Storing and accessing complexed Key of dictionary - in nested dict or delimited key? C#

Question

I wonder what is better from performance side? I have big data that I work with. Sometimes Dict inside dict inside dict ...... Initializing that nested dict is more complicated. I prefer to store the keys by a delimited char:

key = key1|key2|...|key3

When I need to access a specific "key" I just split the key and searching for the specific keys. What is better?

What is the type of the key1, key2 and key3? Are these strings? Also is the number of keys of this nested structure the same for all entries, or some entries have more keys than others? — Theodor Zoulias, Nov 22 '22 at 08:34
in AWS DynamoDb for example they give best practices for building a key (primary key) so it is unique in high probability. Look for it on the web or take a look here https://aws.amazon.com/blogs/database/choosing-the-right-dynamodb-partition-key/ — MC LinkTimeError, Nov 22 '22 at 08:52
@TheodorZoulias - keys are strings. Number of keys is always same — axcelenator, Nov 22 '22 at 09:22

Theodor Zoulias · Accepted Answer · 2022-11-25T04:10:33.960

Every time you are concatenating strings to use as a unified key, you are creating garbage for the garbage collector. This has a performance impact that is not trivial to measure, because you dοn't pay the price instantly. You are paying it when the garbage collection kicks in, and has more work to do than usual. Concatenating strings is also not a bullet proof approach, because it is based on the assumption that the delimiter (the pipe character in your example), is never contained in any of the subkeys.

My suggestion is to use value tuples as keys. For example in case each key is composed of 3 string parts, you could declare your dictionary like this:

Dictionary<(string, string, string), MyItem> dictionary = new();
//...
dictionary.Add(("key1", "key2", "key3"), new MyItem());
//...
if (dictionary.TryGetValue(("key1", "key2", "key3"), out var item)) //...

This will search for the keys with case sensitivity. In case you prefer your searches to be case insensitive, you could create a custom comparer that implements the IEqualityComparer<(string, string, string)> interface, implement accordingly the Equals and GetHashCode methods (look here for an example), and pass an instance to the constructor of the dictionary. You'll probably have to use the StringComparer.OrdinalIgnoreCase in the implementation.

Another option is to install the Nito.Comparers package, and create your comparer fluently like this:

var myComparer = Nito.Comparers.EqualityComparerBuilder
    .For<(string, string, string)>()
    .EquateBy(x => x.Item1, StringComparer.OrdinalIgnoreCase)
    .ThenEquateBy(x => x.Item2, StringComparer.OrdinalIgnoreCase)
    .ThenEquateBy(x => x.Item3, StringComparer.OrdinalIgnoreCase);

Storing and accessing complexed Key of dictionary - in nested dict or delimited key? C#

1 Answers1