CSY/reowolf Changeset - a52ba09256ad · Centrum Wiskunde & Informatica (CWI)

Changeset - a52ba09256ad

Parent rev.

Child rev.

[Not reviewed]

0 2 0

MH - 4 years ago 2021-04-23 10:26:53
henger@cwi.nl

WIP on compiler rearchitecting

2 files changed with 102 insertions and 8 deletions:

src/protocol/parser/pass_definitions.rs

src/protocol/parser/token_parsing.rs

0 comments (0 inline, 0 general)

src/protocol/parser/pass_definitions.rs

➞

Show inline comments

@@ @@ -620,195 +620,212 @@ impl PassDefinitions { @@
             } else if token == TokenKind::OpenSquare {
                 let subject = result;
                 let from_index = self.consume_expression(module, iter, ctx)?;
                 // Check if we have an indexing or slicing operation
                 next = iter.next();
                 if Some(TokenKind::DotDot) = next {
                     iter.consume();
                     let to_index = self.consume_expression(module, iter, ctx)?;
                     let end_span = consume_token(&module.source, iter, TokenKind::CloseSquare)?;
                     span.end = end_span.end;
                     result = ctx.heap.alloc_slicing_expression(|this| SlicingExpression{
                         this, span, subject, from_index, to_index,
                         parent: ExpressionParent::None,
                         concrete_type: ConcreteType::default()
                     }).upcast();
                 } else if Some(TokenKind::CloseSquare) {
                     let end_span = consume_token(&module.source, iter, TokenKind::CloseSquare)?;
                     span.end = end_span.end;
                     result = ctx.heap.alloc_indexing_expression(|this| IndexingExpression{
                         this, span, subject,
                         index: from_index,
                         parent: ExpressionParent::None,
                         concrete_type: ConcreteType::default()
                     }).upcast();
                 } else {
                     return Err(ParseError::new_error_str_at_pos(
                         &module.source, iter.last_valid_pos(), "unexpected token: expected ']' or '..'"
                     ));
+                }
             } else {
                 debug_assert_eq!(token, TokenKind::Dot);
                 let subject = result;
                 let (field_text, field_span) = consume_ident(&module.source, iter)?;
                 let field = if field_text == b"length" {
                     Field::Length
                 } else {
                     let value = ctx.pool.intern(field_text);
                     let identifier = Identifier{ value, span: field_span };
                     Field::Symbolic(FieldSymbolic{ identifier, definition: None, field_idx: 0 });
                 };
                 result = ctx.heap.alloc_select_expression(|this| SelectExpression{
                     this, span, subject, field,
                     parent: ExpressionParent::None,
                     concrete_type: ConcreteType::default()
                 }).upcast();
+            }
             next = iter.next();
+        }
         Ok(result)
+    }
     fn consume_primary_expression(
         &mut self, module: &Module, iter: &mut TokenIter, ctx: &mut PassCtx
     ) -> Result<ExpressionId, ParseError> {
         let next = iter.next();
         let result;
         if next == Some(TokenKind::OpenParen) {
             // Expression between parentheses
             iter.consume();
             result = self.consume_expression(module, iter, ctx)?;
             consume_token(&module.source, iter, TokenKind::CloseParen)?;
         } else if next == Some(TokenKind::OpenCurly) {
             // Array literal
             let (start_pos, mut end_pos) = iter.next_positions();
             let mut expressions = Vec::new();
             consume_comma_separated(
                 TokenKind::OpenCurly, TokenKind::CloseCurly, &module.source, iter,
                 |source, iter| self.consume_expression(module, iter, ctx),
                 &mut expressions, "an expression", "a list of expressions", Some(&mut end_pos)
             )?;
             // TODO: Turn into literal
             result = ctx.heap.alloc_array_expression(|this| ArrayExpression{
                 this,
                 span: InputSpan::from_positions(start_pos, end_pos),
                 elements: expressions,
                 parent: ExpressionParent::None,
                 concrete_type: ConcreteType::default(),
             }).upcast();
         } else if next == Some(TokenKind::Integer) {
             let (literal, span) = consume_integer_literal(&module.source, iter, &mut self.buffer)?;
             result = ctx.heap.alloc_literal_expression(|this| LiteralExpression{
                 this, span,
                 value: Literal::Integer(LiteralInteger{ unsigned_value: literal, negated: false }),
                 parent: ExpressionParent::None,
                 concrete_type: ConcreteType::default(),
             }).upcast();
         } else if next == Some(TokenKind::String) {
             let (text, span) = consume_string_literal(&module.source, iter, &mut self.buffer)?;
             let span = consume_string_literal(&module.source, iter, &mut self.buffer)?;
             let interned = ctx.pool.intern(self.buffer.as_bytes());
             result = ctx.heap.alloc_literal_expression(|this| LiteralExpression{
                 this, span,
                 value: Literal::String(interned),
                 parent: ExpressionParent::None,
                 concrete_type: ConcreteType::default(),
             }).upcast();
         } else if next == Some(TokenKind::Character) {
             let (character, span) = consume_character_literal(&module.source, iter)?;
             result = ctx.heap.alloc_literal_expression(|this| LiteralExpression{
                 this, span,
                 value: Literal::Character(character),
                 parent: ExpressionParent::None,
                 concrete_type: ConcreteType::default(),
             }).upcast();
         } else if next == Some(TokenKind::Ident) {
             // May be a variable, a type instantiation or a function call. If we
             // have a single identifier that we cannot find in the type table
             // then we're going to assume that we're dealign with a variable.
+        }
         Ok(result)
+    }
     //--------------------------------------------------------------------------
     // Expression Utilities
     //--------------------------------------------------------------------------
     #[inline]
     fn consume_generic_binary_expression<
         M: Fn(Option<TokenKind>) -> Option<BinaryOperator>,
         F: Fn(&mut PassDefinitions, &Module, &mut TokenIter, &mut PassCtx) -> Result<ExpressionId, ParseError>
     >(
         &mut self, module: &Module, iter: &mut TokenIter, ctx: &mut PassCtx, match_fn: M, higher_precedence_fn: F
     ) -> Result<ExpressionId, ParseError> {
         let mut result = higher_precedence_fn(self, module, iter, ctx)?;
         while let Some(operation) = match_fn(iter.next()) {
             let span = iter.next_span();
             iter.consume();
             let left = result;
             let right = higher_precedence_fn(self, module, iter, ctx)?;
             result = ctx.heap.alloc_binary_expression(|this| BinaryExpression{
                 this, span, left, operation, right,
                 parent: ExpressionParent::None,
                 concrete_type: ConcreteType::default()
             }).upcast();
+        }
         Ok(result)
+    }
+}
 /// Consumes a type. A type always starts with an identifier which may indicate
 /// a builtin type or a user-defined type. The fact that it may contain
 /// polymorphic arguments makes it a tree-like structure. Because we cannot rely
 /// on knowing the exact number of polymorphic arguments we do not check for
 /// these.
 // TODO: @Optimize, and fix spans if needed
 fn consume_parser_type(
     source: &InputSource, iter: &mut TokenIter, symbols: &SymbolTable, heap: &Heap, poly_vars: &[Identifier],
     cur_scope: SymbolScope, wrapping_definition: DefinitionId, allow_inference: bool
 ) -> Result<ParserType, ParseError> {
     struct Entry{
         element: ParserTypeElement,
         depth: i32,
+    }
     fn insert_array_before(elements: &mut Vec<Entry>, depth: i32, span: InputSpan) {
         let index = elements.iter().rposition(|e| e.depth == depth).unwrap();
         elements.insert(index, Entry{
             element: ParserTypeElement{ full_span: span, variant: ParserTypeVariant::Array },
             depth,
         });
+    }
     // Most common case we just have one type, perhaps with some array
     // annotations.
     let element = consume_parser_type_ident(source, iter, symbols, heap, poly_vars, cur_scope, wrapping_definition, allow_inference)?;
     if iter.next() != Some(TokenKind::OpenAngle) {
         let mut num_array = 0;
         while iter.next() == Some(TokenKind::OpenSquare) {
             iter.consume();
             consume_token(source, iter, TokenKind::CloseSquare)?;
             num_array += 1;
+        }
         let array_span = element.full_span;
         let mut elements = Vec::with_capacity(num_array + 1);
         for _ in 0..num_array {
             elements.push(ParserTypeElement{ full_span: array_span, variant: ParserTypeVariant::Array });
+        }
         elements.push(element);
         return Ok(ParserType{ elements });
     };
     // We have a polymorphic specification. So we start by pushing the item onto
     // our stack, then start adding entries together with the angle-brace depth
     // at which they're found.
     let mut elements = Vec::new();
     elements.push(Entry{ element, depth: 0 });
     // Start out with the first '<' consumed.
     iter.consume();
     enum State { Ident, Open, Close, Comma };
     let mut state = State::Open;
     let mut angle_depth = 1;
     loop {
         let next = iter.next();
         match state {
             State::Ident => {

src/protocol/parser/token_parsing.rs

➞

Show inline comments

             return Err(ParseError::new_error_at_pos(
                 source, iter.last_valid_pos(),
                 format!("expected a '{}', or {}", close_delim.token_chars(), item_name_and_article)
             ));
+        }
         consumer_fn(source, iter)?;
         next = iter.next();
         had_comma = next == Some(TokenKind::Comma);
         if had_comma {
             iter.consume();
+        }
+    }
     Ok(true)
+}
 /// Consumes a comma-separated list and expected the opening and closing
 /// characters to be present. The returned array may still be empty
 pub(crate) fn consume_comma_separated<T, F>(
     open_delim: TokenKind, close_delim: TokenKind, source: &InputSource, iter: &mut TokenIter,
     consumer_fn: F, target: &mut Vec<T>, item_name_and_article: &'static str,
     list_name_and_article: &'static str, close_pos: Option<&mut InputPosition>
 ) -> Result<(), ParseError>
     where F: Fn(&InputSource, &mut TokenIter) -> Result<T, ParseError>
+{
     let first_pos = iter.last_valid_pos();
     match maybe_consume_comma_separated(
         open_delim, close_delim, source, iter, consumer_fn, target,
         item_name_and_article, close_pos
     ) {
         Ok(true) => Ok(()),
         Ok(false) => {
             return Err(ParseError::new_error_at_pos(
                 source, first_pos,
                 format!("expected {}", list_name_and_article)
             ));
         },
         Err(err) => Err(err)
+    }
+}
 /// Consumes an integer literal, may be binary, octal, hexadecimal or decimal,
 /// and may have separating '_'-characters.
 pub(crate) fn consume_integer_literal(source: &InputSource, iter: &mut TokenIter, buffer: &mut String) -> Result<(u64, InputSpan), ParseError> {
     if Some(TokenKind::Integer) != iter.next() {
         return Err(ParseError::new_error_str_at_pos(source, iter.last_valid_pos(), "expected an integer literal"));
+    }
     let integer_span = iter.next_span();
     iter.consume();
     let integer_text = source.section_at_span(integer_span);
     // Determine radix and offset from prefix
     let (radix, input_offset, radix_name) =
         if integer_text.starts_with(b"0b") || integer_text.starts_with(b"0B") {
             // Binary number
             (2, 2, "binary")
         } else if integer_text.starts_with(b"0o") || integer_text.starts_with(b"0O") {
             // Octal number
             (8, 2, "octal")
         } else if integer_text.starts_with(b"0x") || integer_text.starts_with(b"0X") {
             // Hexadecimal number
             (16, 2, "hexadecimal")
         } else {
             (10, 0, "decimal")
         };
     // Take out any of the separating '_' characters
     buffer.clear();
     for char_idx in input_offset..integer_text.len() {
         let char = integer_text[char_idx];
         if char == b'_' {
             continue;
+        }
         if !char.is_ascii_digit() {
             return Err(ParseError::new_error_at_span(
                 source, integer_span,
                 format!("incorrectly formatted {} number", radix_name)
             ));
+        }
         buffer.push(char::from(char));
+    }
     // Use the cleaned up string to convert to integer
     match u64::from_str_radix(&buffer, radix) {
         Ok(number) => Ok((number, integer_span)),
         Err(_) => Err(ParseError::new_error_at_span(
             source, integer_span,
             format!("incorrectly formatted {} number", radix_name)
         )),
+    }
+}
 /// Consumes a character literal. We currently support a limited number of
 /// backslash-escaped characters
 pub(crate) fn consume_character_literal(source: &InputSource, iter: &mut TokenIter, buffer: &mut String) -> Result<char, ParseError> {
 pub(crate) fn consume_character_literal(
     source: &InputSource, iter: &mut TokenIter
 ) -> Result<(char, InputSpan), ParseError> {
     if Some(TokenKind::Character) != iter.next() {
         return Err(ParseError::new_error_str_at_pos(source, iter.last_valid_pos(), "expected a character literal"));
+    }
-    let char_span = iter.next_span();
+    let span = iter.next_span();
     iter.consume();
     let char_text = source.section_at_span(char_span);
     let char_text = source.section_at_span(span);
     if !char_text.is_ascii() {
         return Err(ParseError::new_error_str_at_span(
             source, span, "expected an ASCII character literal"
         ));
+    }
     match char_text.len() {
 => return Err(ParseError::new_error_str_at_span(source, span, "too little characters in character literal")),
 => {
             // We already know the text is ascii, so just throw an error if we have the escape
             // character.
             if char_text[0] == b'\\' {
                 return Err(ParseError::new_error_str_at_span(source, span, "escape character without subsequent character"));
+            }
             return Ok((char_text[0] as char, span));
         },
 => {
             if char_text[0] == b'\\' {
                 let result = parse_escaped_character(char_text[1])?;
                 return Ok((result, span))
+            }
         },
         _ => {}
+    }
-    //
+    return Err(ParseError::new_error_str_at_span(source, span, "too many characters in character literal"))
+}
 /// Consumes a string literal. We currently support a limited number of
 /// backslash-escaped characters.
 pub(crate) fn consume_string_literal(source: &InputSource, iter: &mut TokenIter, buffer: &mut String) -> Result<(>
 /// backslash-escaped characters. Note that the result is stored in the
 /// buffer.
 pub(crate) fn consume_string_literal(
     source: &InputSource, iter: &mut TokenIter, buffer: &mut String
 ) -> Result<InputSpan, ParseError> {
     if Some(TokenKind::String) != iter.next() {
         return Err(ParseError::new_error_str_at_pos(source, iter.last_valid_pos(), "expected a string literal"));
+    }
     buffer.clear();
     let span = iter.next_span();
     iter.consume();
     let text = source.section_at_span(span);
     if !text.is_ascii() {
         return Err(ParseError::new_error_str_at_span(source, span, "expected an ASCII string literal"));
+    }
     buffer.reserve(text.len());
     let mut was_escape = false;
     for idx in 0..text.len() {
         let cur = text[idx];
         if cur != b'\\' {
             if was_escape {
                 let to_push = parse_escaped_character(cur)?;
                 buffer.push(to_push);
             } else {
                 buffer.push(cur as char);
+            }
             was_escape = false;
         } else {
             was_escape = true;
+        }
+    }
     Ok(span)
+}
 fn parse_escaped_character(v: u8) -> Result<char, ParseError> {
     let result = match v {
         b'r' => '\r',
         b'n' => '\n',
         b't' => '\t',
         b'0' => '\0',
         b'\\' => '\\',
         b'\'' => '\'',
         b'"' => '"',
         v => return Err(ParseError::new_error_at_span(
             source, span, format!("unexpected escaped character '{}'", v)
         )),
     };
     Ok(result)
+}
 pub(crate) fn consume_pragma<'a>(source: &'a InputSource, iter: &mut TokenIter) -> Result<(&'a [u8], InputPosition, InputPosition), ParseError> {
     if Some(TokenKind::Pragma) != iter.next() {
         return Err(ParseError::new_error_str_at_pos(source, iter.last_valid_pos(), "expected a pragma"));
+    }
     let (pragma_start, pragma_end) = iter.next_positions();
     iter.consume();
     Ok((source.section_at_pos(pragma_start, pragma_end), pragma_start, pragma_end))
+}
 pub(crate) fn has_ident(source: &InputSource, iter: &mut TokenIter, expected: &[u8]) -> bool {
     peek_ident(source, iter).map_or(false, |section| section == expected)
+}
 pub(crate) fn peek_ident<'a>(source: &'a InputSource, iter: &mut TokenIter) -> Option<&'a [u8]> {
     if Some(TokenKind::Ident) == iter.next() {
         let (start, end) = iter.next_positions();
         return Some(source.section_at_pos(start, end))
+    }
     None
+}
 /// Consumes any identifier and returns it together with its span. Does not
 /// check if the identifier is a reserved keyword.
 pub(crate) fn consume_any_ident<'a>(
     source: &'a InputSource, iter: &mut TokenIter
 ) -> Result<(&'a [u8], InputSpan), ParseError> {
     if Some(TokenKind::Ident) != iter.next() {
         return Err(ParseError::new_error_str_at_pos(source, iter.last_valid_pos(), "expected an identifier"));
+    }
     let (ident_start, ident_end) = iter.next_positions();
     iter.consume();
     Ok((source.section_at_pos(ident_start, ident_end), InputSpan::from_positions(ident_start, ident_end)))
+}
 /// Consumes a specific identifier. May or may not be a reserved keyword.
 pub(crate) fn consume_exact_ident(source: &InputSource, iter: &mut TokenIter, expected: &[u8]) -> Result<InputSpan, ParseError> {
     let (ident, pos) = consume_any_ident(source, iter)?;
     if ident != expected {
         debug_assert!(expected.is_ascii());
         return Err(ParseError::new_error_at_pos(
             source, iter.last_valid_pos(),
             format!("expected the text '{}'", &String::from_utf8_lossy(expected))
         ));
+    }
     Ok(pos)
+}
 /// Consumes an identifier that is not a reserved keyword and returns it
 /// together with its span.
 pub(crate) fn consume_ident<'a>(
     source: &'a InputSource, iter: &mut TokenIter
 ) -> Result<(&'a [u8], InputSpan), ParseError> {
     let (ident, span) = consume_any_ident(source, iter)?;
     if is_reserved_keyword(ident) {
         return Err(ParseError::new_error_str_at_span(source, span, "encountered reserved keyword"));
+    }
     Ok((ident, span))
+}
 /// Consumes an identifier and immediately intern it into the `StringPool`
 pub(crate) fn consume_ident_interned(
     source: &InputSource, iter: &mut TokenIter, ctx: &mut PassCtx
 ) -> Result<Identifier, ParseError> {
     let (value, span) = consume_ident(source, iter)?;
     let value = ctx.pool.intern(value);
     Ok(Identifier{ span, value })
+}
 fn is_reserved_definition_keyword(text: &[u8]) -> bool {
     match text {
         KW_STRUCT | KW_ENUM | KW_UNION | KW_FUNCTION | KW_PRIMITIVE | KW_COMPOSITE => true,
         _ => false,
+    }
+}
 fn is_reserved_statement_keyword(text: &[u8]) -> bool {
     match text {
         KW_IMPORT | KW_AS |
         KW_STMT_CHANNEL | KW_STMT_IF | KW_STMT_WHILE |
         KW_STMT_BREAK | KW_STMT_CONTINUE | KW_STMT_GOTO | KW_STMT_RETURN |
         KW_STMT_SYNC | KW_STMT_ASSERT | KW_STMT_NEW => true,
         _ => false,
+    }
+}
 fn is_reserved_expression_keyword(text: &[u8]) -> bool {
     match text {
         KW_LET |
         KW_LIT_TRUE | KW_LIT_FALSE | KW_LIT_NULL |
         KW_FUNC_GET | KW_FUNC_PUT | KW_FUNC_FIRES | KW_FUNC_CREATE | KW_FUNC_LENGTH => true,
         _ => false,
+    }
+}

0 comments (0 inline, 0 general)