Jira (PUP-3272) Remove support for yaml on the network

[Henrik Lindberg (JIRA)]

Title: Message Title

Henrik Lindberg commented on an issue   Re: Remove support for yaml on the network The exercise of removing yaml on the network and removal of SafeYaml proved to be a bit difficult. The code paths for serialization and deserialization are quite different which makes the code base time consuming to navigate and understand. Writing goes via ZAML (and monkey patched classes) to YAML (Psych). Reading used to go via SafeYaml via Psych, now directly via Psych. The act of deserializing is named intern which makes it very hard to find the places that perform deserialization (since there are numerous calls to intern to create symbols) Serialization is handled by code in the "network" namespace but is also used for serialization that is not "over the network". Safe Yaml produces hashes whereas Psych Yaml produces object directly. Some(?, at least one Resource::Status) optimizes serialization to serialize a TagSet as an Array. When Safe Yaml is no longer used the logic in the class is not triggered since Psych sets the values directly with instance_variable_set. The TagSet class lies in its to_data_hash and returns an Array !!! Psych has a callback that can take over the setting for all attributes called init_with that is given a Psych::Coder that can be asked for a Hash using the map method. (Older methods yaml_new, yaml_initialize have either been removed or are deprecated, we have some traces of these old methods in our code base - for Ruby 1.8.7 ?) So, how to fix this mess? If we want to keep the optimization (avoiding serialization of the list of tags as a Ruby object (causes bloat)) we can: refactor the logic for TagSet which is a very thin wrapper around a Hash used as a Set. It is typically used via inclusion of the module Tagging - but it is still complicated and affects quite a large portion of the functionality since Tagging is used in many places - not certain if all serialization of a TagSet uses the same optimization though, and if such a change breaks other logic / serialization format/API. Accept that classes that serialize/deserialize and need processing on deserialization must implement the init_with method Remove the optimization and accept the bloat (unclear if this breaks things since serialization of Status is now different) Rewrite / Refactor how serialization takes place to get better separation of concerns (each object should not have to support different kinds of serialization/deserialization like it does now) - this is naturally much more work. Add Comment   Puppet / PUP-3272 Remove support for yaml on the network YAML on the network is either dangerous or slow. It is dangerous without safe_yaml and slow with it. We've had it deprecated for a while now. Puppet 4 is the time to completely remove it. We can remove:   * vendored safe_yaml   * serializing and deserializing yaml   * special parsing for query parameters that start with {{\-\-\-}}   * more? This message was sent by Atlassian JIRA (v6.1.4#6159-sha1:44eaede) -- You received this message because you are subscribed to the Google Groups "Puppet Bugs" group. To unsubscribe from this group and stop receiving emails from it, send an email to puppet-bugs+unsubscr...@googlegroups.com. To post to this group, send email to puppet-bugs@googlegroups.com. Visit this group at http://groups.google.com/group/puppet-bugs. For more options, visit https://groups.google.com/d/optout.