JSON Integration Adapter

Project Goal

To create an adapter for the NexJ Core similar in methodology and design of existing message adapters (XML, Fixed, CSV etc.) but using the JSON format

Contributors

• Grace Batumbya
• Brian Lim

Current Status

Changes being implemented as requested from first code review.

Code Review 1 Results

Code review took place on August 9th, 2011 at NexJ with Grace Batumbya, Brian Lim and Andrew Borzenko in attendance.

Major changes include:

1. Allowing modes for all message parts, not just the root.
2. Refactoring methods to check for != right condition instead of checking for the wrong condition to throw an exception. (whitelist instead of blacklist)
3. Formatting all primitives before writing or storing them, particularly timestamps and other format sensitive strings.
4. Creating a RootJSONMessagePartMapping extending JSONMessagePartMapping

Code Review Notes

Issue	Description	Status	Priority
Exception TODO	Complete exceptions marked by TODO	Incomplete
Use Hashholder instead of HashSet	Use HashHolder in the XMLJSONMessagePartLoader instead of HashSet and whenever appropriate replace HashSet with HashHolder	Incomplete
Change 0x000 for byte	For root modes, change 0xXXX to a cast to byte. Also do not use leading zeroes.	Incomplete
Document methods fully in javadoc style	Document every method fully in javadoc style. Also document parameters, and use full English sentences. Make sure special settings like root mode or mode are well documented enough to use	Incomplete
Remove/Use InvocationContextAware	Do not implement the invocation context aware interface if not used by the class. Or, use the interface for formatting of primitives.	Incomplete
Name methods XXXPart, not XXXNode	Rename methods to be XXXPart, for example as writeCompositeMessagePart, instead of writeCompositeMessageNode.	Incomplete
Format primitive values before outputting to writers.	Use the StringFormatter to format primitive values before outputting to the writer. For example, timestamps must be formatted before outputting to the writer.	Incomplete
Make sure exceptions output the correct message part.	Make sure exceptions output the correct message part, for example the inner part, rather than always the root part. More specific is better.	Incomplete
Wrap fail	Wrap all instances of fail with IntegrationException.	Incomplete
Change design document to note that MessageTable only supports one message.	Design document should reflect all special settings of JSON parser and formatter. Specifically, any normally standard features which are not implemented by the JSON parser/formatter should be noted.	Incomplete
Override JSONParse	Override parse method of JSONParser in the JSONMessageParser class. Do this to avoid changing JSONParser.	Incomplete
Use primitive formatter when reading.	When reading information in from a stream in the formatter, use primitive formatter to determine the correct format of a primitive.	Incomplete
Remove spaces after new Object[]{(remove space here)}	Remove spaces after the curly brackets for all exceptions.	Incomplete
Add a newline after creating each object or after each decision construct.	Add newlines to conform to NexJ coding style.	Incomplete
Remove StringParsre from JSONMessagePartMapping	Either remove StringParser from the mapping, or use it in formatting primitives. Do not leave it in if it is not used.	Incomplete
Create class called RootJSONMapping	Create a class called RootJSONMapping for each CompositeMessagePart. This root JSON mapping will handle the "mode" for each CompositeMessagePart.	Incomplete
Allow root modes for each CompositeMessagePart	Rename root mode to mode, so that each CompositeMessagePart can use every type of mode.	Incomplete
Rename m_sFormat	Rename the variable m_sFormat if not used (or use it and keep the name).	Incomplete
Space before break keyword.	Put a space before all break keywords.	Incomplete
Pass enumerations into formatMessageRoot	Do not use member/class variables for storing information unless absolutely necessary. Prefer to pass parameters around so that objects do not grow large as the parser/formatter iterates.	Incomplete

Project Phases

JSON Integration Adapter Phases

Phase I. Research (DONE)

1. Complete Fundamentals of NexJ Studio tutorial
2. Complete NexJ Integration tutorial
3. Install NexJ Studio Express from source

Phase II. Design Proposal (DONE)

1. Receive general approval for project
2. Receive approval for JSON encoding options
3. Receive approval for JSON formatting options
4. Final Project Proposal

Phase III. Message Formatter (DONE)

1. Add JSON to .XSD base types
2. Create sample JSON .message
3. Parse sample JSON .message with JUnit Test
4. Create first set of unit tests for Message Formatter
5. Code Message Formatter first draft
6. Resolve issues with first set of unit tests
7. Create second set of unit tests
8. Resolve issues with second set of unit tests
9. Refactor and code message formatter

Phase IV. Message Parser (DONE)

1. Discover how to connect JSON RPC parser with Message Parser
2. Create first set of trivial unit tests for Message Parser
3. Code Message Parser skeleton
4. Pass first set of trivial unit tests
5. Create second set of unit tests (non-trivial) for Message Parser
6. Code Message Formatter first draft
7. Resolve issues with second set of unit tests
8. Refactor

Phase V. Integration and Stress Tests (IN PROGRESS)

1. Test Message Formatter and Message Parser with integration and stress tests
2. Test through NexJ Studio Express GUI (manual testing)

Phase VI. Internal Code Review

1. Internally review code at CDOT

Phase VII. Optimization

1. Find more code optimizations relevant to Enterprise environments

Phase VIII. First Code Review at NexJ

1. Date TBA

Phase IX. Code Rewrite

1. Rewrite code to conform with NexJ suggestions and standards

Phase X. Second Code Review at NexJ

1. Date TBA

Project Repository

BitBucket : https://bitbucket.org/b_lim/nexj-express-json-integration-adapter/

Technical Notes

Definitions

Term	Description
Formatter	Turns data into JSON
Parser	Turns JSON into data
Message	The metadata to be used when formatting and parsing a message. The grammar. It is a tree-like structure.
MessageInstance	The information to be parsed or formatted

Because the Integration layer allows a Model to interact with external applications asynchronously, a description / grammar of the data to be exchanged between the two systems is defined using a Message in the Model.

• Formatting is turning internal NexJ representation of a MessageInstance into JSON. The formatter is given a Message (the grammar), MessageInstance (the data), and a pipe. This accomplished by walking the Message tree-like structure while validating the data in the MessageInstance and streaming JSON format to the pipe. If the data in the MessageInstance does not conform to the Message, the formatter throws an exception.

• Parsing is turning JSON representation of a MessageInstance into an internal NexJ representation of a MessageInstance. The parser is given a Message (the grammar) and an input stream. First the input stream is parsed for JSON using a JSONParser which returns either an Object or an Array. The formatter walks the Message building the structure of the MessageInstance and adding validated values in the object returned by the JSONParser to the MessageInstance. If the data from the input stream does not match the expected input at any time, the parser throws an exception.

Message Fundamentals

Message - Messages can contain values or other messages. The red nodes are messages, the green nodes are values.

Internally, messages are Transfer Objects. To determine if a node is a message, use instanceof on CompositeMessagePartInstance after retrieving the MessagePart. For example,

public void format(TransferObject tobj, Message message, Output out) throws IntegrationException
{
...
CompositeMessagePart root = message.getRoot(); // Gets the root of the message.
Iterator it = root.getPartIterator();
while (it.hasNext())
{
  part = (MessagePart)it.next());
  if (part instanceof CompositeMessagePartInstance)
  {
    // This part is a message
  }
  else if (part instanceof PrimitiveMessagePart)
  {
    // This part is a value
  }
...
}
...

NOTE: THE MESSAGE PARAMETER DOES NOT CONTAIN THE MESSAGE VALUES. THE MESSAGE PARAMETER CONTAINS THE MESSAGE METADATA. THE TRANSFER OBJECT CONTAINS THE MESSAGE VALUES WHICH MUST BE VALIDATED AGAINST THE MESSAGE TO ENSURE CORRECT FORMAT.

MessagePart

Parts of a message. Messages can contain values or other messages. NexJ Express has two types of message parts, CompositeMessagePart and PrimitiveMessagePart.

CompositeMessagePart

implementation is CompositeMessagePartInstance . The relationship between CompositeMessagePartInstance and PrimitiveMessagePart with the above picture is as follows - CompositeMessagePartInstance are messages (the red nodes) and PrimitiveMessagePart are values (the green nodes). To determine multiplicity of MessageParts, use isCollection() method of MessagePart. Note that multiplicity of the above screenshot are all [0..1], that is zero or one instance. Possible node multiplicities:

[0..1] 

zero or one instance (i.e. an optional entry)

[1..1] 

exactly one instance (i.e. required)

[1..N] 

collection of at least one, at most N instances

[0..0] 

collection of zero to infinite instances (displayed as [*])

[1..0] 

collection of at least one, possibly unlimited instances

XMLJSONMessageMappingLoader

Used by the framework to autoload JSONMessagePartMapping for each of the message parts.

JSONMessagePartMapping

Each node in the above picture may has a corresponding JSONMessagePartMapping. Each node may have its own mapping, with its own values initialized in XMLJSONMessageMappingLoader. In order to get the mapping, first cast MessagePart to a concrete class such as CompositeMessagePartInstance or PrimitiveMessagePart, then use part.getMapping(). The purpose of the mapping is metadata for each node.

JSONMessageFormatter

Used to turn messages into JSON format.

JSONMessageParser

Used to turn JSON into a message.

Algorithm

format(TransferObject tobj, Message message, Output out) Algorithm for format method is as follows

1. Retrieve the root of the message
2. Iterate through each of the parts of the root
   1. If the part is a CompositeMessagePartInstance, it is a message node
   2. If the part is a PrimitiveMessagePart, it is a value node
   3. Validate the part against the MessagePartMapping. Different types of validation must be done if the part is a Composite versus if it is a Primitive.
   4. If the part is a CompositeMessagePartInstance, recursively call the format method with TransferObject set to the part. Suggested to overload the format method to format(TransferObject tobj, MessagePart message, Output out) and pass in the part, since retrieving a message root with getRoot() will always get the highest root of the message but what you want is the parent.
   5. If the part is a PrimitiveMessagePart, write the message part to the output.

Resources

JSON RFC : http://www.ietf.org/rfc/rfc4627.txt
Introduction to NexJ Studio Express : https://www.projects.openhealthtools.org/sf/go/doc1771?nav=1
NexJ Developer's Guide
NexJ Integration Fundamentals
Open Health Tools Integration Platform https://www.projects.openhealthtools.org/sf/projects/oht_aip/