Last month we took an initial look at the open-source Zope application server. In particular, we saw how you can use Zope's DTML (dynamic template markup language) tags to create simple dynamic sites, as well as how you can manage a web site using nothing more than a web browser.
But anyone who has worked with DTML knows that it ceases to be wonderful when you want to create something relatively complex. DTML is best when it is used sparingly or when its functionality is obvious; writing pages of DTML that contain a half-dozen nested conditional (<dtml-if>) tags quickly becomes unreadable and difficult to maintain, not to mention very nonmodular.
Another problem is that DTML exists inside of individual documents, rather than in a central location. If we want to reuse functionality in multiple places, then we must copy our DTML methods and documents. This means that when we want to add or change some functionality, we must go through each of the copies and modify them as well.
The solution to this problem is the Zope product. Each Zope product is actually an object class (or a set of classes) that can be instantiated any number of times in our web site.
This month, we look at Zope products, which form the core of Zope's flexibility. After installing and working with some existing products, we will write our own simple product in Python.
A Zope product is a package of code, graphics and DTML that provides a piece of reusable functionality. For example, if we were interested in creating a simple page that displays the current time, we could create a DTML document:
<p>It is now <dtml-var name="ZopeTime" fmt="fCommon">.</p>
But what if we want to expand our page, displaying a weather forecast retrieved via HTTP from another server? DTML is not the answer here; even if we could use it to create our custom functionality, the result would be difficult to manage, as well as ugly to write. Because Zope products are written in Python, they can use any Python module they like, displaying their output in HTML or any other compatible format.
Because each product is treated as a single entity, we can install and remove them as a single unit even if it defines and uses a number of classes. However, this doesn't mean that each product stands on its own; on the contrary, it is possible for one product to use functionality provided by another product.
In addition to products and DTML documents, Zope provides two other means for creating dynamic content: Python scripts (implemented by a product, no less) allow us to write and use small Python programs within Zope. We also can create, edit and use new products using a system known as ZClasses. ZClasses allow you to create new products (and their associated classes) using nothing more than your web browser and DTML.
While these four options provide a great deal of flexibility, deciding which one to use can sometimes be difficult for beginning Perl programmers. Beehive's The Book of Zope, which I review in this issue of Linux Journal, suggests using ZClasses at the beginning of a project, migrating the code to a full-fledged product after everyone has agreed upon a design. The more complex your functionality is, the more likely it is that you will want to use or write a product rather than rely on DTML and Python scripts.
You can run almost every aspect of Zope products via the Zope management screen, which you can reach via the /manage URL of your Zope server. Click on the control panel link in the left-hand frame to bring up the Zope control panel and on the Product Management link in the main frame to bring up the product management screen.
You should see a list of Zope products, along with a button marked “Add product” at the top of the screen. Products that you can modify through the Web (including ZClasses, which we briefly mentioned above) are identified with an open box, whereas standard Zope products have a closed-box icon. A closed box simply means that you cannot modify the product itself via the Web. However, most products will let you customize them by setting one or more properties via a web-based interface. But the product itself remains unchanged, unless you modify the source code.
Each product is actually a directory under your Zope installation directory in lib/python/Products. The Sessions product is under lib/python/Products/Sessions, while the Transience product is in lib/python/Products/Transience. (I installed Zope under /usr/local/zope/ on my system, so Sessions is actually in /usr/local/zope/lib/python/Products/Sessions/.) A product directory contains Python code, text files and directories, including:
__init__.py: this is what Zope scans and executes when it loads your module. Among other things, the initialize method in __init__.py invokes context.registerClass, which (as its name implies) tells Zope that your product exists, what text to display in the Add menu on the /manage screen (with the meta_type parameter) and how to create a new instance of your product when the Add button has been pressed (with the constructors parameter).
README.txt: as its name implies, this is the README file for a particular product. Clicking on a product name from within the control panel will display a README tab, among others. This tab allows you to look at README.txt without having to look at the filesystem. If the product directory contains no file named README.txt, then no README tab will appear at the top of the screen.
version.txt: this file contains the name and current version number of your product, separated by minus signs (-). Version 1.2.3 of the product Foo thus will have a version.txt with the following contents: Foo-1-2-3. This version information is displayed in the control panel.
Help files: a product may contain a help directory, which contains the text displayed by Zope when you click on the help link. Help files are often written using structured text, a minimalist formatting system similar in spirit to Perl's POD documentation system. Structured text is easy to write with a simple text editor and equally easy to read with a standard Linux tool like less.
Zope only looks at the current list of products when it starts up. This means that if you install a new product, you will need to restart your Zope server. This is done most easily from within the control panel.
Now that we have seen what a typical product may contain, we will install a product by downloading it from the Zope web site, unpack it within lib/python/Products and restart Zope. If all goes well, our newly installed product should then appear in our control panel screen. Moreover, we will be able to create new instances of this product anywhere we want in our web hierarchy.
For example, let's create a Slashdot clone using the Squishdot product for Zope. Our first task is to retrieve a copy of Squishdot from www.zope.org/Products. Squishdot is listed under the Feedback category, among others, and probably will be one of the first products listed. Click on the links that lead to a downloadable version of Squishdot; the latest version as of this writing is 1.3.0. Notice how even a product of moderate complexity is relatively small; the Squishdot version that I downloaded was a little more than 256KB.
To install Squishdot, we must unpack it into lib/python/Products. Assuming that we place newly downloaded files in /downloads, this means that we can unpack Squishdot in the following way:
# Set this to your Zope home export ZOPE=/usr/local/zope # Switch into the products directory cd $ZOPE/lib/python/Products # Unpack Squishdot into the current directory tar -zxvf /downloads/Squishdot-1-3-0.tar.gz
Older Zope products expect to be unpacked from the Zope root directory, rather than from within lib/python/Products. Unfortunately, there does not seem to be any obvious way to know how a product was packaged without looking at it:
tar -ztvf /downloads/ProductName.tar.gz
If each filename begins with the lib/python/Products pathname, then you will want to switch into $ZOPE, rather than $ZOPE/lib/python/Products, before unpacking the product.
Unpacking the archive is all we need to do in order to install Squishdot. However, Zope only looks for products when it starts up; we must restart the server before we can create instances of Squishdot on our system. The best way to do that is to click on the Restart button from within the control panel. Don't panic if your browser complains that the server is no longer running after you click on Restart, or if you see an obscure-looking Python exception backtrace after clicking on the Restart button. Rather, wait several seconds before clicking again on the control panel link in the left-hand frame, and it should work.
You can check to see if your product has been added by returning to the Product Management page within the control panel. If the newly installed product (Squishdot, in this case) does not appear on the list, double-check that it was unpacked correctly and that the permissions allow the Zope user access to the product's files.
At this point, we should be able to create a new Squishdot site by moving to the root (/) directory of the Zope server, selecting Squishdot site from the selection list and clicking on Add. This invokes the methods named in the constructors parameter to context.registerClass, invoked by the initialize function in Squishdot's __init__.py.
And indeed, we could move ahead and create our Squishdot site at this point. But Squishdot uses the Zope MailHost object (which represents an SMTP server) to send e-mail notifications. If you have not yet created and defined a MailHost, the Squishdot configuration screen will remind you to do so.
When Squishdot looks for a MailHost, it begins its search in the current directory. If it does not find a MailHost object, the search continues up the directory tree, stopping when Squishdot reaches / or when it finds a MailHost object. While this might appear to be a simple issue, it demonstrates the concept of acquisition, which is central to Zope. Moreover, it means that different Squishdot sites can send e-mail via different SMTP servers, simply by creating more than one MailHost object. Indeed, we can define a global default MailHost in /, overriding it as necessary by placing additional MailHost objects in subdirectories. The concept of acquisition permeates Zope and means that we can define or redefine nearly anything—MailHosts, users, headers and stylesheets—at a local level.
In this particular case, we will create an instance of MailHost in the / directory by choosing MailHost from the new product list and clicking on Add. Because a MailHost object represents an SMTP server, the configuration of this object is pretty straightforward, requiring that we enter the name of our Zope server's SMTP server. Most Linux machines run their own mail servers, so “localhost” is probably a reasonable value.
The mandatory ID field is used to identify this MailHost uniquely within the current directory, which is why Zope uses IDs to identify objects in URLs uniquely. Just as a filename is a unique identifier within a directory, a Zope object ID is a unique identifier within a folder or other object. The optional Title field is meant for humans, rather than for the underlying Zope server; if it is defined, an object's title is displayed from within the Zope server interface.
After you have created your MailHost object, you will be returned to the main Zope management screen for /. You should see your new MailHost object (represented with a small envelope icon), along with any title that you defined, in the list of objects.
We are now ready to create our Squishdot site. Add a new Squishdot site object using the selection list and Add button in the upper right-hand corner, choose an ID (i.e., URL pathname), optional title and mailhost, and then select some other basic parameters for your Squishdot site. For example, I chose an ID of atf and otherwise left the configuration options with their default values.
To enter my Squishdot site, I now tell my web browser to display http://localhost:8080/atf/. Zope receives this request for /atf and sees that we are referring to a Squishdot object. Zope then asks this object to display itself. Sure enough, we see an introductory screen that looks something like Slashdot but is powered by Zope.
We can create as many Squishdot sites as we might like, keeping in mind that every new site must have its own unique ID. In this way, we can set up one moderated site, one unmoderated site and another internal site for our organization's own uses—each with its own URL, potentially protected with its own set of users and groups.
To modify the Squishdot site, simply append /manage to the name of the object you want to modify, as in http://localhost:8080/atf/manage. This invokes Zope's management system on our Squishdot site. Using tabs at the top of the screen, you can modify nearly any parameter having to do with Squishdot, from moderation rules to the color of the text in which the site name is displayed.
This month we discussed Zope products and saw how to download, install and configure products on our system. While products are inherently more complex than simple DTML pages, their centralized code and additional flexibility make them more suitable for serious tasks than DTML.
Next month we will look at how we can write our own Zope products using a combination of Python and DTML.
Reuven M. Lerner is a consultant specializing in web/database technologies. His book, Core Perl, was published by Prentice Hall in January 2002. He lives with his wife and daughter in Modi'in, Israel.