Streaming parsers are hard to work with, but sometimes you need to parse a really big file. This module gives you the best of both worlds. You give it a specific node to look for, and it will return each of those nodes as an object, one at a time, without loading the whole document into memory at once.
Uses node-expat for fast(est) xml processing.
npm install xml-object-stream
Let's say we have a file, hugePersonDirectory.xml, that looks something like this:
<root>
<people>
<person>...</person>
<person>...</person>
<person>...</person>
<person>...</person>
<person>...</person>
</people>
<root>
You want to do something with each person object, but you can't load them all into memory at once.
xml = require 'xml-object.stream'
fs = require 'fs'
readStream = fs.createReadStream 'hugePersonDirectory.xml'
parser = xml.parse readStream
parser.each 'person', (person) ->
# do something with the person!
The parser emits some streaming events
parser.on 'end', ->
parser.on 'error', (err) ->
parser.on 'close', ->
You can pause and resume it if the xml parser gets too far ahead of your processing
parser.pause()
# then when you catch back up
parser.resume()
Since the parser takes any read stream, you can use it to parse urls without saving them to disk.
Nodes are converted to objects. For the following xml:
<person id="asdf123">
<firstName>Bob</firstName>
<lastName>Wilson</lastName>
<employee id="asdf123"/>
<note author="Joe">Bob is a poor worker</note>
<note author="Jim">Bob spends all his time parsing xml</note>
</person>
You can access attributes with the $
property
person.$.id == "asdf123"
You can access the last child of a given name by its name. Text is accessed with $text
person.firstName.$text == "Bob"
Node names are available under the $name
property
person.$name == "person"
Every child node is put into the $children
array
notes = person.$children.filter (child) ->
return (child.$name is "note")
notes[0].$.author == "Joe"
exports.parse = (nodeReadStream, [options]) ->
# returns a Parser
class Parser
# calls cb each time it finds a node with that name
each: (nodeName, cb) ->
# bind to 'end', 'error', and 'close'
on: (eventName, cb) ->
# pause or resume the read stream to let you processor catch up
pause: ->
resume: ->
{
# removes all namespace information from the node names
stripNamespaces: true
}