Native JSON Types (WIP)

Creating and Manipulating JSON

Creating json and jsonarray values from strings

The simplest way to populate a json or jsonarray variable is to make use of implicit casting from string values to the json and jsonarray types. With the multiline string declaration this can be achieved as follows:

json jsonPerson = /%
	{
		"name": "Jack",
		"surname": "Marques",
		"phoneNumber": "555-6162"
	}
%/;

jsonarray jsonPeople = /%
	[
		{
            "name": "Jack",
            "surname": "Marques",
            "phoneNumber": "555-6162"
        },
        {
            "name": "Bevin",
            "surname": "Sharp",
            "phoneNumber": "555-0123"
        }
    ]
%/;

Note, that no compile time validation is done to validate that the values being converted represents correctly formatted JSON. This means that if any such conversion issues were to arise, it will only be at runtime. Any conversion errors will result in an error message being logged to the Helium Logging Service:

< {"id":"ca7963c5-4d5d-44e9-ba65-dcbb68127b9f","key":"Runtime Error","value":"[PersonResourceV2:23] {\n            \"name\": \"Jack\",\n            \"surname\": \"Marques\",\n            \"phoneNumber\": \"555-6162\n        } could not be converted to a com.mezzanine.program.web.core.type.JsonType@577db227 value","millis":1536562229362,"appId":"a7c76024-b379-49e4-a3c0-bd2ef30d28ab","appUserId":null}

In addition to the above conversion error, the result of the conversion will be null. This might lead to unexpected null pointer exceptions if not handled appropriately in DSL apps.

Although casting from string values to json and jsonarray might be quick and useful in some specific cases, its uses might be limited and, if the values are constructed "by hand", there is a high probability of undesirable runtime errors.

A better approach for constructing JSON values in a DSL app, is to use the supplied json manipulation built-in functions namely jsonPut and jsonGet.

Manipulating json values using jsonPut and jsonGet

In order to use jsonPut and jsonGet a json variable needs to be declared and instantiated. If the variable is not instantiated, any subsequent operations might lead to null pointer exceptions. For instantiation, the implicit casting between string and json can again be utilised:

json jsonPersonObject = "{}";

jsonPut can be used to add fields and values to a json variable:

jsonPersonObject.jsonPut("name", "John");
jsonPersonObject.jsonPut("surname", "Smith");
jsonPersonObject.jsonPut("luckyNumber", 13);
jsonPersonObject.jsonPut("dob", Date:fromString("1936-05-02"));
jsonPersonObject.jsonPut("age", (Date:daysBetween(Date:fromString("1936-05-02"), Mez:now()))/365);
jsonPersonObject.jsonPut("gender", GENDER.Male);
jsonPersonObject.jsonPut("height", 1.83);
jsonPersonObject.jsonPut("militaryService", true);

The arguments required for jsonPut is firstly the name of the json field, followed by the value for that field.

json values can also be nested:

// JSON representing contact details
json jsonContactDetails = "{}";
jsonContactDetails.jsonPut("phoneNumber", "555-6162");
jsonContactDetails.jsonPut("emailAddress", "john.smith@gmail.com");
 
// Adding contact details JSON to a JSON person
jsonPersonObject.jsonPut("contactDetails", jsonContactDetails);

Similarly, jsonGet can be used to retrieve values:

string name = jsonPersonObject.jsonGet("name");
string surname = jsonPersonObject.jsonGet("surname");
string phoneNumber = jsonPersonObject.jsonGet("phoneNumber");
int luckyNumber = jsonPersonObject.jsonGet("luckyNumber");
date dob = jsonPersonObject.jsonGet("dob");
int age = jsonPersonObject.jsonGet("age");
GENDER gender = jsonPersonObject.jsonGet("gender");
decimal height = jsonPersonObject.jsonGet("height");
bool milService = jsonPersonObject.jsonGet("militaryService");
json contactDetails = jsonPersonObject.jsonGet("contactDetails");

In the code segment above, note the implicit casting for the values being retrieved.

To reference a value that is nested more that one level deep, for example a person's contact detail phone number from the example above, one needs to make use of multiple jsonGet statements and intermediate variables. Chaining references to jsonGet is not valid syntax in the Helium DSL.

// First we get the contact details json object and assign it to an intermediate variable
json contactDetails = jsonPersonObject.jsonGet("contactDetails");
 
// Now we can get specific field values for the contact details
string phoneNum = contactDetails.jsonGet("phoneNumber");
string emailAddr = contactDetails.jsonGet("emailAddress");

Manipulating jsonarray values

In the above example we added a single contact detail object to a person where each field represents a channel of communication. Consider an example where we want to add multiple contact details for a person. For example, a collection of emergency contacts.

At present, no built-in functions are provided to help with the construction of jsonarray variables. Implicit casting between any primitive array and jsonarray and between arrays of json and jsonarray is, however, provided:

// Create parent as emergency contact
json parentEmergencyContact = "{}";
parentEmergencyContact.jsonPut("description", "parent");
parentEmergencyContact.jsonPut("mobileNumber", "27763300000");
 
// Create spouse as emergency contact
json spouseEmergencyContact = "{}";
spouseEmergencyContact.jsonPut("description", "spouse");
spouseEmergencyContact.jsonPut("mobileNumber", "27763300001");
 
// Create a json[] of the above
json[] emergencyContacts;
emergencyContacts.append(parentEmergencyContact);
emergencyContacts.append(spouseEmergencyContact);
 
// Cast it to jsonarray so we can add it to the json object
jsonarray emergencyContactsArray = emergencyContacts;
 
// Add it to the person object
jsonPersonObject.jsonPut("emergencyContacts", emergencyContactsArray);

To retrieve the individual values one can make use of jsonGet and implicit casting to json[]:

// Retrieve the emergency contacts
jsonarray retrievedEmergencyContacts = jsonPersonObject.jsonGet("emergencyContacts");
 
// Cast back to json[]
json[] convertedEmergencyContacts = retrievedEmergencyContacts;
 
// Iterate over or reference as normal DSL collection
foreach(json currentEmergencyContact: convertedEmergencyContacts) {
	.
	.
	.
}
 
json firstEmergencyContact = convertedEmergencyContacts.get(0);

Note that casting from jsonarray to json[] is a relatively expensive operation and if possible, it should be avoided for large arrays.

The example below describes the same concepts as discussed above.

json pet1 = /%
	{
		"name": "Jasmine",
		"type": "Dog"
	}
%/;
    
json pet2 = /%
	{
		"name": "Markus",
		"type": "Mole-rat"
	}
%/;
 
json[] petsArray;
petsArray.append(pet1);
petsArray.append(pet2);
 
// json[] converted to jsonarray
jsonarray petsJsonArray = petsArray;
 
// jsonarray converted json[]
json[] convertedPetsArray = petsJsonArray;

In the above examples the implicit casting between json[] and jsonarray is shown. Casting to primitive arrays is also supported:

int[] numbersPrimitiveArray;
numbersPrimitiveArray.append(1);
numbersPrimitiveArray.append(2);
numbersPrimitiveArray.append(3);
 
// int[] converted to jsonarray
jsonarray numbersJsonArray = numbersPrimitiveArray;
 
// jsonarray converted back to int[]
int[] numbersPrimitiveArray2 = numbersJsonArray;

// string representing a number array in json format converted to jsonarray
jsonarray primeNumbersJsonArray = /%[2, 3, 5, 7, 11, 13, 17, 19, 23, 29]%/;
 
// jsonarray converted to int[]
int[] primeNumbersPrimitiveArray = primeNumbersJsonArray;

Reading json and jsonarray values from the app schema

json and jsonarray types can also be used for object attributes. These values are represented in the app schema by the PostgresSQL jsonb type and can be retrieved in any way that other attribute values can.

No compile time validation is done when converting string values to json or jsonarray. This means that if any incorrectly formatted JSON is used, it will result in a runtime error. These errors will be logged and can be inspected using the Helium Logging Service.

Helium does not support chaining of the jsonGet built-in function. To get values for a json field that is more than one level deep in a top level json object, multiple jsonGet statements and intermediate variables need to be used.

JSON Values as Object Attributes

Lets consider a different example where a Person object is created to represent most of the attributes of a person while including a single json attribute to represent the contact details of a person:

@NotTracked
persistent object Person {
    string name;
    string surname;
    date dob;
    json contactDetails;
}

Currently integration between Helium and Journey does not support JSON types. Note the inclusion of the NotTracked annotation above to avoid any possible syncing for this object. This is compulsory for any persistent object that contains a json or jsonarray attribute. If the NotTracked annotation is not included for persistent objects containing json or jsonarray attributes a compiler error will generated:

Loading source files...
Compiling the application...
|--------------------------------------------------------------------------------------------------------------------------------------------|
| The module "Web" generated the following errors                                                                                            |
|--------------------------------------------------------------------------------------------------------------------------------------------|
| #     | Message                                                                                                                            |
|--------------------------------------------------------------------------------------------------------------------------------------------|
| 1     | [Person:2] The custom object Person is persistent, thus needs to be marked with @NotTracked if json attribute types are declared.  |
|--------------------------------------------------------------------------------------------------------------------------------------------|

Inspecting the database table created by Helium for the Person object, reveals the contactdetails column represented by the jsonb type in PostgreSQL:

helium-app-1=# \d person
                                  Table "snapshot_1535373223392_001.person"
     Column     |            Type             |                          Modifiers                           
----------------+-----------------------------+--------------------------------------------------------------
 _id_           | uuid                        | not null
 _tstamp_       | timestamp without time zone | not null
 person         | text                        | 
 _tx_id_        | bigint                      | not null default txid_current()
 _change_type_  | __he_obj_change_type__      | not null default 'create'::__he_obj_change_type__
 _change_seq_   | bigint                      | not null default nextval('__he_sync_change_seq__'::regclass)
 name           | text                        | 
 surname        | text                        | 
 dob            | date                        | 
 contactdetails | jsonb                       | 
Indexes:
    "person_pkey" PRIMARY KEY, btree (_id_)
    "__idx_he_sync_person_txid__" btree (_tx_id_)

Any method of populating attributes in general is also applicable to JSON attributes. This includes the use of database functions.

Helium Journey integration does not currently support objects with json or jsonarray attributes. For this reason, any persistent object with json or jsonarray attributes needs to be annotated with NotTracked. Failure to do so will result in a compiler error.

Description