The Mason Book

At its heart, Mason is simply a mechanism for embedding Perl code into plain text. It is only one of many such mechanisms that all do more or less the same thing. However, Mason represents a particular set of choices about how this embedding should be done, and many people have found that the way Mason does things is very straightforward and extremely conducive to getting jobs done.

In this chapter we'll introduce you to some of Mason's key features and strengths, show you a couple of examples of how to use Mason, and talk about some alternatives to Mason. After reading this chapter, you should have a fairly good idea of how Mason relates to its peers and what kinds of tasks you can accomplish with Mason.

The most common application of Mason is in building large dynamic web sites, and this book focuses mostly on web site building. Mason is broadly applicable to any situation in which fine control over document content is required, however, such as generating mail-merged form letters, creating custom configuration file sets, and even building dynamic GIF images based on varying input parameters. We intend to give you enough facility with Mason that after reading this book, you can imagine Mason-based solutions to problems we haven't ever thought of.

Before we get into the details of Mason and comparisons with its alternatives, we'll just briefly mention some of its guiding design principles. Mason was designed to help you build, organize, and maintain large web sites or other groups of dynamically generated documents. It cooperates fully with Perl, leveraging all the solutions and techniques that Perl developers have come to depend on and that have made Perl such a powerful and widespread tool. It encourages thinking about your site in structural terms rather than as a collection of procedural scripts or modules. All of these things are conducive to getting your job done effectively, letting you concentrate on your goals while Mason takes care of the details.

A First Example

To help make this discussion a little more concrete (this thing is called Mason, after all), let's look at an example. We'll give more in-depth treatment to the details of Mason's syntax later in the book; these examples are just to put some Mason code in front of your eyes and show you what it looks like.

The following code is a complete chunk of Mason code, called a component:

When Mason runs this code, the output is:

We'll talk more about the details of component syntax in Chapter 2, but two basic elements in the preceding example deserve mention here. The first is that any line that begins with a % character tells Mason that the line contains Perl code. The Perl code can be any syntactically correct Perl -- Mason doesn't care what it is or what it does. In this case, it simply sets the value of a variable that will be used later in the component.

The other element in the previous Mason component is the substitution tag, denoted by the sequence <% %>. Mason will evaluate the contents of any such tag and insert the result into the surrounding text. In this case, the variable $planet evaluates to World, and the output of the entire component is Hello, World! Note that any text that isn't a special Mason construct simply becomes part of the output of the component.

These two lines are relatively simple and not particularly exciting, but they should give you a taste for how Mason code looks in its simplest form.

The Main Features of Mason

There are more templating systems written in Perl than you could possibly keep in your head all at once. To help you make sense of Mason's place in the world, this section presents Mason's most important and distinctive features. By the end of this section, you should see that Mason pushes the boundaries of the term " templating system," with lots of features aimed at helping you manage the larger tasks of site design and maintenance.

Components: Modular Design Elements

As we mentioned before, the basic unit of Mason code is called a component. It is a chunk of Mason code that can accept input parameters and generate output text. An important feature of Mason is that any component may call any other component at any point during its execution, much like a Perl subroutine calling another Perl subroutine. Because of this feature, a component may represent a single web page, a part of a web page (like a side navigation bar), or even a shared utility function that generates no output of its own. This separation of design elements allows you to use Mason as a sort of glorified server-side include (SSI) mechanism, as in Example 1-1, Example 1-2, and Example 1-3. Executing mainpage.mas will produce a full page of HTML with the header and footer inserted in place.

Example 1-3 introduces the component call tag syntax, <& &>, which is used to call another component and insert its output into the surrounding text. The component tag can also accept arguments, which in this case can help unify site design by moving the page header text into the header.mas component.

The header.mas component in Example 1-4 now accepts an argument called $head that contains the text that should get inserted into the <h1> tags. A component's arguments are declared by using an <%args> block, which you'll see in more detail later in the book. The $head argument becomes an honest-to-goodness Perl variable that can be used throughout the rest of the component. It's lexically scoped in the header.mas component using Perl's my() function.

The footer.mas component in Example 1-5 is fairly straightforward. It just provides a link to the document root.

In the mainpage.mas component in Example 1-6, the arguments are passed to the header.mas component by using standard Perl syntax (i.e., commas, quotes, and the => operator). In fact, any Perl syntax for passing a list can be used, because the argument list is specified in real Perl syntax.

Mason will take the list of arguments passed to the header.mas component and assign the proper values to the variables specified in the <%args> block.

Object-Style Component Inheritance

Aside from the fact that there's a little bit of Perl thrown into the mix for passing parameters, the examples we've seen don't really show anything that you couldn't do using standard server-side include (SSI) techniques. In fact, the usage demonstrated in these examples is relatively uncommon in building Mason sites, because there are better ways to get the job done. One of the greatest features of Mason is that components can inherit behavior from other components, much like classes and objects in an object-oriented hierarchy.¹ Typically, each component will inherit from a single component called the autohandler . The autohandler implements general behavior for all components, such as the content of headers and footers. Individual components implement specific behavior, such as the body text of the individual pages.

Using component inheritance, we can rewrite Example 1-4 through Example 1-6 in a more common Mason idiom, as shown in Example 1-7 and Example 1-8.

Notice that the header and footer are now both all in one file, the autohandler. Visually, this helps unify the page content, because tags like <html> and <body> that are opened in the header are closed in the same file. The other important difference here is that mainpage.mas no longer has to call the header and footer components explicitly, but rather Mason calls the parent component automatically and it wraps its header and footer around the main content. The page header is now specified by an attributes block, one of Mason's object-oriented mechanisms. An attribute is a component property that inherits via Mason's component inheritance chain.

There are zillions of other uses for Mason's inheritance mechanism, which will be further explored in Chapter 5.

Intelligent Caching Mechanisms

Anyone who has built any dynamically generated web sites knows that sometimes certain portions of a site can take longer to generate and serve than you want to make your users wait. Furthermore, portions of a site might be only "semidynamic," meaning that their content changes periodically but stays static for a long time between changes. Alternatively, as might happen on a news site or for an online poll, content may change continually, but a lag time of a few minutes in updating the content would be acceptable if it improves site performance. For cases like these, Mason provides a very sophisticated caching mechanism that you can use to control how often the output of a component is rebuilt. You can base the expiration decision on time, on certain key parameters like username or content ID, or on an explicit agent that decides when specific data has expired.

The caching mechanism can be used for the output of a component, for an arbitrary block of text, or for any Perl data structure you might want to cache. The first-class support for caching is one of Mason's most endearing qualities, and you'll learn to appreciate it the first time it saves you from spending hours optimizing sluggish code.

To aid overall performance, Mason also has an intelligent internal caching mechanism. During execution, Mason turns each component into Perl source code on disk, then compiles the Perl code into bytecode, then executes the bytecode to produce the component's output. It would be a waste of computing resources to repeat this cycle every time a component needs to be executed, so Mason caches at each stage. As an aid to rapid development, Mason will check your components' modification times and invalidate its cache when you make changes to your components, ensuring that any changes you make to your site take effect immediately. When your site moves from development to production, you probably won't be making frequent changes to your site, so you can disable the freshness checks in order to improve your site's responsiveness .

Integration with Apache and mod_perl

As mentioned before, the most common use of Mason is in building large, dynamic, data-driven web sites. The most popular web server around is Apache, and one of Apache's most powerful features is mod_perl, which lets you use the full power of Perl within the Apache server process. Therefore, it should come as no surprise that Mason is designed to cooperate fully with mod_perl. Mason comes with drop-in mod_perl handlers that let Apache serve your Mason components directly. It lets you take advantage of the sophisticated decision-making mechanisms that Apache has evolved to support, such as custom authentication methods, content negotiation, and dynamic URL rewriting. Mason's caching mechanism and other performance considerations are designed specifically for the task of serving dynamic content efficiently and with enough flexibility to let you design creative solutions to your specific problems. Although Mason lets you build a site without relying very much on assumptions about the server environment, learning about mod_perl and Apache's request cycle can help you use Mason to create slick and powerful features.

Alternatives to Mason

Much of this chapter so far may have sounded like a sales pitch, because we wanted you to know about Mason's biggest strengths so you could have some solid reasons for reading the rest of this book. However, you should also be aware that there are many alternatives to using Mason, and an awareness of these alternatives will help you form an accurate picture of the context for which each system was created. It will also help you decide which system to use for each individual project, because no single system was designed to be a solution to all the problems you might encounter. Just as importantly, different people find that different systems suit the way they think better than others do.

There are generally two kinds of systems that people consider to be alternatives to Mason: lightweight solutions and heavyweight solutions. Lightweight solutions generally have the goal of being small and fast and leave much of the major work up to you. They are often simple templating modules like Text::Template or HTML::Template or even homegrown templating schemes. Using templates is certainly a good idea, and it is one of the core ideas in Mason itself. However, when designing an entire site, you're usually going to need some more sophisticated system that helps you manage your site-building resources; if you choose a templating-only solution, you'll probably end up writing this management code yourself. You may have a good idea of what such a system would entail only after writing and maintaining dozens of complicated web sites, so you'd likely spend more time working on your management code than on building your sites. This is the main trade-off with lightweight solutions: you gain flexibility because you can manage your site however you want, but since the burden rests entirely on you, you might end up preferring to use a tool that handles many of these management issues for you.

By contrast, heavyweight solutions implement several layers on top of their templating capabilities. Despite some disagreement on proper use of the term, "application server" is often used to describe such heavyweight systems. Each anticipates the typical needs of a large, sophisticated web site and provides methods for dealing with these situations cleanly. A heavyweight system will typically have support for integrating a site with a database, working with HTML and URLs, preserving state from one request to the next, caching often-used resources, and dealing with error conditions. Each heavyweight solution is tailored to different site requirements and makes different assumptions about the best ways to deal with them.

Good solutions, such as Mason, Embperl, and Apache::ASP, also help you organize your site code in a way that lets you think about your site's structure in an organized way, both on the small scale and the large scale. They help you design a site that is easy to build and easy to maintain.

Mason itself is sort of a unique offering. It sits somewhere in the middle between lightweight and heavyweight solutions, though it leans a bit toward the heavy. It doesn't directly provide support for database connections, HTML munging, or sessions, but it makes it so easy for you to use regular Perl tools for these purposes that you'll never miss the functionality. If you do decide that Mason is missing a feature you really need, it's easy to add functionalities to Mason and use them just as if they were built in. In a sense, Mason's main strengths lie in the ways it lets you interface various parts of your site with one another and with outside resources.

Consider these design goals as you read the following descriptions. When possible, we have worked with the authors of these systems to make sure the descriptions highlight each system's best features. Keep in mind, though, that our list of alternatives is by no means exhaustive. There are countless other solutions. We have tried to pick the most popular solutions, but for more information on any product you might want to use, read its documentation, find an appropriate mailing list in which to ask questions, and make your own decision.

Embperl

Of the systems presented in this chapter, Embperl may be the most similar one to Mason. Embperl is one of the oldest heavyweight systems that is still in widespread use. It has been used for several years under the name HTML::Embperl , but recent beta releases have switched the module name to just Embperl. Its author, Gerald Richter, is generally very responsive to bug reports and feature requests.

Embperl is targeted specifically toward generating HTML and has several "magical" HTML-manipulation features. For instance, HTML tables can be autogenerated by using the special Embperl variables $row , $col , and $cnt :

This means that Embperl does some scanning of your HTML and your Perl code to decide when you mean to use its magical generation. Some people find this assistance useful, and others would prefer to manage the generation themselves (the approach that Mason takes). The equivalent Mason code would require an explicit loop:

Notice that the Embperl delimiters for embedded Perl code are based on square brackets like [+ +]. This is so that Embperl code can be written by using a WYSIWYG HTML editor that might get confused by angle brackets and treat them as HTML code. Embperl takes this even further and lets you write Perl code that is HTML-escaped by the editor, like so:

The text < will be converted to < by Embperl before execution.

Notice also that Embperl uses its own custom loop control syntax rather than Perl's built-in loop control. There is experimental support in Version 1.2b2 and higher for using native Perl loops, and it will be part of the stable feature set in Embperl Version 2.

Embperl has a feature called EmbperlObjects, which is an inheritance system similar to Mason's autohandler functionality. It also offers integrated support for preserving state between requests with Apache::Session via another special variable, %udat. This can be very handy; see Chapter 12 for how you can accomplish a similar effect with Mason.

Variables you create in Embperl are usually dynamically scoped as either local or global variables (not lexical my() variables), which can sometimes be a bit unnerving in a persistent environment like mod_perl. Fortunately, after each request, Embperl will clean up any variables you created during the request. Variables declared with local() or my() in Embperl are usually scoped to the substitution tag enclosing them. Contrast this to Mason, in which the default scope of localized variables is the entire component. Because of this, it is more common to see my() in Mason code and global variables in Embperl code.

Embperl Version 1.x also provides some support for using the Safe.pm module, which can provide some protection against executing malicious Perl code. This strategy is somewhat dubious, however, because the Safe module is fairly easily defeated and because we hope your page designers aren't going to be trying to sabotage your servers in the first place. Moreover, when using Safe mode, you won't be able to access outside resources like databases, so it might not even be an option in the first place. Because of these problems, support for Safe mode has been removed in Embperl Version 2.

Embperl 2 will add several new features, including support for XML/XSLT processing, custom template syntax, and a highly customizable data caching and output filtering system.

Apache::ASP

Apache::ASP , by Joshua Chamas, is a package for writing Active Server Pages under mod_perl. It is fairly mature (its initial release was in 1998), and several years of development and active use by the Perl community have created a feature set that includes several useful extensions to the Microsoft standard.

With Apache::ASP, instead of using VBScript or JScript for the dynamic portions of the pages, you use Perl.² The ASP feature set has been neatly translated to Perl, so you have access to all of ASP's built-in features such as session management (using a custom session manager, not Apache::Session), email integration, and the ASP object model.

The Perl embedding syntax is very simple: <% %> tags get wrapped around Perl control structures or miscellaneous Perl instructions, and <%= %> tags are wrapped around Perl expressions whose values you wish to insert into the surrounding HTML. For example:

The Mason equivalent is very similar:

The output of both examples is:

Because it is built on the ASP model, Apache::ASP is a natural choice when porting an ASP/IIS application to the Apache platform. You will need to translate the scripting language from VBScript or JScript into Perl, but the overall structure of the site should remain unchanged.

Besides the standard ASP feature set, Apache::ASP supports several additional features, including extra event handlers, XML/XSLT processing, and component output caching. It also supports a cookieless mode for maintaining session data, which can be very handy for end users unwilling or unable to store session cookies. The XSLT support is particularly interesting, since combining it with the output caching features means that you can use dynamic XSLT transformations in web publishing, a technique that might be computationally prohibitive without caching.

For more information on Apache::ASP, please visit http://www.apache-asp.org/.

HTML::Template

Sam Tregar's HTML::Template module falls into the lightweight category. Its chief goal is to allow site builders to separate a site's HTML from its Perl controlling code, and it enforces this division quite strictly. No Perl code is ever embedded within templates, and control structures like if, include, and various loops are all implemented by custom HTML::Template tags. Any variables to be interpolated into the HTML template are explicitly fed to the template by the controlling Perl code.

Philosophically, the reason for this strict division of HTML and Perl is so that a division of labor can be enforced in an organization, with HTML designers and Perl coders working independently, not worrying about whether they're editing the same files. It is also possible to apply the same controlling code to several different templates, so that designers can create different look-and-feel templates without involving programmers.

HTML::Template is generally very fast at filling in HTML templates, especially if you use its just-in-time compiler HTML::Template::JIT. It also provides a caching mechanism, somewhat similar to Mason's, that allows for caching templates in private memory, on disk, in memory shared between processes, or in combinations thereof. Unlike Mason, HTML::Template's caching handles the caching only of templates themselves, not of their output or arbitrary data.

To make this actually do something, you need to write Perl code to call the template. For this template, we might write something like this:

Note that the top layer for the system is a Perl script. Some people love this, some people hate it. You may know by now which category you fall into.

Text::Template

Text::Template , written by Mark-Jason Dominus, is a lightweight solution similar to HTML::Template, but has some philosophical differences. First, it does not assume that the template contains HTML but is designed to work with any kind of text in the template (in truth, HTML::Template can work with arbitrary text, too, but it was designed specifically to work with HTML). Second, it uses native Perl control structures instead of its own custom macros, so Perl programmers have less to keep in their heads. This has the effect of breaking down the barrier that HTML::Template maintains between code and HTML, so in an environment in which designers and programmers need exclusive control over their own work products, HTML::Template may be the better choice.

Like Embperl, Text::Template also supports code sequestering via the Safe.pm module, with the same caveats as mentioned earlier. Template variables can be passed explicitly to the template for substitution or drawn from Perl variables in a specified Perl package. Text::Template also allows the user to customize what delimiters are used to indicate the special Perl code or variable substitution sections. The default delimiters are curly braces:

Text::Template was first released in 1995 and has undergone many revisions in its life cycle. It is considered a mature product, and its author is very responsive to comments and questions about it. There is also a low-volume mailing list for users of Text::Template, the details of which can be found in the module documentation.

Template Toolkit

In order to combat the proliferation of zillions of templating modules that all look and act similar but contain maddening and meaningless differences, the Template Toolkit package by Andy Wardley aims to be the only templating package you'll ever need and offers an extremely full-featured templating system. Like HTML::Template, it uses a large set of custom control macro tags like IF , SWITCH, FOREACH, and so on, as well as more advanced tags like FILTER , TRY and CATCH, and MACRO. It can either allow embedded Perl as Text::Template does or disallow it as HTML::Template does. Like Mason, it lets you build a site in terms of modular components. There is also an Apache::Template module that facilitates building a mod_perl-based site out of Template Toolkit components.

An example of Template Toolkit usage follows:

Because of its many features and its well-supported development, the Template Toolkit is becoming the modern standard for standalone templates. It is quite different in philosophy from Mason, though, so we do not consider it to be in direct competition as a tool for the same situations we'd use Mason for.

PHP

PHP (http://www.php.net/) is pretty far removed from Mason, but we mention it here because of its popularity. PHP is another mechanism for embedding functionality into web pages, but PHP does not use Perl as its scripting language. The name PHP can refer variously to the embedding system itself, the scripting language, or the interpreter that renders the HTML pages.

It is important to understand some of the properties of PHP before deciding to use it. One of the design goals of PHP is to be as simple as possible to install and start using, and in some cases this means that features that experienced Perl programmers rely on are not present. For instance, PHP lacks support for private namespaces, there is no way to create three-tiered applications that separate business logic and presentation code, and there is no mechanism for creating reusable code modules. The Apache mod_php module is only a content generation module, so it cannot cooperate with other request phases in the same way Mason can cooperate with mod_perl's authentication or filename translation phases.³ Importantly, although there is a lot of user-contributed code in the PHP world, it cannot match the breadth and depth of Perl's CPAN. It has often been said that the CPAN is Perl's "killer app," and programmers most appreciate the CPAN when they least expect it.

Finally, although you can theoretically use PHP for general-purpose programming, it wasn't designed for that. PHP is typically used only for embedding PHP code into templates, whereas Perl is a full-featured programming language used for more purposes than any single programmer could imagine. While this does make PHP well-suited for the common tasks of web scripting, it may be limiting. For instance, a certain Perl programming friend was recently contracted to write a "simple shopping cart system" that had one small addition: it had to do some horribly complex optics calculations. For situations like these, a general-purpose programming language like Perl can be quite handy.

Philosophy

The final criterion in evaluating an application platform is peace of mind: your peace of mind as a developer, your company's peace of mind as a provider of services and maintainer of site resources, and your users' peace of mind to enjoy (or suffer through) your final product.⁴ If you are continually frustrated by your development tools and environment, you will almost certainly not be able to create a satisfactory product.

Mason was created to help you with tasks when you need help and get out of your way when you don't. It was designed to cooperate with Perl to the fullest extent possible. The Mason templating language is extremely simple but full featured and leverages Perl to do all the things that a programming language is good at. Mason is not explicitly tied to the HTML markup language, but it was designed to function well in an HTML environment. Mason's design choices were made to encourage a structural approach to site building rather than a procedural approach, and this often makes building large sites with a consistent feel very simple.

If you're interested in learning more about alternatives to Mason, you might be interested in another book from O'Reilly and Associates, Inc., CGI Programming with Perl, written by Scott Guelich, Shishir Gundavaram, and Gunther Birznieks. It contains more extensive discussions of many of the modules mentioned here.

Getting Started with Mason

To help you try out the various examples you'll see throughout the book, we'll walk you through the typical installation and configuration steps. First we'll describe a fairly bare-bones standalone installation that can be used without a web server, and then we'll describe the more common situation, which is to install Mason for use in conjunction with the Apache web server and its mod_perl-embedded Perl interpreter.

A Standalone Installation

Mason can be installed just like any other typical Perl module and used as a toolset for templating. You'll need these ingredients:

You'll also need several other Perl modules that Mason depends on, such as Exception::Class, Class::Container, and Params::Validate. Since it can get fairly tedious to follow all these dependencies yourself, you may want to use the CPAN.pm module to help automate the process. In this case, you can start the CPAN shell with the command perl -MCPAN -e shell, and then type install HTML::Mason at the prompt.

A mod_perl Installation

Most of the examples in this book will assume that

Mason is being used in a web context. Most people using Mason also use it in conjunction with Apache and mod_perl. The main ingredients involved in setting up this paradigm are:

Trying It Out

Once Mason is installed, you can write a simple test program like the one in Example 1-9 to show that it's actually working.

When you run this program, you should see either Greetings, Earthlings or Greetings, Martians. The code between the <% %> tags is just regular Perl code, not anything to do specifically with Mason. It randomly selects an element from the two-element list ("Earthlings", "Martians"). Of course, we'll explain the details of how the rest of the program works later in this book.

If you've installed mod_perl and configured Apache as we described earlier, then you should test out your setup now. Assuming the first configuration option we showed you, you can place the same component at the top level of your web server's documents directory, as shown in Example 1-10.

When you request the URL http://www.example.com/handshake.html from your server, you should again see a random selection of either Greetings, Earthlings or Greetings, Martians.⁵ Assuming you've gotten Mason installed properly and everything's working right, congratulations. If you've run into a problem that you can't figure out, several sources of help are available. First check out the Frequently Asked Questions file at http://www.masonhq.com/docs/faq/, as it contains answers to the most commonly encountered problems. If you don't get any satisfaction there, try searching the archives of the Mason mailing list, available at http://www.masonhq.com/resources/mailing_lists.html, or post a message to mason-users@lists.sourceforge.net. Try to be as specific as possible about your problems if you go this route.

Finally, a word of advice as you read this book -- try out all the examples! You'll have a lot more fun and you'll retain the techniques better if you take a hands-on approach. And of course, if you find any errors in the examples, be sure to let us know.

Footnotes

1. A caveat is necessary when using the term "object-oriented" to describe Mason's content-wrapping and inheritance schemes, because Mason merely borrows some ideas from the object-oriented world rather than employing Perl's (or any other language's) built-in object-oriented inheritance. -- Return.

2. A third-party PerlScript option is also available, but it is not as widely used as VBScript or JScript. -- Return.

3. Although Mason is also just a content generation module, it cooperates with the other request phases by virtue of mod_perl's support for sharing information among the request phases. -- Return.

4. An excellent exposition on the importance of peace of mind when designing and maintaining systems is Zen and the Art of Motorcycle Maintenance by Robert M. Pirsig. -- Return.

5. Throughout this book, we'll use the fictional domain example.com for many of our examples. Unless you own this domain (and you don't, since it's reserved for examples in books like this one), you should replace it with the real name of your web server. -- Return.

Here

There

Chapter 1: Introduction