json-transformer

This small project is an attempt to reduce the complexity of maintaining multiple versions of JSON documents.

It includes following:

• A CSS-based query language for querying JSON nodes.
• A mutable AST for transforming documents based on queries.
• A Groovy-based DSL for writing transformations.

A list of transformations supported:

• Add a JSON value to a node.
• Rename a JSON key.
• Delete a node from the tree.
• Move a JSON key/value pair up or sideways in the tree
• Merge a JSON node with another node.
• Partition a JSON node into more nodes.

Crash course

Given a JSON document:

{
    "destination": {
        "value": 42
    },
    "subtree": {       
        "source": "Jon Snow dies in Season 5",
        "foo": "bar",
        "baz": "botch",
        "stop": "the shitty examples"
    }
}

Selectors

You can select all nodes with the name stop by writing:

def document = JsonDocument.parse(...) // some input file
document.select("stop")

The result will be an instance of JsonNodes, that has a reference to all the matching nodes.

You can write some more complex selectors:

document.select("subtree source")

This will select all nodes with the name source that have a ancestor with the name subtree.

Following selectors are supported:

• name gets all nodes matching the name.
• .class gets all the nodes that have a child @class with a value (or array containing) the classname.
• #id gets all the nodes that have a child @id with the value id.

Combinations of selectors are supported:

• ancestorSelector selector gets all nodes matching selector that also have a ancestor matching ancestorSelector.
• parentSelector > selector gets all nodes matching selector that have a parent matching parentSelector.
• [keyName] gets all nodes that have a child named keyName.
• selector[keyName^=Prefix] gets all nodes matching selector that also have a child named keyName whose value starts with Prefix.
• selector[keyName*=Substring] gets all nodes matching selector that also have a child named keyName whose value contains Substring.
• selector[keyName=SomeName] gets all nodes matching selector that also have a child named keyName whose value equals SomeName.

Some transformations

Besides the selector language, the framework allows transformation of JSON trees.

An example:

document.transform("source")
    .moveTo("destination")
    .apply()
    
document.transform("stop")
    .renameTo("we love")
    .apply()
    
document.transform("subree")
    .deleteChild("foo")
    .deleteChild("baz")
    .apply()      

The resulting JSON will be:

{
    "destination": {
        "source": "Jon Snow dies in Season 5",
        "value": 42
    },
    "subtree": {
        "we love": "the shitty examples"
    }
}

Partitioning a subtree

The partition function splits a node into one or more sibling nodes.

document.transform("subtree")
    .partition([["original_source", "source"], ["random_stuff", "foo", "baz", "stop"]])
    .apply()

The resulting JSON:

{
    "destination": {
        "value": 42
    },
    "original_source": {
            "source": "Jon Snow dies in Season 5"
    },
    "random_stuff": {
        "foo": "bar",
        "baz": "botch",
        "stop": "the shitty examples"
    },
    "subtree": {}
}

Transformations in a method

You will need a JsonDocument instance.

This can be created either by parsing the file with the included BaseNodeParser or by using the JacksonConverter class to convert a Jackson-based JSON tree.

The JsonDocument class has a method for helping out with this:

def document = JsonDocument.parse(some_input_stream)

document.transform(...) // Ready for transformation

Defining transformations as Groovy scripts

It's possible to define transformation as .groovy files.

The syntax is as follows:

version 1
comment "Renames the 'name' property to 'billy' and adds a dog named bingo."
 
transformations {
    transform("name")
            .renameTo("billy")
            .apply()
 
    transform("person")
            .addJson("dog", '{"type":"dog", "name": "bingo"}')
            .apply()
}

The script defines a version, a comment and a closure containing the transformations. The closure is executed with a JsonDocument as delegate.

To run these transformations you have to use an instance of the VersionControl class. Following code shows the steps needed to do so:

def control = new VersionControl(new File("/path/to/directory/with/transformation scripts"))
control.apply(document, desiredVersionNumber)

Every script in the directory is read. The version number used in the apply method will be the last transformation that is applied.