YAML - TypeError：无法转储匿名模块

Question

在 application_controller 的一个动作中，如果我们尝试：

p request.env.to_yaml

我会收到这个错误：

    TypeError: can't dump anonymous module: #<Module:0x007fee26e34ad8>
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:267:in `visit_Module'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:102:in `accept'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:447:in `block in dump_ivars'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:445:in `each'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:445:in `dump_ivars'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:124:in `visit_Object'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:102:in `accept'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:447:in `block in dump_ivars'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:445:in `each'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:445:in `dump_ivars'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:124:in `visit_Object'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:102:in `accept'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:292:in `block in visit_Hash'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:290:in `each'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:290:in `visit_Hash'
    from /Users/twer/.rvm/rubies/ruby-1.9.3-p194/lib/ruby/1.9.1/psych/visitors/yaml_tree.rb:102:in `accept'

我的问题是：如何将request.env 序列化为 yaml？

实际上，我应该将 request.env 传递给delayed_job 并发送电子邮件，但我得到了这个错误，因为delayed_job 需要将对象序列化到数据库中。

Answer 1

physic 的代码听起来有点奇怪，但是

request.env.instance_eval "def name; 'some_name'; end"

可能会奏效。酷，哼？

Answer 2

问题是，request.env 哈希有很多嵌套对象（尤其是模块），无法转换为 yaml。诀窍是删除那些无法转换的哈希部分。

tmp_env = request.env.clone
tmp_env.delete "action_dispatch.routes"
tmp_env.delete "action_controller.instance"
tmp_env["action_dispatch.remote_ip"] = tmp_env["action_dispatch.remote_ip"].to_s 
p tmp_env.to_yaml # now it works

我们首先克隆原始的env 哈希，以免意外修改它。然后我们从副本中删除这些键，这会导致错误。

tmp_env["action_dispatch.routes"] 包含对 ActionDispatch::Routing::RouteSet 对象中未命名模块的引用，这是导致错误的原因。我们最好删除它。

tmp_env["action_controller.instance"] 包含对原始 env-hash 的引用（我们无法转换）。删除它。

最后tmp_env["action_dispatch.remote_ip"] 看起来像一个字符串（当检查它时），但它是一个ActionDispatch::RemoteIp::GetIp 实例。它包含对原始env 哈希的另一个引用。我们将其转换为字符串，因为我不知道您以后是否对那个键感兴趣。

此外，您可以删除更多键以减小 yaml 输出的大小。但是，这应该可以在不抛出您遇到的错误的情况下工作。更精简的解决方案是从一个空的 Hash 开始，并且只复制您在 yaml 输出中真正需要的键。

用 ruby​​ 1.9.3 和 rails 3.2.13 测试

Answer 3

这是我根据 tessi 的例子得出的结论：

module RequestSerializationHelper
  ::SerializableRequest = Struct.new(
    :env,
    :filtered_parameters,
    :fullpath,
    :headers,
    :request_method,
    :remote_ip
  )

  ## From http://stackoverflow.com/questions/7604153/rails-2-3-14-how-to-serialise-an-actioncontrollerrequest-object
  ## with additional modifications

  # build a serializable Struct that out of the given request object, which looks like a real request
  def make_request_serializable(request)
    serializable_request = ::SerializableRequest.new
    serializable_request.env = request.env.clone
    serializable_request.filtered_parameters = request.filtered_parameters.clone if request.respond_to? :filtered_parameters
    serializable_request.fullpath = request.fullpath
    serializable_request.headers = request.respond_to?(:headers) ? request.headers.clone : {}
    serializable_request.request_method = request.request_method

    delete_identified_unserializable_values(serializable_request.env)
    delete_identified_unserializable_values(serializable_request.headers)

    # Some jobs want this, so set it after it's been converted to a string in the env
    serializable_request.remote_ip = serializable_request.env["action_dispatch.remote_ip"]

    # automatically delete anything left that's non-serializable.  If we end up deleting
    # too much and breaking something, here's where to debug it based on info in warning
    delete_unidentified_unserializable_values :env, serializable_request.env
    delete_unidentified_unserializable_values :headers, serializable_request.headers

    serializable_request
  end

  def delete_identified_unserializable_values(hash)
    hash.delete "async.callback"
    hash.delete "action_dispatch.backtrace_cleaner"
    hash.delete "action_dispatch.cookies"
    hash.delete "action_dispatch.request.accepts"
    hash.delete "action_dispatch.routes"
    hash.delete "action_dispatch.logger"
    hash.delete "action_controller.instance"
    hash.delete "rack.input"
    hash.delete "rack.errors"
    hash.delete "rack.session"
    hash.delete "rack.session.options"
    hash["action_dispatch.remote_ip"] = hash["action_dispatch.remote_ip"].to_s
    hash.delete "warden"
    hash.delete_if { |key, _| key =~ /^rack-cache/ }
  end

  private

  def delete_unidentified_unserializable_values(hash_name, hash)
    hash.each do |key, value|
      begin
        serialized = value.to_yaml
        YAML.load(serialized)
      rescue => e
        warning = "RequestSerializationHelper: Automatically removing un(re)serializable entry in " +
          "'#{hash_name}' for key '#{key}' and value '#{value}'.  Exception was: '#{e}'"
        Rails.logger.warn(warning)
        hash.delete key
      end
    end
  end
end