I'm making a Ruby Sinatra application that uses mongomapper and most of my responses will be in the JSON form.
Confusion
Now I've come across a number of different things that have to do with JSON.
The Std-lib 1.9.3 JSON class: http://www.ruby-doc.org/stdlib-1.9.3/libdoc/json/rdoc/JSON.html
The JSON Gem: http://flori.github.io/json/
ActiveSupport JSON http://api.rubyonrails.org/classes/ActiveSupport/JSON.html because I'm using MongoMapper which uses ActiveSupport.
What works
I'm using a single method to handle responses:
def handleResponse(data, haml_path, haml_locals)
case true
when request.accept.include?("application/json") #JSON requested
return data.to_json
when request.accept.include?("text/html") #HTML requested
return haml(haml_path.to_sym, :locals => haml_locals, :layout => !request.xhr?)
else # Unknown/unsupported type requested
return 406 # Not acceptable
end
end
the line:
return data.to_json
works when data is an instance of one of my MongoMapper model classes:
class DeviceType
include MongoMapper::Document
plugin MongoMapper::Plugins::IdentityMap
connection Mongo::Connection.new($_DB_SERVER_CNC)
set_database_name $_DB_NAME
key :name, String, :required => true, :unique => true
timestamps!
end
I suspect in this case the to_json method comes somewhere from ActiveSupport and is further implemented in the mongomapper framework.
What doesn't work
I'm using the same method to handle errors too. The error class I'm using is one of my own:
# Superclass for all CRUD errors on a specific entity.
class EntityCrudError < StandardError
attr_reader :action # :create, :update, :read or :delete
attr_reader :model # Model class
attr_reader :entity # Entity on which the error occured, or an ID for which no entity was found.
def initialize(action, model, entity = nil)
#action = action
#model = model
#entity = entity
end
end
Of course, when calling to_json on an instance of this class, it doesn't work. Not in a way that makes perfect sense: apparantly this method is actually defined. I've no clue where it would come from. From the stack trace, apparently it is activesupport:
Unexpected error while processing request: object references itself
object references itself
/home/id833541/.rvm/gems/ruby-1.9.3-p392/gems/activesupport-3.2.13/lib/active_support/json/encoding.rb:75:in `check_for_circular_references'
/home/id833541/.rvm/gems/ruby-1.9.3-p392/gems/activesupport-3.2.13/lib/active_support/json/encoding.rb:46:in `encode'
/home/id833541/.rvm/gems/ruby-1.9.3-p392/gems/activesupport-3.2.13/lib/active_support/json/encoding.rb:246:in `block in encode_json'
/home/id833541/.rvm/gems/ruby-1.9.3-p392/gems/activesupport-3.2.13/lib/active_support/json/encoding.rb:246:in `each'
/home/id833541/.rvm/gems/ruby-1.9.3-p392/gems/activesupport-3.2.13/lib/active_support/json/encoding.rb:246:in `map'
/home/id833541/.rvm/gems/ruby-1.9.3-p392/gems/activesupport-3.2.13/lib/active_support/json/encoding.rb:246:in `encode_json'
But where is this method actually defined in my class?
The question
I will need to override the method in my class like this:
# Superclass for all CRUD errors on a specific entity.
class EntityCrudError < StandardError
def to_json
#fields to json
end
end
But I don't know how to proceed. Given the 3 ways mentioned at the top, what's the best option for me?
As it turned out, I didn't need to do anything special.
I had not suspected this soon enough, but the problem is this:
class EntityCrudError < StandardError
..
attr_reader :model # Model class
..
end
This field contains the effective model class:
class DeviceType
..
..
end
And this let to circular references. I now replaced this with just the class name, which will do for my purposes. Now to_json doesn't complain anymore and I'm happy too :)
I'm still wondering what's the difference between all these JSON implementations though.
Related
Information on what's going on here in ruby: http://coderrr.wordpress.com/2008/03/11/constant-name-resolution-in-ruby/
Doesn't help me solve my problem.. but it at least explains they 'why'
I've written the following method:
# delegate to a user permission serializer specific to the given object
# - if a serializer is not found for the given object, check the superclass
#
# #raise [NameError] if none of object, or it's superclasses have a known
# user permission serializer
# #return [UserPermission::*Serializer] returns serialized object
def self.serialized_for(object, user, klass: nil, recursion_limit: 5)
object_class = klass ? klass : object.class
# use demodulize to chop off the namespace and get the generic object name
object_name = object_class.name.demodulize
# bulid serializer name
name = "::UserPermission::#{object_name}Serializer"
begin
binding.pry
permissions = object.user_permissions(user)
return name.constantize.new(permissions)
rescue NameError => e
raise e if recursion_limit < 1
# try with super class
UserPermission.serialized_for(
object,
user,
klass: object_class.superclass,
recursion_limit: recursion_limit - 1
)
end
end
The goal is to be able to retrieve the serializer of any subclass, provided the subclass has a superclass with a serializer already defined. (I'm using ActiveModelSerializers, but that's not important here).
My problem is that I'm receiving a non-namespaced class when name.constantize runs.
My existing classes:
UserPermission
UserPermission::ProposalSerializer
PresentationSerializer < ActiveModel::Serializer
Presentation < Proposal
Proposal < ActiveRecord::Base
What I'm expecting to happen, is that when I call UserPermission.serialized_for with a Presentation, that name.constantize tries to give me a ::UserPermission::PresentationSerializer and then throw a NameError because the class doesn't exist.
What I'm getting instead is ::PresentationSerializer, which is no good - used for a different purpose.
Here is what I came up with for replicating the issue in irb:
(maybe the above context is an overly complicated explanation of this):
class NameSpace; end
class NameSpace::Klass; end
class Klass; end
class SubKlass < Klass; end
Object.const_get "::NameSpace::SubKlass"
=> SubKlass
Object.const_get("::NameSpace").const_get("SubKlass")
=> SubKlass
eval("NameSpace::SubKlass")
(eval):1: warning: toplevel constant SubKlass referenced by NameSpace::SubKlass
=> SubKlass
Is there a way I can constantize "::NameSpace::SubKlass" such that I get a NameError due to NameSpace::SubKlass not existing?
P.S.: I hope the context helps.
Edit: found another problem:
UserPermission::Template < UserPermission::Proposal
UserPermission::Template.superclass
=> Proposal
should be UserPermission::Proposal
UserPermission::Proposal
(pry):9: warning: toplevel constant Proposal referenced by UserPermission::Proposal
=> Proposal
UserPermission::Proposal is a class. So... this is a big problem. o.o
I'm using Ruby 2.1.0
Do not define your classes and modules the short-hand way. You run into scoping issues.
module UserPermission
class Proposal
end
end
module UserPermission
class Template < Proposal
end
end
UserPermission::Template.superclass
# => UserPermission::Proposal
I am using Rails 3.2.13 and postgress.
I am getting below error only in production server
NoMethodError (undefined method `unserialized_value' for "--- []\n":String):
app/controllers/blogs_controller.rb:159:in `content_generators'
I am serializing Array to store it in db. Below is code.
Controller
class BlogsController < ApplicationController
def content_generators
#blog = Blog.find(params[:id])
#users = #blog.content_generators.map do |id|
User.find(id)
end
end
end
Model
class Blog < ActiveRecord::Base
serialize :post_access, Array
serialize :content_generators, Array
attr_accessible :post_access, :content_generators
end
Migration
class AddContentgeneratorsToBlog < ActiveRecord::Migration
def change
add_column :blogs, :content_generators, :string, :default => [].to_yaml
end
end
I have already used serialization. You can see post_access is serialized. And that works perfect.
But now when I added another column content_generators it starts breaking.
Thanks for your help in advance.
Since you are using postgresql I strongly recommend using the built in array functionality:
# Gemfile
gem 'postgres_ext'
class MyMigration
def change
add_column :my_table, :that_array_column, :text, array: true, default: []
end
end
Then remove the serialize calls in your model and that's it. PG serialized array's behave exactly the same as YAML serialized ones on the model, except the db supports some query methods on them.
I've got the following to_json in my view:
<%=raw #forums.to_json(:methods => [:topic_count, :last_post]) %>;
and in the model I've got the following method:
def last_post
post = ForumPost.joins(:forum_topic).where(:forum_topics => {:forum_id => self.id}).order("forum_posts.created_at DESC").first
return post
end
However ForumPost contains a relation to a ForumTopic (belongs_to :forum_topic) and I want to include this ForumTopic in my json so I end up with {..., "last_post":{..., "forum_topic":{...}...}, ...}. How can I accomplish this?
Usually it's better to keep that kind of declaration in the Model. You can override the as_json method in your models. This approach allows you to control the the serialization behavior for your entire object graph, and you can write unit tests for this stuff.
In this example I'm guessing that your #forums variable refers to a model called Forum, if i'm wrong hopefully you still get the idea.
class Forum
...
def as_json(options={})
super(methods: [:topic_count, :last_post])
end
end
class ForumPost
...
def as_json(options={})
super(methods: [:forum_topic])
end
end
I'm using data_mapper/sinatra and trying to create some attributes with attr_accessor. The following example code:
require 'json'
class Person
include DataMapper::Resource
property :id, Serial
property :first_name, String
attr_accessor :last_name
end
ps = Person.new
ps.first_name = "Mike"
ps.last_name = "Smith"
p ps.to_json
produces this output:
"{\"id\":null,\"first_name\":\"Mike\"}"
Obviously I would like for it to give me both the first and last name attributes. Any ideas on how to get this to work in the way one would expect so that my json has all of the attributes?
Also, feel free to also explain why my expectation (that I'd get all of the attributes) is incorrect. I'm guessing some internal list of attributes isn't getting the attr_accessor instance variables added to it or something. But even so, why?
Datamapper has it’s own serialization library, dm-serializer, that provides a to_json method for any Datamapper resource. If you require Datamapper with require 'data_mapper' in your code, you are using the data_mapper meta-gem that requires dm-serializer as part of it’s set up.
The to_json method provided by dm-serializer only serializes the Datamapper properties of your object (i.e. those you’ve specified with property) and not the “normal” properties (that you’ve defined with attr_accessor). This is why you get id and first_name but not last_name.
In order to avoid using dm-serializer you need to explicitly require those libraries you need, rather than rely on data_mapper. You will need at least dm-core and maybe others.
The “normal” json library doesn’t include any attributes in the default to_json call on an object, it just uses the objects to_s method. So in this case, if you replace require 'data_mapper' with require 'dm-core', you will get something like "\"#<Person:0x000001013a0320>\"".
To create json representations of your own objects you need to create your own to_json method. A simple example would be to just hard code the attributes you want in the json:
def to_json
{:id => id, :first_name => first_name, :last_name => last_name}.to_json
end
You could create a method that looks at the attributes and properties of the object and create the appropriate json from that instead of hardcoding them this way.
Note that if you create your own to_json method you could still call require 'data_mapper', your to_json will replace the one provided by dm-serializer. In fact dm-serializer also adds an as_json method that you could use to create the combined to_json method, e.g.:
def to_json
as_json.merge({:last_name => last_name}).to_json
end
Thanks to Matt I did some digging and found the :method param for dm-serializer's to_json method. Their to_json method was pretty decent and was basically just a wrapper for an as_json helper method so I overwrote it by just adding a few lines:
if options[:include_attributes]
options[:methods] = [] if options[:methods].nil?
options[:methods].concat(model.attributes).uniq!
end
The completed method override looks like:
module DataMapper
module Serializer
def to_json(*args)
options = args.first
options = {} unless options.kind_of?(Hash)
if options[:include_attributes]
options[:methods] = [] if options[:methods].nil?
options[:methods].concat(model.attributes).uniq!
end
result = as_json(options)
# default to making JSON
if options.fetch(:to_json, true)
MultiJson.dump(result)
else
result
end
end
end
end
This works along with an attributes method I added to a base module I use with my models. The relevant section is below:
module Base
def self.included(base)
base.extend(ClassMethods)
end
module ClassMethods
def attr_accessor(*vars)
#attributes ||= []
#attributes.concat vars
super(*vars)
end
def attributes
#attributes || []
end
end
def attributes
self.class.attributes
end
end
now my original example:
require 'json'
class Person
include DataMapper::Resource
include Base
property :id, Serial
property :first_name, String
attr_accessor :last_name
end
ps = Person.new
ps.first_name = "Mike"
ps.last_name = "Smith"
p ps.to_json :include_attributes => true
Works as expected, with the new option parameter.
What I could have done to selectively get the attributes I wanted without having to do the extra work was to just pass the attribute names into the :methods param.
p ps.to_json :methods => [:last_name]
Or, since I already had my Base class:
p ps.to_json :methods => Person.attributes
Now I just need to figure out how I want to support collections.
First off, I am not using Rails. I am using Sinatra for this project with Active Record.
I want to be able to override either to_json or as_json on my Model class and have it define some 'default' options. For example I have the following:
class Vendor < ActiveRecord::Base
def to_json(options = {})
if options.empty?
super :only => [:id, :name]
else
super options
end
end
end
where Vendor has more attributes than just id and name. In my route I have something like the following:
#vendors = Vendor.where({})
#vendors.to_json
Here #vendors is an Array vendor objects (obviously). The returned json is, however, not invoking my to_json method and is returning all of the models attributes.
I don't really have the option of modifying the route because I am actually using a modified sinatra-rest gem (http://github.com/mikeycgto/sinatra-rest).
Any ideas on how to achieve this functionality? I could do something like the following in my sinatra-rest gem but this seems silly:
#PLURAL.collect! { |obj| obj.to_json }
Try overriding serializable_hash intead:
def serializable_hash(options = nil)
{ :id => id, :name => name }
end
More information here.
If you override as_json instead of to_json, each element in the array will format with as_json before the array is converted to JSON
I'm using the following to only expose only accessible attributes:
def as_json(options = {})
options[:only] ||= self.class.accessible_attributes.to_a
super(options)
end