Analyzer/fixer for converting Regex use to source generator

[stephentoub]

Suggested severity: Info
Suggested category: Performance

For new Regex(...) and Regex.StaticMethod(...) involving patterns, options, and timeouts known at compile time, we should provide an analyzer and fixer that will transform the regex into its corresponding RegexGenerator form, including fixing up the call sites.

Flag

• Use of Regex constructor or static method with inline literals like:

// Flag object creation invocations
Regex myRegex = new Regex("ab|cd", RegexOptions.None, TimeSpan.FromSeconds(2));
// Flag static method invocations
var match = Regex.Match(input, "ab|cd", RegexOptions.None, TimeSpan.FromSeconds(2));

• Use of Regex constructor or static method when using constant values like:

const string pattern = "ab|cd";
const RegexOptions options = RegexOptions.None;

Regex myRegex = new Regex(pattern, options, TimeSpan.FromSeconds(2));
var match = Regex.Match(input, pattern, options, TimeSpan.FromSeconds(2));

Don't Flag

• Use of non-constant variables as parameters like:

public void Foo(RegexOptions parameterOptions)
{
    string nonConstantPattern = "ab|cd";

    Regex myRegex = new Regex(nonConstantPattern , parameterOptions, TimeSpan.FromSeconds(2));
    var match = Regex.Match(input, nonConstantPattern , parameterOptions, TimeSpan.FromSeconds(2));
}

• Options contain NonBacktracking (since this option is not supported by source generator yet) like:

Regex myRegex = new Regex("ab|cd", RegexOptions.NonBacktracking);

Fixer

The fixer will mainly do three things:

• Add partial keyword (if it's not there already) to the type (or types for nested types) which contains the invocation that generated the diagnostic
• Declare a new private static partial method with the RegexGenerator attribute and parameters so the source generator generates the Regex type.
• Change the invocation that produced the diagnostic to use the new static partial method instead.

[ghost]

Tagging subscribers to this area: @dotnet/area-system-text-regularexpressions
See info in area-owners.md if you want to be subscribed.

Issue Details

---

For new Regex(...) and Regex.StaticMethod(...) involving patterns known at compile time, we should provide an analyzer and fixer that will transform the regex into its corresponding RegexGenerator form, including fixing up the call sites

Author:	stephentoub
Assignees:	-
Labels:	api-suggestion, area-System.Text.RegularExpressions
Milestone:	7.0.0

[joperezr]

I already started looking into this on Thursday, so assigning to me.

[meziantou]

I've implemented a similar analyzer/fixer on my side. You can take inspiration from it if you want Analyzer, Fixer, demo

[stephentoub]

I've implemented a similar analyzer/fixer on my side.

Oh, cool, thanks for sharing!

[joperezr]

@meziantou as @stephentoub pointed out, we really appreciate you sharing your work, I took a look and it indeed looks very similar to what we want to have, except that we want to make sure it ships on the same assembly as our source generator. We still need to get the analyzer approved, but once we do, would you be interested in contributing a PR for this analyzer?

[meziantou]

We still need to get the analyzer approved, but once we do, would you be interested in contributing a PR for this analyzer?

Once it is approved, and if I have time at that moment, I can contribute to this analyzer

[terrajobst]

Video

• Looks good as proposed

[joperezr]

Hey @meziantou, this has now been approved, are you still interested on contributing an implementation? If not I already have an initial prototype but wanted to check with you first before I finished it and send out the PR review.

[meziantou]

@joperezr I'm on vacation for 2 weeks, so I won't be able to contribute at the moment. If you already have a prototype, you can create a PR. I will review the change when I get back to work and check if I can add something.

[joperezr]

Sounds good, enjoy your time off.