When using Psych.load to deserialize a YAML document, the document is translated to an intermediary AST. That intermediary AST is then translated in to a Ruby object graph.
In the opposite direction, when using Psych.dump, the Ruby object graph is translated to an intermediary AST which is then converted to a YAML document.
Psych::Nodes contains all of the classes that make up the nodes of a YAML AST. You can manually build an AST and use one of the visitors (see Psych::Visitors) to convert that AST to either a YAML document or to a Ruby object graph.
Here is an example of building an AST that represents a list with one scalar:
# Create our nodes stream = Psych::Nodes::Stream.new doc = Psych::Nodes::Document.new seq = Psych::Nodes::Sequence.new scalar = Psych::Nodes::Scalar.new('foo') # Build up our tree stream.children << doc doc.children << seq seq.children << scalar
The stream is the root of the tree. We can then convert the tree to YAML:
stream.to_yaml => "---\n- foo\n"
Or convert it to Ruby:
stream.to_ruby => [["foo"]]
A valid YAML AST must have one Psych::Nodes::Stream at the root. A Psych::Nodes::Stream node must have 1 or more Psych::Nodes::Document nodes as children.
Psych::Nodes::Document nodes must have one and only one child. That child may be one of:
Psych::Nodes::Sequence and Psych::Nodes::Mapping nodes may have many children, but Psych::Nodes::Mapping nodes should have an even number of children.
All of these are valid children for Psych::Nodes::Sequence and Psych::Nodes::Mapping nodes:
Psych::Nodes::Scalar and Psych::Nodes::Alias are both terminal nodes and should not have any children.