Clarify the positions of the lexer and parser
The lexer and parser use unclear names to indicate their positions in the
source code. I propose the following renamings.
Lexer:
```
pos -> next_pos # it's actually the next pos!
last_pos -> pos # it's actually the current pos!
curr -> ch # the current char
curr_is -> ch_is # tests the current char
col (unchanged) # the current column
```
parser
```
- last_span -> prev_span # the previous token's span
- last_token_kind -> prev_token_kind # the previous token's kind
- LastTokenKind -> PrevTokenKind # ditto (but the type)
- token (unchanged) # the current token
- span (unchanged) # the current span
```
Things to note:
- This proposal removes all uses of "last", which is an unclear word because it
could mean (a) previous, (b) final, or (c) most recent, i.e. current.
- The "current" things (ch, col, token, span) consistently lack a prefix. The
"previous" and "next" things consistently have a prefix.
`opaque::Decoder::read_str` is very hot within `rustc` due to its use in
the reading of crate metadata, and it currently returns a `String`. This
commit changes it to instead return a `Cow<str>`, which avoids a heap
allocation.
This change reduces the number of calls to `malloc` by almost 10% in
some benchmarks.
This is a [breaking-change] to libserialize.
The two branches of this `if` compute the same value. This commit gets
rid of the first branch, which makes this calculation identical to the
one in scan_block_comment().
First, assert! is redundant w.r.t. the unwrap() immediately afterwards.
Second, `byte_offset_diff` is effectively computed as
`current_byte_offset + ch.len_utf8() - current_byte_offset` (with `next`
as an intermediate) which is silly and can be simplified.
Allow more non-inline modules in blocks
Currently, non-inline modules without a `#[path]` attribute are not allowed in blocks.
This PR allows non-inline modules that have an ancestor module with a `#[path]` attribute, provided there is not a nearer ancestor block.
For example,
```rust
fn main() {
#[path = "..."] mod foo {
mod bar; //< allowed by this PR
fn f() {
mod bar; //< still an error
}
}
}
```
Fixes#36772.
r? @nikomatsakis
This applies the HIR changes from the previous commits to the AST, and
is thus a syntax-[breaking-change]
Renames `PatKind::Vec` to `PatKind::Slice`, since these are called slice
patterns, not vec patterns. Renames `TyKind::Vec`, which represents the
type `[T]`, to `TyKind::Slice`. Renames `TyKind::FixedLengthVec` to
`TyKind::Array`.
I am using `ThinAttributes` rather than a vector for attributes
attached to generics, since I expect almost all lifetime and types
parameters to not carry any attributes.
Adds a `ProcMacro` form of syntax extension
This commit adds syntax extension forms matching the types for procedural macros 2.0 (RFC #1566), these still require the usual syntax extension boiler plate, but this is a first step towards proper implementation and should be useful for macros 1.1 stuff too.
Supports both attribute-like and function-like macros.
Note that RFC #1566 has not been accepted yet, but I think there is consensus that we want to head in vaguely that direction and so this PR will be useful in any case. It is also fairly easy to undo and does not break any existing programs.
This is related to #35957 in that I hope it can be used in the implementation of macros 1.1, however, there is no direct overlap and is more of a complement than a competing proposal. There is still a fair bit of work to do before the two can be combined.
r? @jseyfried
cc @alexcrichton, @cgswords, @eddyb, @aturon
This commit adds syntax extension forms matching the types for procedural macros 2.0 (RFC #1566), these still require the usual syntax extension boiler plate, but this is a first step towards proper implementation and should be useful for macros 1.1 stuff too.
Supports both attribute-like and function-like macros.
Optimize the parser's last token handling.
The parser currently makes a heap copy of the last token in four cases:
identifiers, paths, doc comments, and commas. The identifier and
interpolation cases are unused, and for doc comments and commas we only
need to record their presence, not their value.
This commit consolidates the last token handling and avoids the
unnecessary copies by replacing `last_token`, `last_token_eof`, and
`last_token_interpolated` with a new field `last_token_kind`. This
simplifies the parser slightly and speeds up parsing on some files by
3--4%.
Overhaul char_lit()
This commit does the following.
- Removes parsing support for '\X12', '\u123456' and '\U12345678' char
literals. These are no longer valid Rust and rejected by the lexer.
(This strange-sounding situation occurs because the parser rescans
char literals to compute their value.)
- Rearranges the function so that all the escaped values are handled in
a single `match`. The error-handling strategy is based on the one used
by byte_lit().
Avoid loading and parsing unconfigured non-inline modules.
For example, `#[cfg(any())] mod foo;` will always compile after this PR, even if `foo.rs` and `foo/mod.rs` do not exist or do not contain valid Rust.
Fixes#36478 and fixes#27873.
r? @nrc
The parser currently makes a heap copy of the last token in four cases:
identifiers, paths, doc comments, and commas. The identifier and
interpolation cases are unused, and for doc comments and commas we only
need to record their presence, not their value.
This commit consolidates the last token handling and avoids the
unnecessary copies by replacing `last_token`, `last_token_eof`, and
`last_token_interpolated` with a new field `last_token_kind`. This
simplifies the parser slightly and speeds up parsing on some files by
3--4%.
This commit does the following.
- Removes parsing support for '\X12', '\u123456' and '\U12345678' char
literals. These are no longer valid Rust and rejected by the lexer.
(This strange-sounding situation occurs because the parser rescans
char literals to compute their value.)
- Rearranges the function so that all the escaped values are handled in
a single `match`, and changes the error-handling to use vanilla
assert!() and unwrap().
This reduces the time taken to run
`rustc -Zparse-only rustc-benchmarks/issue-32278-big-array-of-strings`
from 0.18s to 0.15s on my machine, and reduces the number of
instructions (as measured by Cachegrind) from 1.34B to 1.01B.
With the change applied, the time to fully compile that benchmark is
1.96s, so this is a 1.5% improvement.
