AutoHotkey Community

[bmcclure]

I'm writing a SAX (Simple API for XML) implementation for AHK, modeled after SAX for Java

I have an input stream of bytes (in this case characters) that I would like to parse.

My question is basically, can anyone help me think of a good way to analyze the XML input character-by-character, turning it into meaningful tokens that I can send off to another function designed to handle them?

Let me start with an example.

An XML file:

[tank]

[bmcclure]

Yes, I'm implementing SAX, and am creating an XMLReader based on the XMLReader interface (to be provided by the XMLReaderFactory class). There is no official SAX API, but I'm using the SAX class reference and have created the appropriate interfaces and classes to match the Java implementation in AHK.

In answer:
1. xpath library won't do what I need, it loads the entire file and then every command essentially re-parses the document. I need something that does one pass over the document from an input stream, without loading the entire document into memory. This will allow documents of essentially unlimited (other than OS limits) size to be parsed efficiently.

2. I wrote an XMLDOM library with COM (that I mostly abandoned in favor of the DOM and SAX implementations I'm writing in pure AHK). Since I'm writing my own library, I don't want to utilize a redundant 3rd party library--I want it all to be native.

3. I plan to use this XMLReader as the back-end parser that constructs the DOM tree in my AHK DOM implementation, so loading the document into DOM and parsing it with SAX from there would create some sort of paradox

SAX and DOM are completely different for different purposes, which is why I'm implementing them both in AHK. SAX is fast, but if you need to access random parts of a document in memory, it's inefficient, and a DOM tree should be built instead. If, on the other hand, you just need to read an XML file and parse its contents, SAX is by far the better choice, as you don't need to load and maintain the entire document tree in memory.

You can use SAX to create a DOM tree (by creating a custom handler), and you can parse a DOM tree using SAX. Different approaches for different needs. For loading files, however, it makes more sense to go the SAX -> DOM route, since SAX can efficiently create the DOM tree that is fully compliant with DOM specifications.

Really, more-so than a question of creating a SAX XMLReader, is just figuring out an efficient way to parse the character stream and decode the elements without having to make multiple passes over any bytes.

Thanks very much for talking through some of this with me
_________________
Ben

My Trac projects
My Wiki
[Broken] - My music