Dealing with single or multiple information on JSON deserialization

Question

It is well known that it is possible to convert a JSON dictionary to an object using Apex. But as I'm trying to do so with a JSON that is returned from a webservice, I'm facing an issue while trying to convert the data to object.

The problem is that I'm having to connect to a legacy webserver that only returns XML, and I have no access to modify it. So to get it as JSON I built a webservice that acts as a bridge between Salesforce and the original webservice. Salesforce calls the bridge, the bridge calls the webservice, the webservice returns XML to the bridge, and the bridge converts the XML to JSON and returns the JSON to Salesforce.

In a certain part of my XML I get a structure like the following:

<activities>
    <activity>activity description here 1</activity>
    <activity>activity description here 2</activity>
    <activity>activity description here 3</activity>
</activities>

Seems only logical that when this is translated to JSON, it becomes something like this:

"activities": {
    "activity": [
      "activity description here 1",
      "activity description here 2",
      "activity description here 3"
    ]
  }

However, sometimes the legacy webservice might return a single "activity", like:

<activities>
    <activity>activity description here 1</activity>
</activities>

And then the JSON loses its list syntax, and becomes:

"activities" {
    "activity" : "activity description here 1"
}

Because of this, I face errors when deserializing the class. Because Salesforce doesn't know when to use a list of values or not. I created the class to deserialize this JSON to, and I cannot use List<String> activities, because it simply doesn't work when a single element is retrieved.

Has anyone faced an issue like this before? How can I solve this problem?

Answer 1

Apex is a strongly-typed language and so if you have the choice, receiving a structured XML response is preferable. Using the WSDL2Apex tool (https://developer.salesforce.com/docs/atlas.en-us.apexcode.meta/apexcode/apex_callouts_wsdl2apex_gen_code.htm) with the WSDL spec of the web service, you can auto-generate classes and methods for querying the web service. Once you receive the response, you can navigate the auto-generated response object for all you need without have to do type checking. You never have to actually do any XML parsing.

If the WSDL specifies it is possible for an element to have multiple entries, the tool will automatically make it a list instead of a single value. So you'll always have reliably typed objects in the response. You end up with a class that contains a bunch of subclasses and a few method definitions. You get both synchronous and asynchronous versions. The synchronous code to use it ends up looking something like:

generatedClass.requestSubclass req = new generatedClass.requestSubclass(); 
generatedClass.responseSubclass resp = req.requestMethod(parameter1, parameter2, parameter3);

The parameters are also based on classes defined in the WSDL, so if the request has structure to it you'd send instance of those classes into the request method. And then resp contains the entire response in a structured class based on the WSDL schema. All the data inside can be retrieved with simple dot notation.

Answer 2

I solved the problem creating a class that parses the content of my response using JSON.deserializeUntyped. When the JSON is passed to the class, I can identify the key, and detect the type of information that the value of the JSON holds. So in the code I have something like this:

if (themap.containsKey('activities')) {
    if (themap.get('activities') instanceof List<Object>) {
        // then we are dealing with a list of values
    }
    else {
        // if we get here, then we have a single element
    }
}

So I end up creating an empty list at the beginning and filling it with either the list of values or just the single value returned. As far as I've searched, this is the only way of dealing with this situation.