Prefer Dictionary<K, V>.TryGetValue() over guarded indexer access

[terrajobst]

Dictionary<K, V>.ContainsKey(key) followed by Dictionary<K, V>.this[key] can be combined into a single TryGetValue call.

Category: Performance

[jeffhandley]

Estimates:

• Analyzer: Medium
• Fixer: Medium

[carlossanlop]

Suggested severity: info

// Before
        public int TryGet(Dictionary<string, int> data, string key)
        {​​
            if (data.ContainsKey(key))
            {
                return data[key];
            }
            return 0;
        }​​
// After
        public int TryGet(Dictionary<string, int> data, string key)
        {​​
            if (data.TryGetValue(key, out int value))
            {
                return value;
            }
            return 0;
        }​​

It may be possible to preserve unrelated code inside the if statement as long as key is used without modifying the dictionary:

// Before
        public int TryGet(Dictionary<string, int> data, string key)
        {​​
            if (data.ContainsKey(key))
            {
                Foo(data[key]); // Unrelated code
                return data[key];
            }
            return 0;
        }​​
// After
        public int TryGet(Dictionary<string, int> data, string key)
        {​​
            if (data.TryGetValue(key, out int value))
            {
                Foo(value); // Unrelated code
                return value;
            }
            return 0;
        }​​

Avoid triggering if the data is modified before returning:

        public int DontTrigger(Dictionary<string, int> data, string key)
        {​​
            if (data.ContainsKey(key))
            {
                data[key] = DoSoemthing(data[key]); // Unrelated code
                return data[key];
            }
            return 0;
        }​​

[terrajobst]

Video

• Looks good
• This is related to Don't guard Dictionary<K, V>.Remove(key) by ContainsKey(key) #33797 and Prefer Dictionary<K, V>.TryAddValue(key) over guarded Add(key) #33799. We should make sure these analyzers don't fight with each other. We might want to put them in the same analyzer and decide which diagnostic is the most appropriate one to raise. They could still generate different IDs.

[CollinAlpert]

I'd like to do this one. @carlossanlop

[carlossanlop]

It's yours, @CollinAlpert . Thanks for offering your help!