SA1642 (constructor summary text) special case for incorrect class name

[sharwell]

When SA1642 is reported, it is always reported for the entire <summary> element. In each of the following scenarios, a more specific location should be used for reporting the warning.

1. When the name of the class is not wrapped in a <see> element, but the text is otherwise correct. For example, the warning reported for the following code should only underline the word Customer:

   /// <summary>
   /// Initializes a new instance of the Customer class.
   /// </summary>
   public Customer() { }

2. As in the first case, but now considering an incorrect class name. For example:

   /// <summary>
   /// Initializes a new instance of the NotTheRightType class.
   /// </summary>
   public Customer() { }

3. When the name of the class specified in the cref attribute does not match the name of the current class, but the text is otherwise correct. In this case, the warning should be reported for the cref value instead of the entire <summary> element. For example:

   /// <summary>
   /// Initializes a new instance of the <see cref="NotTheRightType"/> class.
   /// </summary>
   public Customer() { }

[sharwell]

I initially assigned @pdelvo since he did a great job with the initial implementation. I'll take a look soon if you don't have time.

[sharwell]

💡 The code fix for each of the above cases should replace the current type reference with the corrected type reference, as opposed to inserting the entire standard prefix.

[sharwell]

❓ I also noticed a case like the following:

Initializes a new instance of the <see cref="MyType"/> with A and B.

Notice that everything is correct except the word "class" has been omitted. Should we special case handling for code that is close but not exactly correct? If so, how close is close?

[pdelvo]

This is getting complicated. Calculating the Levenshtein distance and having an upper limit would make this rule almost useless. If I have a spelling mistake in there the distance would be very small and I would not get a diagnostic. But we could add a few special cases like the one you mentioned.

[sharwell]

I tend to agree but wanted to ask the question publicly before just deciding 👍

[sharwell]

Doh. I thought I made a new issue for the "closeness" question.

[sharwell]

The incorrect class name situation is common and relatively simple to detect/correct.

[pdelvo]

I am working on this. It is getting a bit complicated but my solution is able to also detect spelling mistakes and changing the comment instead of adding a new one

[sharwell]

Well that would be cool 👍 It's OK if it doesn't work out too.