Until now, most of Mozilla's Webmaking experiments that involve writing HTML (lovebomb.me, Storything, X-Ray Goggles, etc) have used
.innerHTML to parse HTML into a DOM. However, this has a number of disadvantages:
It's difficult to map an element in the generated DOM back to its location in the original HTML source code. This is useful in a variety of scenarios such as Jessica Klein's lovebomb mockups and webmaker tutorials.
It's completely insecure. For our webmaker initiatives that really need to scale, we'd like to have more fine-grained control over what kinds of elements and attributes are allowed in code.
It can be difficult to help the user pinpoint errors in their code, e.g. mismatching tags.
Slowparse is intended to solve all of these problems. It parses HTML and uses a host DOM implementation to create the DOM representation of the page. Each generated DOM node has an expando property called
parseInfo which provides metadata mapping the location of the node in the original source code. It also provides detailed error feedback so that users can easily pinpoint why their code isn't working the way they intend.
For more information, see:
- A live demo to see how Slowparse interprets your code in real-time.
- The error reporting specification to get an idea of what kinds of errors can be reported, and what their human-friendly descriptions are.
- The annotated source code for slowparse.js, tree-inspectors.js, and spec/errors.jquery.js.
- The test suite.
- The blog post Learning and Grammatical Forgiveness, which explains why Slowparse isn't an HTML/CSS validator.