这个 C# 字典初始化如何正确？

Question

I stumbled upon the following and I'm wondering why it didn't raise a syntax error.

var dict = new Dictionary<string, object>
{
    ["Id"] = Guid.NewGuid(),
    ["Tribes"] = new List<int> { 4, 5 },
    ["MyA"] = new Dictionary<string, object>
    {
        ["Name"] = "Solo",
        ["Points"] = 88
    }
    ["OtherAs"] = new List<Dictionary<string, object>>
    {
        new Dictionary<string, object>
        {
            ["Points"] = 1999
        }
    }
};

Notice that the "," is missing between the "MyA", and "OtherAs".

This is where the confusion happens:

1. The code compiles.
2. The final dictionary "dict" contains only three elements: "Id", "Tribes", and "MyA".
3. The value for all except "MyA" are correct,
4. "MyA" takes the declared value for "OtherAs", while its original value is ignored.

Why isn't this illegal? Is this by design?

Answer 1

The missing comma makes all the difference. It causes the indexer ["OtherAs"] to be applied on this dictionary:

new Dictionary<string, object>
{
    ["Name"] = "Solo",
    ["Points"] = 88
}

So essentially you're saying:

new Dictionary<string, object>
{
    ["Name"] = "Solo",
    ["Points"] = 88
}["OtherAs"] = new List<Dictionary<string, object>>
{
    new Dictionary<string, object>
    {
        ["Points"] = 1999
    }
};

Note that this is an assignment expression ( x = y ). Here x is dictionary with "Name" and "Points", indexed with "OtherAs" and y is the List<Dictionary<string, object>> . An assignment expression evaluates to the value being assigned ( y ), which is the list of dictionaries.

The result of this whole expression is then assigned to the key "MyA", which is why "MyA" has the list of dictionaries.

You can confirm that this is what's happening by changing the type of the dictionary x :

new Dictionary<int, object>
{
    [1] = "Solo",
    [2] = 88
}
// compiler error saying "can't convert string to int"
// so indeed this indexer is applied to the previous dictionary
["OtherAs"] = new List<Dictionary<string, object>>
{
    new Dictionary<string, object>
    {
        ["Points"] = 1999
    }
}

Here is your code, but reformatted and some parentheses added to illustrated how the compiler has parsed it:

["MyA"] 
= 
(
    (
        new Dictionary<string, object>
        {
            ["Name"] = "Solo",
            ["Points"] = 88
        }["OtherAs"] 
    )
    = 
    (
        new List<Dictionary<string, object>>
        {
            new Dictionary<string, object>
            {
                ["Points"] = 1999
            }
        }
    )
)

Answer 2

What's happening here is that you are creating a dictionary and then indexing into it. The result of the indexer/assignment expression is then returned and that is what is getting assigned into the MyA dictionary slot.

This:

["MyA"] = new Dictionary<string, string> 
{
   ["Name"] = "Solo",
   ["Points"] = "88" 
}
["OtherAs"] = new List<Dictionary<string, object>>
{
   new Dictionary<string, object>
   {
       ["Points"] = 1999
   }
}

Can be split out into the following psuedo-code:

var temp = new Dictionary<string, object>
{ 
   ["Name"] = "Solo", 
   ["Points"] = 88 
};
// indexed contains result of assignment
var indexed = temp["OtherAs"] = new List<Dictionary<string, object>>
{
   new Dictionary<string, object>
   {
      ["Points"] = 1999
   }
};
// value is set to result of assignment from previous step
["MyA"] = indexed;
// temp is discarded

The result of assigning to the indexer of the second dictionary is returned (the assignment returns the value assigned/right hand side) That dictionary is a temporary local that just "disappears into the ether". The result of the indexer (the list of dictionaries) is what is placed into the main dictionary in the end.

This is a weird case, which is made easier to fall into due to the use of object as the type of the dictionary values.