Better "expected results" in ADA

A common failure in ADA is the result of poor result formatting. In my experience we can significantly improve on this by having the generation use a regular expression for validation rather than the similarity. This is because the system will often use truncated results (which is good from a readability perspective). For example:

{
    "agentPoolProfiles": [
        {
            "name": "nodepool1",
            "count": 1,
            "vmSize": "Standard_DS2_v2",
            "maxPods": 110,
            "osType": "Linux"
        }
    ],
    "dnsPrefix": "AKSClusterxxx-dns",
    "provisioningState": "Succeeded",
    "resourceGroup": "AKSLabResourceGroupxxx",
    "name": "AKSClusterxxx",
    "type": "Microsoft.ContainerService/ManagedClusters",
    ...
} 

However, this is not valid JSON and thus the validation step fails.

In this case, and in many like it, what we are actually looking for is "provisioningState": "Succeeded". Therefore for responses like this one we can use:

<!-- expected_similarity=.*Succeeded -->

On a similar note, it seems the tool gives a default similarity value of 0.3 which, for the most part, is too low to be reliable.