Parse doc comments

This commit is contained in:
Richard Feldman 2020-05-23 21:42:30 -04:00
parent 4d8cbc4811
commit 3dac73d8d0
5 changed files with 138 additions and 67 deletions

View file

@ -294,6 +294,7 @@ pub enum AssignedField<'a, Val> {
pub enum CommentOrNewline<'a> {
Newline,
LineComment(&'a str),
DocComment(&'a str),
}
#[derive(Clone, Debug, PartialEq)]

View file

@ -202,6 +202,13 @@ pub fn space1<'a>(min_indent: u16) -> impl Parser<'a, &'a [CommentOrNewline<'a>]
spaces(true, min_indent)
}
#[derive(Debug, Clone, Copy)]
enum LineState {
Normal,
Comment,
DocComment,
}
#[inline(always)]
fn spaces<'a>(
require_at_least_one: bool,
@ -213,85 +220,129 @@ fn spaces<'a>(
let mut space_list = Vec::new_in(arena);
let mut chars_parsed = 0;
let mut comment_line_buf = String::new_in(arena);
let mut is_parsing_comment = false;
let mut line_state = LineState::Normal;
let mut state = state;
let mut any_newlines = false;
for ch in chars {
chars_parsed += 1;
if is_parsing_comment {
match ch {
' ' => {
// If we're in a line comment, this won't affect indentation anyway.
state = state.advance_without_indenting(1)?;
match line_state {
LineState::Normal => {
match ch {
' ' => {
// Don't check indentation here; it might not be enough
// indentation yet, but maybe it will be after more spaces happen!
state = state.advance_spaces(1)?;
}
'\n' => {
// No need to check indentation because we're about to reset it anyway.
state = state.newline()?;
comment_line_buf.push(ch);
}
'\n' => {
state = state.newline()?;
// Newlines only get added to the list when they're outside comments.
space_list.push(Newline);
// This was a newline, so end this line comment.
space_list.push(LineComment(comment_line_buf.into_bump_str()));
comment_line_buf = String::new_in(arena);
any_newlines = true;
}
'#' => {
// Check indentation to make sure we were indented enough
// before this comment began.
state = state
.check_indent(min_indent)
.map_err(|(fail, _)| (fail, original_state.clone()))?
.advance_without_indenting(1)?;
is_parsing_comment = false;
}
nonblank => {
// Chars can have btye lengths of more than 1!
state = state.advance_without_indenting(nonblank.len_utf8())?;
// We're now parsing a line comment!
line_state = LineState::Comment;
}
nonblank => {
return if require_at_least_one && chars_parsed <= 1 {
// We've parsed 1 char and it was not a space,
// but we require parsing at least one space!
Err(unexpected(nonblank, 0, state.clone(), state.attempting))
} else {
// First make sure we were indented enough!
//
// (We only do this if we've encountered any newlines.
// Otherwise, we assume indentation is already correct.
// It's actively important for correctness that we skip
// this check if there are no newlines, because otherwise
// we would have false positives for single-line defs.)
if any_newlines {
state = state
.check_indent(min_indent)
.map_err(|(fail, _)| (fail, original_state))?;
}
comment_line_buf.push(nonblank);
Ok((space_list.into_bump_slice(), state))
};
}
}
}
} else {
match ch {
' ' => {
// Don't check indentation here; it might not be enough
// indentation yet, but maybe it will be after more spaces happen!
state = state.advance_spaces(1)?;
}
'\n' => {
// No need to check indentation because we're about to reset it anyway.
state = state.newline()?;
LineState::Comment => {
match ch {
' ' => {
// If we're in a line comment, this won't affect indentation anyway.
state = state.advance_without_indenting(1)?;
// Newlines only get added to the list when they're outside comments.
space_list.push(Newline);
match comment_line_buf.chars().next() {
Some('#') => {
// This is a comment begining with `## ` - that is,
// a doc comment.
//
// (The space is important; otherwise, this is not
// a doc comment, but rather something like a
// big separator block, e.g. ############)
line_state = LineState::DocComment;
any_newlines = true;
}
'#' => {
// Check indentation to make sure we were indented enough
// before this comment began.
state = state
.check_indent(min_indent)
.map_err(|(fail, _)| (fail, original_state.clone()))?
.advance_without_indenting(1)?;
// We're now parsing a line comment!
is_parsing_comment = true;
}
nonblank => {
return if require_at_least_one && chars_parsed <= 1 {
// We've parsed 1 char and it was not a space,
// but we require parsing at least one space!
Err(unexpected(nonblank, 0, state.clone(), state.attempting))
} else {
// First make sure we were indented enough!
//
// (We only do this if we've encountered any newlines.
// Otherwise, we assume indentation is already correct.
// It's actively important for correctness that we skip
// this check if there are no newlines, because otherwise
// we would have false positives for single-line defs.)
if any_newlines {
state = state
.check_indent(min_indent)
.map_err(|(fail, _)| (fail, original_state))?;
// This is now the beginning of the doc comment.
comment_line_buf.clear();
}
_ => {
comment_line_buf.push(ch);
}
}
}
'\n' => {
state = state.newline()?;
Ok((space_list.into_bump_slice(), state))
};
// This was a newline, so end this line comment.
space_list.push(LineComment(comment_line_buf.into_bump_str()));
comment_line_buf = String::new_in(arena);
line_state = LineState::Normal;
}
nonblank => {
// Chars can have btye lengths of more than 1!
state = state.advance_without_indenting(nonblank.len_utf8())?;
comment_line_buf.push(nonblank);
}
}
}
LineState::DocComment => {
match ch {
' ' => {
// If we're in a doc comment, this won't affect indentation anyway.
state = state.advance_without_indenting(1)?;
comment_line_buf.push(ch);
}
'\n' => {
state = state.newline()?;
// This was a newline, so end this doc comment.
space_list.push(DocComment(comment_line_buf.into_bump_str()));
comment_line_buf = String::new_in(arena);
line_state = LineState::Normal;
}
nonblank => {
// Chars can have btye lengths of more than 1!
state = state.advance_without_indenting(nonblank.len_utf8())?;
comment_line_buf.push(nonblank);
}
}
}
}