What is the addressing issue with my complex document update() in MongoDB?

I have been unable to reach into my MongoDB collection and change a value in a complex document. I have tried more variations than the one example shown below, all sorts of variations, but they fail.

I want to change the Value of the Key "air" from "rain" to "clear". In real life, I will not know that the current Value of the Key "air" is "rain".

Note, I am not using the MongoDB _id Object and would like to accomplish this without using it.

3 documents in the weatherSys collection:

    {
    "_id" : ObjectId("58a638fb1831c61917f921c5"),
    "SanFrancisco" : [
        { "sky" : "grey" },
        { "air" : "rain" },
        { "ground" : "wet" }
      ]
    }
    {
    "_id" : ObjectId("58a638fb1831c61917f921c6"),
    "LosAngeles" : [
        { "sky" : "grey" },
        { "air" : "rain" },
        { "ground" : "wet" }
      ]
    }
    {
    "_id" : ObjectId("58a638fb1831c61917f921c7"),
    "SanDiego" : [
        { "sky" : "grey" },
        { "air" : "rain" },
        { "ground" : "wet" }
      ]
    }

var docKey   = "LosAngeles";
var subKey   = "air";
var newValue = "clear";

var query       = {};
//var queryKey  = docKey + ".$";  
query[query]    = subKey;                    // query = { }

var set     = {};
var setKey  = docKey + ".0." + subKey;
set[setKey] = newValue;                      // set = { "weather.0.air" : "clear" }

db.collection('weatherSys').update(query, { $set: set }, function(err, result) {
  if (err) throw err;
});

UPDATE-1: Ok, so I was hoping I could find a layout a bit simpler than you had suggested but I failed. Everything I tried was not addressable at the "air" Key level. So I copy and pasted your exact JSON into my collection and ran it. I'm using MongoChef to manipulate and test the collection.

Here is my new layout drived from pasting your JSON in 3 times to create 3 documents:

When I then attempted to update the "San Francisco" document's "air" key I got an unexpected result. Rather than updating "air":"dry" it created a new "air" key in the "San Francisco" Object:

So I thought ok, lets try the update again and see what happens:

As you can see it updated the "air" key that it had previously created. I could fight this out and try to make it work "my" way but I just want it to work so I reconfigure my collection layout again, along the lines of what is "working":

And run the update again:

Then I verify it by running the update again:

It works, I am updating properly in a multi-document environment. So this is my current working collection layout:

I have a couple of questions about this-

1. I am using the top level Key "weather" in every document. It adds nothing to the information within the document. Is there a layout design change that would not necessitate that Key and the overhead it brings along?

2. Lets say I have to use the "weather" key. Its value is an array, but that array only has one element, the Object which contains the Keys: city, sky, air, and ground. Does addressing necessitate the use of an array with only one element? Or could I get rid of it. Instead of "weather":[{}] could the design be "weather":{} or would I get into non addressability issues again?

3. It appears I can now update() any of the Values for the Keys: air, sky, and ground, but what is the find() structure to say READ the Value of the Key "ground" in one of the documents?

----> OK, I think I've got this question #3- db.weatherSys.find({ "weather.city" : "San Francisco" }, { "weather.ground": 1 })

4. In the original collection layout that you had suggested, could you explain to me why it did not update as you and I had expected but instead created a new the "city" object?

A lot here. I appreciate your sticking with it.

You can't use positional operator for querying the array by its key.

You can access the weather array by index, but that means you know the array index.

For example if you want to update air element value in weather array.

db.collection('weatherSys').update( {}, { $set: { "weather.1.air" : "clear"} } );

Update:

Unfortunately, I can't see any way to update the values without knowing the array index for key.

You don't need query object as your keys are unique .

db.collection('weatherSys').update( {}, { $set: { "SanFrancisco.1.air" : "clear"} } );

or

Other variant if you want to make sure the key exists.

db.collection('weatherSys').update( { "SanFrancisco": { $exists: true } }, { $set: { "SanFrancisco.1.air" : "clear"} } );

Not sure if you can but if you can update your structure to below.

{
    "_id" : ObjectId("58a638fb1831c61917f921c5"),
    "weather" : [
        { 
          "city": "LosAngeles",
          "sky" : "grey" ,
          "air" : "rain" ,
          "ground" : "wet" 
        }
    ]
}

You can now use $positional operator for update.

db.collection('weatherSys').update( {"weather.city":"LosAngeles"}, { $set: { "weather.$.air" : "clear"} } );

I am using the top level Key "weather" in every document. It adds nothing to the information within the document. Is there a layout design change that would not necessitate that Key and the overhead it brings along?

The only layout that I can think of is promoting all the embedded properties to the top level. Sorry, not sure why I didn't think of this the first time around. Sometimes you just need a right question to get the right answer.

{
    "_id" : ObjectId("58a638fb1831c61917f921c5"),
    "city": "LosAngeles",
    "sky" : "grey",
    "air" : "rain",
    "ground" : "wet" 
}

All the updates will be simply top level updates.

db.collection('weatherSys').update( {"city":"LosAngeles"}, { $set: { "air" : "clear"} } );

Lets say I have to use the "weather" key. Its value is an array, but that array only has one element, the Object which contains the Keys: city, sky, air, and ground. Does addressing necessitate the use of an array with only one element? Or could I get rid of it. Instead of "weather":[{}] could the design be "weather":{} or would I get into non addressability issues again?

N/A if you are okay with first suggestion.

It appears I can now update() any of the Values for the Keys: air, sky, and ground, but what is the find() structure to say READ the Value of the Key "ground" in one of the documents?

db.weatherSys.find({ "city" : "San Francisco" }, { "ground": 1 })

In the original collection layout that you had suggested, could you explain to me why it did not update as you and I had expected but instead created a new the "city" object?

That is a copy paste error. I meant to suggest the working layout you have right now. Updated my previous layout.