Support binary interpolation addition

[333fred]

This adds support for treating binary addition of interpolated strings as if they were a single string, for the purposes of conversions to custom handler types and lowering. I implemented this by delaying binding of binary additions composed entirely of interpolated strings until use, and adjusting all the points that currently expect a string to instead expect either an interpolated string or a binary addition of interpolated strings. @dotnet/roslyn-compiler for review: I'd suggest reviewing commit-by-commit. Commit 1 implements the basic changes just for interpolated strings used as strings, and the second commit adds support for conversions. Closes #54584.

Test plan: #51499

[333fred]

/cc @stephentoub fyi.

[stephentoub]

Yay. Thanks.

[333fred]

@dotnet/roslyn-compiler for a second review.

[jcouv]

Looking

[jcouv]

CustomHandler c = " + expression + @";

Not blocking this PR, but we should also think about deconstruction declaration. It should behave like local declarations.
For example: (Handler x, var y) = ($"...." + $"...", 42); should behave like Handler x = $"..." + $"...";.

---

In reply to: 887677366

---

In reply to: 887677366

---

In reply to: 887677366

---

In reply to: 887677366 #Closed

---

Refers to: src/Compilers/CSharp/Test/Semantic/Semantics/InterpolationTests.cs:11710 in 0f97ce1. [](commit_id = 0f97ce1, deletion_comment = False)

[jcouv]

CustomHandler c = " + expression + @";

Do we have tests for various groupings with parens: ($"" + $"") + $"" vs. $"" + ($"" + $"")?

---

In reply to: 887677366

---

Refers to: src/Compilers/CSharp/Test/Semantic/Semantics/InterpolationTests.cs:11710 in 0f97ce1. [](commit_id = 0f97ce1, deletion_comment = False)

[jcouv]

CustomHandler c = " + expression + @";

I'm having some trouble getting an overall picture from this PR. If we're converting a suitable unconverted binary op, what are the desired type and converted type for the binary op and for the interpolated string operands? I don't have a clear idea what I should expect.
Similarly, what is the binary operator symbol we'll surface in the semantic model?
Do we have tests for those?

---

In reply to: 887718786

---

Refers to: src/Compilers/CSharp/Test/Semantic/Semantics/InterpolationTests.cs:11710 in 0f97ce1. [](commit_id = 0f97ce1, deletion_comment = False)

[jcouv]

Done with review (iteration 3). Haven't yet much looked at tests

[333fred]

CustomHandler c = " + expression + @";

what are the desired type and converted type for the binary op and for the interpolated string operands? I don't have a clear idea what I should expect.

The of the operator and all interpolated string components is string, with a converted type of the destination type. This is verified by VerifyInterpolatedStringExpression. I can add binary operator symbol validation to that method as well.

---

In reply to: 887936166

---

Refers to: src/Compilers/CSharp/Test/Semantic/Semantics/InterpolationTests.cs:11710 in 0f97ce1. [](commit_id = 0f97ce1, deletion_comment = False)

[333fred]

CustomHandler c = " + expression + @";

Deconstruction and tuples appear to just work, actually, so I'll add a couple of tests to cover.

---

In reply to: 888491370

---

Refers to: src/Compilers/CSharp/Test/Semantic/Semantics/InterpolationTests.cs:11710 in 0f97ce1. [](commit_id = 0f97ce1, deletion_comment = False)

[333fred]

@chsienki @jcouv thanks for the feedback. It should be addressed in the latest revision. I'll push another commit later tonight adding the expression tree exception to this PR as well, it's on my home machine and I forgot to push it yesterday.

[jcouv]

    }

Glad that deconstruction just worked out.
Not blocking this PR, but I was hoping for some explicit tests of GetConstantValue to capture the behavior you described (constant value of interpolated strings vs. that of binary ops in different conversion scenarios).

---

Refers to: src/Compilers/CSharp/Test/Semantic/Semantics/InterpolationTests.cs:12045 in cbac96b. [](commit_id = cbac96b, deletion_comment = False)

[jcouv]

Done with review pass (iteration 5). One question left unaddressed

[jcouv]

LGTM Thanks (iteration 6). Please follow-up on a test that is affected (remaining open issue)

[chsienki]

Nit: can we get rid of the commented code? either enter the function and return node, or remove.

[333fred]

I intentionally left this in to leave a place for the issue comment.

[jcouv]

Will this target 17.0p3? Branch should change if so