JSON Schema Walker
A system that visits all schema objects in a JSON Schema document and makes callbacks before and/or after visiting all of the current schema object's subschemas.
Currently, there are several vocabularies to choose from, which must be used with the two walking functions. As the vocabulary concept is finalized in JSON Schema draft-08, this package will be reworked to support it directly.
Both walkers accept one or two callback functions, but differ slightly in their behavior. In both cases, there are "path" or "parentPath" concepts which are arrays of the JSON property names or array indexes followed to reach the current schema.
Note that walker callbacks are expected to modify the schema structure in place, so clone a copy if you need the original as well.
schemaWalk(schema, preFunc, postFunc, vocabulary)
This simply calls the preFunc callback (if there is one),
subschemaWalk wth an empty parent path of
then calls the postFunc callback.
subschemaWalk(scheam, preFunc, postFunc, parentPath, vocabulary)
Sometimes it is useful to start somewhere within a document,
subschemaWalk accomplishes by taking a parent path.
This is also how the recursive calls are made once you start
a walk- the parent path for each subsequent call is calculated
during the walk.
In particular, handling the root schema differently is common,
subschemaWalk direclty on the root with an empty
 allows skipping the root schema.
postFunc call can be null, or both can be passed for the same walk. They are expected to take the following
as positional parameters:
- The schema object (or boolean) being processed
- The path to the schema from its parent
- The parent schema object, if any
- The path from the schema document root to the parent
Paths are arrays of property names (strings) and array indices (integers) that are analogous to JSON Pointers. Given
["foo", 42] as the path from the root to the parent object, and
["bar"] as the path from the parent object to the schema being processed, the equivalent JSON Pointer fragment for the parent would be
"#/foo/42, while for the schema being processed it would be
The equivalent Relative JSON Pointer from the parent to the schema being processed would be
An empty path is represented as an empty array
. For a pointer from the root, this is equivalent to
# as a JSON Pointer fragment.
For an empty path from the parent to the schema being processed (meaning that both are the same, which happens for the root schema), this is equivalent to
0 as a Relative JSON Pointer.
getVocabulary(schema, defaultVocabulary) This returns the walker vocabulary object based on the schema's
$schema keyword, or returns the default that you pass in if
$schema is not present or not recognized. The current default if none is provided is Draft-07 JSON Hyper-Schema.
Once draft-08 is published with a formalized concept of a schema vocabulary, the vocabulary implementation for this package will likely be substantially reworked.