Support comments between (* and *)

Rakefile

@@ -14,7 +14,7 @@ directory 'build/boot'
 directory 'build/valid'
 
 def compile(input, output)
-  sh ENV.fetch('CC', 'gcc'), '-nostdlib', '-o', output, input
+  sh ENV.fetch('CC', 'gcc'), '-nostdlib', '-fpie', '-g', '-o', output, input
 end
 
 def run(exe)
@@ -39,43 +39,16 @@ end
 
 desc 'Convert previous stage language into the current stage language'
 task :convert do
-  File.open('boot/stage8.elna', 'w') do |current_stage|
+  File.open('boot/stage9.elna', 'w') do |current_stage|
-    File.readlines('boot/stage7.elna').each do |line|
+    File.readlines('boot/stage8.elna').each do |line|
-      if line == ".section .bss\n"
+      comment_match = /^(\s*)#(.*)/.match line
-        current_stage << <<~SECTION
-          const
-          	symbol_builtin_name_int := "Int";
-          	symbol_builtin_name_word := "Word";
-          	symbol_builtin_name_pointer := "Pointer";
-          	symbol_builtin_name_char := "Char";
-          	symbol_builtin_name_bool := "Bool";
 
-          	# Every type info starts with a word describing what type it is.
+      if comment_match.nil?
-		#
-		# PRIMITIVE_TYPE = 1
-		#
-          	# Primitive types have only type size.
-          	symbol_builtin_type_int := S(1, 4);
-          	symbol_builtin_type_word := S(1, 4);
-          	symbol_builtin_type_pointer := S(1, 4);
-          	symbol_builtin_type_char := S(1, 1);
-          	symbol_builtin_type_bool := S(1, 1);
-
-		# Info objects start with a word describing its type.
-		#
-		# INFO_TYPE = 1
-		#
-		# Type info has the type it belongs to.
-          	symbol_type_info_int := S(1, @symbol_builtin_type_int);
-          	symbol_type_info_word := S(1, @symbol_builtin_type_word);
-          	symbol_type_info_pointer := S(1, @symbol_builtin_type_pointer);
-          	symbol_type_info_char := S(1, @symbol_builtin_type_char);
-          	symbol_type_info_bool := S(1, @symbol_builtin_type_bool);
-        SECTION
-      elsif line == ".section .data\n"
-        current_stage << "var\n"
-      elsif !(line == ".section .text\n" || line == ".globl _start\n")
         current_stage << line
+      elsif comment_match[2].empty?
+        current_stage << "\n"
+      else
+        current_stage << "#{comment_match[1]}(* #{comment_match[2].strip} *)\n"
       end
     end
   end

boot/stage8.elna

@@ -4,7 +4,9 @@
 
 # Stage 8 compiler.
 #
-
+# - Procedure calls in expressions.
+# - Comments between (* and *) are supported. These are still single line
+#   comments and they should be on a separate line.
 const
 	symbol_builtin_name_int := "Int";
 	symbol_builtin_name_word := "Word";
@@ -34,14 +36,13 @@ const
 	symbol_type_info_char := S(1, @symbol_builtin_type_char);
 	symbol_type_info_bool := S(1, @symbol_builtin_type_bool);
 
-# When modifiying also change the read size in the entry point procedure.
-.type source_code, @object
-source_code: .zero 81920
-
-.type compiler_strings, @object
-compiler_strings: .zero 8192
-
 var
+	source_code: Array;
+	compiler_strings: Array;
+	symbol_table_global: Array;
+	symbol_table_local: Array;
+	classification: Array;
+
 	compiler_strings_position: Pointer := @compiler_strings;
 	compiler_strings_length: Word := 0;
 	source_code_position: Pointer := @source_code;
@@ -130,7 +131,8 @@ begin
 	mv a1, a0
 	# STDIN.
 	li a0, 0
-	li a7, 63 # SYS_READ.
+	# SYS_READ.
+	li a7, 63
 	ecall
 end;
 
@@ -145,7 +147,8 @@ begin
 	mv a1, a0
 	# STDOUT.
 	li a0, 1
-	li a7, 64 # SYS_WRITE.
+	# SYS_WRITE.
+	li a7, 64
 	ecall
 end;
 
@@ -323,7 +326,8 @@ begin
 
 .read_token_loop:
 	lw t0, 0(sp)
-	lb t0, (t0) # Current character.
+	# Current character.
+	lb t0, (t0)
 
 	# First we try to read a derictive.
 	# A derictive can contain a dot and characters.
@@ -425,44 +429,6 @@ begin
 	return v88
 end;
 
-proc _compile_section();
-begin
-	# Print and skip the ".section" (8 characters) directive and a space after it.
-	_write_token(9);
-	_advance_token();
-
-	# Read the section name.
-	_read_token();
-	addi a0, a0, 1
-
-	_write_token();
-	_advance_token();
-end;
-
-# Prints and skips a line.
-proc _skip_comment();
-begin
-	la t0, source_code_position
-	lw t1, (t0)
-
-.skip_comment_loop:
-	# Check for newline character.
-	lb t2, (t1)
-	li t3, '\n'
-	beq t2, t3, .skip_comment_end
-
-	# Advance the input string by one byte.
-	addi t1, t1, 1
-	sw t1, (t0)
-
-	goto .skip_comment_loop;
-
-.skip_comment_end:
-	# Skip the newline.
-	addi t1, t1, 1
-	sw t1, (t0)
-end;
-
 # Prints and skips a line.
 proc _compile_line();
 begin
@@ -601,6 +567,9 @@ begin
 	li t1, '"'
 	beq a0, t1, .compile_term_string_literal
 
+	li t1, '_'
+	beq a0, t1, .compile_term_call
+
 	_is_digit(v0);
 	bnez a0, .compile_term_integer_literal
 
@@ -630,6 +599,11 @@ begin
 	_compile_string_literal();
 	goto .compile_term_end;
 
+.compile_term_call:
+	_compile_call();
+	_write_z("\nmv t0, a0\n\0");
+	goto .compile_term_end;
+
 .compile_term_variable:
 	_compile_variable_expression();
 	goto .compile_term_end;
@@ -1038,6 +1012,8 @@ end;
 proc _compile_procedure_body();
 begin
 .compile_procedure_body_loop:
+	_skip_empty_lines();
+
 	# 3 is "end" length.
 	_memcmp(source_code_position, "end", 3);
 
@@ -1091,7 +1067,8 @@ begin
 	_advance_token(5);
 
 	_read_token();
-	sw a0, 0(sp) # Save the procedure name length.
+	# Save the procedure name length.
+	sw a0, 0(sp)
 
 	# Write .type _procedure_name, @function.
 	_write_z(".type \0");
@@ -1136,13 +1113,41 @@ begin
 .skip_newlines_end:
 end;
 
+# Prints and skips a line.
+proc _skip_comment();
+begin
+	la t0, source_code_position
+	lw t1, (t0)
+
+.skip_comment_loop:
+	# Check for newline character.
+	lb t2, (t1)
+	li t3, '\n'
+	beq t2, t3, .skip_comment_end
+
+	# Advance the input string by one byte.
+	addi t1, t1, 1
+	sw t1, (t0)
+
+	goto .skip_comment_loop;
+
+.skip_comment_end:
+	# Skip the newline.
+	addi t1, t1, 1
+	sw t1, (t0)
+end;
+
 # Skip newlines and comments.
 proc _skip_empty_lines();
 begin
-.skip_empty_lines_loop:
+.skip_empty_lines_rerun:
 	la t0, source_code_position
 	lw t0, (t0)
-	lb t0, (t0)
+	sw t0, 0(sp)
+
+.skip_empty_lines_loop:
+	lw t2, 0(sp)
+	lb t0, (t2)
 
 	li t1, '#'
 	beq t0, t1, .skip_empty_lines_comment
@@ -1150,15 +1155,35 @@ begin
 	li t1, '\n'
 	beq t0, t1, .skip_empty_lines_newline
 
+	li t1, '\t'
+	beq t0, t1, .skip_empty_lines_tab
+
+	li t1, '('
+	bne t0, t1, .skip_empty_lines_end
+	addi t2, t2, 1
+	lb t0, (t2)
+	li t1, '*'
+	beq t0, t1, .skip_empty_lines_comment
+
 	goto .skip_empty_lines_end;
 
 .skip_empty_lines_comment:
+	la t0, source_code_position
+	lw t1, 0(sp)
+	sw t1, (t0)
 	_skip_comment();
-	goto .skip_empty_lines_loop;
+	goto .skip_empty_lines_rerun;
 
 .skip_empty_lines_newline:
-	_advance_token(1);
+	la t0, source_code_position
-	goto .skip_empty_lines_loop;
+	lw t1, 0(sp)
+	addi t1, t1, 1
+	sw t1, (t0)
+	goto .skip_empty_lines_rerun;
+
+.skip_empty_lines_tab:
+	v0 := v0 + 1;
+	goto .skip_empty_lines_loop
 
 .skip_empty_lines_end:
 end;
@@ -1295,12 +1320,6 @@ begin
 
 	_advance_token(1);
 
-	la t0, source_code_position
-	lw t0, (t0)
-	lb t0, (t0)
-	li t1, '#'
-	beq t0, t1, .compile_const_part_loop
-
 	_compile_constant_declaration();
 	goto .compile_const_part_loop;
 
@@ -1358,18 +1377,16 @@ begin
 	_write_z(".section .data\n\0");
 
 .compile_var_part_loop:
+	_skip_empty_lines();
+
 	la t0, source_code_position
 	lw t0, (t0)
 	lb t0, (t0)
 
-	li t1, 'p'
-	beq t0, t1, .compile_var_part_end
-
 	li t1, '\t'
 	beq t0, t1, .compile_var_part_declaration
 
-	_compile_line();
+	goto .compile_var_part_end;
-	goto .compile_var_part_loop;
 
 .compile_var_part_declaration:
 	_advance_token(1);
@@ -1383,64 +1400,25 @@ end;
 proc _compile_module();
 begin
 	_compile_const_part();
-	_write_z(".section .bss\n\0");
+	_skip_empty_lines();
-
-.compile_module_bss:
-	la t0, source_code_position
-	lw t0, (t0)
-	lb t0, (t0)
-	li t1, 'v'
-	beq t0, t1, .compile_module_code
-
-	li t1, 'p'
-	beq t0, t1, .compile_module_code
-
-	_compile_line();
-	goto .compile_module_bss;
-
-.compile_module_code:
 	_compile_var_part();
+
 	_write_z(".section .text\n\0");
 .compile_module_loop:
-	_skip_newlines();
+	_skip_empty_lines();
 
 	la t0, source_code_position
 	lw t0, (t0)
 	lb t0, (t0)
 	beqz t0, .compile_module_end
-	li t1, '#'
-	beq t0, t1, .compile_module_comment
-
-	# 8 is ".section" length.
-	_memcmp(source_code_position, ".section", 8);
-	beqz a0, .compile_module_section
 
 	# 5 is "proc " length. Space is needed to distinguish from "procedure".
 	_memcmp(source_code_position, "proc ", 5);
 	beqz a0, .compile_module_procedure
 
-	# 6 is ".globl" length.
-	_memcmp(source_code_position, ".globl", 6);
-	beqz a0, .compile_module_global
-
 	# Not a known token, exit.
 	goto .compile_module_end;
 
-.compile_module_section:
-	_compile_section();
-
-	goto .compile_module_loop;
-
-.compile_module_global:
-	_compile_line();
-
-	goto .compile_module_loop;
-
-.compile_module_comment:
-	_skip_comment();
-
-	goto .compile_module_loop;
-
 .compile_module_procedure:
 	_compile_procedure();
 
@@ -1490,14 +1468,512 @@ begin
 	ecall
 end;
 
+# Inserts a symbol into the table.
+#
+# Parameters:
+# a0 - Symbol pointer.
+# a1 - Symbol name length.
+# a2 - Symbol name pointer.
+# a3 - Symbol table.
+proc _symbol_table_enter();
+begin
+	# The first word in the symbol table is its length, get it.
+	lw a0, 76(sp)
+	lw a0, (a0)
+	sw a0, 0(sp)
+
+	# Calculate the offset for the new symbol.
+	v4 := v0 * 4;
+	v4 := v4 + 4;
+	v4 := v76 + 4;
+
+	_memcpy(v4, @v80, 12);
+
+	# Increment the symbol table length.
+	v0 := v0 + 1;
+	lw t0, 0(sp)
+	lw t1, 76(sp)
+	sw t0, (t1)
+end;
+
+proc _symbol_table_build();
+begin
+	_symbol_table_enter(@symbol_type_info_int, 3, symbol_builtin_name_int, @symbol_table_global);
+	_symbol_table_enter(@symbol_type_info_word, 4, symbol_builtin_name_word, @symbol_table_global);
+	_symbol_table_enter(@symbol_type_info_pointer, 7, symbol_builtin_name_pointer, @symbol_table_global);
+	_symbol_table_enter(@symbol_type_info_char, 4, symbol_builtin_name_char, @symbol_table_global);
+	_symbol_table_enter(@symbol_type_info_bool, 4, symbol_builtin_name_bool, @symbol_table_global);
+end;
+
+#
+# Classification table assigns each possible character to a group (class). All
+# characters of the same group a handled equivalently.
+#
+# Classification:
+#
+# TransitionClass = (
+#   transitionClassInvalid = 1,
+#   transitionClassDigit = 2,
+#   transitionClassAlpha = 3,
+#   transitionClassSpace = 4,
+#   transitionClassColon = 5,
+#   transitionClassEquals = 6,
+#   transitionClassLeftParen = 7,
+#   transitionClassRightParen = 8,
+#   transitionClassAsterisk = 9,
+#   transitionClassUnderscore = 10,
+#   transitionClassSingle = 11,
+#   transitionClassHex = 12,
+#   transitionClassZero = 13,
+#   transitionClassX = 14,
+#   transitionClassEof = 15,
+#   transitionClassDot = 16,
+#   transitionClassMinus = 17,
+#   transitionClassSingleQuote = 18,
+#   transitionClassDoubleQuote = 19,
+#   transitionClassGreater = 20,
+#   transitionClassLess = 21,
+#   transitionClassOther = 22
+# );
+# TransitionState = (
+#   transitionStateStart = 1,
+#   transitionStateColon = 2,
+#   transitionStateIdentifier = 3,
+#   transitionStateDecimal = 4,
+#   transitionStateGreater = 5,
+#   transitionStateMinus = 6,
+#   transitionStateLeftParen = 7,
+#   transitionStateLess = 8,
+#   transitionStateDot = 9,
+#   transitionStateComment = 10,
+#   transitionStateClosingComment = 11,
+#   transitionStateCharacter = 12,
+#   transitionStateString = 13,
+#   transitionStateLeadingZero = 14,
+#   transitionStateDecimalSuffix = 15,
+#   transitionStateEnd = 16
+# );
+# Transition = record
+#   action: TransitionAction;
+#   next_state: TransitionState
+# end;
+# TransitionAction = (
+#   none = 1,
+#   accumulate = 2,
+#   skip = 3,
+#   single = 4,
+#   eof = 5,
+#   finalize = 6,
+#   composite = 7,
+#   key_id = 8,
+#   integer = 9,
+#   delimited = 10
+# );
+
+# Assigns some value to at array index.
+#
+# Parameters:
+# a0 - Array pointer.
+# a1 - Index (word offset into the array).
+# a2 - Data to assign.
+proc _assign_at();
+begin
+	v0 := v84 + -1;
+	v0 := v0 * 4;
+	v0 := v88 + v0;
+
+	lw t0, 0(sp)
+	lw t1, 80(sp)
+	sw t1, (t0)
+end;
+
+proc _initialize_classification();
+begin
+	_assign_at(@classification, 1, 15);
+	_assign_at(@classification, 2, 1);
+	_assign_at(@classification, 3, 1);
+	_assign_at(@classification, 4, 1);
+	_assign_at(@classification, 5, 1);
+	_assign_at(@classification, 6, 1);
+	_assign_at(@classification, 7, 1);
+	_assign_at(@classification, 8, 1);
+	_assign_at(@classification, 9, 1);
+	_assign_at(@classification, 10, 4);
+	_assign_at(@classification, 11, 4);
+	_assign_at(@classification, 12, 1);
+	_assign_at(@classification, 13, 1);
+	_assign_at(@classification, 14, 4);
+	_assign_at(@classification, 15, 1);
+	_assign_at(@classification, 16, 1);
+	_assign_at(@classification, 17, 1);
+	_assign_at(@classification, 18, 1);
+	_assign_at(@classification, 19, 1);
+	_assign_at(@classification, 20, 1);
+	_assign_at(@classification, 21, 1);
+	_assign_at(@classification, 22, 1);
+	_assign_at(@classification, 23, 1);
+	_assign_at(@classification, 24, 1);
+	_assign_at(@classification, 25, 1);
+	_assign_at(@classification, 26, 1);
+	_assign_at(@classification, 27, 1);
+	_assign_at(@classification, 28, 1);
+	_assign_at(@classification, 29, 1);
+	_assign_at(@classification, 30, 1);
+	_assign_at(@classification, 31, 1);
+	_assign_at(@classification, 32, 1);
+	_assign_at(@classification, 33, 4);
+	_assign_at(@classification, 34, 11);
+	_assign_at(@classification, 35, 19);
+	_assign_at(@classification, 36, 22);
+	_assign_at(@classification, 37, 22);
+	_assign_at(@classification, 38, 11);
+	_assign_at(@classification, 39, 11);
+	_assign_at(@classification, 40, 18);
+	_assign_at(@classification, 41, 7);
+	_assign_at(@classification, 42, 8);
+	_assign_at(@classification, 43, 9);
+	_assign_at(@classification, 44, 11);
+	_assign_at(@classification, 45, 11);
+	_assign_at(@classification, 46, 17);
+	_assign_at(@classification, 47, 16);
+	_assign_at(@classification, 48, 11);
+	_assign_at(@classification, 49, 13);
+	_assign_at(@classification, 50, 2);
+	_assign_at(@classification, 51, 2);
+	_assign_at(@classification, 52, 2);
+	_assign_at(@classification, 53, 2);
+	_assign_at(@classification, 54, 2);
+	_assign_at(@classification, 55, 2);
+	_assign_at(@classification, 56, 2);
+	_assign_at(@classification, 57, 2);
+	_assign_at(@classification, 58, 2);
+	_assign_at(@classification, 59, 5);
+	_assign_at(@classification, 60, 11);
+	_assign_at(@classification, 61, 21);
+	_assign_at(@classification, 62, 6);
+	_assign_at(@classification, 63, 20);
+	_assign_at(@classification, 64, 22);
+	_assign_at(@classification, 65, 11);
+	_assign_at(@classification, 66, 3);
+	_assign_at(@classification, 67, 3);
+	_assign_at(@classification, 68, 3);
+	_assign_at(@classification, 69, 3);
+	_assign_at(@classification, 70, 3);
+	_assign_at(@classification, 71, 3);
+	_assign_at(@classification, 72, 3);
+	_assign_at(@classification, 73, 3);
+	_assign_at(@classification, 74, 3);
+	_assign_at(@classification, 75, 3);
+	_assign_at(@classification, 76, 3);
+	_assign_at(@classification, 77, 3);
+	_assign_at(@classification, 78, 3);
+	_assign_at(@classification, 79, 3);
+	_assign_at(@classification, 80, 3);
+	_assign_at(@classification, 81, 3);
+	_assign_at(@classification, 82, 3);
+	_assign_at(@classification, 83, 3);
+	_assign_at(@classification, 84, 3);
+	_assign_at(@classification, 85, 3);
+	_assign_at(@classification, 86, 3);
+	_assign_at(@classification, 87, 3);
+	_assign_at(@classification, 88, 3);
+	_assign_at(@classification, 89, 3);
+	_assign_at(@classification, 90, 3);
+	_assign_at(@classification, 91, 3);
+	_assign_at(@classification, 92, 11);
+	_assign_at(@classification, 93, 22);
+	_assign_at(@classification, 94, 11);
+	_assign_at(@classification, 95, 11);
+	_assign_at(@classification, 96, 10);
+	_assign_at(@classification, 97, 22);
+	_assign_at(@classification, 98, 12);
+	_assign_at(@classification, 99, 12);
+	_assign_at(@classification, 100, 12);
+	_assign_at(@classification, 101, 12);
+	_assign_at(@classification, 102, 12);
+	_assign_at(@classification, 103, 12);
+	_assign_at(@classification, 104, 3);
+	_assign_at(@classification, 105, 3);
+	_assign_at(@classification, 106, 3);
+	_assign_at(@classification, 107, 3);
+	_assign_at(@classification, 108, 3);
+	_assign_at(@classification, 109, 3);
+	_assign_at(@classification, 110, 3);
+	_assign_at(@classification, 111, 3);
+	_assign_at(@classification, 112, 3);
+	_assign_at(@classification, 113, 3);
+	_assign_at(@classification, 114, 3);
+	_assign_at(@classification, 115, 3);
+	_assign_at(@classification, 116, 3);
+	_assign_at(@classification, 117, 3);
+	_assign_at(@classification, 118, 3);
+	_assign_at(@classification, 119, 3);
+	_assign_at(@classification, 120, 3);
+	_assign_at(@classification, 121, 14);
+	_assign_at(@classification, 122, 3);
+	_assign_at(@classification, 123, 3);
+	_assign_at(@classification, 124, 22);
+	_assign_at(@classification, 125, 11);
+	_assign_at(@classification, 126, 22);
+	_assign_at(@classification, 127, 11);
+	_assign_at(@classification, 128, 1);
+
+	v0 := 129;
+
+# Set the remaining 129 - 256 bytes to transitionClassOther.
+.initialize_classification_loop:
+	_assign_at(@classification, v0, 22);
+	v0 := v0 + 1;
+
+	lw t0, 0(sp)
+	li t1, 257
+	blt t0, t1, .initialize_classification_loop
+end;
+
+# Parameters:
+# a0 - Current state (first index into transitions table).
+# a1 - Transition (second index into transitions table)..
+# a2 - Action to assign.
+# a3 - Next state to assign.
+proc _set_transition();
+begin
+	# Transitions start at offset in classification array. Save the transitions start in v0.
+	la t0, classification
+	addi t0, t0, 256
+	sw t0, 0(sp)
+
+	# Each state is 8 bytes long (2 words: action and next state).
+	# There are 16 transition classes, so a transition 8 * 16 = 128 bytes long.
+
+	v4 := v88 + -1;
+	v4 := v4 * 128;
+
+	v8 := v84 + -1;
+	v8 := v8 * 8;
+
+	v12 := v0 + v4;
+	v12 := v12 + v8;
+
+	lw t0, 12(sp)
+	lw t1, 80(sp)
+	lw t2, 76(sp)
+	sw t1, (t0)
+	addi t0, t0, 4
+	sw t2, (t0)
+end;
+
+# Parameters:
+# a0 - Current state (Transition state enumeration).
+# a1 - Default action (Callback).
+# a2 - Next state (Transition state enumeration).
+proc _set_default_transition();
+begin
+	_set_transition(v88, 1, v84, v80);
+	_set_transition(v88, 2, v84, v80);
+	_set_transition(v88, 3, v84, v80);
+	_set_transition(v88, 4, v84, v80);
+	_set_transition(v88, 5, v84, v80);
+	_set_transition(v88, 6, v84, v80);
+	_set_transition(v88, 7, v84, v80);
+	_set_transition(v88, 8, v84, v80);
+	_set_transition(v88, 9, v84, v80);
+	_set_transition(v88, 10, v84, v80);
+	_set_transition(v88, 11, v84, v80);
+	_set_transition(v88, 12, v84, v80);
+	_set_transition(v88, 13, v84, v80);
+	_set_transition(v88, 14, v84, v80);
+	_set_transition(v88, 15, v84, v80);
+	_set_transition(v88, 16, v84, v80);
+	_set_transition(v88, 17, v84, v80);
+	_set_transition(v88, 18, v84, v80);
+	_set_transition(v88, 19, v84, v80);
+	_set_transition(v88, 20, v84, v80);
+	_set_transition(v88, 21, v84, v80);
+	_set_transition(v88, 22, v84, v80);
+end;
+
+
+# The transition table describes transitions from one state to another, given
+# a symbol (character class).
+#
+# The table has m rows and n columns, where m is the amount of states and n is
+# the amount of classes. So given the current state and a classified character
+# the table can be used to look up the next state.
+#
+# Each cell is a word long.
+# - The least significant byte of the word is a row number (beginning with 0).
+#   It specifies the target state. "ff" means that this is an end state and no
+#   transition is possible.
+# - The next byte is the action that should be performed when transitioning.
+#   For the meaning of actions see labels in the lex_next function, which
+#   handles each action.
+proc _initialize_transitions();
+begin
+	# Start state.
+	_set_transition(1, 1, 1, 16);
+	_set_transition(1, 2, 2, 4);
+	_set_transition(1, 3, 2, 3);
+	_set_transition(1, 4, 3, 1);
+	_set_transition(1, 5, 2, 5);
+	_set_transition(1, 6, 4, 16);
+	_set_transition(1, 7, 2, 7);
+	_set_transition(1, 8, 4, 16);
+	_set_transition(1, 9, 4, 16);
+	_set_transition(1, 10, 2, 3);
+	_set_transition(1, 11, 4, 16);
+	_set_transition(1, 12, 2, 3);
+	_set_transition(1, 13, 2, 14);
+	_set_transition(1, 14, 2, 3);
+	_set_transition(1, 15, 5, 16);
+	_set_transition(1, 16, 2, 9);
+	_set_transition(1, 17, 2, 6);
+	_set_transition(1, 18, 2, 12);
+	_set_transition(1, 19, 2, 13);
+	_set_transition(1, 20, 2, 5);
+	_set_transition(1, 21, 2, 8);
+	_set_transition(1, 22, 1, 16);
+
+	# Colon state.
+	_set_default_transition(2, 6, 16);
+	_set_transition(2, 6, 7, 16);
+
+	# Identifier state.
+	_set_default_transition(3, 8, 16);
+	_set_transition(3, 2, 2, 3);
+	_set_transition(3, 3, 2, 3);
+	_set_transition(3, 10, 2, 3);
+	_set_transition(3, 12, 2, 3);
+	_set_transition(3, 13, 2, 3);
+	_set_transition(3, 14, 2, 3);
+
+	# Decimal state.
+	_set_default_transition(4, 9, 16);
+	_set_transition(4, 2, 2, 4);
+	_set_transition(4, 3, 2, 15);
+	_set_transition(4, 10, 1, 16);
+	_set_transition(4, 12, 2, 15);
+	_set_transition(4, 13, 2, 4);
+	_set_transition(4, 14, 2, 15);
+
+	# Greater state.
+	_set_default_transition(5, 6, 16);
+	_set_transition(5, 6, 7, 16);
+
+	# Minus state.
+	_set_default_transition(6, 6, 16);
+	_set_transition(6, 20, 7, 16);
+
+	# Left paren state.
+	_set_default_transition(7, 6, 16);
+	_set_transition(7, 9, 2, 10);
+
+	# Less state.
+	_set_default_transition(8, 6, 16);
+	_set_transition(8, 6, 7, 16);
+	_set_transition(8, 20, 7, 16);
+
+	# Hexadecimal after 0x.
+	_set_default_transition(9, 6, 16);
+	_set_transition(9, 16, 7, 16);
+
+	# Comment.
+	_set_default_transition(10, 2, 10);
+	_set_transition(10, 9, 2, 11);
+	_set_transition(10, 15, 1, 16);
+
+	# Closing comment.
+	_set_default_transition(11, 2, 10);
+	_set_transition(11, 1, 1, 16);
+	_set_transition(11, 8, 10, 16);
+	_set_transition(11, 9, 2, 11);
+	_set_transition(11, 15, 1, 16);
+
+	# Character.
+	_set_default_transition(12, 2, 12);
+	_set_transition(12, 1, 1, 16);
+	_set_transition(12, 15, 1, 16);
+	_set_transition(12, 18, 10, 16);
+
+	# String.
+	_set_default_transition(13, 2, 13);
+	_set_transition(13, 1, 1, 16);
+	_set_transition(13, 15, 1, 16);
+	_set_transition(13, 19, 10, 16);
+
+	# Leading zero.
+	_set_default_transition(14, 9, 16);
+	_set_transition(14, 2, 1, 16);
+	_set_transition(14, 3, 1, 16);
+	_set_transition(14, 10, 1, 16);
+	_set_transition(14, 12, 1, 16);
+	_set_transition(14, 13, 1, 16);
+	_set_transition(14, 14, 1, 16);
+
+	# Digit with a character suffix.
+	_set_default_transition(15, 9, 16);
+	_set_transition(15, 3, 1, 16);
+	_set_transition(15, 2, 1, 16);
+	_set_transition(15, 12, 1, 16);
+	_set_transition(15, 13, 1, 16);
+	_set_transition(15, 14, 1, 16);
+end;
+
+proc _lexer_get_state();
+begin
+	# Lexer state is saved after the transition tables. The offset is 256 + 16 * 22.
+	v0 := @classification;
+	v4 := 16 * 22;
+	v0 := v0 + 256;
+
+	return v0 + v4
+end;
+
+# Gets pointer to the current source text.
+proc _lexer_get_current();
+begin
+	_lexer_get_state();
+	sw a0, 0(sp)
+
+	return v0 + 4
+end;
+
+# Resets the lexer state for reading the next token.
+proc _lexer_reset();
+begin
+	# Transition start state is 1.
+	_lexer_get_state();
+	li t0, 1
+	sw t0, (a0)
+	sw a0, 0(sp)
+
+	# Text pointer to the beginning of the currently read token.
+	_lexer_get_current();
+	la t0, source_code_position
+	lw t0, (t0)
+	sw t0, (a0)
+
+	# Initial length of the token is 0.
+	addi t0, t0, 4
+	sw zero, (t0)
+end;
+
+# One time lexer initialization.
+proc _lexer_initialize();
+begin
+	_initialize_classification();
+	_initialize_transitions();
+end;
+
 # Entry point.
 proc _start();
 begin
+	_lexer_initialize();
+	_symbol_table_build();
+
 	# Read the source from the standard input.
 	# Second argument is buffer size. Modifying update the source_code definition.
 	_read_file(@source_code, 81920);
 	_compile();
 
 	_exit(0);
-
 end;