Josh Haberman

I’m asking this question, not answering it!

While you’re sitting tight waiting for Gazelle 0.2, I have a challenge I’m putting to my readers. I want my program gzlparse to parse some input text and output the parse tree in some useful format. What is the most useful format for visualizing a parse tree?

I want both a good text-based format and a good graphical format, if possible. For text formats there’s:

• XML (ugh. I didn’t want to say it, but I knew someone else would if I didn’t first. I’ll probably support it, but I’ll put ambivalent emoticons in the source code).
• S-expressions. Maybe I’ll win over some LISPers.
• ??

A good useful text format would be nice, but a good graphical format could be groundbreaking. I could always draw it as a tree, but I’m wondering if there isn’t something better. Something that keeps the text in its original format, but uses color or borders or something like that to represent the parse tree structure.

Here’s an example of the kind of visualization I think is really great and innovative. It’s the way that Lurker displays an email thread:

What’s so brilliant about this view is that it shows you both time-order of the messages and complete threading information in an attractive way. Of course, this is simpler than a parse tree, and such a nice view of a parse tree might not be possible. But what I’d really love to see is a parse tree visualization that:

• kept the original text recognizable (viewing it purely as a tree throws away the original text formatting completely)
• shows the parse tree structure somehow
• major bonus: can be rendered in a browser using a DOM. like, would allow me to write JavaScript to create this DOM inside the browser.

If this were possible, then I could write a web-based syntax analyzing text box that parsed your text as you typed it and showed you beautiful graphical representations of the parse. Something like this extremely awesome interactive regex visualizer, but for full context-free grammars. Or something like what ANTLRWorks provides, but on the web.

That would be SO HOT.

I’m really excited about the interest the Gazelle manual has generated! Thanks for checking it out, and for your feedback.

I got a little scared when people said they were going to start trying Gazelle out now, because it immediately made me think of how many things I’ve been meaning to fix before I unleashed it on anybody else! But on the other hand, it gives me all the more motivation to get it to a point where other people can try it out. And I’m never more motivated or get as much done as when I know people are waiting for me!

So here’s my line for the moment. Don’t try out Gazelle just yet. There are too many things for me to fix at the moment that I know are broken. But I want to fix those things ASAP and get Gazelle 0.2 out the door, so I can finally have a release that I can recommend people try out.

When will Gazelle 0.2 come? I’m hoping no more than a month. Here’s the target feature set:

• complete Strong-LL(k) lookahead support. I have the code to generate Strong-LL(k) lookahead, I just need to support this at the bytecode and runtime stage.
• a command-line compiler program (gzlc) that takes reasonable options and is simple enough to use by reading its --help
• a “tour” section for the manual
• a command-line program (gzlparse) that can output the parse tree in a useful format, so you can see how Gazelle parses your input text.
• a test suite, so that when people report bugs I can add the bugs to the test suite and not regress. this will be important for keeping my sanity.
• (stretch): make Gazelle self-hosting, so that the parser is more robust and easier to understand than the hand-written recursive descent parser I’m currently using. I don’t want people to have to deal with corner-case parser bugs.