General cleanup

assembler.go

@@ -10,10 +10,11 @@ import (
 )
 
 const (
-	WORD    = iota // [A-Za-z_-]+
+	WORD        = iota // [A-Za-z_-]+
-	NUMBER         // [0-9]+
+	INSTRUCTION        // Instruction word
-	NEWLINE        // \n
+	NUMBER             // [0-9]+
-	ERROR          // Error
+	NEWLINE            // \n
+	ERROR              // Error
 )
 
 type location struct {
@@ -22,9 +23,10 @@ type location struct {
 }
 
 type token struct {
-	location location // Position of the token
+	location    location // Position of the token
-	value    string   // Text content of the token
+	value       string   // Text content of the token
-	kind     int      // Kind of the token
+	instruction int      // INSTRUCTION ID
+	kind        int      // Kind of the token
 }
 
 type tokenizer struct {
@@ -57,19 +59,62 @@ func isWordTail(c byte) bool {
 
 // -----------------------------------------------------------------------------
 
+const (
+	IHALT = iota * 100
+	IADD
+	ISUBTRACT
+	ISTORE
+	_
+	ILOAD
+	IBRANCH
+	IBRANCH_IF_ZERO
+	IBRANCH_IF_POSITIVE
+	IIO
+)
+
+const (
+	_ = iota
+	IO_INPUT
+	IO_OUTPUT
+)
+
+var instructions = map[string]int{
+	"HLT": IHALT,
+	"COB": IHALT,
+	"ADD": IADD,
+	"SUB": ISUBTRACT,
+	"STA": ISTORE,
+	"LDA": ILOAD,
+	"BRA": IBRANCH,
+	"BRZ": IBRANCH_IF_ZERO,
+	"BRP": IBRANCH_IF_POSITIVE,
+	"INP": IIO + IO_INPUT,
+	"OUT": IIO + IO_OUTPUT,
+	"DAT": 0,
+}
+
+// -----------------------------------------------------------------------------
+
 func (t *tokenizer) send(start location, kind int) {
-	t.tokens <- token{start, string(t.value), kind}
+	tok := token{start, strings.ToUpper(string(t.value)), 0, kind}
+	if kind == WORD {
+		if instr, found := instructions[tok.value]; found {
+			tok.kind = INSTRUCTION
+			tok.instruction = instr
+		}
+	}
+	t.tokens <- tok
 	t.value = []byte{}
 }
 
 // XXX: the handling could probably be simplified by extending the "byte"
 //   to also include a special out-of-band value for errors
 func (t *tokenizer) peek() (byte, error) {
-	buf, err := t.reader.Peek(1)
+	if buf, err := t.reader.Peek(1); err != nil {
-	if err != nil {
 		return '?', err
+	} else {
+		return buf[0], nil
 	}
-	return buf[0], err
 }
 
 func (t *tokenizer) eat() (byte, error) {
@@ -170,13 +215,11 @@ func tokenize(r io.Reader, tokens chan<- token) {
 		reader:   bufio.NewReader(r),
 	}
 	for {
-		err := t.step()
+		if err := t.step(); err == io.EOF {
-		if err == io.EOF {
 			break
-		}
+		} else if err != nil {
-		if err != nil {
 			t.tokens <- token{t.location, fmt.Sprintf("line %d, column %d: %s",
-				t.location.line, t.location.column, err.Error()), ERROR}
+				t.location.line, t.location.column, err.Error()), 0, ERROR}
 			break
 		}
 	}
@@ -185,44 +228,12 @@ func tokenize(r io.Reader, tokens chan<- token) {
 
 // -----------------------------------------------------------------------------
 
-const (
-	IHALT = iota
-	IADD
-	ISUBTRACT
-	ISTORE
-	ILOAD
-	_
-	IBRANCH
-	IBRANCH_IF_ZERO
-	IBRANCH_IF_POSITIVE
-	IINPUT
-	IOUTPUT
-	IDATA
-)
-
-var instructions = map[string]int{
-	"HLT": IHALT,
-	"COB": IHALT,
-	"ADD": IADD,
-	"SUB": ISUBTRACT,
-	"STA": ISTORE,
-	"LDA": ILOAD,
-	"BRA": IBRANCH,
-	"BRZ": IBRANCH_IF_ZERO,
-	"BRP": IBRANCH_IF_POSITIVE,
-	"INP": IINPUT,
-	"OUT": IOUTPUT,
-	"DAT": IDATA,
-}
-
 type instruction struct {
 	id     int    // What instruction this is
 	target string // Label name
 	number int    // Immediate value
 }
 
-// -----------------------------------------------------------------------------
-
 type assembler struct {
 	tokens chan token     // Where tokens come from
 	output []instruction  // The assembled program
@@ -236,37 +247,22 @@ func (a *assembler) step() (bool, error) {
 	}
 
 	// TODO: add token location information to returned errors
-
 	switch token.kind {
 	case WORD:
-		canonical := strings.ToUpper(token.value)
+		if _, dup := a.labels[token.value]; dup {
-		instr, found := instructions[canonical]
+			return false, fmt.Errorf("Duplicate label: %s", token.value)
-
-		// Not found in the instruction list
-		// Assume it is a label
-		if !found {
-			if _, dup := a.labels[canonical]; dup {
-				return false, fmt.Errorf("Duplicate label: %s", canonical)
-			}
-			a.labels[canonical] = len(a.output)
-
-			token, ok = <-a.tokens
-			if !ok {
-				return false, errors.New("Unexpected end of file")
-			}
-			if token.kind != WORD {
-				return false, errors.New("Expected word")
-			}
-
-			// XXX: it might be better to classify this in the lexer
-			canonical = strings.ToUpper(token.value)
-			instr, found = instructions[canonical]
-		}
-		if !found {
-			return false, fmt.Errorf("Unknown instruction: %s", canonical)
 		}
+		a.labels[token.value] = len(a.output)
 
-		instrHolder := instruction{id: instr}
+		if token, ok = <-a.tokens; !ok {
+			return false, errors.New("Unexpected end of file")
+		}
+		if token.kind != INSTRUCTION {
+			return false, errors.New("Expected instruction name after label")
+		}
+		fallthrough
+	case INSTRUCTION:
+		instrHolder := instruction{id: token.instruction}
 
 		token, ok := <-a.tokens
 		eol := false
@@ -274,6 +270,7 @@ func (a *assembler) step() (bool, error) {
 		case token.kind == WORD:
 			instrHolder.target = strings.ToUpper(token.value)
 		case token.kind == NUMBER:
+			// TODO: we should check the number
 			instrHolder.number, _ = strconv.Atoi(token.value)
 		case token.kind == ERROR:
 			return false, errors.New(token.value)
@@ -289,9 +286,7 @@ func (a *assembler) step() (bool, error) {
 			token, ok := <-a.tokens
 			switch {
 			case !ok:
-				break
 			case token.kind == NEWLINE:
-				break
 			case token.kind == ERROR:
 				return false, errors.New(token.value)
 			default:
@@ -313,11 +308,9 @@ func Assemble(r io.Reader) (code []int16, err error) {
 	go tokenize(r, a.tokens)
 
 	for {
-		cont, err := a.step()
+		if cont, err := a.step(); err != nil {
-		if err != nil {
 			return nil, err
-		}
+		} else if !cont {
-		if !cont {
 			break
 		}
 	}
@@ -327,17 +320,10 @@ func Assemble(r io.Reader) (code []int16, err error) {
 		if i >= len(code) {
 			return nil, errors.New("Program too long")
 		}
-		n := x.id * 100
+		n := x.id
-		// XXX: this also stinks
-		if x.id == IDATA {
-			n = 0
-		}
-		// XXX: we should be able to handle the strange INP and OUT better
 		switch {
-		case x.id == IINPUT:
+		case x.id%100 != 0:
-			n = 901
+			// TODO: we could complain that arguments aren't allowed
-		case x.id == IOUTPUT:
-			n = 902
 		case len(x.target) != 0:
 			// Resolve targets to code locations
 			if resolved, ok := a.labels[x.target]; !ok {

machine.go

@@ -20,39 +20,41 @@ func Run(code []int16) error {
 	pc := 0
 	for pc >= 0 && pc < len(code) {
 		i := code[pc]
+		arg := i % 100
 		pc++
-		switch i / 100 {
+
+		switch i - arg {
 		case IHALT:
 			return nil
 		case IADD:
-			accumulator += code[i%100]
+			accumulator += code[arg]
 		case ISUBTRACT:
-			accumulator -= code[i%100]
+			accumulator -= code[arg]
 		case ISTORE:
-			code[i%100] = accumulator
+			code[arg] = accumulator
 		case ILOAD:
-			accumulator = code[i%100]
+			accumulator = code[arg]
 		case IBRANCH:
 			pc = int(i % 100)
 		case IBRANCH_IF_ZERO:
 			if accumulator == 0 {
-				pc = int(i % 100)
+				pc = int(arg)
 			}
 		case IBRANCH_IF_POSITIVE:
 			if accumulator > 0 {
-				pc = int(i % 100)
+				pc = int(arg)
 			}
-		default:
+		case IIO:
-			switch i {
+			switch arg {
-			case 901:
+			case IO_INPUT:
 				fmt.Printf("Input: ")
 				fmt.Scanf("%d\n", &accumulator)
-			case 902:
+			case IO_OUTPUT:
 				fmt.Printf("Output: %d\n", accumulator)
-			default:
-				e := fmt.Sprintf("Unsupported instruction %d at %d", i, pc)
-				return errors.New(e)
 			}
+		default:
+			e := fmt.Sprintf("Unsupported instruction %d at %d", i, pc)
+			return errors.New(e)
 		}
 	}
 	return errors.New("Program counter ran away")