Source file src/fmt/format.go

     1  // Copyright 2009 The Go Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  package fmt
     6  
     7  import (
     8  	"strconv"
     9  	"unicode/utf8"
    10  )
    11  
    12  const (
    13  	ldigits = "0123456789abcdefx"
    14  	udigits = "0123456789ABCDEFX"
    15  )
    16  
    17  const (
    18  	signed   = true
    19  	unsigned = false
    20  )
    21  
    22  // flags placed in a separate struct for easy clearing.
    23  type fmtFlags struct {
    24  	widPresent  bool
    25  	precPresent bool
    26  	minus       bool
    27  	plus        bool
    28  	sharp       bool
    29  	space       bool
    30  	zero        bool
    31  
    32  	// For the formats %+v %#v, we set the plusV/sharpV flags
    33  	// and clear the plus/sharp flags since %+v and %#v are in effect
    34  	// different, flagless formats set at the top level.
    35  	plusV  bool
    36  	sharpV bool
    37  }
    38  
    39  // A fmt is the raw formatter used by Printf etc.
    40  // It prints into a buffer that must be set up separately.
    41  type fmt struct {
    42  	buf *buffer
    43  
    44  	fmtFlags
    45  
    46  	wid  int // width
    47  	prec int // precision
    48  
    49  	// intbuf is large enough to store %b of an int64 with a sign and
    50  	// avoids padding at the end of the struct on 32 bit architectures.
    51  	intbuf [68]byte
    52  }
    53  
    54  func (f *fmt) clearflags() {
    55  	f.fmtFlags = fmtFlags{}
    56  }
    57  
    58  func (f *fmt) init(buf *buffer) {
    59  	f.buf = buf
    60  	f.clearflags()
    61  }
    62  
    63  // writePadding generates n bytes of padding.
    64  func (f *fmt) writePadding(n int) {
    65  	if n <= 0 { // No padding bytes needed.
    66  		return
    67  	}
    68  	buf := *f.buf
    69  	oldLen := len(buf)
    70  	newLen := oldLen + n
    71  	// Make enough room for padding.
    72  	if newLen > cap(buf) {
    73  		buf = make(buffer, cap(buf)*2+n)
    74  		copy(buf, *f.buf)
    75  	}
    76  	// Decide which byte the padding should be filled with.
    77  	padByte := byte(' ')
    78  	if f.zero {
    79  		padByte = byte('0')
    80  	}
    81  	// Fill padding with padByte.
    82  	padding := buf[oldLen:newLen]
    83  	for i := range padding {
    84  		padding[i] = padByte
    85  	}
    86  	*f.buf = buf[:newLen]
    87  }
    88  
    89  // pad appends b to f.buf, padded on left (!f.minus) or right (f.minus).
    90  func (f *fmt) pad(b []byte) {
    91  	if !f.widPresent || f.wid == 0 {
    92  		f.buf.write(b)
    93  		return
    94  	}
    95  	width := f.wid - utf8.RuneCount(b)
    96  	if !f.minus {
    97  		// left padding
    98  		f.writePadding(width)
    99  		f.buf.write(b)
   100  	} else {
   101  		// right padding
   102  		f.buf.write(b)
   103  		f.writePadding(width)
   104  	}
   105  }
   106  
   107  // padString appends s to f.buf, padded on left (!f.minus) or right (f.minus).
   108  func (f *fmt) padString(s string) {
   109  	if !f.widPresent || f.wid == 0 {
   110  		f.buf.writeString(s)
   111  		return
   112  	}
   113  	width := f.wid - utf8.RuneCountInString(s)
   114  	if !f.minus {
   115  		// left padding
   116  		f.writePadding(width)
   117  		f.buf.writeString(s)
   118  	} else {
   119  		// right padding
   120  		f.buf.writeString(s)
   121  		f.writePadding(width)
   122  	}
   123  }
   124  
   125  // fmtBoolean formats a boolean.
   126  func (f *fmt) fmtBoolean(v bool) {
   127  	if v {
   128  		f.padString("true")
   129  	} else {
   130  		f.padString("false")
   131  	}
   132  }
   133  
   134  // fmtUnicode formats a uint64 as "U+0078" or with f.sharp set as "U+0078 'x'".
   135  func (f *fmt) fmtUnicode(u uint64) {
   136  	buf := f.intbuf[0:]
   137  
   138  	// With default precision set the maximum needed buf length is 18
   139  	// for formatting -1 with %#U ("U+FFFFFFFFFFFFFFFF") which fits
   140  	// into the already allocated intbuf with a capacity of 68 bytes.
   141  	prec := 4
   142  	if f.precPresent && f.prec > 4 {
   143  		prec = f.prec
   144  		// Compute space needed for "U+" , number, " '", character, "'".
   145  		width := 2 + prec + 2 + utf8.UTFMax + 1
   146  		if width > len(buf) {
   147  			buf = make([]byte, width)
   148  		}
   149  	}
   150  
   151  	// Format into buf, ending at buf[i]. Formatting numbers is easier right-to-left.
   152  	i := len(buf)
   153  
   154  	// For %#U we want to add a space and a quoted character at the end of the buffer.
   155  	if f.sharp && u <= utf8.MaxRune && strconv.IsPrint(rune(u)) {
   156  		i--
   157  		buf[i] = '\''
   158  		i -= utf8.RuneLen(rune(u))
   159  		utf8.EncodeRune(buf[i:], rune(u))
   160  		i--
   161  		buf[i] = '\''
   162  		i--
   163  		buf[i] = ' '
   164  	}
   165  	// Format the Unicode code point u as a hexadecimal number.
   166  	for u >= 16 {
   167  		i--
   168  		buf[i] = udigits[u&0xF]
   169  		prec--
   170  		u >>= 4
   171  	}
   172  	i--
   173  	buf[i] = udigits[u]
   174  	prec--
   175  	// Add zeros in front of the number until requested precision is reached.
   176  	for prec > 0 {
   177  		i--
   178  		buf[i] = '0'
   179  		prec--
   180  	}
   181  	// Add a leading "U+".
   182  	i--
   183  	buf[i] = '+'
   184  	i--
   185  	buf[i] = 'U'
   186  
   187  	oldZero := f.zero
   188  	f.zero = false
   189  	f.pad(buf[i:])
   190  	f.zero = oldZero
   191  }
   192  
   193  // fmtInteger formats signed and unsigned integers.
   194  func (f *fmt) fmtInteger(u uint64, base int, isSigned bool, verb rune, digits string) {
   195  	negative := isSigned && int64(u) < 0
   196  	if negative {
   197  		u = -u
   198  	}
   199  
   200  	buf := f.intbuf[0:]
   201  	// The already allocated f.intbuf with a capacity of 68 bytes
   202  	// is large enough for integer formatting when no precision or width is set.
   203  	if f.widPresent || f.precPresent {
   204  		// Account 3 extra bytes for possible addition of a sign and "0x".
   205  		width := 3 + f.wid + f.prec // wid and prec are always positive.
   206  		if width > len(buf) {
   207  			// We're going to need a bigger boat.
   208  			buf = make([]byte, width)
   209  		}
   210  	}
   211  
   212  	// Two ways to ask for extra leading zero digits: %.3d or %03d.
   213  	// If both are specified the f.zero flag is ignored and
   214  	// padding with spaces is used instead.
   215  	prec := 0
   216  	if f.precPresent {
   217  		prec = f.prec
   218  		// Precision of 0 and value of 0 means "print nothing" but padding.
   219  		if prec == 0 && u == 0 {
   220  			oldZero := f.zero
   221  			f.zero = false
   222  			f.writePadding(f.wid)
   223  			f.zero = oldZero
   224  			return
   225  		}
   226  	} else if f.zero && f.widPresent {
   227  		prec = f.wid
   228  		if negative || f.plus || f.space {
   229  			prec-- // leave room for sign
   230  		}
   231  	}
   232  
   233  	// Because printing is easier right-to-left: format u into buf, ending at buf[i].
   234  	// We could make things marginally faster by splitting the 32-bit case out
   235  	// into a separate block but it's not worth the duplication, so u has 64 bits.
   236  	i := len(buf)
   237  	// Use constants for the division and modulo for more efficient code.
   238  	// Switch cases ordered by popularity.
   239  	switch base {
   240  	case 10:
   241  		for u >= 10 {
   242  			i--
   243  			next := u / 10
   244  			buf[i] = byte('0' + u - next*10)
   245  			u = next
   246  		}
   247  	case 16:
   248  		for u >= 16 {
   249  			i--
   250  			buf[i] = digits[u&0xF]
   251  			u >>= 4
   252  		}
   253  	case 8:
   254  		for u >= 8 {
   255  			i--
   256  			buf[i] = byte('0' + u&7)
   257  			u >>= 3
   258  		}
   259  	case 2:
   260  		for u >= 2 {
   261  			i--
   262  			buf[i] = byte('0' + u&1)
   263  			u >>= 1
   264  		}
   265  	default:
   266  		panic("fmt: unknown base; can't happen")
   267  	}
   268  	i--
   269  	buf[i] = digits[u]
   270  	for i > 0 && prec > len(buf)-i {
   271  		i--
   272  		buf[i] = '0'
   273  	}
   274  
   275  	// Various prefixes: 0x, -, etc.
   276  	if f.sharp {
   277  		switch base {
   278  		case 2:
   279  			// Add a leading 0b.
   280  			i--
   281  			buf[i] = 'b'
   282  			i--
   283  			buf[i] = '0'
   284  		case 8:
   285  			if buf[i] != '0' {
   286  				i--
   287  				buf[i] = '0'
   288  			}
   289  		case 16:
   290  			// Add a leading 0x or 0X.
   291  			i--
   292  			buf[i] = digits[16]
   293  			i--
   294  			buf[i] = '0'
   295  		}
   296  	}
   297  	if verb == 'O' {
   298  		i--
   299  		buf[i] = 'o'
   300  		i--
   301  		buf[i] = '0'
   302  	}
   303  
   304  	if negative {
   305  		i--
   306  		buf[i] = '-'
   307  	} else if f.plus {
   308  		i--
   309  		buf[i] = '+'
   310  	} else if f.space {
   311  		i--
   312  		buf[i] = ' '
   313  	}
   314  
   315  	// Left padding with zeros has already been handled like precision earlier
   316  	// or the f.zero flag is ignored due to an explicitly set precision.
   317  	oldZero := f.zero
   318  	f.zero = false
   319  	f.pad(buf[i:])
   320  	f.zero = oldZero
   321  }
   322  
   323  // truncateString truncates the string s to the specified precision, if present.
   324  func (f *fmt) truncateString(s string) string {
   325  	if f.precPresent {
   326  		n := f.prec
   327  		for i := range s {
   328  			n--
   329  			if n < 0 {
   330  				return s[:i]
   331  			}
   332  		}
   333  	}
   334  	return s
   335  }
   336  
   337  // truncate truncates the byte slice b as a string of the specified precision, if present.
   338  func (f *fmt) truncate(b []byte) []byte {
   339  	if f.precPresent {
   340  		n := f.prec
   341  		for i := 0; i < len(b); {
   342  			n--
   343  			if n < 0 {
   344  				return b[:i]
   345  			}
   346  			wid := 1
   347  			if b[i] >= utf8.RuneSelf {
   348  				_, wid = utf8.DecodeRune(b[i:])
   349  			}
   350  			i += wid
   351  		}
   352  	}
   353  	return b
   354  }
   355  
   356  // fmtS formats a string.
   357  func (f *fmt) fmtS(s string) {
   358  	s = f.truncateString(s)
   359  	f.padString(s)
   360  }
   361  
   362  // fmtBs formats the byte slice b as if it was formatted as string with fmtS.
   363  func (f *fmt) fmtBs(b []byte) {
   364  	b = f.truncate(b)
   365  	f.pad(b)
   366  }
   367  
   368  // fmtSbx formats a string or byte slice as a hexadecimal encoding of its bytes.
   369  func (f *fmt) fmtSbx(s string, b []byte, digits string) {
   370  	length := len(b)
   371  	if b == nil {
   372  		// No byte slice present. Assume string s should be encoded.
   373  		length = len(s)
   374  	}
   375  	// Set length to not process more bytes than the precision demands.
   376  	if f.precPresent && f.prec < length {
   377  		length = f.prec
   378  	}
   379  	// Compute width of the encoding taking into account the f.sharp and f.space flag.
   380  	width := 2 * length
   381  	if width > 0 {
   382  		if f.space {
   383  			// Each element encoded by two hexadecimals will get a leading 0x or 0X.
   384  			if f.sharp {
   385  				width *= 2
   386  			}
   387  			// Elements will be separated by a space.
   388  			width += length - 1
   389  		} else if f.sharp {
   390  			// Only a leading 0x or 0X will be added for the whole string.
   391  			width += 2
   392  		}
   393  	} else { // The byte slice or string that should be encoded is empty.
   394  		if f.widPresent {
   395  			f.writePadding(f.wid)
   396  		}
   397  		return
   398  	}
   399  	// Handle padding to the left.
   400  	if f.widPresent && f.wid > width && !f.minus {
   401  		f.writePadding(f.wid - width)
   402  	}
   403  	// Write the encoding directly into the output buffer.
   404  	buf := *f.buf
   405  	if f.sharp {
   406  		// Add leading 0x or 0X.
   407  		buf = append(buf, '0', digits[16])
   408  	}
   409  	var c byte
   410  	for i := 0; i < length; i++ {
   411  		if f.space && i > 0 {
   412  			// Separate elements with a space.
   413  			buf = append(buf, ' ')
   414  			if f.sharp {
   415  				// Add leading 0x or 0X for each element.
   416  				buf = append(buf, '0', digits[16])
   417  			}
   418  		}
   419  		if b != nil {
   420  			c = b[i] // Take a byte from the input byte slice.
   421  		} else {
   422  			c = s[i] // Take a byte from the input string.
   423  		}
   424  		// Encode each byte as two hexadecimal digits.
   425  		buf = append(buf, digits[c>>4], digits[c&0xF])
   426  	}
   427  	*f.buf = buf
   428  	// Handle padding to the right.
   429  	if f.widPresent && f.wid > width && f.minus {
   430  		f.writePadding(f.wid - width)
   431  	}
   432  }
   433  
   434  // fmtSx formats a string as a hexadecimal encoding of its bytes.
   435  func (f *fmt) fmtSx(s, digits string) {
   436  	f.fmtSbx(s, nil, digits)
   437  }
   438  
   439  // fmtBx formats a byte slice as a hexadecimal encoding of its bytes.
   440  func (f *fmt) fmtBx(b []byte, digits string) {
   441  	f.fmtSbx("", b, digits)
   442  }
   443  
   444  // fmtQ formats a string as a double-quoted, escaped Go string constant.
   445  // If f.sharp is set a raw (backquoted) string may be returned instead
   446  // if the string does not contain any control characters other than tab.
   447  func (f *fmt) fmtQ(s string) {
   448  	s = f.truncateString(s)
   449  	if f.sharp && strconv.CanBackquote(s) {
   450  		f.padString("`" + s + "`")
   451  		return
   452  	}
   453  	buf := f.intbuf[:0]
   454  	if f.plus {
   455  		f.pad(strconv.AppendQuoteToASCII(buf, s))
   456  	} else {
   457  		f.pad(strconv.AppendQuote(buf, s))
   458  	}
   459  }
   460  
   461  // fmtC formats an integer as a Unicode character.
   462  // If the character is not valid Unicode, it will print '\ufffd'.
   463  func (f *fmt) fmtC(c uint64) {
   464  	r := rune(c)
   465  	if c > utf8.MaxRune {
   466  		r = utf8.RuneError
   467  	}
   468  	buf := f.intbuf[:0]
   469  	w := utf8.EncodeRune(buf[:utf8.UTFMax], r)
   470  	f.pad(buf[:w])
   471  }
   472  
   473  // fmtQc formats an integer as a single-quoted, escaped Go character constant.
   474  // If the character is not valid Unicode, it will print '\ufffd'.
   475  func (f *fmt) fmtQc(c uint64) {
   476  	r := rune(c)
   477  	if c > utf8.MaxRune {
   478  		r = utf8.RuneError
   479  	}
   480  	buf := f.intbuf[:0]
   481  	if f.plus {
   482  		f.pad(strconv.AppendQuoteRuneToASCII(buf, r))
   483  	} else {
   484  		f.pad(strconv.AppendQuoteRune(buf, r))
   485  	}
   486  }
   487  
   488  // fmtFloat formats a float64. It assumes that verb is a valid format specifier
   489  // for strconv.AppendFloat and therefore fits into a byte.
   490  func (f *fmt) fmtFloat(v float64, size int, verb rune, prec int) {
   491  	// Explicit precision in format specifier overrules default precision.
   492  	if f.precPresent {
   493  		prec = f.prec
   494  	}
   495  	// Format number, reserving space for leading + sign if needed.
   496  	num := strconv.AppendFloat(f.intbuf[:1], v, byte(verb), prec, size)
   497  	if num[1] == '-' || num[1] == '+' {
   498  		num = num[1:]
   499  	} else {
   500  		num[0] = '+'
   501  	}
   502  	// f.space means to add a leading space instead of a "+" sign unless
   503  	// the sign is explicitly asked for by f.plus.
   504  	if f.space && num[0] == '+' && !f.plus {
   505  		num[0] = ' '
   506  	}
   507  	// Special handling for infinities and NaN,
   508  	// which don't look like a number so shouldn't be padded with zeros.
   509  	if num[1] == 'I' || num[1] == 'N' {
   510  		oldZero := f.zero
   511  		f.zero = false
   512  		// Remove sign before NaN if not asked for.
   513  		if num[1] == 'N' && !f.space && !f.plus {
   514  			num = num[1:]
   515  		}
   516  		f.pad(num)
   517  		f.zero = oldZero
   518  		return
   519  	}
   520  	// The sharp flag forces printing a decimal point for non-binary formats
   521  	// and retains trailing zeros, which we may need to restore.
   522  	if f.sharp && verb != 'b' {
   523  		digits := 0
   524  		switch verb {
   525  		case 'v', 'g', 'G', 'x':
   526  			digits = prec
   527  			// If no precision is set explicitly use a precision of 6.
   528  			if digits == -1 {
   529  				digits = 6
   530  			}
   531  		}
   532  
   533  		// Buffer pre-allocated with enough room for
   534  		// exponent notations of the form "e+123" or "p-1023".
   535  		var tailBuf [6]byte
   536  		tail := tailBuf[:0]
   537  
   538  		hasDecimalPoint := false
   539  		sawNonzeroDigit := false
   540  		// Starting from i = 1 to skip sign at num[0].
   541  		for i := 1; i < len(num); i++ {
   542  			switch num[i] {
   543  			case '.':
   544  				hasDecimalPoint = true
   545  			case 'p', 'P':
   546  				tail = append(tail, num[i:]...)
   547  				num = num[:i]
   548  			case 'e', 'E':
   549  				if verb != 'x' && verb != 'X' {
   550  					tail = append(tail, num[i:]...)
   551  					num = num[:i]
   552  					break
   553  				}
   554  				fallthrough
   555  			default:
   556  				if num[i] != '0' {
   557  					sawNonzeroDigit = true
   558  				}
   559  				// Count significant digits after the first non-zero digit.
   560  				if sawNonzeroDigit {
   561  					digits--
   562  				}
   563  			}
   564  		}
   565  		if !hasDecimalPoint {
   566  			// Leading digit 0 should contribute once to digits.
   567  			if len(num) == 2 && num[1] == '0' {
   568  				digits--
   569  			}
   570  			num = append(num, '.')
   571  		}
   572  		for digits > 0 {
   573  			num = append(num, '0')
   574  			digits--
   575  		}
   576  		num = append(num, tail...)
   577  	}
   578  	// We want a sign if asked for and if the sign is not positive.
   579  	if f.plus || num[0] != '+' {
   580  		// If we're zero padding to the left we want the sign before the leading zeros.
   581  		// Achieve this by writing the sign out and then padding the unsigned number.
   582  		if f.zero && f.widPresent && f.wid > len(num) {
   583  			f.buf.writeByte(num[0])
   584  			f.writePadding(f.wid - len(num))
   585  			f.buf.write(num[1:])
   586  			return
   587  		}
   588  		f.pad(num)
   589  		return
   590  	}
   591  	// No sign to show and the number is positive; just print the unsigned number.
   592  	f.pad(num[1:])
   593  }
   594  

View as plain text