Implement grammar parser

This implements the grammar struct together with
some other helpful structures like char-set and character.

The grammar struct parses a grammar file and also generates
the FIRST and FOLLOWS tables.

1 files changed, 272 insertions, 0 deletions

diff --git a/src/grammar.zig b/src/grammar.zig
new file mode 100644
index 0000000..0f25095
--- /dev/null
+++ b/src/grammar.zig
@@ -0,0 +1,272 @@
+const std = @import("std");
+
+const Character = @import("character.zig").Character;
+const RuleList = @import("rule-list.zig");
+
+const Grammar = struct {
+	const Self = @This();
+
+	rules: [26]RuleList,
+
+	pub fn init(buffer: []const u8, allocator: std.mem.Allocator) !Self {
+		var lines = std.mem.tokenizeScalar(u8, buffer, '\n');
+
+		var self: Self = undefined;
+		errdefer self.deinit(allocator);
+
+		for (&self.rules, 'A'..) |*rule, name| {
+			rule.* = RuleList.init(
+				@truncate(name),
+				allocator
+			);
+		}
+
+		while (lines.next()) |line| {
+			const name, const rhs = try RuleList.parse_rule(line, allocator);
+			try self.rule_by_name(name).add_rule(rhs);
+		}
+
+		self.generate_first();
+		self.generate_follows();
+
+		return self;
+	}
+
+	pub fn deinit(self: *Self, allocator: std.mem.Allocator) void {
+		for (&self.rules) |*rule| {
+			rule.deinit(allocator);
+		}
+
+		self.* = undefined;
+	}
+
+	pub inline fn rule_by_name(self: *Self, name: u8) *RuleList {
+		return &self.rules[name - 'A'];
+	}
+
+	pub inline fn entry_point(self: *Self) *RuleList {
+		return self.rule('S');
+	}
+
+	fn generate_first(self: *Self) void {
+		var modified = true;
+
+		while (modified) {
+			modified = false;
+
+			for (&self.rules) |*rule| {
+				for (rule.rhs.items) |rhs| {
+					switch (rhs[0]) {
+						.terminal => |t| {
+							modified = modified or !rule.first.is_set(t);
+							rule.first.set(t);
+						},
+
+						.non_terminal, .epsilon => {
+							if (rule.first.is_set(Character.EPSILON)) continue;
+							var insert_epsilon = true;
+
+							for (rhs) |*char| {
+								switch (char.*) {
+									.terminal => |t| {
+										modified = modified or !rule.first.is_set(t);
+										rule.first.set(t);
+										insert_epsilon = false;
+										break;
+									},
+									.non_terminal => |n| {
+										const first = self.rule_by_name(n).first;
+										for (first.charset, 0..) |child_first, index| {
+											if (index == Character.EPSILON) continue;
+
+											modified = modified or (!rule.first.is_set(index) and child_first);
+											rule.first.set_if(index, child_first);
+										}
+
+										if (!first.is_set(Character.EPSILON)) {
+											insert_epsilon = false;
+											break;
+										}
+									},
+									.epsilon => {},
+								}
+							}
+
+							if (insert_epsilon) {
+								modified = true;
+								rule.first.set(Character.EPSILON);
+							}
+						},
+					}
+				}
+			}
+		}
+	}
+
+	fn generate_follows(self: *Self) void {
+		self.rule_by_name('S').follows.set(Character.END);
+		var modified = true;
+
+		while (modified) {
+			modified = false;
+
+			for (&self.rules) |*parent_rules| {
+				for (parent_rules.rhs.items) |rule| {
+					for (0..rule.len) |character_index| {
+
+						switch (rule[character_index]) {
+							.non_terminal => |n| {
+								const current_rule = self.rule_by_name(n);
+								const ends_with_epsilon = brk: {
+									for (character_index + 1..rule.len) |peek_index| {
+										switch (rule[peek_index]) {
+											.terminal => |t| {
+												modified = modified or !current_rule.follows.is_set(t);
+												current_rule.follows.set(t);
+												break :brk false;
+											},
+											.non_terminal => |peek_n| {
+												const peek_rule = self.rule_by_name(peek_n);
+												for (peek_rule.first.charset, 0..) |is_set, index| {
+													if (Character.EPSILON == index) continue;
+
+													modified = modified or (!current_rule.follows.is_set(index) and is_set);
+													current_rule.follows.set_if(index, is_set);
+												}
+
+												if (!peek_rule.first.is_set(Character.EPSILON)) {
+													break :brk false;
+												}
+											},
+											.epsilon => {},
+										}
+									}
+									break :brk true;
+								};
+
+								if (ends_with_epsilon) {
+									for (parent_rules.follows.charset, 0..) |is_set, char_index| {
+										modified = modified or (!current_rule.follows.is_set(char_index) and is_set);
+										current_rule.follows.set_if(char_index, is_set);
+									}
+								}
+							},
+							else => {}
+						}
+
+					}
+				}
+			}
+		}
+	}
+};
+
+test "expr" {
+	const text =
+				\\S -> B A
+				\\A -> + B A
+				\\A -> _
+				\\B -> D C
+				\\C -> * D C
+				\\C -> _
+				\\D -> ( S )
+				\\D -> a
+				\\D -> b
+				;
+
+	const allocator = std.testing.allocator;
+	var grammar = try Grammar.init(text, allocator);
+	defer grammar.deinit(allocator);
+
+	try grammar.rule_by_name('S').first.expect(&[_]u8 { '(', 'a', 'b' });
+	try grammar.rule_by_name('A').first.expect(&[_]u8 { '+', Character.EPSILON });
+	try grammar.rule_by_name('B').first.expect(&[_]u8 { '(', 'a', 'b' });
+	try grammar.rule_by_name('C').first.expect(&[_]u8 { '*', Character.EPSILON });
+	try grammar.rule_by_name('D').first.expect(&[_]u8 { '(', 'a', 'b' });
+
+	try grammar.rule_by_name('S').follows.expect(&[_]u8 { ')', Character.END });
+	try grammar.rule_by_name('A').follows.expect(&[_]u8 { ')', Character.END });
+	try grammar.rule_by_name('B').follows.expect(&[_]u8 { '+', ')', Character.END });
+	try grammar.rule_by_name('C').follows.expect(&[_]u8 { '+', ')', Character.END });
+	try grammar.rule_by_name('D').follows.expect(&[_]u8 { '*', '+', ')', Character.END });
+}
+
+test "sample 0" {
+	const text =
+				\\S -> A C B
+				\\S -> C b b
+				\\S -> B a
+				\\A -> d a
+				\\A -> B C
+				\\B -> g
+				\\B -> _
+				\\C -> h
+				\\C -> _
+				;
+
+	const allocator = std.testing.allocator;
+	var grammar = try Grammar.init(text, allocator);
+	defer grammar.deinit(allocator);
+
+	try grammar.rule_by_name('S').first.expect(&[_]u8 { 'd', 'g', 'h', 'b', 'a', Character.EPSILON });
+	try grammar.rule_by_name('A').first.expect(&[_]u8 { 'd', 'g', 'h', Character.EPSILON });
+	try grammar.rule_by_name('B').first.expect(&[_]u8 { 'g', Character.EPSILON });
+	try grammar.rule_by_name('C').first.expect(&[_]u8 { 'h', Character.EPSILON });
+
+	try grammar.rule_by_name('S').follows.expect(&[_]u8 { Character.END });
+	try grammar.rule_by_name('A').follows.expect(&[_]u8 { 'g', 'h', Character.END });
+	try grammar.rule_by_name('B').follows.expect(&[_]u8 { 'a', 'g', 'h', Character.END });
+	try grammar.rule_by_name('C').follows.expect(&[_]u8 { 'b', 'g', 'h', Character.END });
+}
+
+test "sample 1" {
+	const text =
+		\\S -> a A B b
+		\\A -> c
+		\\A -> _
+		\\B -> d
+		\\B -> _
+		;
+
+	const allocator = std.testing.allocator;
+	var grammar = try Grammar.init(text, allocator);
+	defer grammar.deinit(allocator);
+
+	try grammar.rule_by_name('S').first.expect(&[_]u8 { 'a' });
+	try grammar.rule_by_name('A').first.expect(&[_]u8 { 'c', Character.EPSILON });
+	try grammar.rule_by_name('B').first.expect(&[_]u8 { 'd', Character.EPSILON });
+
+	try grammar.rule_by_name('S').follows.expect(&[_]u8 { Character.END });
+	try grammar.rule_by_name('A').follows.expect(&[_]u8 { 'b', 'd' });
+	try grammar.rule_by_name('B').follows.expect(&[_]u8 { 'b' });
+}
+
+test "sample 2" {
+	const text =
+		\\S -> A B
+		\\S -> e D a
+		\\A -> a b
+		\\A -> c
+		\\B -> d C
+		\\C -> e C
+		\\C -> _
+		\\D -> f D
+		\\D -> _
+		;
+
+	const allocator = std.testing.allocator;
+	var grammar = try Grammar.init(text, allocator);
+	defer grammar.deinit(allocator);
+
+	try grammar.rule_by_name('S').first.expect(&[_]u8 { 'a', 'c', 'e' });
+	try grammar.rule_by_name('A').first.expect(&[_]u8 { 'a', 'c' });
+	try grammar.rule_by_name('B').first.expect(&[_]u8 { 'd' });
+	try grammar.rule_by_name('C').first.expect(&[_]u8 { 'e', Character.EPSILON });
+	try grammar.rule_by_name('D').first.expect(&[_]u8 { 'f', Character.EPSILON });
+
+	try grammar.rule_by_name('S').follows.expect(&[_]u8 { Character.END });
+	try grammar.rule_by_name('A').follows.expect(&[_]u8 { 'd' });
+	try grammar.rule_by_name('B').follows.expect(&[_]u8 { Character.END });
+	try grammar.rule_by_name('C').follows.expect(&[_]u8 { Character.END });
+	try grammar.rule_by_name('D').follows.expect(&[_]u8 { 'a' });
+}