. Mason may also be used in ways that don't require a component root at all, such as from a standalone perl script. Since the focus of this book is on building sites, we will generally assume that there is a component root unless we mention otherwise.

It is very important to understand that component paths, like URL paths, always use the forward slash (/) as their directory separator, no matter what operating system Mason is running on. In other words, a component path can be thought of as a unique identifier for a particular component, in much the same way that a URL is a unique identifier for a particular resource. Also much like a URL, a component path usually corresponds to a file on disk with a related path, but it needn't necessarily.

Basic Component Syntax

Mason parses components by taking the text of a component and translating it into actual Perl code. This Perl code, when executed, creates a new HTML::Mason::Component object. This object, in turn, can be used to generate the text originally found in the component. In a sense, this inverts the component, turning it from text with embedded Perl into Perl with embedded text.

The markup language Mason uses can give certain parts of the component special semantics, just like any other markup language such as XML or HTML. In this case, the syntax is used to tell Mason that certain parts of the component's text represent either Perl code, special instructions for Mason, or in some cases both.

The markup language used for Mason components contains a simple tag to do in-place substitution of Perl expressions, a way to mark a single line as being a line of Perl, and a set of block tags, most of which contain Perl code that is given a special meaning based on the particular tag being used (see Table 2-1).

Tag Name Contains
<% ... %> Substitution Perl that is evaluated and sent as output
% ... Perl line A single line of Perl code 2
<%perl> ... </%perl> Perl block Perl code
<& ... &> Component call A call to another component, possibly with arguments
<%init> ... </%init> init block Perl code that executes before the main body of the component
<%args> ... </%args> args block A component's input argument declarations
Table 2-1. A portion of Mason's markup language

Substitution Tags: <% %>

The simplest kind of Mason tag is the substitution tag, used to insert the results of a Perl expression into your text. This tag is quite similar to those found in other templating systems. A simple example might look like this:

  % $cd_count = 207; # this is embedded Perl
  You have <% $cd_count %> CDs.

The output of this example would be:

  You have 207 CDs.

The contents of the tag are evaluated in a list context and joined together just as if they had been passed as arguments to Perl's built-in print() function.

It is possible, and often desirable, to put more complicated Perl expressions into your substitution tags. For example, to handle plurals properly, the second line in the previous example could be rewritten as:

  You have <% $cd_count %> CD<% $cd_count != 1 ? 's': '' %>

This could output any of the following, depending on the value of the $cd_count variable:

  You have 207 CDs.
  You have 1 CD.
  You have 0 CDs.

The contents of the substitution tag are evaluated as Perl code, so whitespace is ignored, meaning <%$cd_count%> would be perfectly valid, though perhaps a bit difficult to read. Our style is to always include whitespace in a substitution tag.

Escaping substitutions

One very useful feature provided by Mason is the ability to escape the contents of a tag before it is sent as output. Escaping is the process of making unsafe characters safe. In a web context, safe means that we do not generate output that could be mistaken for HTML. In addition, we may need to do URL-style escaping as well.

Substitution escaping is indicated with a pipe (|) followed by one or more escape flags placed before the close of the tag. Currently, there are three valid escape flags, h for HTML entity escaping (i.e., > into &gt; ), u for URI escaping (i.e., > into %3E ), and n for no escaping. The HTML and URI escape flags can be combined (i.e., hu) or used separately. An example might look like:

  Name: <% $name | h %>
  Homepage: <a href="redirect?url=<% $homepage | u %>">

The HTML escaping mode escapes the string using the HTML::Entities module, which means that all control and high-bit characters are escaped, as well as the greater-than and l ess-than signs (< and >), the ampersand (&), and the double quote character (").

HTML escaping is particularly useful when you're populating a page with data from an external data source like a database. For instance, consider the following code:

  <textarea name="foo"><% $foo_data %></textarea>

If $foo_data contains the string </textarea>, your HTML will be broken. Guard against this possibility by escaping the output:

  <textarea name="foo"><% $foo_data | h %></textarea>

Mason uses HTML::Entities internally but does not provide a way to tell HTML::Entities not to escape certain characters. By default, HTML::Entities assumes that you are using the ISO-8859-1 character set and escapes characters accordingly. If you are generating text for another character set, such as Big5, you will need to override the way this escaping is done.

The URI escaping mode escapes any character besides alphanumerics, the underscore ( _ ), dash ( - ), and period ( . ).

The "no escape" escaping mode is used when you have set a default escaping mode via the default_escape_flags parameter (see Chapter 6 for details). The n flag turns off the default escaping for the substitution tag in which it is used.

If you want to escape using a different mode than the default, you can combine the n escape with another flag, for example:

  # default is 'u'
  <% $contains_html | nh %>

The use of spaces around the pipe is optional.

The purist will note that $variable | h is perfectly valid Perl syntax for obtaining the value of $variable bitwise OR'ed against the output of the h subroutine (or perhaps the bareword string h ), and therefore this valid Perl construct has a different meaning in <% %> tags than it has in other Perl environments. If you really mean to do the bitwise OR (in which case we strongly suspect you really shouldn't mean to), a workaround looks like this:

  <% ($variable | h) %>

No doubt this will cause much consternation among those who write code that involves OR-ing together variables and the output of subroutines with single character names, who are being made second-class citizens in the Mason world. Sorry, but we're standing firm here.

As of version 1.14, Mason supports user-defined escapes so that you can create your own escaping flags, or override existing flags. This is not documented in the book at present, so please see the documentation for the HTML::Mason::Interp module in a recent release of Mason for details.

Embedded Perl: % Lines and <%perl> Blocks

There are two ways to embed Perl code into text with Mason. The first, the Perl line, is a line that starts with a percent sign ( % ). The rest of that line (up to the newline character) is interpreted as Perl code. This percent sign cannot be preceded by any horizontal whitespace such as spaces or tabs. A typical use of these lines is to implement Perl control structures. For example:

  % foreach my $person (@people) {
    Name: <% $person->{name} %>
    Age: <% $person->{age} %>
    Height: <% $person->{height} %>
    Weight: <% $person->{weight} %>
  % }

You can put any valid piece of Perl code on these lines. It is possible to use a Perl line for a larger chunk of code too -- the previous code could have been equivalently written like the following:

  % foreach my $person (@people) {
  %   print "Name: ", $person->{name}, "\n";
  %   print "Age: ", $person->{age}, "\n";
  %   print "Height: ", $person->{height}, "\n";
  %   print "Weight: ", $person->{weight}, "\n";
  % }

If you have more than a few lines of Perl code in a row, however, it is probably best to use a Perl block instead. A Perl block is equivalent to a bunch of Perl lines in a row. It begins with the start tag <%perl> and ends with the end tag </%perl>. The contents of these blocks may be any valid Perl code.

You may want to use this tag if you need to do some data processing in the midst of your text. For example:

  <%perl>
   my @words = $sentence =~ /\b(\S+)\b/g;
   my @small_words = grep { length $_ <= 3 } @words;
   my @big_words = grep { length $_ > 3 } @words;
  </%perl>
  There were <% scalar @words %> in the sentence.
  The big words were:
  % foreach my $word (@big_words) {
    <% $word %>
  % }
  The small words were:
  % foreach my $word (@small_words) {
    <% $
  word %>
  % }

Calling Other Components: <& &> Tags

One of the most powerful features in Mason is the ability of one component to execute another, causing the called component's output to appear inside the calling component's output. The called component can, in turn, call other components, and so on. There are several ways to call components, but the simplest way is via the ampersand tag, like this:

  <html>
  <head>
  <title>The Goober Guide</title>
  </head>
  <body>
  <h1>Welcome to The Goober Guide!</h1>
  <& menu &>
  ...
  </body>
  </html>

The menu component might contain a navigation bar used on all the pages for a site. Other example calls might look like this:

  <& /path/to/menu &>
  <& $component &>
  <& menu, width => 640, admin => 1 &>

These calls illustrate several facets of Mason's component call tag. First, the component can be specified either directly using its name in plain text or indirectly as the result of Perl expression like $component in the example. In addition, component calls can take arguments (like width and admin in the third example) just like a Perl subroutine -- internally, they actually are subroutines.

How does Mason figure out which component calls are specified directly and which indirectly? It applies some simple parsing rules. In a component call tag, if the first nonwhitespace character is a letter, number, underscore ( _ ), slash ( / ), or period ( . ), Mason assumes that this text is a plain text component path rather than a Perl expression. In that case, everything up to the first comma or end of the tag (&>), whichever comes first, is assumed to be a string specifying the component path. Anything after a comma, if present, will be considered a list of arguments to pass to the called component.

If the first nonwhitespace character is something else, it is assumed that the component call contains a Perl expression (perhaps a variable or function call) whose value indicates the desired component.

These rules may seem a little arcane, but they manage to capture most people's expectations pretty well. Most of the time you can just specify the component in the most natural way, and it will just work. If you want to use a Perl expression (the "indirect" syntax) starting with one of the special characters mentioned in the previous paragraph, however, it is necessary to do something to force Mason to see it as Perl. An easy way to do this is to wrap the Perl expression in parentheses or to prefix it with Perl's no-op unary plus operator (+). For example:

  <& ( component_path_returner( ) ) &>
  <& +component_path_returner( ) &>

An alternative to the <& &> syntax for calling other components is the $m->comp() method. The $m variable contains the HTML::Mason::Request object for the current request, and you may use the $m->comp() method in Perl code just as you would use a <& &> tag in the component body.3 In fact, in the current version of Mason, the <&&> tag is implemented internally with the $m->comp() method. So the following two lines are equivalent to each other:

  <& menu, width => 640, admin => 1 &>
  % $m->comp('menu', width => 640, admin => 1);

Notice how we used a Perl line, starting with a %, to embed the $m->comp() call in the component.

In this section we have been intentionally vague about how a Perl expression "specifies" a component. There are two ways it may do so: it may either evaluate to a string that gives the path to the component or evaluate to a component object,4 rather than a path, and that object will then be executed. There are a few idioms in which this is useful, but they're used fairly rarely, and you'll mostly call components by their paths.

Components called with content

With Version 1.10, Mason introduced support for a powerful new construct, which we call "components called with content." Using this construct, it is possible to pass a Mason content block as part of a component call. Here is an example:

  <%args>
   $name
  </%args>
  <&| /i18n/itext, lang => $lang &>
  %# The bits in here will be available from $m->content in the /i18/text
    <en>Hello, <% $name %>.  These words are in English.</en>
    <fr>Bonjour, <% $name %>, ces mots sont franE<#xC3>E<#xA7>ais.</fr>
    <pig>Ellohay <% substr($name,2) . substr($name,0,1) . 'ay' %>, 
         esethay ordsway areyay inyay Igpay Atinlay.</pig>
  </&>

Presumably, we expect the /i18n/itext component to filter this text so that only the correct language, as specified in $lang, is used. The /i18n/itext component would probably look something like this, using the $m->content() method to retrieve the content block:

  <% $text %>
  <%args>
   $lang
  </%args>
  <%init>
   my ($text) = $m->content =~ m{<$lang>(.+?)</$lang>}s;
  </%init>

The content block gets executed when $m->content() is called, but it still has access to the variables as declared in the original component, such as $name. Components called with content will be covered in more depth in Chapter 5.

Other Named Blocks

Mason has a variety of other named blocks. These all have the same start and end tag syntax as <%perl> blocks, and most of them contain plan Perl. However, these other blocks are interpreted as having special meanings by Mason.

If any of these blocks, or a <%perl> block, is immediately followed by a newline, then that newline is discarded from the text output. This is a convenience to prevent you from having to do this all over the place:

  <%args>
   ...
  </%args><%init>
   ...
  </%init><%perl>
   ...
  </%perl>This is the start of the component ...

<%init> blocks

This is one of the most commonly used Mason blocks. The Perl code it contains is run before any other code except for code in <%once> or <%shared> blocks. It is run every time the component is called.

Using this block achieves the same effect as putting a <%perl> block at the top of a component but may be aesthetically more pleasing, because it allows you to isolate code at the bottom of a component, out of the way of the component's main body.

The <%init> block is typically used for doing things like checking arguments, creating objects, or retrieving data from a database. The variables created here are used in substitutions and perl lines throughout the rest of the component.

  It is currently <% $temp %> degrees.
  <%init>
   my ($temp) = $dbh->selectrow_array("SELECT temperature FROM current_weather");
  </%init>

<%args> blocks

As we have mentioned, components can take a variety of arguments, from either an external source (an HTTP request, for example) or an internal one (one component calling another).

It is usually desirable to declare the names and datatypes of the arguments that a component expects, as well as default values for these arguments, if they have any. This is done via the <%args> block. A typical block might look like this:

  <%args>
   $color
   $size  => 20  # A default value
   @items => ( 1, 2, 'something else' )
   %pairs => ( key1 => 1, key2 => 'value' )
  </%args>

This example demonstrates all the syntax possibilities for this block. First of all, we have argument types and names. The valid types are scalar, array, and hash, represented by their corresponding Perl sigil ($, @, or %), exactly as would be expected.

It is possible to give an argument a default value to be used if none is provided when the component is called. Any argument without a default is considered a required argument. Calling a component without specifying all its required arguments will cause a fatal exception to be thrown.

An argument's default can refer to an earlier argument, so this is completely legal:

  <%args>
   $x
   $y => $x * 2 > 20 ? 50 : 100
  </%args>

While this block looks as if it contains Perl, it is important to realize that its syntax is actually something unique to Mason. Importantly, lines should not end with a semicolon or comma, and each variable definition must be on a single line.

It is possible to have comments both after an argument declaration and on their own line. Comments start with the # character and continue to the end of the line, just as in Perl. Blank lines are also allowed.

<%filter> blocks

A <%filter> block is called after a component has finished running. It is given the entire output of the component in the $_ variable, and any changes to this variable are reflected in the output of the component. For example, this filter uppercases all of the component's output:

  <%filter>
   s/(\w+)/\U$1/g
  </%filter>

<%once> blocks

This block is executed whenever the component is loaded into memory. It is executed before any other block (including an <%init> block). Any variables declared here remain in existence (and in scope) until the component is flushed from memory or the Perl interpreter running Mason shuts down, whichever comes first. The <%once> section is useful for things like creating database handles or instantiating large, resource-intensive objects.

  The universe is this big: <% $size %>
  <%once>
   my $size = calculate_size_of_universe( );
  </%once>

<%cleanup> blocks

The cleanup block is executed right before the component exits and is the counterpart to the <%init> block. It is useful if you have created resources -- such as circular references -- that need to be freed. Technically, it is the same as placing a <%perl> block at the end of a component.

  <%init>
   my $resource = get_a_resource( );
  </%init>
  
  ... do something interesting with that resource
  
  <%cleanup>
   $resource->dispose;
  </%cleanup>

Since cleanup code tends to be put at the end of the component anyway, <%cleanup> blocks aren't very common. Their chief advantage is that their name is cleanup .

Cleanup blocks are not executed if the component dies or aborts.

<%text> blocks

The contents of this block are output exactly as they are, without any parsing. This if useful if you need to write a component containing text about Mason. For example:

  <%text>
   Substitution tags look like this: <% $var %>.
  </%text>

<%doc> blocks

This block is intended for use by component authors for documentation purposes. Its contents are completely ignored. In the future Mason may do something more useful with them.

  <%doc>
  =head1 My Story
  
  This is the part where I tell you what the component does.  But I'd
  rather tell you a story about my childhood.  When I was but a
  child, my mother said to me ...
  </%doc>

As you can see, there's no reason not to use POD (Perl's Plain Old Documentation markup language) in these blocks, and you can even run perldoc on a component file.

<%flags> and <%attr> blocks

These two blocks share the same syntax and are used to declare one or more key/value pairs. The key can contain only letters, numbers, and the underscore character ( _ ). The value can be any Perl expression whose results can fit into a scalar (such as a number, string, reference, or undef ).

As in the <%args> block, the syntax in these blocks looks like Perl, but it is not. First, you cannot end a line with a comma or semicolon. Second, the whole key/value pair must be on a single line.

The difference between these two is that the <%flags> block may contain only official Mason flags, which are used to affect the component's behavior. Currently, there is only one flag defined, inherit. This is used to specify the component's parent component. Component inheritance is discussed in Chapter 3.

The <%attr> block may contain any keys that you want, as the variables defined in this block are not used by Mason but may be used in your code. Its contents are available by calling the object's attr() method and giving the desired key as the argument. See Chapter 5 for the details.

  <%flags>
   inherit => '/some/other/component'
  </%flags>
  
  <%attr>
   color => "I'm so blue"
   size => 'mucho grande'
  </%attr>
  
  My color: <% $m->base_comp->attr('color') %>

There is one other important difference between flags and attributes: flags refer to only the current component, whereas attributes are part of Mason's inheritance scheme, discussed in Chapter 5.

<%def> and <%method> blocks

These two blocks use a syntax slightly different from any other Mason block because their contents are, in turn, components. The <%def> block contains a subcomponent, an embedded component that can be called via the normal Mason component calling syntax. A <%method> block also contains an embedded component, but one that may be inherited by a component's children. <%def> and <%method> blocks require a name in the initial tag. In the following example, a subcomponent named .make_a_link is defined:

  <%def .make_a_link>
   <a href="<% $url %>"><% $text %></a>
   <%args>
    $path
    %query => ( )
    $text
   </%args>
   <%init>
    my $url = ...
   </%init>
  </%def>

The name of a subcomponent or method may contain alphanumerics, underscores ( _ ), dashes ( - ), or periods ( . ). Customarily, a period is the first character of subcomponent names, in order to distinguish them from nonembedded components. Methods generally do not follow this convention; they have names without leading periods.

The main difference between subcomponents and methods is simply that subcomponents are visible only within the component in which they are defined, whereas methods are visible outside of the component and can be inherited via Mason's component inheritance mechanism. Subcomponents and methods are covered in Chapter 5.

<%shared> blocks

This block also contains Perl code. Code in this block is executed once per request, before the <%init> block, but unlike in an <%init> block, the variables declared in this block are in scope both in the component's main body and in any subcomponents or methods it may contain. This is useful for sharing a common chunk of code between all the parts of a single component. The uses of this block are discussed in Chapter 5.

Escaping a Newline

When using Mason, you may find that you want to suppress a newline in your text. A typical example is this:

  <pre>
  I am
  % if ($height < 5) {
   not
  % } elsif ( $height < 5.75 ) {
   not very
  % } elsif ( $height > 6.25 ) {
   very
  % }
  tall
  </pre>

This will generate the following output if $height is less than 5:

  <pre>
  I am
   not
  tall
  </pre>

The newlines in the output are not desirable but are unavoidable because of the need for the Perl code to exist on separate lines. Mason therefore provides the ability to get rid of a newline simply by preceding it with a backslash ( \ ).

If we rewrote the preceding example with escaped newlines, it would look like this:

  <pre>
  I am\
  % if ($height < 5) {
   not\
  % } elsif ( $height < 5.75 ) {
   not very\
  % } elsif ( $height > 6.25 ) {
   very\
  % }
  tall
  </pre>

Given this, the output for a $height less than 5 would then be:

  <pre>
  I am not tall
  </pre>

This example could be redone on a single line using multiple <%perl> blocks, but it would be pretty hideous looking.

Component Arguments

Most components will expect to receive named arguments, and these can be passed in one of two ways. Components can receive arguments as the result of external requests, such as those via HTTP, or they can receive arguments when they are called from another component. These arguments are available in the called component via several mechanisms. But from a component's perspective, how it is called is largely irrelevant.

<%args> Block Revisited

Since we are talking about arguments, it is worth revisiting the <%args> block discussed previously. This block is used to declare the arguments that a component expects. In addition, it can also be used to specify a default value if none is given when the component is called.

The block we used earlier was:

  <%args>
   $color
   $size => 20
   @items => ( 1, 2, 'something else' )
   %pairs => ( key1 => 1, key2 => 'value' )
  </%args>

This says, in English, that this component expects two scalars, one named color , which is mandatory, and one named size , which is not mandatory and defaults to 20. It also expects an array named items , which defaults to (1, 2, 'something else') and a hash named pairs, which defaults to (key1 => 1, key2 => 'value' ). Neither of these latter two arguments is mandatory.

These arguments are all available in your component as lexically scoped variables. For example, your component will have a lexically scoped $color variable available. You do not need to declare it anywhere but in the <%args> block.

If a mandatory argument (one with no default) is not provided in the call to the component, an exception is thrown. If an argument with a default is not given a value, the default is transparently assigned to the variable. Just to be clear, we will explicitly note that undef is a valid value for an argument. It is the absence of an argument that causes the exception.

%ARGS

In addition to any lexically scoped variables created via their declaration in an <%args> block, each component body also has a lexically scoped hash called %ARGS. This hash contains all of the arguments with which the component was called.

One point of confusion for those new to Mason is the difference between %ARGS and the <%args> block. The %ARGS hash contains the arguments exactly as they were passed to a component, whether or not they are declared in the <%args> block. The keys of the %ARGS hash do not contain the Perl sigils ($, @, or %). An argument declared as $color in the <%args> block would therefore be available via $ARGS{color}. Any assignment of defaults by the <%args> block is not visible in %ARGS; the values are given exactly as they were passed.

In addition, the %ARGS hash is always present,5 but the <%args> block is optional.

If you are expecting input with a large number of similarly named items, such as input1 , input2 , and so on through input20 , declaring all of them in an <%args> block may be a bit unwieldy. In this case, the %ARGS hash can be quite handy.

%ARGS is also useful if you expect arguments with names that cannot be used for Perl variables. For example, when submitting a web form by clicking on an image named submit , the browser will generate two additional form values, called submit.x and submit.y . You cannot have a Perl variable named $submit.x, so the only way to get at this argument is to check $ARGS{'submit.x'}.

There are other ways to retrieve the arguments passed to a component, which are discussed in Chapter 4.

%ARGS Versus @_

The Mason tradition has always been to use named arguments. However, for simple components, you may prefer to use @_ to access the arguments, just as in Perl subroutines. There are several caveats here. If your component contains an <%args> section, Mason expects it to receive an even number of arguments in @_ so that it can assign @_ to %ARGS. If it receives an odd number of arguments, a fatal error will occur. But regardless of how arguments are passed, @_ is always available in components.

So the following pieces of code are near-identical when a component receives an even number of arguments:

  % foreach (sort %ARGS) {
    <% $_ %>
  % }
  
  % foreach (sort @_) {
    <% $_ %>
  % }

Argument Examples

Let's take a look at a number of scenarios involving

argument passing, first via an HTTP URL query string and then via an internal component call. Then we will see how this interacts with the component's <%args> block and the %ARGS hash.

Arguments submitted via POST and GET requests are treated in exactly the same way, and if both are present they are merged together before the component is called.

Let's assume that the component being called contains this <%args> block:

  <%args>
   $colors
   @colors
   %colors
  </%args>

For each example, we show you two ways to call that component. The first is via an HTTP query string, which is how a component is called to generate a web page. The second is via a component call tag, as a component would be called from another Mason component.

  • /some/component?colors=blue <& /some/component, colors => 'blue' &>

    In both cases, $colors is the string "blue" and @colors is a single-element array containing ('blue'). In addition, $ARGS{colors} would be the string "blue" as well.

    This component will die when it is called, however, because Mason does not allow you to assign an odd number of elements to a hash, so the assignment to %colors is fatal.

  • /some/component?colors=blue&colors=red&colors=green <& /some/component, colors => [ 'blue', 'red', 'green' ] &>

    Again the URL and internal example give the same result. The $colors variables contains a reference to a three-element array, ['blue','red','green']. This time, $ARGS{colors} contains the same three-element array reference as $colors and the @colors array contains a three-element array with those same elements.

    Again, assigning an odd number of elements to the %colors hash causes a fatal error.

  • /some/component?colors=blue&colors=cyan&colors=green&colors=mint <& /some/component, colors => [ 'blue', 'cyan', 'green', 'mint' ] &>

    Now, $colors contains a reference to a four-element array, and the @colors array has four elements as well. Finally, the assignment to %colors works without an error and will result in a hash containing ('blue'=>'cyan','green'=>'mint' ). $ARGS{colors} contains the same array reference as $colors.

  • <& /some/component, colors => { blue => 'cyan', green => 'mint' } &>

    This set of arguments isn't representable with a query string, because there's no way to indicate that the arguments are structured in a hash via a web request.

    In this call, $colors contains a reference to a hash, not an array, though the @colors array contains four elements, just as in the previous example. The %colors hash is likewise the same as the previous example. Now, the $ARGS{colors} hash entry contains a hash reference.

This discrepancy in how hash assignments are treated, depending on the way a call is made, is probably not too important because Mason simply does the right thing based on the contents of %args. You declare %colors as an argument, and as long as an even number of colors elements are passed in, you get a hash.

Arguments via Component Calls

When calling another component that expects named

arguments, it is important to remember that arrays and hashes need to be passed as references. For example, a component named /display with an <%args> block like this:

  <%args>
   @elements
   %labels
  </%args>

Should be called like this:

  <& /display, elements => \@some_data, labels => \%data_labels &>

Mason will do the right thing and translate the references back into an array and a hash in the /display component.

Arguments via HTTP Requests

When using Mason to make a web application, you must understand the details of how external HTTP requests are converted into component calls. Specifically, we are interested in how query string and POST parameters are converted into arguments.

These requests are expected to be in the standard name/value pair scheme used by most web interfaces. If a parameter is given only once (i.e., component?foo=1&bar=2), it will be present in the %ARGS hash as a simple scalar, regardless of how it is declared in the <%args> section.

If a parameter is declared as a scalar ($foo) but given multiple values (i.e., component?foo=1&foo=2), the $foo parameter will end up containing a reference to an array, as will $ARGS{foo}. Future versions of Mason may provide the ability to coerce these arguments into specific data structures.

If a parameter is declared as an array (@foo), it will contain zero or more values depending on what is in the query string and/or POST data. A hash is treated more or less like an array, except that giving a parameter declared as a hash an odd number of values will cause a fatal error.

One caution: the key/value associations in a declared hash are determined by the order of the input. Let's assume we have a component with this <%args> block:

  <%args>
   %foo
  </%args>

A request for component?foo=1&foo=2 will result in a different hash from component?foo=2&foo=1. This isn't generally a problem because you can usually control the order of the arguments by their position in an HTML form. However, neither the HTTP or HTML specifications specify that a client needs to respect this ordering when submitting the form, and, even if it were, some browsers would probably screw it up eventually.6 It's not a great idea, therefore, to use hashes as arguments in a top-level component that may be called via an HTTP request generated by a form. When you can control the query string yourself, this is not a problem.

Component Return Values

So far, we know three ways to call c omponents: by using the inline component call tag (<& &>), by using the $m->comp() method, or via a URL. When using a component call tag, the called component's output is placed exactly where the tag was. When a component is called via a URL, its output is sent to the client. The $m->comp() tag offers an additional channel of component output: the return value. By default, Mason components return undef. If you want to return something else, you can add an explicit return() statement inside that component, such as this:

  <%init>
   my $size = 20;
   return $size;
  </%init>

Perl's return() function will end processing of the component, and any values specified will be the return value of $m->comp(). Since Perl's normal rules of scalar/list context apply, a component may return either a scalar or a list.

Special Globals

All Mason components are given access to certain special variables. We have already discussed %ARGS, which is lexically scoped for each component. Mason also has a few special global variables available.

$m

This variable is an HTML::Mason::Request object, which has a number of methods that allow you to do things such as retrieve information on the current request, call other components, or affect the flow of execution. This object is discussed in detail in Chapter 4.

$r

If Mason is running under mod_perl (as is the case in most Mason setups), all components also have access to the Apache request object via the global variable $r. Mason's special hooks into mod_perl are covered in Chapter 7.

Sample Component

The component shown in Example 2-1 is part of our sample site and the focus of Chapter 8. The component here is responsible for displaying news about the site. It is called news.mas and is not intended to standalone by itself, but rather to form one part of a complete page.

It demonstrates a typical small Mason component. Its <%init> block does some very simple work to figure out the time that the file was last altered, and then it turns that time into a human-readable string.

Example 2-1. news.mas
  <table width="100%" cellspacing="0" cellpadding="5">
   <tr>
    <td class="heading"><h2 class="headline">What's New?</h2></td>
   </tr>
   <tr>
    <td>
     <p>
     The whole site, at this point.
     </p>
  
     <p>
     <em>Last modified: <% $last_mod %></em>
     </p>
    </td>
   </tr>
  </table>
  <%init>
   my $comp_time = (stat $m->current_comp->source_file)[9];
   my $last_mod =
       Time::Piece->strptime( $comp_time, '%s' )->strftime( '%B %e, %Y %H:%M' );
  </%init>

No single component can demonstrate all of Mason's features, so if you're curious to see more, browse some of the components shown in Chapter 8.

Footnotes

1. For the curious, these issues are covered in Chapter 3, Chapter 5, and Chapter 12. -- Return.

2. The percent sign (%) must occur at the beginning of the line. -- Return.

3. The HTML::Mason::Request object provides access to several properties and methods concerning the currently executing chain of components. It is treated in detail in Chapter 4. -- Return.

4. Component objects are returned by several of the HTML::Mason::Request and HTML::Mason::Interp methods, covered in detail in Chapter 4 and Chapter 6. -- Return.

5. Unless the component is called with an odd number of arguments. See the next section for details on this exception. -- Return.

6. We know of no browsers that actually screw it up, but surely there must be some out there. Browsers have a history of simply making up their own unique behaviors, even when there is a specification. -- Return.


Table of Contents | Foreword | Preface
Chapters: 1 2 3 4 5 6 7 8 9 10 11 12
Appendices: A B C D
Glossary | Colophon | Copyright

These HTML pages were created by running this script against the pseudo-POD source.