Jazimov
Jazimov

Reputation: 13302

Querying JSON Nested Arrays with Linq, JSON.NET, C#

This post is an effort to ask a more direct question related to my other recent post (Picking Out Simple Properties from Hierarchical JSON Part II):

Given nested JSON document, as such:

{
  "Array1": {
    "Array1A": [
      { "Item1": "1" },
      { "Item2": "2" },
      { "Item3": "3" }
    ],
    "Array1B": [
      { "Item1": "1" },
      { "Item2": "2" },
      { "Item3": "3" }
    ]
  },
  "Array2": {
    "Array2A": [
      { "Item1": "1" },
      { "Item2": "2" },
      { "Item3": "3" }
    ]
  },
  "Array3": {
    "Array3A": [
      { "Item1": "1" },
      { "Item2": "2" },
      { "Item3": "3" }
    ],
    "Array3B": [
      { "Item1": "1" },
      { "Item2": "2" },
      {
        "Array3B1": [
          { "Item1": "1" },
          { "Item2": "2" },
          { "Item3": "3" }
        ]
      }
    ],
    "Array3C": [
      { "Item1": "1" },
      { "Item2": "2" },
      { "Item3": "3" }
    ]
  }
}

(Note: Above was validated with JSLint.)

Notice that the JSON is dynamic--I do not know in advance how many arrays there will be or how deep the arrays will nest.

Goal: My goal is to represent each array (that is, Array1, Array2, Array3, Array3A, Array3B, and Array3B1) in a List<JObject> object. Each item in the list will be a collection of JProperty objects containing that array's string properties. Because the List doesn't itself model hierarchical data, I would need to add a synthetic property to each List<JObject> item that references the parent of that array. Thus, the parent of Array1 is an empty string; Array2 is Array1, Array3 is Array2, Array3A is Array3, Array3B is Array3, and Array 3B1 is Array3B...

Question: 1. How can I use C# Linq to create a List<JObject> object that looks like this:

list[0]: 
{"Name":"Array1","Parent":""}

list[1]:
{"Name":"Array1A","Item1":"1","Item2":"2","Item3":"3","Parent":"Array1"}

list[2]:
{"Name":"Array1B","Item1":"1","Item2":"2","Item3":"3","Parent":"Array1"}

list[3]:
{"Name":"Array2","Parent":""}

list[4]:
{"Name":"Array2A","Item1":"1","Item2":"2","Item3":"3","Parent":"Array2"}

list[5]:
{"Name":"Array3","Parent":""}

list[6]:
{"Name":"Array3A","Item1":"1","Item2":"2","Item3":"3","Parent":"Array3"}

list[7]:
{"Name":"Array3B","Item1":"1","Item2":"2","Parent":"Array3"}

list[8]:
{"Name":"Array3B1","Item1":"1","Item2":"2","Item3":"3","Parent":"ArrayB"}

list[9]:
{"Name":"Array3C","Item1":"1","Item2":"2","Item3":"3","Parent":"Array3"}

Please note that:

Upvotes: 2

Views: 7210

Answers (1)

Brian Rogers
Brian Rogers

Reputation: 129827

How about something like this:

List<JObject> list = 
    JObject.Parse(json)
           .Descendants()
           .Where(jt => jt.Type == JTokenType.Property && ((JProperty)jt).Value.HasValues)
           .Cast<JProperty>()
           .Select(prop =>
           {
               var obj = new JObject(new JProperty("Name", prop.Name));
               if (prop.Value.Type == JTokenType.Array)
               {
                   var items = prop.Value.Children<JObject>()
                                         .SelectMany(jo => jo.Properties())
                                         .Where(jp => jp.Value.Type == JTokenType.String);
                   obj.Add(items);
               }
               var parentName = prop.Ancestors()
                                    .Where(jt => jt.Type == JTokenType.Property)
                                    .Select(jt => ((JProperty)jt).Name)
                                    .FirstOrDefault();
               obj.Add("Parent", parentName ?? "");
               return obj;
           })
           .ToList();

Fiddle: https://dotnetfiddle.net/FMxzls

If you're not that familiar with LINQ-to-JSON, here's how it breaks down:

  1. Parse the json string into a JObject

    JObject.Parse(json)
    
  2. From that JObject, get all of its descendant JTokens

           .Descendants()
    
  3. Filter that list to only JProperties whose values have children

           .Where(jt => jt.Type == JTokenType.Property && ((JProperty)jt).Value.HasValues)
    
  4. Cast the JTokens to JProperties to make them easier to work with in the next step

           .Cast<JProperty>()
    
  5. Now, for each JProperty we selected, transform it as follows:

           .Select(prop =>
           {
    
  6. Create a new JObject and add the JProperty's name as the Name property of the new object

               var obj = new JObject(new JProperty("Name", prop.Name));
    
  7. If the value of the JProperty is an array...

               if (prop.Value.Type == JTokenType.Array)
               {
    
  8. Get all the direct children of the array which are JObjects

                   var items = prop.Value.Children<JObject>()
    
  9. From those JObjects, get all the JProperties

                                         .SelectMany(jo => jo.Properties())
    
  10. Filter those JProperties to include only the ones whose values are strings)

                                         .Where(jp => jp.Value.Type == JTokenType.String);
    
  11. Add these item JProperties to the new JObject we created earlier

                    obj.Add(items);
                }
    
  12. Next, find the first ancestor JProperty of the current JProperty and get its name

                var parentName = prop.Ancestors()
                                     .Where(jt => jt.Type == JTokenType.Property)
                                     .Select(jt => ((JProperty)jt).Name)
                                     .FirstOrDefault();
    
  13. Add the parent name to the JObject we're are building; use an empty string if there was no parent

                obj.Add("Parent", parentName ?? "");
    
  14. Continue with the next transform

                return obj;
            })
    
  15. Lastly put all the JObjects we built into a list.

            .ToList();
    

Upvotes: 9

Related Questions