carbon-lang/executable_semantics/syntax/lexer.lpp

/*
Part of the Carbon Language project, under the Apache License v2.0 with LLVM
Exceptions. See /LICENSE for license information.
SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
*/

%{

#include <cstdlib>

#include "common/check.h"
#include "common/string_helpers.h"
#include "executable_semantics/common/tracing_flag.h"
#include "executable_semantics/syntax/parse_and_lex_context.h"
#include "executable_semantics/syntax/parser.h"
#include "llvm/ADT/StringExtras.h"
%}

/* Turn off legacy bits we don't need. */
%option noyywrap nounput nodefault noinput

%option reentrant

/* Lexing a token immediately after consuming some whitespace. */
%s AFTER_WHITESPACE
/*
 * Lexing a token immediately after consuming an operand-ending token:
 * a closing bracket, identifier, or literal.
 */
%s AFTER_OPERAND

AND                  "and"
ARROW                "->"
AUTO                 "auto"
AWAIT                "__await"
BOOL                 "Bool"
BREAK                "break"
CASE                 "case"
CHOICE               "choice"
CLASS                "class"
COLON                ":"
COLON_BANG           ":!"
COMMA                ","
CONTINUATION         "__continuation"
CONTINUATION_TYPE    "__Continuation"
CONTINUE             "continue"
DEFAULT              "default"
DOUBLE_ARROW         "=>"
ELSE                 "else"
EQUAL                "="
EQUAL_EQUAL          "=="
FALSE                "false"
FN                   "fn"
FNTY                 "fnty"
IF                   "if"
LEFT_CURLY_BRACE     "{"
LEFT_PARENTHESIS     "("
LEFT_SQUARE_BRACKET  "["
MATCH                "match"
MINUS                "-"
NOT                  "not"
OR                   "or"
PERIOD               "."
PLUS                 "+"
RETURN               "return"
RIGHT_CURLY_BRACE    "}"
RIGHT_PARENTHESIS    ")"
RIGHT_SQUARE_BRACKET "]"
RUN                  "__run"
SEMICOLON            ";"
SLASH                "/"
STRING               "String"
TRUE                 "true"
TYPE                 "Type"
UNDERSCORE           "_"
VAR                  "var"
WHILE                "while"

identifier            [A-Za-z_][A-Za-z0-9_]*
sized_type_literal    [iuf][1-9][0-9]*
integer_literal       [0-9]+
horizontal_whitespace [ \t\r]
whitespace            [ \t\r\n]
one_line_comment      \/\/[^\n]*\n
operand_start         [(A-Za-z0-9_"]

/* Single-line string literals should reject vertical whitespace. */
string_literal        \"([^\\\"\n\v\f\r]|\\.)*\"

%{
  // This macro is expanded immediately before each action specified below.
  //
  // Advances the current token position by yyleng columns without changing
  // the line number, and takes us out of the after-whitespace / after-operand
  // state.
  # define YY_USER_ACTION \
      context.current_token_position.columns(yyleng); \
      if (YY_START == AFTER_WHITESPACE || \
          YY_START == AFTER_OPERAND) { \
        BEGIN(INITIAL); \
      }

  #define SIMPLE_TOKEN(name) \
      Carbon::Parser::make_##name(context.current_token_position);

  #define ARG_TOKEN(name, arg) \
      Carbon::Parser::make_##name(arg, context.current_token_position);
%}

%%

%{
  // Code run each time yylex is called.

  // Begin with an empty token span starting where its previous end was.
  context.current_token_position.step();
%}

{AND}                 { return SIMPLE_TOKEN(AND); }
{ARROW}               { return SIMPLE_TOKEN(ARROW); }
{AUTO}                { return SIMPLE_TOKEN(AUTO); }
{AWAIT}               { return SIMPLE_TOKEN(AWAIT); }
{BOOL}                { return SIMPLE_TOKEN(BOOL); }
{BREAK}               { return SIMPLE_TOKEN(BREAK); }
{CASE}                { return SIMPLE_TOKEN(CASE); }
{CHOICE}              { return SIMPLE_TOKEN(CHOICE); }
{CLASS}               { return SIMPLE_TOKEN(CLASS); }
{COLON_BANG}          { return SIMPLE_TOKEN(COLON_BANG); }
{COLON}               { return SIMPLE_TOKEN(COLON); }
{COMMA}               { return SIMPLE_TOKEN(COMMA); }
{CONTINUATION_TYPE}   { return SIMPLE_TOKEN(CONTINUATION_TYPE); }
{CONTINUATION}        { return SIMPLE_TOKEN(CONTINUATION); }
{CONTINUE}            { return SIMPLE_TOKEN(CONTINUE); }
{DEFAULT}             { return SIMPLE_TOKEN(DEFAULT); }
{DOUBLE_ARROW}        { return SIMPLE_TOKEN(DOUBLE_ARROW); }
{ELSE}                { return SIMPLE_TOKEN(ELSE); }
{EQUAL_EQUAL}         { return SIMPLE_TOKEN(EQUAL_EQUAL); }
{EQUAL}               { return SIMPLE_TOKEN(EQUAL); }
{FALSE}               { return SIMPLE_TOKEN(FALSE); }
{FNTY}                { return SIMPLE_TOKEN(FNTY); }
{FN}                  { return SIMPLE_TOKEN(FN); }
{IF}                  { return SIMPLE_TOKEN(IF); }
{LEFT_PARENTHESIS}    { return SIMPLE_TOKEN(LEFT_PARENTHESIS); }
{LEFT_CURLY_BRACE}    { return SIMPLE_TOKEN(LEFT_CURLY_BRACE); }
{LEFT_SQUARE_BRACKET} { return SIMPLE_TOKEN(LEFT_SQUARE_BRACKET); }
{MATCH}               { return SIMPLE_TOKEN(MATCH); }
{MINUS}               { return SIMPLE_TOKEN(MINUS); }
{NOT}                 { return SIMPLE_TOKEN(NOT); }
{OR}                  { return SIMPLE_TOKEN(OR); }
{PERIOD}              { return SIMPLE_TOKEN(PERIOD); }
{PLUS}                { return SIMPLE_TOKEN(PLUS); }
{RETURN}              { return SIMPLE_TOKEN(RETURN); }
{RUN}                 { return SIMPLE_TOKEN(RUN); }
{SEMICOLON}           { return SIMPLE_TOKEN(SEMICOLON); }
{SLASH}               { return SIMPLE_TOKEN(SLASH); }
{STRING}              { return SIMPLE_TOKEN(STRING); }
{TRUE}                { return SIMPLE_TOKEN(TRUE); }
{TYPE}                { return SIMPLE_TOKEN(TYPE); }
{UNDERSCORE}          { return SIMPLE_TOKEN(UNDERSCORE); }
{VAR}                 { return SIMPLE_TOKEN(VAR); }
{WHILE}               { return SIMPLE_TOKEN(WHILE); }

 /* More modern Bisons provide make_EOF. */
<<EOF>>               { return SIMPLE_TOKEN(END_OF_FILE); }

{RIGHT_PARENTHESIS} {
  BEGIN(AFTER_OPERAND);
  return SIMPLE_TOKEN(RIGHT_PARENTHESIS);
}
{RIGHT_CURLY_BRACE} {
  BEGIN(AFTER_OPERAND);
  return SIMPLE_TOKEN(RIGHT_CURLY_BRACE);
}
{RIGHT_SQUARE_BRACKET} {
  BEGIN(AFTER_OPERAND);
  return SIMPLE_TOKEN(RIGHT_SQUARE_BRACKET);
}

 /*
  * For a `*` operator, we look at whitespace and local context to determine the
  * arity and fixity. There are two ways to write a binary operator:
  *
  * 1) Whitespace on both sides.
  * 2) Whitespace on neither side, and the previous token is considered to be
  *    the end of an operand, and the next token is considered to be the start
  *    of an operand.
  *
  * Otherwise, the operator is unary, but we also check for whitespace to help
  * the parser enforce the rule that whitespace is not permitted between the
  * operator and its operand, leading to three more cases:
  *
  * 3) Whitespace before (but implicitly not after, because that would give a
  *    longer match and hit case 1): this can only be a prefix operator.
  * 4) Whitespace after and not before: this can only be a postfix operator.
  * 5) No whitespace on either side (otherwise the longest match would take us
  *    to case 4): this is a unary operator and could be either prefix or
  *    postfix.
  */

 /* `*` operator case 1: */
<AFTER_WHITESPACE>"*"{whitespace}+ {
  BEGIN(AFTER_WHITESPACE);
  return SIMPLE_TOKEN(BINARY_STAR);
}
 /* `*` operator case 2: */
<AFTER_OPERAND>"*"/{operand_start} {
  return SIMPLE_TOKEN(BINARY_STAR);
}
 /* `*` operator case 3: */
<AFTER_WHITESPACE>"*" {
  return SIMPLE_TOKEN(PREFIX_STAR);
}
 /* `*` operator case 4: */
<INITIAL,AFTER_OPERAND>"*"{whitespace}+ {
  BEGIN(AFTER_WHITESPACE);
  return SIMPLE_TOKEN(POSTFIX_STAR);
}
 /* `*` operator case 5: */
<INITIAL,AFTER_OPERAND>"*" {
  return SIMPLE_TOKEN(UNARY_STAR);
}

{sized_type_literal} {
  return ARG_TOKEN(sized_type_literal, yytext);
}

{identifier} {
  BEGIN(AFTER_OPERAND);
  return ARG_TOKEN(identifier, yytext);
}

{integer_literal} {
  BEGIN(AFTER_OPERAND);
  int val;
  CHECK(llvm::to_integer(yytext, val));
  return ARG_TOKEN(integer_literal, val);
}

{string_literal} {
  llvm::StringRef str(yytext);
  CHECK(str.consume_front("\"") && str.consume_back("\""));
  std::optional<std::string> unescaped = Carbon::UnescapeStringLiteral(str);
  if (unescaped == std::nullopt) {
    if (Carbon::tracing_output) {
      // Print a newline because tracing prints an incomplete line
      // "Reading a token: ".
      llvm::errs() << "\n";
    }
    FATAL_COMPILATION_ERROR(context.SourceLoc())
        << "Invalid escaping in string: " << yytext;
  }
  return ARG_TOKEN(string_literal, *unescaped);
}

{one_line_comment} {
  // Advance end by 1 line, resetting the column to zero.
  context.current_token_position.lines(1);
  // Make the span empty by setting start to end.
  context.current_token_position.step();
}

{horizontal_whitespace}+ {
  // Make the span empty by setting start to end.
  context.current_token_position.step();
  BEGIN(AFTER_WHITESPACE);
}

\n+ {
  // Advance end by yyleng lines, resetting the column to zero.
  context.current_token_position.lines(yyleng);
  // Make the span empty by setting start to end.
  context.current_token_position.step();
  BEGIN(AFTER_WHITESPACE);
}

. {
  if (Carbon::tracing_output) {
    // Print a newline because tracing prints an incomplete line
    // "Reading a token: ".
    llvm::errs() << "\n";
  }
  FATAL_COMPILATION_ERROR(context.SourceLoc()) << "invalid character '\\x"
            << llvm::toHex(llvm::StringRef(yytext, 1)) << "' in source file.";
}

%%