Json::Streamer

Status and support

• ✔ stable
• ✔ supported
• ✖ no ongoing development

You are viewing the README of version v2.1.0. You can find other releases here.

Branch	Status
Release
Development

This gem will basically spare you the need to define your own callbacks (i.e. implement an actual JSON parser using start_object, end_object, key, value, etc.).

Streaming is useful for

• big files that do not fit in the memory (or you'd rather avoid the risk)
• files read in chunks (e.g. arriving over network)
• cases where you expect some issue with the file (e.g. losing connection to source, invalid data at some point) but would like to get as much data as possible anyway

This gem is aimed at making streaming as easy and convenient as possible.

Highly depends on the event generator. Out of the box the gem uses JSON::Stream. It was chosen because it's a pure Ruby parser with no runtime dependencies. You can use any custom event generator, such as Yajl::FFI which is dependent on the native YAJL library and is ~10 times faster. See the Custom event generators chapter.

I did not measure the performance of my implementation on top of these libraries.

The gem's single runtime dependency is JSON::Stream. It is only loaded if the default event generator is used.

Add this line to your application's Gemfile:

gem 'json-streamer'

And then execute:

$ bundle

Or install it yourself as:

$ gem install json-streamer

require 'json/streamer'

Since v1.2.0

file_stream = File.open('data.json', 'r')
chunk_size = 500 # defaults to 1000

streamer = Json::Streamer.parser(file_io: file_stream, chunk_size: chunk_size)

# Level zero yields the full JSON, first level yields data within the JSON 1-by-1, etc.
streamer.get(nesting_level:1) do |object|
    p object
end

Input:

{
    "object1": "first_level_value",
    "object2": {}
}

Output:

"first_level_value"
{}

streamer.get(key:'desired_key') do |object|
    p object
end

Input:

{
    "obj1" : {
        "desired_key" : "value1"
    },
    "desired_key" : "value2",
    "obj2" : {
        "desired_key" : {
            "desired_key" : "value3"
        }
    }
}

Output:

"value1"
"value2"
"value3"
{"desired_key" => "value3"}

streamer.get(nesting_level:1, yield_values:false) do |object|
    p object
end

Input:

{
    "obj1" : {},
    "key" : "value"
}

Output:

{}

Since v1.3.0

streamer.get(nesting_level:0, symbolize_keys: true) do |object|
    p object
end

Input:

{
    "obj1" : {"key" : "value"}
}

Output:

{:obj1=>{:key=>"value"}}

# Get a JsonStreamer object that provides access to the parser
# but does not start processing immediately
streamer = Json::Streamer.parser

streamer.get(nesting_level:1) do |object|
    p object
end

Then later in your EventMachine handler:

def receive_data(data)
    streamer << data
end

Since v2.1.0

require "yajl/ffi"

Json::Streamer.parser(event_generator: Yajl::FFI::Parser.new)

Any parser can be used that provides the right events. The gem is tested with Yajl::FFI and JSON::Stream.

Since v2.0.0

Custom conditions provide ultimate control over what to yield.

The Conditions API exposes 3 callbacks:

• yield_value
• yield_array
• yield_object

Each of them may be redefined. They are called once the corresponding data (value, array or object) is available. They should return whether the data should be yielded for the outside. They receive the data and the aggregator as parameters.

The aggregator exposes data about the current state of the partly parsed JSON such as:

• level - current level
• key - current key
• value - current value
• key_for_level(level) - key for custom level
• value_for_level(level) - value for custom level
• get - the raw data (in a custom format)

Example usage (inspired by this issue):

conditions = Json::Streamer::Conditions.new
conditions.yield_value = ->(aggregator:, value:) { false }
conditions.yield_array = ->(aggregator:, array:) { false }
conditions.yield_object = lambda do |aggregator:, object:|
    aggregator.level.eql?(2) && aggregator.key_for_level(1).eql?('items1')
end

streamer.get_with_conditions(conditions) do |object|
    p object
end

Input:

{
  "other": "stuff",
  "items1": [
    {
      "key1": "value"
    },
    {
      "key2": "value"
    }
  ],
  "items2": [
    {
      "key3": "value"
    },
    {
      "key4": "value"
    }
  ]
}

Output:

{"key1"=>"value"}
{"key2"=>"value"}

Since v2.1.0

When not passed a block both get and get_with_conditions return an enumerator of the requested objects. When passed a block they return an empty enumerator. This means that when not passed a block the requested objects will accumulate in memory.

Without block

objects = streamer.get(nesting_level:1)
p objects

Input:

{
    "object1": "first_level_value",
    "object2": {}
}

Output:

["first_level_value", {}]

With block

unyielded_objects = streamer.get(nesting_level:1) { |object| do_something(object) }
p unyielded_objects

Input:

{
    "object1": "first_level_value",
    "object2": {}
}

Output:

[]

Check the unit tests for more examples (spec/streamer_spec.rb).

One streamer object handles one set of conditions. For multiple conditions create multiple streamers. For more details see this discussion.

Pre v1.2.0

This functionality is deprecated but kept for compatibility reasons.

# Same as Json::Streamer.parser
streamer = Json::Streamer::JsonStreamer.new

# Same as streamer << data
streamer.parser << data

This project is built around known use-cases. If you have one that isn't covered don't hesitate to open an issue and start a discussion.

Bug reports and pull requests are welcome on GitHub at https://github.com/thisismydesign/json-streamer. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the Contributor Covenant code of conduct.

This project follows C-Hive guides for code style, way of working and other development concerns.

The gem is available as open source under the terms of the MIT License.