I am trying to write a recursive C# function that will operate on generic dictionaries of the form IDictionary<string, T> where T is either another IDictionary<string, T> or a string.
My first failed attempt looked something like this:
public string HandleDict(IDictionary<string, string> dict){
// handle the leaf-node here
}
public string HandleDict<T>(IDictionary<string, IDictionary<string, T>> dict){
// loop through children
foreach(string key in dict.Keys){
HandleDict(dict[key]);
}
}
I also tried variants of HandleDict<T>(IDictionary<string, T>) where T : IDictionary<string, T> but that also doesn't quite work. I know that what I want to do could be achieved through a custom class as the parameter instead of dictionaries, and that my attempts are conceptually flawed (there's no end to the recursion in the generic function). I would still like to know if there is actually a way to do what I want here using just generic IDictionaries.
In order to store either a Dictionary<string, string> or a Dictionary<string,??> as a value in a dictionary, you'd need to declare the dictionary to be of type Dictionary<string, object>.
public string HandleDict(IDictionary<string, object> dict)
{
foreach (KeyValuePair<string, object> kvp in dict)
{
Then you'd need to check for each value if it's a string or another Dictionary<string, string> or another Dictionary<string, object> -- C# performs overload resolution at compile-time, not at run-time!
var leaf = kvp.Value as IDictionary<string, string>;
if (leaf != null)
{
HandleDict(leaf);
}
else
{
HandleDict((IDictionary<string, object>)kvp.Value);
}
}
}
Recursion would end when a leaf is reached, because the HandleDict overload for Dictionary<string, string> does not continue the recursion.
However, as you'll note, there's a lot of type checking and casting going on, and a Dictionary<string, object> smells badly -- it essentially defeats the purpose of generics.
I suggest you have a look at the Visitor Pattern instead.
The C# type system is not powerful enough to represent the sort of "higher" generic types that you need to do this in the type system proper. Consider Haskell or F# if you need a language with a richer generic type system.
It is possible to do what you want in C# 4 using the 'dynamic' feature, but I'd recommend against it. Doing that basically starts up the compiler again at runtime. Rather than resorting to runtime type analysis I'd spend some time revisiting the decision to use dictionaries in the first place. I suspect there are better data structures for your actual task at hand. Describe that task and we'll see if there are better data structures.
Related
I'm looking to have a error raise to prevent a build if there are duplicate keys in my static Dictionary.
My current Dictionary below
public static readonly Dictionary<string, string> Fruits = new Dictionary<string, string>
{
{"Sobeys", "Apples"},
{"NoFrills", "Oranges"}
}
But lets say someone accidentally changes Sobeys to be Nofrills, I would like a compiler error to be raised to prevent anything to be done until that duplicate key is resolved. May I ask is that possible? If so how abouts would I do that?
public static readonly Dictionary<string, string> Fruits = new Dictionary<string, string>
{
{"NoFrills", "Apples"},
{"NoFrills", "Oranges"}
}
Nope, that's a runtime thing. As soon as the class is loaded into memory it will throw an exception (which are inherently runtime).
You could add a custom check through a diagnostic analyzer but that would be a lot of pain for very little gain. I suggest you just keep it as it is and keep an eye on any exceptions with a new deployment. You can always add a comment to make it clear to other devs that keys have to be unique but where do you stop? A developer is supposed to know basic framework rules like that.
You could also use, for example, an enum as the key which would make it clear at devtime when you try to add something that already exists. Another option is to refer to const string fields although it remains somewhat brittle.
You can also use the following hack (I do not recommend it): convert anonymous type to dictionary.
Anonymous types do not allow duplicate property names.
Example:
Dictionary<string, string> Fruits = ToDictionary(
new
{
Sobeys = "Apples",
NoFrills = "Oranges"
}
);
But this approach has the following limitation: you can have only valid identifiers as keys in your dictionary.
How to implement ToDictionary method is described there:
In c# convert anonymous type into key/value array?
C# compiler does not provide such possibility.
You will get an Exception at runtime.
To solve this, you have to implement your own pre-processing checks or handle runtime exceptions.
I would like an compiler error to be raised
It is not an error. That's why the compiler wan't issue a compile-time error. Collection initializer are syntactic sugar and are translated into respective Add() method (or indexer in case of the alternate syntax) calls.
Your best option is to rely on some 3rd party code analysis tool that can spot such potential errors.
I wanted to add a KeyValuePair<T,U> to a Dictionary<T, U> and I couldn't. I have to pass the key and the value separately, which must mean the Add method has to create a new KeyValuePair object to insert, which can't be very efficient. I can't believe there isn't an Add(KeyValuePair<T, U>) overload on the Add method. Can anyone suggest a possible reason for this apparent oversight?
You can use the IDictionary<TKey,TValue> interface which provides the Add(KeyValuePair<TKey,TValue>) method:
IDictionary<int, string> dictionary = new Dictionary<int, string>();
dictionary.Add(new KeyValuePair<int,string>(0,"0"));
dictionary.Add(new KeyValuePair<int,string>(1,"1"));
Backup a minute...before going down the road of the oversight, you should establish whether creating a new KeyValuePair is really so inefficient.
First off, the Dictionary class is not internally implemented as a set of key/value pairs, but as a bunch of arrays. That aside, let's assume it was just a set of KeyValuePairs and look at efficiency.
The first thing to notice is that KeyValuePair is a structure. The real implication of that is that it has to be copied from the stack to the heap in order to be passed as a method parameter. When the KeyValuePair is added to the dictionary, it would have to be copied a second time to ensure value type semantics.
In order to pass the Key and Value as parameters, each parameter may be either a value type or a reference type. If they are value types, the performance will be very similar to the KeyValuePair route. If they are reference types, this can actually be a faster implementation since only the address needs to be passed around and very little copying has to be done. In both the best case and worst case, this option is marginally better than the KeyValuePair option due to the increased overhead of the KeyValuePair struct itself.
There is such a method – ICollection<KeyValuePair<K, T>>.Add but as it is explicitly implemented you need to cast your dictionary object to that interface to access it.
((ICollection<KeyValuePair<KeyType, ValueType>>)myDict).Add(myPair);
See
List of Explicit Interface Implementations on Dictionary<K, T>'s documentation page (you'll need to scroll down).
Explicit member implementation
The page on this method includes an example.
Should somebody really want to do this, here is an Extension
public static void Add<T, U>(this IDictionary<T, U> dic, KeyValuePair<T, U> KVP)
{
dic.Add(KVP.Key, KVP.Value);
}
but i would recommend to not do this if there is no real need to do this
Unless I'm mistaken, .NET 4.5 and 4.6 adds the ability to add a KeyValuePair to a Dictionary. (If I'm wrong, just notify me and I'll delete this answer.)
https://msdn.microsoft.com/en-us/library/cc673027%28v=vs.110%29.aspx
From the above link, the relevant piece of information is this code example:
public static void Main()
{
// Create a new dictionary of strings, with string keys, and
// access it through the generic ICollection interface. The
// generic ICollection interface views the dictionary as a
// collection of KeyValuePair objects with the same type
// arguments as the dictionary.
//
ICollection<KeyValuePair<String, String>> openWith =
new Dictionary<String, String>();
// Add some elements to the dictionary. When elements are
// added through the ICollection<T> interface, the keys
// and values must be wrapped in KeyValuePair objects.
//
openWith.Add(new KeyValuePair<String,String>("txt", "notepad.exe"));
openWith.Add(new KeyValuePair<String,String>("bmp", "paint.exe"));
openWith.Add(new KeyValuePair<String,String>("dib", "paint.exe"));
openWith.Add(new KeyValuePair<String,String>("rtf", "wordpad.exe"));
...
}
As can be seen, a new object of type Dictionary is created and called openWith. Then a new KVP object is created and added to openWith using the .Add method.
just because the enumerator for the Dictionary class returns a KeyValuePair, does not mean that is how it is implemented internally.
use IDictionary if you really need to pass KVP's because you've already got them in that format. otherwise use assignment or just use the Add method.
What would be wrong with just adding it into your project as an extension?
namespace System.Collection.Generic
{
public static class DictionaryExtensions
{
public static void AddKeyValuePair<K,V>(this IDictionary<K, V> me, KeyValuePair<K, V> other)
{
me.Add(other.Key, other.Value);
}
}
}
I'm not 100% sure, but I think the internal implementation of a Dictionary is a Hash-table, which means key's are converted to hashes to perform quick look ups.
Have a read here if you want to know more about hashtables
http://en.wikipedia.org/wiki/Hash_table
Hi im doing a school assignment, and I need to convert this JAVA code to C#
private Map<ItemID, ProductDescription> descriptions = new HashMap()<ItemID, ProductDescription>;
Is it possible to make a straight conversion?
I've already decided to make ItemID into an int, and ProductDescription is a class.
Yes, of course you can.
Please look into following examples:
IDictionary<int, string> h = new Dictionary<int, string>();
h.Add(1, "a");
h.Add(2, "b");
h.Add(3, "c");
SortedList<int, string> s = new SortedList<int, string>();
s.Add(1, "a");
s.Add(2, "b");
I think this is what you are looking for.
You could use a Dictionary<int, ProductDescription> instead.
Dictionary<TKey, TValue> Class
Represents a collection of keys and values. The key must be unique.
private Dictionary<ItemID, ProductDescription> descriptions = new Dictionary<ItemID, ProductDescription>();
The hasmap indeed allows for one null key entry. In the (rare?) case you would need this I'd simply create a special ItemID and use that for the null key.
You could ofcourse make a dictionary descendant with null key support, but that would be overdoing it imho ;-)
Yes, just replace HashMap with Dictionary. You might want to type the variable as an IDictionary (in the same spirit as the Java code), but that's not strictly necessary.
Yes, You can do the conversion using a Dictionary instead of HashMap. And of course it is more effective to get the idea of each code segment and convert. Trying to convert line by line is not recommended since you may miss a better way that can be used to resolve the problem.
There are many options.
There is an
Hashtable in C#
KeyValuePair So it can be List<KeyValuePair<T,U>>
Dictionary //Preferred
This is a good match but,
private IDictionary<ItemID, ProductDescription> descriptions
= new Dictionary<ItemID, ProductDescription>();
Note
HashMap will accept null key values, where as Dictionary will not.
If you really want to support null key values, I'd like to see you reasoning before attempting a perfect .Net HashMap implementation.
I wonder if there is an implementation of IDictionary interface in the standard library that has property IsFixedSize == true.
In my code, I am using a Dictionary indexed by an enum, like this:
Dictionary<EMyEnum, string> myDictionary = new Dictionary<EMyEnum, string>();
I always have values in the dictionary for all possible values of MyEnum, using a code like this:
foreach (EMyEnum myEnum in Enum.GetValues(typeof(EMyEnum)))
{
m_myDictionary.Add(myEnum, "");
}
so my dictionary actually has a fixed size. It feels kind of 'wrong' to be using Dictionary for it and I am wondering if there is perhaps a better type I am missing.
I was looking in System.Collections.Generic and System.Collections namespaces but could not find anything suitable.
In response to one of the questions, the enum is just a bunch of values, like this:
enum EMyEnum
{
meValue1,
meValue2,
meValue3
}
EDIT:
I have accepted supermem's answer as it is most to the point of the original question. To elaborate my original concern, an ordinary dictionary can have one of the keys/items deleted and then any code that relies on the fact that iterating over keys of the dictionary also iterates over all values of enum will be broken.
I understand that I can create a fixed size array but casting enums to ints is not that pretty, besides it would only work for contiguous enums (good enough in my case).
I think there is a good case for Dictionary that has enum values as keys and it is a pity that such a class does not ship with .Net.
.Net 4.5 actually introduces a new interface that sounds like it may be what you need, IReadOnlyDictionary<TKey, TValue>.
Also, IReadOnlyList<T>
No, there are no fixed size IDictionary implementations shipping with the CLR. (Except for the ReadOnly* sort, but not applicable to your problem.)
You can roll your own, but unless you have profiled this code and determined that the use of Dictionary presents a meaningful bottleneck (IMO, very rare), I'd say to keep it as is. You also have the benefit of readability as Dictionary is very well understood and easy to grok.
If you are using the simple enum declaration and not explicitly assigning values to the elements, you could use just an array:
Array values = Enum.GetValues(typeof(EMyEnum));
string[] myDictionary = new string[values.Length];
foreach (EMyEnum myEnum in values)
{
myDictionary[(int)myEnum] = "";
}
You could even wrap this approach in a custom class that treat the enum type as indexer, if you are bothered by those explicit casts to int everywhere:
class EnumDictionary<EnumType, ValueType>
{
private readonly ValueType[] array = new ValueType[Enum.GetValues(typeof(EnumType)).Length];
public ValueType this[EnumType index]
{
get { return array[Convert.ToInt32(index)]; }
set { array[Convert.ToInt32(index)] = value; }
}
}
EnumDictionary<EMyEnum, string> enumDictionary = new EnumDictionary<EMyEnum, string>();
foreach ( EMyEnum myEnum in Enum.GetValues(typeof(EnumType))
enumDictionary[myEnum] = "";
Now, if this is better than using Dictionary or not is a completely different discussion.
I have:
IDictionary<string, IDictionary<string, IList<long>>> OldDic1;
(just for illustration purposes, it is instantiated and has values - somewhere else)
Why can I do this: ?
Dictionary<string, IDictionary<string, IList<long>>> dic1 =
OldDic1 as Dictionary<string, IDictionary<string, IList<long>>>;
Basically dic1 after executing this line has all the values from OldDic1; works.
However when I do this:
Dictionary<string, Dictionary<string, List<long>>> dic1 =
OldDic1 as Dictionary<string, Dictionary<string, List<long>>>;
I get null, it is the same as casting except it doesn't crash and instead it returns null. So the question is why I can't cast it from the interfaces to types? is there solution, other then changing how it is stored in the first place?
You can only re-cast the outermost interface/class name, not the generic parameters. The reason your second cast doesn't work is the same reason you can't cast from one array type to another even if you can cast the "contained" objects. Here's why:
List<object> objects;
List<string> strings;
objects = strings as List<object>;
// Uh oh, that's not a string!
objects.Add(42);
objects = new List<object> { "The", "answer", "is", 42 };
// Uh oh, now strings contains an integer?
strings = objects as List<string>;
The same thing would happen in your second case. You could add some kind of IDictionary to OldDic1 which is not actually a Dictionary, and then dic1 would blow up. It would have a non-Dictionary value. Ruh roh!
So, when you have containers you can change from IList<X> to List<X> and back as long as X is identical for each.
The behavior is related to the as keyword in C#. A Dictionary is not the same thing as an IDictionary.
If you were casting the other way, you may be able to get it to work in the next version of .NET which has increased support for covariance and contravariance.
The solution you might want to determine why you need to cast to the concrete Dictionary/List, and if it's required, then change the storage of the type.