Make hir ProjectionKind more precise

[arora-aman]

This commit also categorizing access as Field, Index, or Subslice.

Ideas are taken from mir::ProjectionElem.

Proposed changes: https://github.com/rust-lang/project-rfc-2229/blob/master/hir-place-target.md

Closes: rust-lang/project-rfc-2229#1,
Closes: rust-lang/project-rfc-2229#2

---

This change is 

[arora-aman]

Leaving some thoughts here so that I don't forget

[arora-aman]

Tests passed: http://csclub.uwaterloo.ca/~a52arora/rust-builds/log_cac4587e-0cdb-477b-942b-0f755d0bb958_commit_51168fc3ff85285ae2247dc889205730df37a068

[arora-aman]

Tests passed: http://csclub.uwaterloo.ca/~a52arora/rust-builds/log_c65e9607-0a9b-4790-8581-6a373a4b6fb3_commit_73c44877cada198333bea0c9a049945308ec791d

[arora-aman]

Tests: http://csclub.uwaterloo.ca/~a52arora/rust-builds/log_commit_4253304f654de2212f5b9ff4cbf94eba0ba337b1_4ae6f54e-8cf1-45d7-9795-fb82c163326c

[Azhng]

In rust-lang/project-rfc-2229#1 it mentioned Downcast projection kind, which is not present in this PR. Should this PR mark that issue as closed ?

Reviewable status: 0 of 2 files reviewed, 1 unresolved discussion (waiting on @arora-aman, @Azhng, @ChrisPardy, @jenniferwills, @logmosier, and @null-sleep)

---

src/librustc_typeck/mem_categorization.rs, line 822 at r1 (raw file):

                // S { f1: p1, ..., fN: pN }

                let variant_index = if let Some(variant_index) =

Shouldn't this be 0 since there's is only one variant for a struct ?

[arora-aman]

@Azhng

Enums can be defined like structures too: https://doc.rust-lang.org/book/ch18-03-pattern-syntax.html#destructuring-enums

Downcast in MIR is to represent using an enum variant

[Azhng]

Also use mc::Result for error handling so it's consistent with rest of the code

Reviewable status: 0 of 2 files reviewed, 1 unresolved discussion (waiting on @arora-aman, @ChrisPardy, @jenniferwills, @logmosier, and @null-sleep)

[arora-aman]

Reviewable status: 0 of 2 files reviewed, 1 unresolved discussion (waiting on @Azhng, @ChrisPardy, @jenniferwills, @logmosier, and @null-sleep)

---

src/librustc_typeck/mem_categorization.rs, line 822 at r1 (raw file):

Previously, Azhng (Archer Zhang) wrote…

Shouldn't this be 0 since there's is only one variant for a struct ?

Enums can be defined like structures too: https://doc.rust-lang.org/book/ch18-03-pattern-syntax.html#destructuring-enums

[arora-aman]

Tests: http://csclub.uwaterloo.ca/~a52arora/rust-builds/log_20922e4f-0acf-407c-94d9-17c8f10842ce

[arora-aman]

@Azhng fixed

[Azhng]

Reviewable status: 0 of 2 files reviewed, 2 unresolved discussions (waiting on @arora-aman, @ChrisPardy, @jenniferwills, @logmosier, and @null-sleep)

---

src/librustc_typeck/mem_categorization.rs, line 786 at r2 (raw file):

                // Any ADT related errors would've been catched when reading the `variant_index`
                let total_fields =
                    self.total_fields_in_adt_variant(pat.hir_id, variant_index, pat.span).unwrap();

I think we can use the ? syntax here instead of unwrapping

---

src/librustc_typeck/mem_categorization.rs, line 816 at r2 (raw file):

                for fp in field_pats {
                    let field_ty = self.pat_ty_adjusted(&fp.pat)?;
                    let field_index = self

The code here that retrieve field_index is almost identical to the code in cat_expr_unadjusted. Maybe refactor it to a helper func ?

[Azhng]

Reviewable status: 0 of 2 files reviewed, 1 unresolved discussion (waiting on @arora-aman, @ChrisPardy, @jenniferwills, @logmosier, and @null-sleep)

[roxelo]

LGTM too

[arora-aman]

Tests passed: http://csclub.uwaterloo.ca/~a52arora/rust-builds/log_commit_a24fc4b7f535cbdb416550bd36d0ca06bd8d01ec_bcada83c-e86e-4c0c-bcd0-bdfe4243cc25

[arora-aman]

Rust PR: rust-lang#74140

[Azhng]

Reviewable status: 0 of 2 files reviewed, 2 unresolved discussions (waiting on @arora-aman, @ChrisPardy, @jenniferwills, @logmosier, and @null-sleep)

---

src/librustc_typeck/mem_categorization.rs, line 799 at r3 (raw file):

                    self.total_fields_in_adt_variant(pat.hir_id, variant_index, pat.span)?;

                for (i, subpat) in subpats.iter().enumerate_and_adjust(total_fields, dots_pos) {

TIL

[arora-aman]

Reviewable status: 0 of 2 files reviewed, 2 unresolved discussions (waiting on @arora-aman, @ChrisPardy, @jenniferwills, @logmosier, and @null-sleep)

---

src/librustc_typeck/mem_categorization.rs, line 799 at r3 (raw file):

                    self.total_fields_in_adt_variant(pat.hir_id, variant_index, pat.span)?;

                for (i, subpat) in subpats.iter().enumerate_and_adjust(total_fields, dots_pos) {

TIL

This is only for the very direct specific type slide of HirPat borrow