Commit 01aa74a93f

Shachaf Ben-Kiki <shachaf@gmail.com>
2021-04-25 01:41:10
parser: Use an operator precedence table
Instead of having one function per precedence level, mirroring the BNF grammar, this uses the "precedence climbing" algorithm with a table of operator precedences (and a few special cases that don't fit into the table as it is). This is a first pass -- it's probably possible to put more of the parser into this form, e.g. to support prefix/suffix operators with precedence, if necessary, or just to simplify the code more. It may also be possible to speed this up by putting more useful information into the tokens during tokenization, to avoid the extra branch on the token in operInfo.
1 parent 37b0574
Changed files (2)
lib/std/zig/parse.zig
@@ -1309,9 +1309,8 @@ const Parser = struct {
         return expr;
     }
 
-    /// Expr <- BoolOrExpr
     fn parseExpr(p: *Parser) Error!Node.Index {
-        return p.parseBoolOrExpr();
+        return p.parseExprPrecedence(0);
     }
 
     fn expectExpr(p: *Parser) Error!Node.Index {
@@ -1323,263 +1322,104 @@ const Parser = struct {
         }
     }
 
-    /// BoolOrExpr <- BoolAndExpr (KEYWORD_or BoolAndExpr)*
-    fn parseBoolOrExpr(p: *Parser) Error!Node.Index {
-        var res = try p.parseBoolAndExpr();
-        if (res == 0) return null_node;
-
-        while (true) {
-            switch (p.token_tags[p.tok_i]) {
-                .keyword_or => {
-                    const or_token = p.nextToken();
-                    const rhs = try p.parseBoolAndExpr();
-                    if (rhs == 0) {
-                        return p.fail(.invalid_token);
-                    }
-                    res = try p.addNode(.{
-                        .tag = .bool_or,
-                        .main_token = or_token,
-                        .data = .{
-                            .lhs = res,
-                            .rhs = rhs,
-                        },
-                    });
-                },
-                else => return res,
-            }
-        }
-    }
-
-    /// BoolAndExpr <- CompareExpr (KEYWORD_and CompareExpr)*
-    fn parseBoolAndExpr(p: *Parser) !Node.Index {
-        var res = try p.parseCompareExpr();
-        if (res == 0) return null_node;
-
-        while (true) {
-            switch (p.token_tags[p.tok_i]) {
-                .keyword_and => {
-                    const and_token = p.nextToken();
-                    const rhs = try p.parseCompareExpr();
-                    if (rhs == 0) {
-                        return p.fail(.invalid_token);
-                    }
-                    res = try p.addNode(.{
-                        .tag = .bool_and,
-                        .main_token = and_token,
-                        .data = .{
-                            .lhs = res,
-                            .rhs = rhs,
-                        },
-                    });
-                },
-                .invalid_ampersands => {
-                    try p.warn(.invalid_and);
-                    p.tok_i += 1;
-                    return p.parseCompareExpr();
-                },
-                else => return res,
-            }
-        }
-    }
+    const Assoc = enum {
+        left,
+        none,
+    };
 
-    /// CompareExpr <- BitwiseExpr (CompareOp BitwiseExpr)?
-    /// CompareOp
-    ///     <- EQUALEQUAL
-    ///      / EXCLAMATIONMARKEQUAL
-    ///      / LARROW
-    ///      / RARROW
-    ///      / LARROWEQUAL
-    ///      / RARROWEQUAL
-    fn parseCompareExpr(p: *Parser) !Node.Index {
-        const expr = try p.parseBitwiseExpr();
-        if (expr == 0) return null_node;
+    const OperInfo = struct {
+        prec: i8,
+        tag: Node.Tag,
+        assoc: Assoc = Assoc.left,
+    };
 
-        const tag: Node.Tag = switch (p.token_tags[p.tok_i]) {
-            .equal_equal => .equal_equal,
-            .bang_equal => .bang_equal,
-            .angle_bracket_left => .less_than,
-            .angle_bracket_right => .greater_than,
-            .angle_bracket_left_equal => .less_or_equal,
-            .angle_bracket_right_equal => .greater_or_equal,
-            else => return expr,
+    // A table of binary operator information. Higher precedence numbers are
+    // stickier. All operators at the same precedence level should have the same
+    // associativity.
+    fn operInfo(tag: Token.Tag) OperInfo {
+        return switch (tag) {
+            .keyword_or => .{ .prec = 10, .tag = .bool_or },
+
+            .keyword_and => .{ .prec = 20, .tag = .bool_and },
+            .invalid_ampersands => .{ .prec = 20, .tag = .bool_and },
+
+            .equal_equal => .{ .prec = 30, .tag = .equal_equal, .assoc = Assoc.none },
+            .bang_equal => .{ .prec = 30, .tag = .bang_equal, .assoc = Assoc.none },
+            .angle_bracket_left => .{ .prec = 30, .tag = .less_than, .assoc = Assoc.none },
+            .angle_bracket_right => .{ .prec = 30, .tag = .greater_than, .assoc = Assoc.none },
+            .angle_bracket_left_equal => .{ .prec = 30, .tag = .less_or_equal, .assoc = Assoc.none },
+            .angle_bracket_right_equal => .{ .prec = 30, .tag = .greater_or_equal, .assoc = Assoc.none },
+
+            .ampersand => .{ .prec = 40, .tag = .bit_and },
+            .caret => .{ .prec = 40, .tag = .bit_xor },
+            .pipe => .{ .prec = 40, .tag = .bit_or },
+            .keyword_orelse => .{ .prec = 40, .tag = .@"orelse" },
+            .keyword_catch => .{ .prec = 40, .tag = .@"catch" },
+
+            .angle_bracket_angle_bracket_left => .{ .prec = 50, .tag = .bit_shift_left },
+            .angle_bracket_angle_bracket_right => .{ .prec = 50, .tag = .bit_shift_right },
+
+            .plus => .{ .prec = 60, .tag = .add },
+            .minus => .{ .prec = 60, .tag = .sub },
+            .plus_plus => .{ .prec = 60, .tag = .array_cat },
+            .plus_percent => .{ .prec = 60, .tag = .add_wrap },
+            .minus_percent => .{ .prec = 60, .tag = .sub_wrap },
+
+            .pipe_pipe => .{ .prec = 70, .tag = .merge_error_sets },
+            .asterisk => .{ .prec = 70, .tag = .mul },
+            .slash => .{ .prec = 70, .tag = .div },
+            .percent => .{ .prec = 70, .tag = .mod },
+            .asterisk_asterisk => .{ .prec = 70, .tag = .array_mult },
+            .asterisk_percent => .{ .prec = 70, .tag = .mul_wrap },
+
+            else => .{ .prec = -1, .tag = ast.Node.Tag.root }, // irrelevant
         };
-        return p.addNode(.{
-            .tag = tag,
-            .main_token = p.nextToken(),
-            .data = .{
-                .lhs = expr,
-                .rhs = try p.expectBitwiseExpr(),
-            },
-        });
     }
 
-    /// BitwiseExpr <- BitShiftExpr (BitwiseOp BitShiftExpr)*
-    /// BitwiseOp
-    ///     <- AMPERSAND
-    ///      / CARET
-    ///      / PIPE
-    ///      / KEYWORD_orelse
-    ///      / KEYWORD_catch Payload?
-    fn parseBitwiseExpr(p: *Parser) !Node.Index {
-        var res = try p.parseBitShiftExpr();
-        if (res == 0) return null_node;
-
-        while (true) {
-            const tag: Node.Tag = switch (p.token_tags[p.tok_i]) {
-                .ampersand => .bit_and,
-                .caret => .bit_xor,
-                .pipe => .bit_or,
-                .keyword_orelse => .@"orelse",
-                .keyword_catch => {
-                    const catch_token = p.nextToken();
-                    _ = try p.parsePayload();
-                    const rhs = try p.parseBitShiftExpr();
-                    if (rhs == 0) {
-                        return p.fail(.invalid_token);
-                    }
-                    res = try p.addNode(.{
-                        .tag = .@"catch",
-                        .main_token = catch_token,
-                        .data = .{
-                            .lhs = res,
-                            .rhs = rhs,
-                        },
-                    });
-                    continue;
-                },
-                else => return res,
-            };
-            res = try p.addNode(.{
-                .tag = tag,
-                .main_token = p.nextToken(),
-                .data = .{
-                    .lhs = res,
-                    .rhs = try p.expectBitShiftExpr(),
-                },
-            });
-        }
-    }
-
-    fn expectBitwiseExpr(p: *Parser) Error!Node.Index {
-        const node = try p.parseBitwiseExpr();
+    fn parseExprPrecedence(p: *Parser, min_prec: i32) Error!Node.Index {
+        var node = try p.parsePrefixExpr();
         if (node == 0) {
-            return p.fail(.invalid_token);
-        } else {
-            return node;
+            return null_node;
         }
-    }
 
-    /// BitShiftExpr <- AdditionExpr (BitShiftOp AdditionExpr)*
-    /// BitShiftOp
-    ///     <- LARROW2
-    ///      / RARROW2
-    fn parseBitShiftExpr(p: *Parser) Error!Node.Index {
-        var res = try p.parseAdditionExpr();
-        if (res == 0) return null_node;
+        var banned_prec: i8 = -1;
 
         while (true) {
-            const tag: Node.Tag = switch (p.token_tags[p.tok_i]) {
-                .angle_bracket_angle_bracket_left => .bit_shift_left,
-                .angle_bracket_angle_bracket_right => .bit_shift_right,
-                else => return res,
-            };
-            res = try p.addNode(.{
-                .tag = tag,
-                .main_token = p.nextToken(),
-                .data = .{
-                    .lhs = res,
-                    .rhs = try p.expectAdditionExpr(),
+            const tok_tag = p.token_tags[p.tok_i];
+            const info = operInfo(tok_tag);
+            if (info.prec < min_prec or info.prec == banned_prec) {
+                break;
+            }
+            const oper_token = p.nextToken();
+            // Special-case handling for "catch" and "&&".
+            switch (tok_tag) {
+                .keyword_catch => {
+                    _ = try p.parsePayload();
                 },
-            });
-        }
-    }
-
-    fn expectBitShiftExpr(p: *Parser) Error!Node.Index {
-        const node = try p.parseBitShiftExpr();
-        if (node == 0) {
-            return p.fail(.invalid_token);
-        } else {
-            return node;
-        }
-    }
-
-    /// AdditionExpr <- MultiplyExpr (AdditionOp MultiplyExpr)*
-    /// AdditionOp
-    ///     <- PLUS
-    ///      / MINUS
-    ///      / PLUS2
-    ///      / PLUSPERCENT
-    ///      / MINUSPERCENT
-    fn parseAdditionExpr(p: *Parser) Error!Node.Index {
-        var res = try p.parseMultiplyExpr();
-        if (res == 0) return null_node;
-
-        while (true) {
-            const tag: Node.Tag = switch (p.token_tags[p.tok_i]) {
-                .plus => .add,
-                .minus => .sub,
-                .plus_plus => .array_cat,
-                .plus_percent => .add_wrap,
-                .minus_percent => .sub_wrap,
-                else => return res,
-            };
-            res = try p.addNode(.{
-                .tag = tag,
-                .main_token = p.nextToken(),
-                .data = .{
-                    .lhs = res,
-                    .rhs = try p.expectMultiplyExpr(),
+                .invalid_ampersands => {
+                    try p.warn(.invalid_and);
                 },
-            });
-        }
-    }
-
-    fn expectAdditionExpr(p: *Parser) Error!Node.Index {
-        const node = try p.parseAdditionExpr();
-        if (node == 0) {
-            return p.fail(.invalid_token);
-        }
-        return node;
-    }
-
-    /// MultiplyExpr <- PrefixExpr (MultiplyOp PrefixExpr)*
-    /// MultiplyOp
-    ///     <- PIPE2
-    ///      / ASTERISK
-    ///      / SLASH
-    ///      / PERCENT
-    ///      / ASTERISK2
-    ///      / ASTERISKPERCENT
-    fn parseMultiplyExpr(p: *Parser) Error!Node.Index {
-        var res = try p.parsePrefixExpr();
-        if (res == 0) return null_node;
+                else => {},
+            }
+            const rhs = try p.parseExprPrecedence(info.prec + 1);
+            if (rhs == 0) {
+                return p.fail(.invalid_token);
+            }
 
-        while (true) {
-            const tag: Node.Tag = switch (p.token_tags[p.tok_i]) {
-                .pipe_pipe => .merge_error_sets,
-                .asterisk => .mul,
-                .slash => .div,
-                .percent => .mod,
-                .asterisk_asterisk => .array_mult,
-                .asterisk_percent => .mul_wrap,
-                else => return res,
-            };
-            res = try p.addNode(.{
-                .tag = tag,
-                .main_token = p.nextToken(),
+            node = try p.addNode(.{
+                .tag = info.tag,
+                .main_token = oper_token,
                 .data = .{
-                    .lhs = res,
-                    .rhs = try p.expectPrefixExpr(),
+                    .lhs = node,
+                    .rhs = rhs,
                 },
             });
-        }
-    }
 
-    fn expectMultiplyExpr(p: *Parser) Error!Node.Index {
-        const node = try p.parseMultiplyExpr();
-        if (node == 0) {
-            return p.fail(.invalid_token);
+            if (info.assoc == Assoc.none) {
+                banned_prec = info.prec;
+            }
         }
+
         return node;
     }
 
lib/std/zig/parser_test.zig
@@ -2796,6 +2796,7 @@ test "zig fmt: precedence" {
         \\    a or b and c;
         \\    (a or b) and c;
         \\    (a or b) and c;
+        \\    a == b and c == d;
         \\}
         \\
     );
@@ -4860,6 +4861,16 @@ test "recovery: missing comma" {
     });
 }
 
+test "recovery: non-associative operators" {
+    try testError(
+        \\const x = a == b == c;
+        \\const x = a == b != c;
+    , &[_]Error{
+        .expected_token,
+        .expected_token,
+    });
+}
+
 test "recovery: extra qualifier" {
     try testError(
         \\const a: *const const u8;