如何从 ASP.NET5 中的 config.json 正确读取嵌套配置值？

Question

I was following some examples for ASP.NET 5 and I got stumbled with how to properly read "nested" configuration values (if that is the proper term).

Here is relevant portion of config.json :

{
    "ApplicationName" : "OwNextApp",
    "AppSettings": {
        "SiteTitle": "OwNext"
    },
}

And relevant portion of HomeController.cs :

public IActionResult About()
{
    var appNestedNameFailed = _config.Get("AppSettings.SiteTitle");
    var appNestedNameSuccess = _config.Get("AppSettings:SiteTitle");
    var appName = _config.Get("ApplicationName");
    ViewBag.Message = string.Format(@"Your 
        APP NAME: {0};
        APP NESTED NAME FAILED: {1}; 
        APP NESTED NAME SUCCESS: {2}", 
            appName, appNestedNameFailed, appNestedNameSuccess);

    return View();
}

Value for appNestedNameFailed is empty (my initial try before research). And appNestedNameSuccess has value; after I did research and found in tests for Configuration (relevant code shown):

// Assert
Assert.Equal("IniValue1", config.Get("IniKey1"));
Assert.Equal("IniValue2", config.Get("IniKey2:IniKey3"));

Can someone explain why is this the case? Why would it make sense to use : over . ? From my interaction with JSON data usually . notation works fine, eg How to access nested json data .

Also, I found similar SO question but this does not give explanation of why : was chosen.

Answer 1

That's the convention that we decided upon when we first created the configuration model. We started with json in mind and : is the delimiter there.

Anyways, if you don't want to worry about those conventions, I recommend using the ConfigurationBinder which binds a configuration to a model (a strong type object). Here are the tests on GitHub that can serve as example.

Answer 2

using Microsoft.Extensions.Configuration;
using System.IO;

IConfigurationRoot configuration = new ConfigurationBuilder()
                .SetBasePath(Directory.GetCurrentDirectory())
                .AddJsonFile("appsettings.json")
                .Build();

var connectionString = configuration.GetValue<string>("ConnectionStrings:DefaultConnection");

// or

var connectionString2= configuration.GetSection("ConnectionStrings").GetSection("DefaultConnection").Value;  

appsettings.json:

{
  "ConnectionStrings": {
    "DefaultConnection": "myconnection"
  },
}

Answer 3

Peering deep into the bowels of the JsonConfigurationFileParser source with blame on for the enter/exit methods that look at:

private void VisitJObject(JObject jObject)
{
    foreach (var property in jObject.Properties())
    {
        EnterContext(property.Name);
        VisitProperty(property);
        ExitContext();
    }
}

private void EnterContext(string context)
{
    _context.Push(context);
    _currentPath = string.Join(":", _context.Reverse());
}

private void ExitContext()
{
    _context.Pop();
    _currentPath = string.Join(":", _context.Reverse());
}

it seems that the ASP.NET team should leave more illuminating check-in comments :).

My best guess is that there could be data stored in the config.json file that would need to have a . in it, whereas : would be less common. For instance:

"AppSettings": {
    "Site.Title": "Is .NET getting faster?"
},

It's a bad example, but it seems reasonable that they wanted to be as "safe" as possible and use something outside of the norm. If you wanted to store a type's full name, that would also be slightly easier without needing to worry about a stray period.

"AppSettings": {
    "ImportantTypeName": "WebApp.CoolStuff.Helpers.AwesomeClass"
},