perf(parse): Reduce throwaway work

toml-rs · Jul 29, 2024 · 4e9b4e8 · 4e9b4e8
1 parent 8be588b
commit 4e9b4e8
Show file tree

Hide file tree

Showing 2 changed files with 15 additions and 24 deletions.
diff --git a/crates/toml_edit/src/parser/strings.rs b/crates/toml_edit/src/parser/strings.rs
@@ -328,7 +328,7 @@ fn ml_literal_body<'i>(input: &mut Input<'i>) -> PResult<&'i str> {
 
 // mll-content = mll-char / newline
 fn mll_content(input: &mut Input<'_>) -> PResult<u8> {
-    alt((one_of(MLL_CHAR), newline)).parse_next(input)
+    alt((one_of(MLL_CHAR), newline.value(b'\n'))).parse_next(input)
 }
 
 // mll-char = %x09 / %x20-26 / %x28-7E / non-ascii

diff --git a/crates/toml_edit/src/parser/trivia.rs b/crates/toml_edit/src/parser/trivia.rs
@@ -50,61 +50,52 @@ pub(crate) const NON_EOL: (u8, RangeInclusive<u8>, RangeInclusive<u8>) =
 pub(crate) const COMMENT_START_SYMBOL: u8 = b'#';
 
 // comment = comment-start-symbol *non-eol
-pub(crate) fn comment<'i>(input: &mut Input<'i>) -> PResult<&'i [u8]> {
+pub(crate) fn comment<'i>(input: &mut Input<'i>) -> PResult<()> {
     (COMMENT_START_SYMBOL, take_while(0.., NON_EOL))
-        .recognize()
+        .void()
         .parse_next(input)
 }
 
 // newline = ( %x0A /              ; LF
 //             %x0D.0A )           ; CRLF
-pub(crate) fn newline(input: &mut Input<'_>) -> PResult<u8> {
-    alt((
-        one_of(LF).value(b'\n'),
-        (one_of(CR), one_of(LF)).value(b'\n'),
-    ))
-    .parse_next(input)
+pub(crate) fn newline(input: &mut Input<'_>) -> PResult<()> {
+    alt((one_of(LF).void(), (one_of(CR), one_of(LF)).void())).parse_next(input)
 }
 pub(crate) const LF: u8 = b'\n';
 pub(crate) const CR: u8 = b'\r';
 
 // ws-newline       = *( wschar / newline )
-pub(crate) fn ws_newline<'i>(input: &mut Input<'i>) -> PResult<&'i str> {
+pub(crate) fn ws_newline<'i>(input: &mut Input<'i>) -> PResult<()> {
     repeat(
         0..,
         alt((newline.value(&b"\n"[..]), take_while(1.., WSCHAR))),
     )
     .map(|()| ())
-    .recognize()
-    .map(|b| unsafe { from_utf8_unchecked(b, "`is_wschar` and `newline` filters out on-ASCII") })
     .parse_next(input)
 }
 
 // ws-newlines      = newline *( wschar / newline )
-pub(crate) fn ws_newlines<'i>(input: &mut Input<'i>) -> PResult<&'i str> {
-    (newline, ws_newline)
-        .recognize()
-        .map(|b| unsafe {
-            from_utf8_unchecked(b, "`is_wschar` and `newline` filters out on-ASCII")
-        })
-        .parse_next(input)
+pub(crate) fn ws_newlines<'i>(input: &mut Input<'i>) -> PResult<()> {
+    (newline, ws_newline).void().parse_next(input)
 }
 
 // note: this rule is not present in the original grammar
 // ws-comment-newline = *( ws-newline-nonempty / comment )
-pub(crate) fn ws_comment_newline<'i>(input: &mut Input<'i>) -> PResult<&'i [u8]> {
+pub(crate) fn ws_comment_newline<'i>(input: &mut Input<'i>) -> PResult<()> {
     (
         repeat(0.., (take_while(0.., WSCHAR), opt(comment), newline)).map(|()| ()),
         take_while(0.., WSCHAR),
     )
-        .recognize()
+        .void()
         .parse_next(input)
 }
 
 // note: this rule is not present in the original grammar
 // line-ending = newline / eof
-pub(crate) fn line_ending<'i>(input: &mut Input<'i>) -> PResult<&'i str> {
-    alt((newline.value("\n"), eof.value(""))).parse_next(input)
+pub(crate) fn line_ending<'i>(input: &mut Input<'i>) -> PResult<()> {
+    alt((newline.value("\n"), eof.value("")))
+        .void()
+        .parse_next(input)
 }
 
 // note: this rule is not present in the original grammar
@@ -143,7 +134,7 @@ mod test {
         ];
         for input in inputs {
             dbg!(input);
-            let parsed = ws_comment_newline.parse(new_input(input));
+            let parsed = ws_comment_newline.recognize().parse(new_input(input));
             assert!(parsed.is_ok(), "{:?}", parsed);
             let parsed = parsed.unwrap();
             assert_eq!(parsed, input.as_bytes());