json-stream

Streaming JSON.parse and stringify

Based on dominictarr's JSONStream, this version has been modified to handle null values as significant. https://www.npmjs.com/package/JSONStream

Details

Note about this particular version. Objects structured:

{ a: 1, b: null } 

would be read in and be missing the b key from the top level object in the original version. For the purposes of reading in objects where the null value is significant, this needed an adjustment.

This version of this library will read in the object with null values correctly. It does mean that it will ignore undefined values however as I replaced the nullish behaviour with undefined behaviour. For purposes of reading in data that was serialized originally from a SQL database, this is preferable. The test cases in the package have been updated to reflect this.

install

npm install @alexrmturner/json-stream

example

var request = require('request')
  , JSONStream = require('JSONStream')
  , es = require('event-stream')

request({url: 'http://isaacs.couchone.com/registry/_all_docs'})
  .pipe(JSONStream.parse('rows.*'))
  .pipe(es.mapSync(function (data) {
    console.error(data)
    return data
  }))

JSONStream.parse(path)

parse stream of values that match a path

  JSONStream.parse('rows.*.doc')

The .. operator is the recursive descent operator from JSONPath, which will match a child at any depth (see examples below).

If your keys have keys that include . or * etc, use an array instead. ['row', true, /^doc/].

If you use an array, RegExps, booleans, and/or functions. The .. operator is also available in array representation, using {recurse: true}. any object that matches the path will be emitted as 'data' (and piped down stream)

If path is empty or null, no 'data' events are emitted.

If you want to have keys emitted, you can prefix your * operator with $: obj.$* - in this case the data passed to the stream is an object with a key holding the key and a value property holding the data.

Examples

query a couchdb view:

curl -sS localhost:5984/tests/_all_docs&include_docs=true

you will get something like this:

{"total_rows":129,"offset":0,"rows":[
  { "id":"change1_0.6995461115147918"
  , "key":"change1_0.6995461115147918"
  , "value":{"rev":"1-e240bae28c7bb3667f02760f6398d508"}
  , "doc":{
      "_id":  "change1_0.6995461115147918"
    , "_rev": "1-e240bae28c7bb3667f02760f6398d508","hello":1}
  },
  { "id":"change2_0.6995461115147918"
  , "key":"change2_0.6995461115147918"
  , "value":{"rev":"1-13677d36b98c0c075145bb8975105153"}
  , "doc":{
      "_id":"change2_0.6995461115147918"
    , "_rev":"1-13677d36b98c0c075145bb8975105153"
    , "hello":2
    }
  },
]}

we are probably most interested in the rows.*.doc

create a Stream that parses the documents from the feed like this:

var stream = JSONStream.parse(['rows', true, 'doc']) //rows, ANYTHING, doc

stream.on('data', function(data) {
  console.log('received:', data);
});
//emits anything from _before_ the first match
stream.on('header', function (data) {
  console.log('header:', data) // => {"total_rows":129,"offset":0}
})

awesome!

In case you wanted the contents the doc emitted:

var stream = JSONStream.parse(['rows', true, 'doc', {emitKey: true}]) //rows, ANYTHING, doc, items in docs with keys

stream.on('data', function(data) {
  console.log('key:', data.key);
  console.log('value:', data.value);
});

You can also emit the path:

var stream = JSONStream.parse(['rows', true, 'doc', {emitPath: true}]) //rows, ANYTHING, doc, items in docs with keys

stream.on('data', function(data) {
  console.log('path:', data.path);
  console.log('value:', data.value);
});

recursive patterns (..)

JSONStream.parse('docs..value') (or JSONStream.parse(['docs', {recurse: true}, 'value']) using an array) will emit every value object that is a child, grand-child, etc. of the docs object. In this example, it will match exactly 5 times at various depth levels, emitting 0, 1, 2, 3 and 4 as results.

{
  "total": 5,
  "docs": [
    {
      "key": {
        "value": 0,
        "some": "property"
      }
    },
    {"value": 1},
    {"value": 2},
    {"blbl": [{}, {"a":0, "b":1, "value":3}, 10]},
    {"value": 4}
  ]
}

JSONStream.parse(pattern, map)

provide a function that can be used to map or filter the json output. map is passed the value at that node of the pattern, if map return non-nullish (anything but null or undefined) that value will be emitted in the stream. If it returns a nullish value, nothing will be emitted.

JSONStream also emits 'header' and 'footer' events, the 'header' event contains anything in the output that was before the first match, and the 'footer', is anything after the last match.

JSONStream.stringify(open, sep, close)

Create a writable stream.

you may pass in custom open, close, and seperator strings. But, by default, JSONStream.stringify() will create an array, (with default options open='[\n', sep='\n,\n', close='\n]\n')

If you call JSONStream.stringify(false) the elements will only be seperated by a newline.

If you only write one item this will be valid JSON.

If you write many items, you can use a RegExp to split it into valid chunks.

JSONStream.stringifyObject(open, sep, close)

Very much like JSONStream.stringify, but creates a writable stream for objects instead of arrays.

Accordingly, open='{\n', sep='\n,\n', close='\n}\n'.

When you .write() to the stream you must supply an array with [ key, data ] as the first argument.

unix tool

query npm to see all the modules that browserify has ever depended on.

curl https://registry.npmjs.org/browserify | JSONStream 'versions.*.dependencies'

numbers

numbers will be emitted as numbers. huge numbers that cannot be represented in memory as javascript numbers will be emitted as strings. cf https://github.com/creationix/jsonparse/commit/044b268f01c4b8f97fb936fc85d3bcfba179e5bb for details.

Acknowlegements

based originally on https://www.npmjs.com/package/JSONStream

this module depends on https://github.com/creationix/jsonparse by Tim Caswell and also thanks to Florent Jaby for teaching me about parsing with: https://github.com/Floby/node-json-streams

license

Dual-licensed under the MIT License or the Apache License, version 2.0