Streaming XML->JS parser that extracts selected components from XML documents one-by-ony as they are found and pipes them downstream. This method ensures a low memory footprint as neither the XML file nor the corresponding JSON have to be fully loaded into memory at any point in time.

Usage:

xmler([selector],[options])
// returns a stream

The selector can be any of the following:

• String => a selected tag has to match the supplied selector exactly
• RegExp => the fullpath (space delimited) has to pass the regex
• Number => pipes down all element (with any child objects) that reside at a particular depth inside the document.
• Function => element selected when function returns true. The function will be called with the following object:
  • tag : name of the tag
  • path : an array of the tags of the full path
  • attr : an object with the attributes of the element
  • depth : the depth of the object within the xml document
• Array => Each element is a criteria - will be evaluated as or

Each selected element will include all children as a json object. Selecting depth 0 for example will just result in the whole document piped down as one json record. Each object (+ children) will contain a property named __attr__ that contains any attributes defined in the XML document. By default the __attr__ object is non-enumerable, meaning that its not visible to screen or Object keys but can still be accessable.

The piped record will include the following properties: tag, path,depth and value with the element+children placed in the value property.

The following options can be defined

• arrayNotation : default false - enforces that all children are within an array, even when there is only one child node
• coerce : default false - coerces numbers and booleans if turned on, both in text and attribute values. If coerce is an object then custom coercion methods can be defined for each element name / attribute name
• textObject : default false - Enforces that each text element is placed under .text property. The default behavior is: if text object the only child of a node:|{text:'xxxx'} is simplified to `'xxxx'.
• valuesOnly : pipes down just the matched record (i.e. the .value of the default record)
• showAttr : default false - show the __att__ object instead of hiding it behind the non-enumerable definition
• mergeAttr : default false - merges the attributes with the children objects (risking collisions when attribute name == children tag name)

Simple example

Given a sample file.xml

<xml>
  <item>
    <description>First</description>
    <price>4.32</price>
  </item>
  <item>
    <description>Second</description>
    <price>5.73</price>
  </item>
</xml>

Running the following:

fs.createReadStream('file.xml')
  .pipe(xmler(1,{coerce:true}))
  .pipe(stream.Transform({
    transform: (d,e,cb) => return console.log(d) && cb();
  }));

will console.log the following records one by one:

{
  tag: 'item',
  path: ['xml','item'],
  depth: 1,
  value: {
    description: 'First',
    price: 4.32
  }
}

{
  tag: 'item',
  path: ['xml','item'],
  depth: 1,
  value: {
  description: 'Second',
    price: 4.73
  }
}

Using .pipe(xmler('item'),{coerce:true}) will extract the same records. Using .pipe(xmler('price',{coerce:true})) will return an array where the values will be 4.32 and 4.37 and the path will be ['xml','item','price']

Real life example

This example streams the full Open Streetmap XML for North America into JSON records:

var request = require('request');
var bz2 = require('unbzip2-stream');
var etl = require('etl');
var xmler = require('xmler');
var fs = require('fs');

// Keep track of linecount and report every second
var count = 0;
setInterval( () => console.log(count),1000);

request('http://download.geofabrik.de/north-america-latest.osm.bz2')
  .pipe(bz2())
  .pipe(xmler(['node','way','relation']))
  .pipe(etl.map(d => {
    count++;
    return JSON.stringify(d);
  }))
  .pipe(fs.createWriteStream('america.json'));

Same example with the records piped to mongo (bulk = 100 records at a time and max concurrent connections = 5)

var mongo = require('mongodb');
var collection = mongo.connect('mongodb://localhost:27017/osm')
      .then(db => db.collection('osm'));

request('http://download.geofabrik.de/north-america-latest.osm.bz2')
  .pipe(bz2())
  .pipe(xmler(['node','way','relation']))
  .pipe(etl.map(d => {
    count++;
    d._id = d.attr.id;
    return d;
  }))
  .pipe(etl.collect(100))
  .pipe(etl.mongo.update(collection,['_id'],{upsert: true, concurrency: 5}));