| <html> | |
| <head> | |
| <title>Validation Framework Overview</title> | |
| </head> | |
| <body> | |
| <p>A framework for calling validators that are build based.</p> | |
| <h2>Approach</h2> | |
| <p>In WTP 3.0 a new validation framework has been added. This was developed at | |
| the request of validator owners and end users. To ease the transition to this | |
| new framework, that approach that has been taken is to support both the old and | |
| the new frameworks for a while. This means that the old validators work exactly | |
| like they used to. They can still plug into the old framework and will continue | |
| to be run by the old framework. This allows validator owners to migrate at their | |
| own pace.</p> | |
| <p>For validator owners that wish to exploit the new framework, it has been | |
| designed n such a way to minimize the transition from the old framework to the | |
| new framework. There is a new extension point that needs to be specified, and a | |
| new abstract class that needs to be extended.</p> | |
| <h3>Benefits</h3> | |
| <p>What benefits does the new framework have to offer? Or asked another way, as | |
| a validator owner, why would I want to switch?</p> | |
| <h4>Performance</h4> | |
| <h5>Tuning</h5> | |
| <p>The main reason for creating a new framework was to improve performance. As | |
| more and more validators are added, validation is consuming a bigger slice of | |
| the resources that it takes to perform a build.</p> | |
| <p>The best performance improvement is eliminating unnecessary work. We | |
| eliminate work, by providing significantly more tuning options. Validation | |
| owners can tune which resources they are called on. This is done through the use | |
| of filter rules. These filter rules are grouped into inclusion and exclusion | |
| groups. If you have multiple inclusion groups they are "anded" together, meaning | |
| that each of the inclusion groups must have at least one successful filter rule. | |
| Exclusion groups are used to specify what doesn't get validated. Exclusion | |
| groups have a higher precedence than inclusion groups. If an exclusion group | |
| excludes a resource than it is excluded no matter what the inclusion groups may | |
| say.</p> | |
| <p>They are several different types of filter rules:</p> | |
| <ul> | |
| <li><strong>File extension</strong> - Rules based on the file extension. For | |
| example, include all the files with extension *.xml.</li> | |
| <li><strong>Folder or File name</strong> - Rules based on the folder or file | |
| name. For example exclude all the resources in the build folder. File names | |
| can be fully qualified, project relative names, e.g. <strong>WebContent/WEB-INF/web.xml</strong>, | |
| or simple names, <strong>web.xml</strong>.</li> | |
| <li><strong>Project nature</strong> - Projects with certain natures can be | |
| either included or excluded.</li> | |
| <li><strong>Facet</strong> - Projects with certain facets can be either | |
| included or excluded.</li> | |
| <li><strong>Content Type </strong>- Resources of a particular content type | |
| can be included or excluded.</li> | |
| </ul> | |
| <p>The filter rules and groups form the building blocks of what gets validated. | |
| </p> | |
| <p>In terms of tuning however, the end user has the final say. The validator | |
| owners specify the default filter rules and groups for their validator, but the | |
| end user through the use of preferences can override these settings. </p> | |
| <h5>Caching</h5> | |
| <p>Another way that performance is improved is through more effective caching. | |
| With the previous framework we were noticing that different validators would be | |
| opening and processing the same file, but that this would happen far enough | |
| apart, that caches weren't effective. Even if files were being cached, the | |
| second (or third, or forth, ...) open would happen far enough a means that way | |
| that the file would have already been evicted from the cache. In the new | |
| framework, the processing order is more tightly controlled. The framework calls | |
| the validator and tells it what to validate. This way files are processed in | |
| order, which means that the chance of them being a cache is much higher. Every | |
| validator that is interested in file A processes it, and then we move on and | |
| every validator that is interested in file B processes it, etc.</p> | |
| <p>Also because state information is kept for the lifetime of the validation | |
| operation, validators can cache expensive operations, like model creation, that | |
| might be needed as part of validating several files. </p> | |
| <h5>Job Control</h5> | |
| <p>With the earlier framework, there were too many jobs/threads running | |
| concurrently. This is now more tightly controlled.</p> | |
| <h4>Dependency Support</h4> | |
| <p>Another new feature is dependency support. A validator can now assert that a | |
| resource is dependant on some other resources. For example that a XML file is | |
| dependant on a XSD file. Now if the XSD file changes the XML can be revalidated. | |
| As the user comes to trust this dependency support, they wont perform as many | |
| clean builds.</p> | |
| <h4>Primary Framework</h4> | |
| <p>This is the primary validation framework. At some point, the earlier | |
| framework will be deprecated and then removed. All the new investment will | |
| happen in this framework. </p> | |
| <p>The framework collateral will continue to be enhanced (junit testcases, | |
| documentation, performance improvements, ...)</p> | |
| </body> | |
| </html> |