Comments Loading...

2007-11-28

In Cincom Smalltalk, you can invoke the XML parser on a URI that happens to be network oriented, such as: (XML.XMLParser on: 'http://www.w3.org' asURI) scanDocument.

Most people who use the XML parser are quick to turn off validation in the hopes that it will parse documents faster because it won't have to download all the entity rules to do the validation. If you're like me, you'd then have written the code:

(XML.XMLParser on: 'http://www.w3.org' asURI) validate: false; scanDocument

However, we quickly discover that even with the validation off, the entities are still downloaded off the internet. This can be detrimental to performance, as the base network code has no caching mechanism - and let's face it, we don't really want to hit the internet randomly when we might not even have internet connectivity.

So over the years, I've found myself doing the same patch over and over to the XML parser to not download external entities if validation is off. This was wrong, very wrong. You do need the entities to resolve names like " correctly. So what you really want to do is take control of how the XML parser gets its external resources.

In VisualWorks 7.6 and future versions of ObjectStudio 8, you can now specify a handler for resolving external entities that is block based so that it's nice and pluggable. For example, if we want to deny all internet access, we can write the following code:

(XML.XMLParser on: 'http://www.w3.org' asURI)
	validate: false;
	entityResolver: (XML.PluggableEntityResolver withBlock: [XML.ResolveEmptyResource]);
	scanDocument

Or if we want to log all external access, we can write the following code:

(XML.XMLParser on: 'http://www.w3.org' asURI)
	validate: false;
	entityResolver: (XML.PluggableEntityResolver withBlock: [:publicID :systemID |
		Core.Transcript cr; show: publicID; tab; show: systemID.
		XML.ResolveDefaultResource]);
	scanDocument

We can selectively choose to allow the regular download code or deny it like in the following:

(XML.XMLParser on: 'http://www.w3.org' asURI)
	validate: false;
	entityResolver: (XML.PluggableEntityResolver withBlock: [:publicID :systemID |
		(UI.Dialog confirm: 'Download the resource: ', systemID, ' ?')
			ifTrue:	[XML.ResolveDefaultResource]
			ifFalse:	[XML.ResolveEmptyResource]]);
	scanDocument

And finally, we can fake a download by returning our own stream - so we could make a local copy of the w3 HTML entities and avoid a download while still resulting in the same parse as if we had hit the internet, eg:

(XML.XMLParser on: 'http://www.w3.org' asURI)
	validate: false;
	entityResolver: (XML.PluggableEntityResolver withBlock: [:publicID :systemID | '' readStream]);
	scanDocument