libxslt

node-libxslt

Node.js bindings for libxslt compatible with libxmljs.

Installation

npm install libxslt

Basic usage

var lixslt = require('libxslt');

var stylesheet = libxslt.parse(stylesheetString);

var params = {
	MyParam: 'my value'
};

// 'params' parameter is optional
stylesheet.apply(documentString, params, function(err, result){
	// err contains any error from parsing the document or applying the stylesheet
	// result is a string containing the result of the transformation
});

Libxmljs integration

Node-libxslt depends on libxmljs in the same way that libxslt depends on libxml. This dependancy makes possible to bundle and to load in memory libxml only once for users of both libraries.

It is possible to work with libxmljs documents instead of strings:

var lixslt = require('libxslt');
var libxmljs = require('libxmljs');

var stylesheetObj = libxmljs.parseXml(stylesheetString);
var stylesheet = libxslt.parse(stylesheetObj);

var document = libxmljs.parseXml(documentString);
stylesheet.apply(document, function(err, result){
	// result is now a libxmljs document containing the result of the transformation
});

This is only useful if you already needed to parse a document before applying the stylesheet for previous manipulations. Or if you wish to be returned a document instead of a string for ulterior manipulations. In these cases you will prevent extraneous parsings and serializations.

Sync or async

The same apply() function can be used in synchronous mode simply by removing the callback parameter. In this case if a parsing error occurs it will be thrown.

var lixslt = require('libxslt');

var stylesheet = libxslt.parse(stylesheetString);

var result = stylesheet.apply(documentString);

The asynchronous function uses the libuv work queue to provide parallelized computation in node.js worker threads. This makes it non-blocking for the main event loop of node.js.

Note that libxmljs parsing doesn't use the work queue, so only a part of the process is actually parallelized.

A small benchmark is available in the project. It has a very limited scope, it uses always the same small transformation a few thousand times. To run it use:

node benchmark.js

This is an example of its results with an intel core i5 3.1GHz:

10000 synchronous apply from parsed doc			 		in 331ms = 30211/s
10000 asynchronous apply in series from parsed doc		in 538ms = 18587/s
10000 asynchronous apply in parallel from parsed doc	in 217ms = 46083/s

Observations:

• it's pretty fast !
• asynchronous is slower when running in series.
• asynchronous can become faster when concurrency is high.

Conclusion:

• use asynchronous by default it will be kinder to your main event loop and is pretty fast anyway.
• use synchronous only if you really want the highest performance and expect low concurrency.
• of course you can also use synchronous simply to reduce code depth. If you don't expect a huge load it will be ok.

Environment compatibility

For now 64bits linux and 32bits windows are confirmed. Other environments are probably ok, but not checked. Please report an issue if you encounter some difficulties.

Node-libxslt depends on node-gyp, you will need to meet its requirements. This can be a bit painful mostly for windows users. The node-gyp version bundled in your npm will have to be greater than 0.13.0, so you might have to follow these instructions to upgrade. There is no system dependancy otherwise, libxslt is bundled in the project.