diff options
| author | Martin Haug <mhaug@live.de> | 2021-11-28 18:18:45 +0100 |
|---|---|---|
| committer | Martin Haug <mhaug@live.de> | 2021-11-28 18:18:45 +0100 |
| commit | e05eb5fda5d1dfeef168b6fc071b20fdbcce2dcd (patch) | |
| tree | 6b4585c065ce21a76784cd0915d9b6e0a414f88d /src | |
| parent | edc686d7384470068858e16f2926cf50f31b2c90 (diff) | |
Code Review: Parser, I can't let you do this
Diffstat (limited to 'src')
| -rw-r--r-- | src/parse/incremental.rs | 449 | ||||
| -rw-r--r-- | src/parse/mod.rs | 10 | ||||
| -rw-r--r-- | src/parse/parser.rs | 14 | ||||
| -rw-r--r-- | src/parse/tokens.rs | 28 | ||||
| -rw-r--r-- | src/source.rs | 16 | ||||
| -rw-r--r-- | src/syntax/mod.rs | 50 |
6 files changed, 227 insertions, 340 deletions
diff --git a/src/parse/incremental.rs b/src/parse/incremental.rs index 9c912aae..8e52c143 100644 --- a/src/parse/incremental.rs +++ b/src/parse/incremental.rs @@ -1,13 +1,57 @@ use std::ops::Range; use std::rc::Rc; -use crate::syntax::{Green, GreenNode, NodeKind, Span}; +use crate::syntax::{Green, GreenNode, NodeKind}; use super::{ parse_atomic, parse_atomic_markup, parse_block, parse_comment, parse_markup, parse_markup_elements, parse_template, TokenMode, }; +/// The conditions that a node has to fulfill in order to be replaced. +/// +/// This can dictate if a node can be replaced at all and if yes, what can take +/// its place. +#[derive(Debug, Copy, Clone, Eq, PartialEq)] +pub enum Postcondition { + /// Changing this node can never have an influence on the other nodes. + Safe, + /// This node has to be replaced with a single token of the same kind. + SameKind(Option<TokenMode>), + /// Changing this node into a single atomic expression is allowed if it + /// appears in code mode, otherwise it is safe. + AtomicPrimary, + /// Changing an unsafe layer node changes what the parents or the + /// surrounding nodes would be and is therefore disallowed. Change the + /// parents or children instead. If it appears in Markup, however, it is + /// safe to change. + UnsafeLayer, + /// Changing an unsafe node or any of its children will trigger undefined + /// behavior. Change the parents instead. + Unsafe, +} + +/// The conditions under which a node can be inserted or remain in a tree. +/// +/// These conditions all search the neighbors of the node and see if its +/// existence is plausible with them present. This can be used to encode some +/// context-free language components for incremental parsing. +#[derive(Debug, Copy, Clone, Eq, PartialEq)] +pub enum Precondition { + /// These nodes depend on being at the start of a line. Reparsing of safe + /// left neighbors has to check this invariant. Otherwise, this node is + /// safe. + AtStart, + /// These nodes depend on not being at the start of a line. Reparsing of + /// safe left neighbors has to check this invariant. Otherwise, this node is + /// safe. + NotAtStart, + /// These nodes must be followed by whitespace. + RightWhitespace, + /// No additional requirements. + None, +} + /// Allows partial refreshs of the [`Green`] node tree. /// /// This struct holds a description of a change. Its methods can be used to try @@ -16,21 +60,21 @@ pub struct Reparser<'a> { /// The new source code, with the change applied. src: &'a str, /// Which range in the old source file was changed. - replace_range: Span, - /// How many characters replaced the text in `replacement_range`. + replace_range: Range<usize>, + /// How many characters replaced the text in `replace_range`. replace_len: usize, } impl<'a> Reparser<'a> { /// Create a new reparser. - pub fn new(src: &'a str, replace_range: Span, replace_len: usize) -> Self { + pub fn new(src: &'a str, replace_range: Range<usize>, replace_len: usize) -> Self { Self { src, replace_range, replace_len } } } impl Reparser<'_> { /// Find the innermost child that is incremental safe. - pub fn reparse(&self, green: &mut GreenNode) -> Result<Range<usize>, ()> { + pub fn reparse(&self, green: &mut GreenNode) -> Option<Range<usize>> { self.reparse_step(green, 0, TokenMode::Markup, true) } @@ -39,30 +83,26 @@ impl Reparser<'_> { green: &mut GreenNode, mut offset: usize, parent_mode: TokenMode, - outermost: bool, - ) -> Result<Range<usize>, ()> { + mut outermost: bool, + ) -> Option<Range<usize>> { let kind = green.kind().clone(); let mode = kind.mode().unwrap_or(parent_mode); - let mut loop_result = None; let mut child_at_start = true; - let last = green.children().len() - 1; + let last = green.children().len().saturating_sub(1); let mut start = None; for (i, child) in green.children_mut().iter_mut().enumerate() { - let child_span = - Span::new(self.replace_range.source, offset, offset + child.len()); + let child_span = offset .. offset + child.len(); // We look for the start in the element but we only take a position // at the right border if this is markup or the last element. // // This is because in Markup mode, we want to examine all nodes // touching a replacement but in code we want to atomically replace. - if child_span.contains(self.replace_range.start) - && (mode == TokenMode::Markup - || self.replace_range.start != child_span.end - || self.replace_range.len() == 0 - || i == last) + if child_span.contains(&self.replace_range.start) + || (mode == TokenMode::Markup + && self.replace_range.start == child_span.end) { start = Some((i, offset)); break; @@ -72,31 +112,21 @@ impl Reparser<'_> { child_at_start = child.kind().is_at_start(child_at_start); } - let (start_idx, start_offset) = start.ok_or(())?; + let (start_idx, start_offset) = start?; + let mut end = None; - for (i, child) in (green.children_mut()[start_idx ..]).iter_mut().enumerate() { - let i = i + start_idx; - let child_span = - Span::new(self.replace_range.source, offset, offset + child.len()); + for (i, child) in green.children_mut().iter_mut().enumerate().skip(start_idx) { + let child_span = offset .. offset + child.len(); // Similarly to above, the end of the edit must be in the node but // if it is at the edge and we are in markup node, we also want its // neighbor! - if child_span.contains(self.replace_range.end) - && (mode != TokenMode::Markup - || self.replace_range.end != child_span.end - || i == last) + if child_span.contains(&self.replace_range.end) + || self.replace_range.end == child_span.end + && (mode != TokenMode::Markup || i == last) { - loop_result = Some(( - start_idx .. i + 1, - Span::new( - self.replace_range.source, - start_offset, - offset + child.len(), - ), - i == last && outermost, - child.kind().clone(), - )); + outermost &= i == last; + end = Some(i); break; } else if mode != TokenMode::Markup || !child.kind().post().markup_safe() { break; @@ -105,28 +135,30 @@ impl Reparser<'_> { offset += child.len(); } - let (child_idx_range, child_span, child_outermost, child_kind) = - loop_result.ok_or(())?; + let end = end?; + let child_idx_range = start_idx .. end + 1; + let child_span = start_offset .. offset + green.children()[end].len(); + let child_kind = green.children()[end].kind().clone(); if child_idx_range.len() == 1 { let idx = child_idx_range.start; let child = &mut green.children_mut()[idx]; + let prev_len = child.len(); - let old_len = child.len(); // First, we try if the child has another, more specific applicable child. if !child_kind.post().unsafe_interior() { - if let Ok(range) = match child { + if let Some(range) = match child { Green::Node(n) => self.reparse_step( Rc::make_mut(n), start_offset, kind.mode().unwrap_or(TokenMode::Code), - child_outermost, + outermost, ), - Green::Token(_) => Err(()), + Green::Token(_) => None, } { let new_len = child.len(); - green.update_child_len(new_len, old_len); - return Ok(range); + green.update_child_len(new_len, prev_len); + return Some(range); } } } @@ -134,28 +166,31 @@ impl Reparser<'_> { debug_assert_ne!(child_idx_range.len(), 0); if mode == TokenMode::Code && child_idx_range.len() > 1 { - return Err(()); + return None; } // We now have a child that we can replace and a function to do so. - let (func, policy) = - child_kind.reparsing_function(kind.mode().unwrap_or(TokenMode::Code)); - let func = func?; + let func = + child_kind.reparsing_function(kind.mode().unwrap_or(TokenMode::Code))?; + let policy = child_kind.post(); + + let len_change = self.replace_len as isize - self.replace_range.len() as isize; + let mut src_span = child_span; + src_span.end = (src_span.end as isize + len_change) as usize; - let src_span = inserted_span(child_span, self.replace_range, self.replace_len); let recompile_range = if policy == Postcondition::AtomicPrimary { src_span.start .. self.src.len() } else { - src_span.to_range() + src_span.clone() }; - let (mut new_children, unterminated) = - func(&self.src[recompile_range], child_at_start).ok_or(())?; + let (mut new_children, terminated) = + func(&self.src[recompile_range], child_at_start)?; // Do not accept unclosed nodes if the old node did not use to be at the // right edge of the tree. - if !child_outermost && unterminated { - return Err(()); + if !outermost && !terminated { + return None; } let insertion = match check_invariants( @@ -164,17 +199,17 @@ impl Reparser<'_> { child_idx_range.clone(), child_at_start, mode, - src_span, + src_span.clone(), policy, ) { - InvariantResult::Ok => Ok(new_children), - InvariantResult::UseFirst => Ok(vec![std::mem::take(&mut new_children[0])]), - InvariantResult::Error => Err(()), + InvariantResult::Ok => Some(new_children), + InvariantResult::UseFirst => Some(vec![std::mem::take(&mut new_children[0])]), + InvariantResult::Error => None, }?; green.replace_child_range(child_idx_range, insertion); - Ok(src_span.to_range()) + Some(src_span) } } @@ -191,7 +226,7 @@ fn check_invariants( child_idx_range: Range<usize>, child_at_start: bool, mode: TokenMode, - src_span: Span, + src_span: Range<usize>, policy: Postcondition, ) -> InvariantResult { let (new_children, ok) = if policy == Postcondition::AtomicPrimary { @@ -206,10 +241,7 @@ fn check_invariants( (use_children, InvariantResult::Ok) }; - let child_mode = old_children[child_idx_range.start] - .kind() - .mode() - .unwrap_or(TokenMode::Code); + let child_mode = old_children[child_idx_range.start].kind().mode().unwrap_or(mode); // Check if the children / child has the right type. let same_kind = match policy { @@ -249,25 +281,22 @@ fn check_invariants( } } - let mut new_at_start = child_at_start; + let mut post_at_start = child_at_start; for child in new_children { - new_at_start = child.kind().is_at_start(new_at_start); + post_at_start = child.kind().is_at_start(post_at_start); } for child in &old_children[child_idx_range.end ..] { if child.kind().is_trivia() { - new_at_start = child.kind().is_at_start(new_at_start); + post_at_start = child.kind().is_at_start(post_at_start); continue; } - match child.kind().pre() { - Precondition::AtStart if !new_at_start => { - return InvariantResult::Error; - } - Precondition::NotAtStart if new_at_start => { - return InvariantResult::Error; - } - _ => {} + let pre = child.kind().pre(); + if pre == Precondition::AtStart && !post_at_start + || pre == Precondition::NotAtStart && post_at_start + { + return InvariantResult::Error; } break; } @@ -276,56 +305,31 @@ fn check_invariants( ok } -/// Create a new span by specifying a span in which a modification happened -/// and how many characters are now in that span. -fn inserted_span(mut source: Span, other: Span, n: usize) -> Span { - if !source.surrounds(other) { - panic!(); - } - - let len_change = n as i64 - other.len() as i64; - source.end = (source.end as i64 + len_change) as usize; - source -} - impl NodeKind { /// Return the correct reparsing function given the postconditions for the /// type. fn reparsing_function( &self, parent_mode: TokenMode, - ) -> ( - Result<fn(&str, bool) -> Option<(Vec<Green>, bool)>, ()>, - Postcondition, - ) { + ) -> Option<fn(&str, bool) -> Option<(Vec<Green>, bool)>> { let policy = self.post(); let mode = self.mode().unwrap_or(parent_mode); match policy { - Postcondition::Unsafe | Postcondition::UnsafeLayer => (Err(()), policy), - Postcondition::AtomicPrimary if mode == TokenMode::Code => { - (Ok(parse_atomic), policy) - } - Postcondition::AtomicPrimary => (Ok(parse_atomic_markup), policy), - Postcondition::SameKind(x) if x == None || x == Some(mode) => { - let parser: fn(&str, bool) -> _ = match self { - NodeKind::Template => parse_template, - NodeKind::Block => parse_block, - NodeKind::LineComment | NodeKind::BlockComment => parse_comment, - _ => return (Err(()), policy), - }; - - (Ok(parser), policy) - } - _ => { - let parser: fn(&str, bool) -> _ = match mode { - TokenMode::Markup if self == &Self::Markup => parse_markup, - TokenMode::Markup => parse_markup_elements, - _ => return (Err(()), policy), - }; - - (Ok(parser), policy) - } + Postcondition::Unsafe | Postcondition::UnsafeLayer => None, + Postcondition::AtomicPrimary if mode == TokenMode::Code => Some(parse_atomic), + Postcondition::AtomicPrimary => Some(parse_atomic_markup), + Postcondition::SameKind(x) if x == None || x == Some(mode) => match self { + NodeKind::Template => Some(parse_template), + NodeKind::Block => Some(parse_block), + NodeKind::LineComment | NodeKind::BlockComment => Some(parse_comment), + _ => None, + }, + _ => match mode { + TokenMode::Markup if self == &Self::Markup => Some(parse_markup), + TokenMode::Markup => Some(parse_markup_elements), + _ => return None, + }, } } @@ -369,10 +373,6 @@ impl NodeKind { | Self::Dots | Self::Arrow => Postcondition::Unsafe, - // These keywords are literals and can be safely be substituted with - // other expressions. - Self::None | Self::Auto => Postcondition::AtomicPrimary, - // These keywords change what kind of expression the parent is and // how far the expression would go. Self::Let @@ -389,45 +389,14 @@ impl NodeKind { | Self::Include | Self::From => Postcondition::Unsafe, - Self::Markup => Postcondition::SameKind(None), - - Self::Space(_) => Postcondition::SameKind(Some(TokenMode::Code)), - - // These are all replaceable by other tokens. - Self::Parbreak - | Self::Linebreak - | Self::Text(_) - | Self::TextInLine(_) - | Self::NonBreakingSpace - | Self::EnDash - | Self::EmDash - | Self::Escape(_) - | Self::Strong - | Self::Emph - | Self::Heading - | Self::Enum - | Self::List - | Self::Raw(_) - | Self::Math(_) => Postcondition::Safe, - // Changing the heading level, enum numbering, or list bullet // changes the next layer. Self::EnumNumbering(_) => Postcondition::Unsafe, - // These are expressions that can be replaced by other expressions. - Self::Ident(_) - | Self::Bool(_) - | Self::Int(_) - | Self::Float(_) - | Self::Length(_, _) - | Self::Angle(_, _) - | Self::Percentage(_) - | Self::Str(_) - | Self::Fraction(_) - | Self::Array - | Self::Dict - | Self::Group => Postcondition::AtomicPrimary, + Self::Error(_, _) | Self::Unknown(_) => Postcondition::Unsafe, + // These are complex expressions which may screw with their + // environments. Self::Call | Self::Unary | Self::Binary @@ -439,11 +408,44 @@ impl NodeKind { // is not atomic. Self::Closure | Self::ClosureParams => Postcondition::UnsafeLayer, + // Missing these creates errors for the parents. + Self::WithExpr | Self::ForPattern | Self::ImportItems => { + Postcondition::UnsafeLayer + } + + // Only markup is expected at the points where it does occur. + Self::Markup => Postcondition::SameKind(None), + + // These can appear everywhere and must not change to other stuff + // because that could change the outer expression. + Self::LineComment | Self::BlockComment => Postcondition::SameKind(None), + // These can appear as bodies and would trigger an error if they // became something else. Self::Template => Postcondition::SameKind(None), Self::Block => Postcondition::SameKind(Some(TokenMode::Code)), + // Whitespace in code mode has to remain whitespace or else the type + // of things would change. + Self::Space(_) => Postcondition::SameKind(Some(TokenMode::Code)), + + // These are expressions that can be replaced by other expressions. + Self::Ident(_) + | Self::Bool(_) + | Self::Int(_) + | Self::Float(_) + | Self::Length(_, _) + | Self::Angle(_, _) + | Self::Percentage(_) + | Self::Str(_) + | Self::Fraction(_) + | Self::Array + | Self::Dict + | Self::Group + | Self::None + | Self::Auto => Postcondition::AtomicPrimary, + + // More complex, but still an expression. Self::ForExpr | Self::WhileExpr | Self::IfExpr @@ -452,15 +454,22 @@ impl NodeKind { | Self::ImportExpr | Self::IncludeExpr => Postcondition::AtomicPrimary, - Self::WithExpr | Self::ForPattern | Self::ImportItems => { - Postcondition::UnsafeLayer - } - - // These can appear everywhere and must not change to other stuff - // because that could change the outer expression. - Self::LineComment | Self::BlockComment => Postcondition::SameKind(None), - - Self::Error(_, _) | Self::Unknown(_) => Postcondition::Unsafe, + // These are all replaceable by other tokens. + Self::Parbreak + | Self::Linebreak + | Self::Text(_) + | Self::TextInLine(_) + | Self::NonBreakingSpace + | Self::EnDash + | Self::EmDash + | Self::Escape(_) + | Self::Strong + | Self::Emph + | Self::Heading + | Self::Enum + | Self::List + | Self::Raw(_) + | Self::Math(_) => Postcondition::Safe, } } @@ -475,50 +484,6 @@ impl NodeKind { } } -/// The conditions that a node has to fulfill in order to be replaced. -/// -/// This can dictate if a node can be replaced at all and if yes, what can take -/// its place. -#[derive(Debug, Copy, Clone, Eq, PartialEq)] -pub enum Postcondition { - /// Changing this node can never have an influence on the other nodes. - Safe, - /// This node has to be replaced with a single token of the same kind. - SameKind(Option<TokenMode>), - /// Changing this node into a single atomic expression is allowed if it - /// appears in code mode, otherwise it is safe. - AtomicPrimary, - /// Changing an unsafe layer node changes what the parents or the - /// surrounding nodes would be and is therefore disallowed. Change the - /// parents or children instead. If it appears in Markup, however, it is - /// safe to change. - UnsafeLayer, - /// Changing an unsafe node or any of its children will trigger undefined - /// behavior. Change the parents instead. - Unsafe, -} - -/// The conditions under which a node can be inserted or remain in a tree. -/// -/// These conditions all search the neighbors of the node and see if its -/// existence is plausible with them present. This can be used to encode some -/// context-free language components for incremental parsing. -#[derive(Debug, Copy, Clone, Eq, PartialEq)] -pub enum Precondition { - /// These nodes depend on being at the start of a line. Reparsing of safe - /// left neighbors has to check this invariant. Otherwise, this node is - /// safe. - AtStart, - /// These nodes depend on not being at the start of a line. Reparsing of - /// safe left neighbors has to check this invariant. Otherwise, this node is - /// safe. - NotAtStart, - /// These nodes must be followed by whitespace. - RightWhitespace, - /// No additional requirements. - None, -} - impl Postcondition { pub fn unsafe_interior(&self) -> bool { match self { @@ -544,6 +509,7 @@ mod tests { use super::*; #[test] + #[rustfmt::skip] fn test_incremental_parse() { #[track_caller] fn test(prev: &str, range: Range<usize>, with: &str, incr: Range<usize>) { @@ -551,12 +517,14 @@ mod tests { let range = source.edit(range, with); assert_eq!(range, incr); - let incr_tree = source.root(); + let incr_tree = source.root().clone(); assert_eq!(parse(source.src()), incr_tree); } // Test simple replacements. - test("hello world", 6 .. 11, "wankers", 5 .. 13); + test("hello world", 6 .. 11, "walkers", 5 .. 13); + test("some content", 0..12, "", 0..0); + test("", 0..0, "do it", 0..5); test("a d e", 1 .. 3, " b c d", 0 .. 8); test("a #f() e", 1 .. 6, " b c d", 0 .. 8); test("{(0, 1, 2)}", 5 .. 6, "11pt", 5 .. 9); @@ -564,53 +532,18 @@ mod tests { test("your thing", 5 .. 5, "a", 4 .. 11); test("a your thing a", 6 .. 7, "a", 2 .. 12); test("{call(); abc}", 7 .. 7, "[]", 0 .. 15); - test("#call() abc", 7 .. 7, "[]", 0 .. 13); - // test( - // "hi\n- item\n- item 2\n - item 3", - // 10 .. 10, - // " ", - // 9 .. 33, - // ); - test( - "#grid(columns: (auto, 1fr, 40%), [*plonk*], rect(width: 100%, height: 1pt, fill: conifer), [thing])", - 16 .. 20, - "none", - 16 .. 20, - ); - test( - "#grid(columns: (auto, 1fr, 40%), [*plonk*], rect(width: 100%, height: 1pt, fill: conifer), [thing])", - 33 .. 42, - "[_gronk_]", - 33 .. 42, - ); - test( - "#grid(columns: (auto, 1fr, 40%), [*plonk*], rect(width: 100%, height: 1pt, fill: conifer), [thing])", - 34 .. 41, - "_bar_", - 34 .. 39, - ); + test("#call() abc", 7 .. 7, "[]", 0 .. 10); + // test("hi\n- item\n- item 2\n - item 3", 10 .. 10, " ", 9 .. 33); + test("#grid(columns: (auto, 1fr, 40%), [*plonk*], rect(width: 100%, height: 1pt, fill: conifer), [thing])", 16 .. 20, "none", 16 .. 20); + test("#grid(columns: (auto, 1fr, 40%), [*plonk*], rect(width: 100%, height: 1pt, fill: conifer), [thing])", 33 .. 42, "[_gronk_]", 33 .. 42); + test("#grid(columns: (auto, 1fr, 40%), [*plonk*], rect(width: 100%, height: 1pt, fill: conifer), [thing])", 34 .. 41, "_bar_", 34 .. 39); test("{let i=1; for x in range(5) {i}}", 6 .. 6, " ", 1 .. 9); - test("{let i=1; for x in range(5) {i}}", 13 .. 14, " ", 13 .. 15); + test("{let i=1; for x in range(5) {i}}", 13 .. 14, " ", 10 .. 32); test("hello {x}", 6 .. 9, "#f()", 5 .. 10); - test( - "this is -- in my opinion -- spectacular", - 8 .. 10, - "---", - 7 .. 12, - ); - test( - "understanding `code` is complicated", - 15 .. 15, - "C ", - 14 .. 22, - ); + test("this is -- in my opinion -- spectacular", 8 .. 10, "---", 7 .. 12); + test("understanding `code` is complicated", 15 .. 15, "C ", 14 .. 22); test("{ let x = g() }", 10 .. 12, "f(54", 0 .. 17); - test( - "a #let rect with (fill: eastern)\nb", - 16 .. 31, - " (stroke: conifer", - 2 .. 34, - ); + test("a #let rect with (fill: eastern)\nb", 16 .. 31, " (stroke: conifer", 2 .. 34); // Test the whitespace invariants. test("hello \\ world", 7 .. 8, "a ", 6 .. 14); @@ -642,18 +575,8 @@ mod tests { test(r"{{let x = z}; a = 1} b", 6 .. 6, "//", 0 .. 24); test("a b c", 1 .. 1, " /* letters */", 0 .. 16); test("a b c", 1 .. 1, " /* letters", 0 .. 16); - test( - "{if i==1 {a} else [b]; b()}", - 12 .. 12, - " /* letters */", - 1 .. 35, - ); - test( - "{if i==1 {a} else [b]; b()}", - 12 .. 12, - " /* letters", - 0 .. 38, - ); + test("{if i==1 {a} else [b]; b()}", 12 .. 12, " /* letters */", 1 .. 35); + test("{if i==1 {a} else [b]; b()}", 12 .. 12, " /* letters", 0 .. 38); test(r#"a ```typst hello``` b"#, 16 .. 17, "", 0 .. 20); test(r#"a ```typst hello```"#, 16 .. 17, "", 2 .. 18); diff --git a/src/parse/mod.rs b/src/parse/mod.rs index 2c421374..2c5afb6b 100644 --- a/src/parse/mod.rs +++ b/src/parse/mod.rs @@ -64,7 +64,7 @@ pub fn parse_markup_elements( /// Parse a template literal. Returns `Some` if all of the input was consumed. pub fn parse_template(source: &str, _: bool) -> Option<(Vec<Green>, bool)> { let mut p = Parser::new(source, TokenMode::Code); - if !matches!(p.peek(), Some(NodeKind::LeftBracket)) { + if !p.at(&NodeKind::LeftBracket) { return None; } @@ -75,7 +75,7 @@ pub fn parse_template(source: &str, _: bool) -> Option<(Vec<Green>, bool)> { /// Parse a code block. Returns `Some` if all of the input was consumed. pub fn parse_block(source: &str, _: bool) -> Option<(Vec<Green>, bool)> { let mut p = Parser::new(source, TokenMode::Code); - if !matches!(p.peek(), Some(NodeKind::LeftBrace)) { + if !p.at(&NodeKind::LeftBrace) { return None; } @@ -252,14 +252,14 @@ fn expr_prec(p: &mut Parser, atomic: bool, min_prec: usize) -> ParseResult { let marker = p.marker(); // Start the unary expression. - match (!atomic).then(|| p.peek().and_then(UnOp::from_token)).flatten() { - Some(op) => { + match p.peek().and_then(UnOp::from_token) { + Some(op) if !atomic => { p.eat(); let prec = op.precedence(); expr_prec(p, atomic, prec)?; marker.end(p, NodeKind::Unary); } - None => primary(p, atomic)?, + _ => primary(p, atomic)?, }; loop { diff --git a/src/parse/parser.rs b/src/parse/parser.rs index 06cb1578..ade9b5df 100644 --- a/src/parse/parser.rs +++ b/src/parse/parser.rs @@ -22,7 +22,7 @@ pub struct Parser<'s> { /// The children of the currently built node. children: Vec<Green>, /// Whether the last group was terminated. - last_group_terminated: bool, + last_terminated: bool, } impl<'s> Parser<'s> { @@ -38,7 +38,7 @@ impl<'s> Parser<'s> { current_start: 0, groups: vec![], children: vec![], - last_group_terminated: true, + last_terminated: true, } } @@ -50,7 +50,7 @@ impl<'s> Parser<'s> { /// End the parsing process and return multiple children. pub fn eject(self) -> Option<(Vec<Green>, bool)> { if self.eof() && self.group_success() { - Some((self.children, self.tokens.was_unterminated())) + Some((self.children, self.tokens.was_terminated())) } else { None } @@ -99,7 +99,7 @@ impl<'s> Parser<'s> { /// remains stuff in the string. pub fn eject_partial(self) -> Option<(Vec<Green>, bool)> { self.group_success() - .then(|| (self.children, self.tokens.was_unterminated())) + .then(|| (self.children, self.tokens.was_terminated())) } /// Whether the end of the source string or group is reached. @@ -244,7 +244,7 @@ impl<'s> Parser<'s> { let group = self.groups.pop().expect("no started group"); self.tokens.set_mode(group.prev_mode); self.repeek(); - self.last_group_terminated = true; + self.last_terminated = true; let mut rescan = self.tokens.mode() != group_mode; @@ -263,7 +263,7 @@ impl<'s> Parser<'s> { rescan = false; } else if required { self.push_error(format_eco!("expected {}", end)); - self.last_group_terminated = false; + self.last_terminated = false; } } @@ -283,7 +283,7 @@ impl<'s> Parser<'s> { /// Check if the group processing was successfully terminated. pub fn group_success(&self) -> bool { - self.last_group_terminated && self.groups.is_empty() + self.last_terminated && self.groups.is_empty() } /// Low-level bump that consumes exactly one token without special trivia diff --git a/src/parse/tokens.rs b/src/parse/tokens.rs index 7be31fe1..836e8cf1 100644 --- a/src/parse/tokens.rs +++ b/src/parse/tokens.rs @@ -13,7 +13,7 @@ use crate::util::EcoString; pub struct Tokens<'s> { s: Scanner<'s>, mode: TokenMode, - has_unterminated: bool, + was_terminated: bool, } /// What kind of tokens to emit. @@ -32,7 +32,7 @@ impl<'s> Tokens<'s> { Self { s: Scanner::new(src), mode, - has_unterminated: false, + was_terminated: true, } } @@ -69,10 +69,10 @@ impl<'s> Tokens<'s> { self.s } - /// Whether the last token was unterminated. + /// Whether the last token was terminated. #[inline] - pub fn was_unterminated(&self) -> bool { - self.has_unterminated + pub fn was_terminated(&self) -> bool { + self.was_terminated } } @@ -259,7 +259,7 @@ impl<'s> Tokens<'s> { ) } } else { - self.has_unterminated = true; + self.was_terminated = false; NodeKind::Error( ErrorPos::End, "expected closing brace".into(), @@ -358,7 +358,7 @@ impl<'s> Tokens<'s> { let remaining = backticks - found; let noun = if remaining == 1 { "backtick" } else { "backticks" }; - self.has_unterminated = true; + self.was_terminated = false; NodeKind::Error( ErrorPos::End, if found == 0 { @@ -406,7 +406,7 @@ impl<'s> Tokens<'s> { display, })) } else { - self.has_unterminated = true; + self.was_terminated = false; NodeKind::Error( ErrorPos::End, if !display || (!escaped && dollar) { @@ -495,7 +495,7 @@ impl<'s> Tokens<'s> { if self.s.eat_if('"') { NodeKind::Str(string) } else { - self.has_unterminated = true; + self.was_terminated = false; NodeKind::Error(ErrorPos::End, "expected quote".into()) } } @@ -503,7 +503,7 @@ impl<'s> Tokens<'s> { fn line_comment(&mut self) -> NodeKind { self.s.eat_until(is_newline); if self.s.peek().is_none() { - self.has_unterminated = true; + self.was_terminated = false; } NodeKind::LineComment } @@ -511,7 +511,7 @@ impl<'s> Tokens<'s> { fn block_comment(&mut self) -> NodeKind { let mut state = '_'; let mut depth = 1; - let mut terminated = false; + self.was_terminated = false; // Find the first `*/` that does not correspond to a nested `/*`. while let Some(c) = self.s.eat() { @@ -519,7 +519,7 @@ impl<'s> Tokens<'s> { ('*', '/') => { depth -= 1; if depth == 0 { - terminated = true; + self.was_terminated = true; break; } '_' @@ -532,10 +532,6 @@ impl<'s> Tokens<'s> { } } - if !terminated { - self.has_unterminated = true; - } - NodeKind::BlockComment } diff --git a/src/source.rs b/src/source.rs index 421412ee..6cca9f75 100644 --- a/src/source.rs +++ b/src/source.rs @@ -12,7 +12,7 @@ use crate::diag::TypResult; use crate::loading::{FileHash, Loader}; use crate::parse::{is_newline, parse, Reparser, Scanner}; use crate::syntax::ast::Markup; -use crate::syntax::{self, Category, GreenNode, RedNode, Span}; +use crate::syntax::{self, Category, GreenNode, RedNode}; use crate::util::PathExt; #[cfg(feature = "codespan-reporting")] @@ -265,7 +265,8 @@ impl SourceFile { /// Edit the source file by replacing the given range. /// - /// This panics if the `replace` range is out of bounds. + /// Returns the range of the section in the new source that was ultimately + /// reparsed. The method panics if the `replace` range is out of bounds. pub fn edit(&mut self, replace: Range<usize>, with: &str) -> Range<usize> { let start = replace.start; self.src.replace_range(replace.clone(), with); @@ -284,9 +285,8 @@ impl SourceFile { .extend(newlines(&self.src[start ..]).map(|idx| start + idx)); // Update the root node. - let span = Span::new(self.id, replace.start, replace.end); - let reparser = Reparser::new(&self.src, span, with.len()); - if let Ok(range) = reparser.reparse(Rc::make_mut(&mut self.root)) { + let reparser = Reparser::new(&self.src, replace, with.len()); + if let Some(range) = reparser.reparse(Rc::make_mut(&mut self.root)) { range } else { self.root = parse(&self.src); @@ -302,12 +302,6 @@ impl SourceFile { let red = RedNode::from_root(self.root.clone(), self.id); syntax::highlight(red.as_ref(), range, &mut f) } - - /// Obtain a reference to the source's root green node. - #[cfg(test)] - pub(crate) fn root(&self) -> Rc<GreenNode> { - self.root.clone() - } } /// The indices at which lines start (right behind newlines). diff --git a/src/syntax/mod.rs b/src/syntax/mod.rs index 9ab530d8..b72e5843 100644 --- a/src/syntax/mod.rs +++ b/src/syntax/mod.rs @@ -48,15 +48,6 @@ impl Green { self.data().len() } - /// Set the length of the node. - pub fn set_len(&mut self, len: usize) { - let data = match self { - Self::Node(node) => &mut Rc::make_mut(node).data, - Self::Token(data) => data, - }; - data.set_len(len); - } - /// Whether the node or its children contain an error. pub fn erroneous(&self) -> bool { match self { @@ -139,11 +130,6 @@ impl GreenNode { &self.children } - /// The node's children, mutably. - pub fn children_mut(&mut self) -> &mut [Green] { - &mut self.children - } - /// The node's metadata. pub fn data(&self) -> &GreenData { &self.data @@ -159,10 +145,15 @@ impl GreenNode { self.data().len() } + /// The node's children, mutably. + pub(crate) fn children_mut(&mut self) -> &mut [Green] { + &mut self.children + } + /// Replaces a range of children with some replacement. /// /// This method updates the `erroneous` and `data.len` fields. - pub fn replace_child_range( + pub(crate) fn replace_child_range( &mut self, child_idx_range: Range<usize>, replacement: Vec<Green>, @@ -187,12 +178,12 @@ impl GreenNode { self.erroneous = self.erroneous || replacement.iter().any(Green::erroneous); self.children.splice(child_idx_range, replacement); - self.data.set_len(self.data.len + new_len - old_len); + self.data.len = self.data.len + new_len - old_len; } /// Update the length of this node given the old and new length of a /// replaced child. - pub fn update_child_len(&mut self, new_len: usize, old_len: usize) { + pub(crate) fn update_child_len(&mut self, new_len: usize, old_len: usize) { self.data.len = self.data.len() + new_len - old_len; self.erroneous = self.children.iter().any(|x| x.erroneous()); } @@ -246,11 +237,6 @@ impl GreenData { pub fn len(&self) -> usize { self.len } - - /// Set the length of the node. - pub fn set_len(&mut self, len: usize) { - self.len = len; - } } impl From<GreenData> for Green { @@ -261,7 +247,7 @@ impl From<GreenData> for Green { impl Debug for GreenData { fn fmt(&self, f: &mut Formatter) -> fmt::Result { - write!(f, "{:?}: {}", self.kind, self.len) + write!(f, "{:?}: {}", &self.kind, self.len) } } @@ -375,11 +361,6 @@ impl<'a> RedRef<'a> { Span::new(self.id, self.offset, self.offset + self.green.len()) } - /// Whether the node or its children contain an error. - pub fn erroneous(self) -> bool { - self.green.erroneous() - } - /// The error messages for this node and its descendants. pub fn errors(self) -> Vec<Error> { if !self.green.erroneous() { @@ -731,19 +712,12 @@ impl NodeKind { /// Whether this is whitespace. pub fn is_whitespace(&self) -> bool { - match self { - Self::Space(_) | Self::Parbreak => true, - _ => false, - } + matches!(self, Self::Space(_) | Self::Parbreak) } /// Whether this is trivia. pub fn is_trivia(&self) -> bool { - match self { - _ if self.is_whitespace() => true, - Self::LineComment | Self::BlockComment => true, - _ => false, - } + self.is_whitespace() || matches!(self, Self::LineComment | Self::BlockComment) } /// Whether this is some kind of error. @@ -765,7 +739,6 @@ impl NodeKind { pub fn mode(&self) -> Option<TokenMode> { match self { Self::Markup - | Self::Space(_) | Self::Linebreak | Self::Parbreak | Self::Text(_) @@ -783,6 +756,7 @@ impl NodeKind { | Self::Raw(_) | Self::Math(_) => Some(TokenMode::Markup), Self::Template + | Self::Space(_) | Self::Block | Self::Ident(_) | Self::LetExpr |
