After seeing a `0`, if it's followed by any of `[0-9]`, `_`, `.`, `e`,
or `E`, we consume all the digits. But in the `.`, `e` and `E` cases
this is pointless because we know there aren't any digits.
Note that at the time of this commit, `unic-emoji-char` seems to have
data tables only up to Unicode 5.0, but Unicode is already newer than
this.
A newer emoji such as `🥺` will not be recognized as an emoji
but older emojis such as `🐱` will.
Use `as_deref` in compiler (but only where it makes sense)
This simplifies some code :3
(there are some changes that are not exacly `as_deref`, but more like "clever `Option`/`Result` method use")
Instead of `ast::Lit`.
Literal lowering now happens at two different times. Expression literals
are lowered when HIR is crated. Attribute literals are lowered during
parsing.
This commit changes the language very slightly. Some programs that used
to not compile now will compile. This is because some invalid literals
that are removed by `cfg` or attribute macros will no longer trigger
errors. See this comment for more details:
https://github.com/rust-lang/rust/pull/102944#issuecomment-1277476773
There are three kinds of "byte" literals: byte literals, byte string
literals, and raw byte string literals. None are allowed to have
non-ASCII chars in them.
Two `EscapeError` variants exist for when that constraint is violated.
- `NonAsciiCharInByte`: used for byte literals and byte string literals.
- `NonAsciiCharInByteString`: used for raw byte string literals.
As a result, the messages for raw byte string literals use different
wording, without good reason. Also, byte string literals are incorrectly
described as "byte constants" in some error messages.
This commit eliminates `NonAsciiCharInByteString` so the three cases are
handled similarly, and described correctly. The `mode` is enough to
distinguish them.
Note: Some existing error messages mention "byte constants" and some
mention "byte literals". I went with the latter here, because it's a
more correct name, as used by the Reference.
It's passed to numerous places where we just need an `is_byte` bool.
Passing the bool avoids the need for some assertions.
Also rename `is_bytes()` as `is_byte()`, to better match `Mode::Byte`,
`Mode::ByteStr`, and `Mode::RawByteStr`.
These have been bugging me for a while.
- `literal_text`: `src` is also used and is shorter and better.
- `first_char`: used even when "first" doesn't make sense; `c` is
shorter and better.
- `curr`: `c` is shorter and better.
- `unescaped_char`: `result` is also used and is shorter and better.
- `second_char`: these have a single use and can be elided.
- Rename `unescape_raw_str_or_raw_byte_str` as
`unescape_raw_str_or_byte_str`, which is more accurate.
- Remove the unused `Mode::in_single_quotes` method.
- Make some assertions more precise, and add a missing one to
`unescape_char_or_byte`.
- Change all the assertions to `debug_assert!`, because this code is
reasonably hot, and the assertions aren't required for memory safety,
and any violations are likely to be sufficiently obvious that normal
tests will trigger them.
`Cursor` keeps track of the position within the current token. But it
uses confusing names that don't make it clear that the "length consumed"
is just within the current token.
This commit renames things to make this clearer.
`Cursor` is currently hidden, and the main tokenization path uses
`rustc_lexer::first_token` which involves constructing a new `Cursor`
for every single token, which is weird. Also, `first_token` also can't
handle empty input, so callers have to check for that first.
This commit makes `Cursor` public, so `StringReader` can contain a
`Cursor`, which results in a simpler structure. The commit also changes
`StringReader::advance_token` so it returns an `Option<Token>`,
simplifying the the empty input case.
From 72 bytes to 12 bytes (on x86-64).
There are two parts to this:
- Changing various source code offsets from 64-bit to 32-bit. This is
not a problem because the rest of rustc also uses 32-bit source code
offsets. This means `Token` is no longer `Copy` but this causes no
problems.
- Removing the `RawStrError` from `LiteralKind`. Raw string literal
invalidity is now indicated by a `None` value within
`RawStr`/`RawByteStr`, and the new `validate_raw_str` function can be
used to re-lex an invalid raw string literal to get the `RawStrError`.
There is one very small change in behaviour. Previously, if a raw string
literal matched both the `InvalidStarter` and `TooManyHashes` cases,
the latter would override the former. This has now changed, because
`raw_double_quoted_string` now uses `?` and so returns immediately upon
detecting the `InvalidStarter` case. I think this is a slight
improvement to report the earlier-detected error, and it explains the
change in the `test_too_many_hashes` test.
The commit also removes a couple of comments that refer to #77629 and
say that the size of these types don't affect performance. These
comments are wrong, though the performance effect is small.
`scan_escape` currently has a fast path (for when the first char isn't
'\\') and a slow path.
This commit changes `scan_escape` so it only handles the slow path, i.e.
the actual escaping code. The fast path is inlined into the two call
sites.
This change makes the code faster, because there is no function call
overhead on the fast path. (`scan_escape` is a big function and doesn't
get inlined.)
This change also improves readability, because it removes a bunch of
mode checks on the the fast paths.
Optimize `rustc_lexer`
The `cursor.first()` method in `rustc_lexer` now calls the `chars.next()` method instead of `chars.nth_char(0)`.
This allows LLVM to optimize the code better. The biggest win is that `eat_while()` is now fully inlined and generates better assembly. This improves the lexer's performance by 35% in a micro-benchmark I made (Lexing all 18MB of code in the compiler directory). But lexing is only a small part of the overall compilation time, so I don't know how significant it is.
Big thanks to criterion and `cargo asm`.
