Make it possible to track the original text range

[ijklam]

Description

Make it possible to track the original text range after #line directive supplied.

Used to support #17519

Checklist

• Test cases added
• Performance benchmarks added in case of performance changes
• Release notes entry updated:

[github-actions]

❗ Release notes required

---

✅ Found changes and release notes in following paths:

Change path	Release notes path	Description
src/Compiler	docs/release-notes/.FSharp.Compiler.Service/9.0.300.md

[Martin521]

I had considered exactly this for #18049 - it would simplify that work considerably. And the current Range design (which just throws away the real (original) range) is really bad.
But I considered it too impactful a change in this basic type, where a lot of compute effort is spent to keep it to 64 bit, and now we would more than double the size.
I am curious about the reaction of the team.

[T-Gro]

This is more than doubling the size of a very low-level structure used throughout most representations used in the compiler.

Considering #line a rather edge feature targeting code generators, I don't see this increase well justified.

Given the #line feature usage, I would rather sacrifice UX when #line is encountered at the referenced feature.

[T-Gro]

I had considered exactly this for #18049 - it would simplify that work considerably. And the current Range design (which just throws away the real (original) range) is really bad. But I considered it too impactful a change in this basic type, where a lot of compute effort is spent to keep it to 64 bit, and now we would more than double the size. I am curious about the reaction of the team.

I think the approached you ended up at #18049 makes the right compromise - the only size/perf hit is when #line is actually used, and regular F# code is not paying for it. The LineMaps data structure could fit here well IMO.

[ijklam]

I had considered exactly this for #18049 - it would simplify that work considerably. And the current Range design (which just throws away the real (original) range) is really bad. But I considered it too impactful a change in this basic type, where a lot of compute effort is spent to keep it to 64 bit, and now we would more than double the size. I am curious about the reaction of the team.

I think the approached you ended up at #18049 makes the right compromise - the only size/perf hit is when #line is actually used, and regular F# code is not paying for it. The LineMaps data structure could fit here well IMO.

Ok, I see it. I will close this and revert it from #17519.