Scanning:
An analogous set of functions scans formatted text to yield
- values. Scan and Scanln read from os.Stdin; Fscan and Fscanln
- read from a specified os.Reader; Sscan and Sscanln read from
- an argument string. By default, tokens are separated by
- spaces. Sscanln, Fscanln and Sscanln stop scanning at a
- newline and require that the items be followed by one; the
- other routines treat newlines as spaces.
+ values. Scan and Scanln read from os.Stdin; Fscan and
+ Fscanln read from a specified os.Reader; Sscan and Sscanln
+ read from an argument string. Sscanln, Fscanln and Sscanln
+ stop scanning at a newline and require that the items be
+ followed by one; the other routines treat newlines as spaces.
Scanf, Fscanf, and Sscanf parse the arguments according to a
format string, analogous to that of Printf. For example, "%x"
%T is not implemented
%e %E %f %F %g %g are all equivalent and scan any floating
point or complex value
+ %s and %v on strings scan a space-delimited token
+
+ Width is interpreted in the input text (%5s means at most
+ five runes of input will be read to scan a string) but there
+ is no syntax for scanning with a precision (no %5.2f, just
+ %5f).
When scanning with a format, all non-empty runs of space
characters (including newline) are equivalent to a single
*/
package fmt
-// BUG: format precision and flags are not yet implemented for scanning.
-
import (
"bytes"
"io"
GetRune() (rune int, err os.Error)
// UngetRune causes the next call to GetRune to return the rune.
UngetRune(rune int)
- // Token returns the next space-delimited token from the input.
+ // Width returns the value of the width option and whether it has been set.
+ // The unit is Unicode code points.
+ Width() (wid int, ok bool)
+ // Token returns the next space-delimited token from the input. If
+ // a width has been specified, the returned token will be no longer
+ // than the width.
Token() (token string, err os.Error)
}
// receiver, which must be a pointer to be useful. The Scan method is called
// for any argument to Scan or Scanln that implements it.
type Scanner interface {
- Scan(ScanState) os.Error
+ Scan(state ScanState, verb int) os.Error
}
// Scan scans text read from standard input, storing successive
// ss is the internal implementation of ScanState.
type ss struct {
- rr readRuner // where to read input
- buf bytes.Buffer // token accumulator
- nlIsSpace bool // whether newline counts as white space
- peekRune int // one-rune lookahead
+ rr readRuner // where to read input
+ buf bytes.Buffer // token accumulator
+ nlIsSpace bool // whether newline counts as white space
+ peekRune int // one-rune lookahead
+ maxWid int // max width of field, in runes
+ widPresent bool // width was specified
+ wid int // width consumed so far; used in accept()
}
func (s *ss) GetRune() (rune int, err os.Error) {
return
}
+func (s *ss) Width() (wid int, ok bool) {
+ return s.maxWid, s.widPresent
+}
+
const EOF = -1
// The public method returns an error; this private one panics.
}
s.nlIsSpace = nlIsSpace
s.peekRune = -1
+ s.maxWid = 0
+ s.widPresent = false
return s
}
// skipSpace skips spaces and maybe newlines
func (s *ss) skipSpace() {
- s.buf.Reset()
for {
rune := s.getRune()
if rune == EOF {
}
}
-// token returns the next space-delimited string from the input.
-// For Scanln, it stops at newlines. For Scan, newlines are treated as
-// spaces.
+// token returns the next space-delimited string from the input. It
+// skips white space. For Scanln, it stops at newlines. For Scan,
+// newlines are treated as spaces.
func (s *ss) token() string {
s.skipSpace()
// read until white space or newline
- for {
+ for nrunes := 0; !s.widPresent || nrunes < s.maxWid; nrunes++ {
rune := s.getRune()
if rune == EOF {
break
var intBits = uint(reflect.Typeof(int(0)).Size() * 8)
var uintptrBits = uint(reflect.Typeof(int(0)).Size() * 8)
var complexError = os.ErrorString("syntax error scanning complex number")
+var boolError = os.ErrorString("syntax error scanning boolean")
+
+// accepts checks the next rune in the input. If it's a byte (sic) in the string, it puts it in the
+// buffer and returns true. Otherwise it return false.
+func (s *ss) accept(ok string) bool {
+ if s.wid >= s.maxWid {
+ return false
+ }
+ rune := s.getRune()
+ if rune == EOF {
+ return false
+ }
+ for i := 0; i < len(ok); i++ {
+ if int(ok[i]) == rune {
+ s.buf.WriteRune(rune)
+ s.wid++
+ return true
+ }
+ }
+ if rune != EOF {
+ s.UngetRune(rune)
+ }
+ return false
+}
// okVerb verifies that the verb is present in the list, setting s.err appropriately if not.
func (s *ss) okVerb(verb int, okVerbs, typ string) bool {
if !s.okVerb(verb, "tv", "boolean") {
return false
}
- tok := s.token()
- b, err := strconv.Atob(tok)
- if err != nil {
- s.error(err)
+ // Syntax-checking a boolean is annoying. We're not fastidious about case.
+ switch s.mustGetRune() {
+ case '0':
+ return false
+ case '1':
+ return true
+ case 't', 'T':
+ if s.accept("rR") && (!s.accept("uU") || !s.accept("eE")) {
+ s.error(boolError)
+ }
+ return true
+ case 'f', 'F':
+ if s.accept("aL") && (!s.accept("lL") || !s.accept("sS") || !s.accept("eE")) {
+ s.error(boolError)
+ }
+ return false
}
- return b
+ return false
}
-// getBase returns the numeric base represented by the verb.
-func (s *ss) getBase(verb int) int {
+// Numerical elements
+const (
+ binaryDigits = "01"
+ octalDigits = "01234567"
+ decimalDigits = "0123456789"
+ hexadecimalDigits = "0123456789aAbBcCdDeEfF"
+ sign = "+-"
+ period = "."
+ exponent = "eE"
+)
+
+// getBase returns the numeric base represented by the verb and its digit string.
+func (s *ss) getBase(verb int) (base int, digits string) {
s.okVerb(verb, "bdoxXv", "integer") // sets s.err
- base := 10
+ base = 10
+ digits = decimalDigits
switch verb {
case 'b':
base = 2
+ digits = binaryDigits
case 'o':
base = 8
+ digits = octalDigits
case 'x', 'X':
base = 16
+ digits = hexadecimalDigits
}
- return base
+ return
+}
+
+// scanNumber returns the numerical string with specified digits starting here.
+func (s *ss) scanNumber(digits string) string {
+ if !s.accept(digits) {
+ s.errorString("expected integer")
+ }
+ for s.accept(digits) {
+ }
+ return s.buf.String()
}
// scanInt returns the value of the integer represented by the next
// token, checking for overflow. Any error is stored in s.err.
func (s *ss) scanInt(verb int, bitSize uint) int64 {
- base := s.getBase(verb)
- tok := s.token()
+ base, digits := s.getBase(verb)
+ s.skipSpace()
+ s.accept(sign) // If there's a sign, it will be left in the token buffer.
+ tok := s.scanNumber(digits)
i, err := strconv.Btoi64(tok, base)
if err != nil {
s.error(err)
// scanUint returns the value of the unsigned integer represented
// by the next token, checking for overflow. Any error is stored in s.err.
func (s *ss) scanUint(verb int, bitSize uint) uint64 {
- base := s.getBase(verb)
- tok := s.token()
+ base, digits := s.getBase(verb)
+ s.skipSpace()
+ tok := s.scanNumber(digits)
i, err := strconv.Btoui64(tok, base)
if err != nil {
s.error(err)
return i
}
-// complexParts returns the strings representing the real and imaginary parts of the string.
-func (s *ss) complexParts(str string) (real, imag string) {
- if len(str) > 2 && str[0] == '(' && str[len(str)-1] == ')' {
- str = str[1 : len(str)-1]
- }
- real, str = floatPart(str)
- // Must now have a sign.
- if len(str) == 0 || (str[0] != '+' && str[0] != '-') {
- s.error(complexError)
- }
- imag, str = floatPart(str)
- if str != "i" {
- s.error(complexError)
- }
- return real, imag
-}
-
-// floatPart returns strings holding the floating point value in the string, followed
-// by the remainder of the string. That is, it splits str into (number,rest-of-string).
-func floatPart(str string) (first, last string) {
- i := 0
+// floatToken returns the floating-point number starting here, no longer than swid
+// if the width is specified. It's not rigorous about syntax because it doesn't check that
+// we have at least some digits, but Atof will do that.
+func (s *ss) floatToken() string {
+ s.buf.Reset()
// leading sign?
- if len(str) > i && (str[0] == '+' || str[0] == '-') {
- i++
- }
+ s.accept(sign)
// digits?
- for len(str) > i && '0' <= str[i] && str[i] <= '9' {
- i++
+ for s.accept(decimalDigits) {
}
- // period?
- if str[i] == '.' {
- i++
- }
- // fraction?
- for len(str) > i && '0' <= str[i] && str[i] <= '9' {
- i++
+ // decimal point?
+ if s.accept(period) {
+ // fraction?
+ for s.accept(decimalDigits) {
+ }
}
// exponent?
- if len(str) > i && (str[i] == 'e' || str[i] == 'E') {
- i++
+ if s.accept(exponent) {
// leading sign?
- if str[i] == '+' || str[i] == '-' {
- i++
- }
+ s.accept(sign)
// digits?
- for len(str) > i && '0' <= str[i] && str[i] <= '9' {
- i++
+ for s.accept(decimalDigits) {
}
}
- return str[0:i], str[i:]
+ return s.buf.String()
+}
+
+// complexTokens returns the real and imaginary parts of the complex number starting here.
+// The number might be parenthesized and has the format (N+Ni) where N is a floating-point
+// number and there are no spaces within.
+func (s *ss) complexTokens() (real, imag string) {
+ // TODO: accept N and Ni independently?
+ parens := s.accept("(")
+ real = s.floatToken()
+ s.buf.Reset()
+ // Must now have a sign.
+ if !s.accept("+-") {
+ s.error(complexError)
+ }
+ // Sign is now in buffer
+ imagSign := s.buf.String()
+ imag = s.floatToken()
+ if !s.accept("i") {
+ s.error(complexError)
+ }
+ if parens && !s.accept(")") {
+ s.error(complexError)
+ }
+ return real, imagSign + imag
}
// convertFloat converts the string to a float value.
if !s.okVerb(verb, floatVerbs, "complex") {
return 0
}
- tok := s.token()
- sreal, simag := s.complexParts(tok)
+ s.skipSpace()
+ sreal, simag := s.complexTokens()
real := atof(s, sreal)
imag := atof(s, simag)
return cmplx(real, imag)
return s.token() // %s and %v just return the next word
}
-// quotedString returns the double- or back-quoted string.
+// quotedString returns the double- or back-quoted string represented by the next input characters.
func (s *ss) quotedString() string {
quote := s.mustGetRune()
switch quote {
// scanOne scans a single value, deriving the scanner from the type of the argument.
func (s *ss) scanOne(verb int, field interface{}) {
+ s.buf.Reset()
var err os.Error
// If the parameter has its own Scan method, use that.
if v, ok := field.(Scanner); ok {
- err = v.Scan(s)
+ err = v.Scan(s, verb)
if err != nil {
s.error(err)
}
return
}
+ if !s.widPresent {
+ s.maxWid = 1 << 30 // Huge
+ }
+ s.wid = 0
switch v := field.(type) {
case *bool:
*v = s.scanBool(verb)
// scan in high precision and convert, in order to preserve the correct error condition.
case *float:
if s.okVerb(verb, floatVerbs, "float") {
- *v = float(s.convertFloat(s.token()))
+ s.skipSpace()
+ *v = float(s.convertFloat(s.floatToken()))
}
case *float32:
if s.okVerb(verb, floatVerbs, "float32") {
- *v = float32(s.convertFloat32(s.token()))
+ s.skipSpace()
+ *v = float32(s.convertFloat32(s.floatToken()))
}
case *float64:
if s.okVerb(verb, floatVerbs, "float64") {
- *v = s.convertFloat64(s.token())
+ s.skipSpace()
+ *v = s.convertFloat64(s.floatToken())
}
case *string:
*v = s.convertString(verb)
v.Elem(i).(*reflect.Uint8Value).Set(str[i])
}
case *reflect.FloatValue:
- v.Set(float(s.convertFloat(s.token())))
+ s.skipSpace()
+ v.Set(float(s.convertFloat(s.floatToken())))
case *reflect.Float32Value:
- v.Set(float32(s.convertFloat(s.token())))
+ s.skipSpace()
+ v.Set(float32(s.convertFloat(s.floatToken())))
case *reflect.Float64Value:
- v.Set(s.convertFloat(s.token()))
+ s.skipSpace()
+ v.Set(s.convertFloat(s.floatToken()))
case *reflect.ComplexValue:
v.Set(complex(s.scanComplex(verb, (*ss).convertFloat)))
case *reflect.Complex64Value:
}
i++ // % is one byte
- // TODO: FLAGS
+ // do we have 20 (width)?
+ s.maxWid, s.widPresent, i = parsenum(format, i, end)
+
c, w := utf8.DecodeRuneInString(format[i:])
i += w
s.scanOne(c, field)
numProcessed++
}
+ if numProcessed < len(a) {
+ s.errorString("too many operands")
+ }
return
}
out interface{}
}
+type ScanfMultiTest struct {
+ format string
+ text string
+ in []interface{}
+ out []interface{}
+ err string
+}
+
type (
renamedBool bool
renamedInt int
float32Val float32
float64Val float64
stringVal string
+ stringVal1 string
bytesVal []byte
complexVal complex
complex64Val complex64
renamedComplex128Val renamedComplex128
)
-// Xs accepts any non-empty run of x's.
-var xPat = testing.MustCompile("x+")
-
+// Xs accepts any non-empty run of the verb character
type Xs string
-func (x *Xs) Scan(state ScanState) os.Error {
- tok, err := state.Token()
+func (x *Xs) Scan(state ScanState, verb int) os.Error {
+ var tok string
+ var c int
+ var err os.Error
+ wid, present := state.Width()
+ if !present {
+ tok, err = state.Token()
+ } else {
+ for i := 0; i < wid; i++ {
+ c, err = state.GetRune()
+ if err != nil {
+ break
+ }
+ tok += string(c)
+ }
+ }
if err != nil {
return err
}
- if !xPat.MatchString(tok) {
+ if !testing.MustCompile(string(verb) + "+").MatchString(tok) {
return os.ErrorString("syntax error for xs")
}
*x = Xs(tok)
ScanTest{"115\n", &renamedBytesVal, renamedBytes([]byte("115"))},
// Custom scanner.
- ScanTest{" xxx ", &xVal, Xs("xxx")},
+ ScanTest{" vvv ", &xVal, Xs("vvv")},
}
var scanfTests = []ScanfTest{
ScanfTest{"%v", "-71\n", &intVal, -71},
ScanfTest{"%d", "72\n", &intVal, 72},
ScanfTest{"%d", "73\n", &int8Val, int8(73)},
- ScanfTest{"%d", "-74\n", &int16Val, int16(-74)},
+ ScanfTest{"%d", "+74\n", &int16Val, int16(74)},
ScanfTest{"%d", "75\n", &int32Val, int32(75)},
ScanfTest{"%d", "76\n", &int64Val, int64(76)},
ScanfTest{"%b", "1001001\n", &intVal, 73},
ScanfTest{"here is\tthe value:%d", "here is the\tvalue:118\n", &intVal, 118},
ScanfTest{"%% %%:%d", "% %:119\n", &intVal, 119},
+ // Corner cases
ScanfTest{"%x", "FFFFFFFF\n", &uint32Val, uint32(0xFFFFFFFF)},
+
+ // Custom scanner.
+ ScanfTest{"%s", " sss ", &xVal, Xs("sss")},
+ ScanfTest{"%2s", "sssss", &xVal, Xs("ss")},
}
var overflowTests = []ScanTest{
ScanTest{"(1-1e500i)", &complex128Val, 0},
}
+var i, j, k int
+var f float
+var s, t string
+var c complex
+var x, y Xs
+
+func args(a ...interface{}) []interface{} { return a }
+
+var multiTests = []ScanfMultiTest{
+ ScanfMultiTest{"", "", nil, nil, ""},
+ ScanfMultiTest{"%d", "23", args(&i), args(23), ""},
+ ScanfMultiTest{"%2s%3s", "22333", args(&s, &t), args("22", "333"), ""},
+ ScanfMultiTest{"%2d%3d", "44555", args(&i, &j), args(44, 555), ""},
+ ScanfMultiTest{"%2d.%3d", "66.777", args(&i, &j), args(66, 777), ""},
+ ScanfMultiTest{"%d, %d", "23, 18", args(&i, &j), args(23, 18), ""},
+ ScanfMultiTest{"%3d22%3d", "33322333", args(&i, &j), args(333, 333), ""},
+ ScanfMultiTest{"%6vX=%3fY", "3+2iX=2.5Y", args(&c, &f), args((3 + 2i), float(2.5)), ""},
+ ScanfMultiTest{"%d%s", "123abc", args(&i, &s), args(123, "abc"), ""},
+
+ // Custom scanner.
+ ScanfMultiTest{"%2e%f", "eefffff", []interface{}{&x, &y}, []interface{}{Xs("ee"), Xs("fffff")}, ""},
+
+ // Errors
+ ScanfMultiTest{"%t", "23 18", []interface{}{&i}, nil, "bad verb"},
+ ScanfMultiTest{"%d %d %d", "23 18", []interface{}{&i, &j}, []interface{}{23, 18}, "too few operands"},
+ ScanfMultiTest{"%d %d", "23 18 27", []interface{}{&i, &j, &k}, []interface{}{23, 18}, "too many operands"},
+}
+
func testScan(t *testing.T, scan func(r io.Reader, a ...interface{}) (int, os.Error)) {
for _, test := range scanTests {
r := strings.NewReader(test.text)
}
}
+// TODO: there's no conversion from []T to ...T, but we can fake it. These
+// functions do the faking. We index the table by the length of the param list.
+var scanf = []func(string, string, []interface{}) (int, os.Error){
+ 0: func(s, f string, i []interface{}) (int, os.Error) { return Sscanf(s, f) },
+ 1: func(s, f string, i []interface{}) (int, os.Error) { return Sscanf(s, f, i[0]) },
+ 2: func(s, f string, i []interface{}) (int, os.Error) { return Sscanf(s, f, i[0], i[1]) },
+ 3: func(s, f string, i []interface{}) (int, os.Error) { return Sscanf(s, f, i[0], i[1], i[2]) },
+}
+
+func TestScanfMulti(t *testing.T) {
+ sliceType := reflect.Typeof(make([]interface{}, 1)).(*reflect.SliceType)
+ for _, test := range multiTests {
+ n, err := scanf[len(test.in)](test.text, test.format, test.in)
+ if err != nil {
+ if test.err == "" {
+ t.Errorf("got error scanning (%q, %q): %q", test.format, test.text, err)
+ } else if strings.Index(err.String(), test.err) < 0 {
+ t.Errorf("got wrong error scanning (%q, %q): %q; expected %q", test.format, test.text, err, test.err)
+ }
+ continue
+ }
+ if test.err != "" {
+ t.Errorf("expected error %q error scanning (%q, %q)", test.err, test.format, test.text)
+ }
+ if n != len(test.out) {
+ t.Errorf("count error on entry (%q, %q): expected %d got %d", test.format, test.text, len(test.out), n)
+ continue
+ }
+ // Convert the slice of pointers into a slice of values
+ resultVal := reflect.MakeSlice(sliceType, n, n)
+ for i := 0; i < n; i++ {
+ v := reflect.NewValue(test.in[i]).(*reflect.PtrValue).Elem()
+ resultVal.Elem(i).(*reflect.InterfaceValue).Set(v)
+ }
+ result := resultVal.Interface()
+ if !reflect.DeepEqual(result, test.out) {
+ t.Errorf("scanning (%q, %q): expected %v got %v", test.format, test.text, test.out, result)
+ }
+ }
+}
+
func TestScanMultiple(t *testing.T) {
- text := "1 2 3"
- r := strings.NewReader(text)
- var a, b, c, d int
- n, err := Fscan(r, &a, &b, &c)
- if n != 3 {
- t.Errorf("Fscan count error: expected 3: got %d", n)
+ var a int
+ var s string
+ n, err := Sscan("123abc", &a, &s)
+ if n != 2 {
+ t.Errorf("Sscan count error: expected 2: got %d", n)
}
if err != nil {
- t.Errorf("Fscan expected no error scanning %q; got %s", text, err)
- }
- text = "1 2 3 x"
- r = strings.NewReader(text)
- n, err = Fscan(r, &a, &b, &c, &d)
- if n != 3 {
- t.Errorf("Fscan count error: expected 3: got %d", n)
- }
- if err == nil {
- t.Errorf("Fscan expected error scanning %q", text)
+ t.Errorf("Sscan expected no error; got %s", err)
}
- text = "1 2 3 x"
- r = strings.NewReader(text)
- n, err = Fscanf(r, "%d %d %d\n", &a, &b, &c, &d)
- if n != 3 {
- t.Errorf("Fscanf count error: expected 3: got %d", n)
- }
- text = "1 2"
- r = strings.NewReader(text)
- n, err = Fscanf(r, "%d %d %d\n", &a, &b, &c, &d)
- if n != 2 {
- t.Errorf("Fscanf count error: expected 2: got %d", n)
- }
- if err == nil {
- t.Errorf("Fscanf expected error scanning %q", text)
+ if a != 123 || s != "abc" {
+ t.Errorf("Sscan wrong values: got (%d %q) expected (123 \"abc\")", a, s)
}
}