...
Run Format

Source file src/bufio/scan.go

Documentation: bufio

     1  // Copyright 2013 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 bufio
     6  
     7  import (
     8  	"bytes"
     9  	"errors"
    10  	"io"
    11  	"unicode/utf8"
    12  )
    13  
    14  // Scanner provides a convenient interface for reading data such as
    15  // a file of newline-delimited lines of text. Successive calls to
    16  // the Scan method will step through the 'tokens' of a file, skipping
    17  // the bytes between the tokens. The specification of a token is
    18  // defined by a split function of type SplitFunc; the default split
    19  // function breaks the input into lines with line termination stripped. Split
    20  // functions are defined in this package for scanning a file into
    21  // lines, bytes, UTF-8-encoded runes, and space-delimited words. The
    22  // client may instead provide a custom split function.
    23  //
    24  // Scanning stops unrecoverably at EOF, the first I/O error, or a token too
    25  // large to fit in the buffer. When a scan stops, the reader may have
    26  // advanced arbitrarily far past the last token. Programs that need more
    27  // control over error handling or large tokens, or must run sequential scans
    28  // on a reader, should use bufio.Reader instead.
    29  //
    30  type Scanner struct {
    31  	r            io.Reader // The reader provided by the client.
    32  	split        SplitFunc // The function to split the tokens.
    33  	maxTokenSize int       // Maximum size of a token; modified by tests.
    34  	token        []byte    // Last token returned by split.
    35  	buf          []byte    // Buffer used as argument to split.
    36  	start        int       // First non-processed byte in buf.
    37  	end          int       // End of data in buf.
    38  	err          error     // Sticky error.
    39  	empties      int       // Count of successive empty tokens.
    40  	scanCalled   bool      // Scan has been called; buffer is in use.
    41  	done         bool      // Scan has finished.
    42  }
    43  
    44  // SplitFunc is the signature of the split function used to tokenize the
    45  // input. The arguments are an initial substring of the remaining unprocessed
    46  // data and a flag, atEOF, that reports whether the Reader has no more data
    47  // to give. The return values are the number of bytes to advance the input
    48  // and the next token to return to the user, plus an error, if any. If the
    49  // data does not yet hold a complete token, for instance if it has no newline
    50  // while scanning lines, SplitFunc can return (0, nil, nil) to signal the
    51  // Scanner to read more data into the slice and try again with a longer slice
    52  // starting at the same point in the input.
    53  //
    54  // If the returned error is non-nil, scanning stops and the error
    55  // is returned to the client.
    56  //
    57  // The function is never called with an empty data slice unless atEOF
    58  // is true. If atEOF is true, however, data may be non-empty and,
    59  // as always, holds unprocessed text.
    60  type SplitFunc func(data []byte, atEOF bool) (advance int, token []byte, err error)
    61  
    62  // Errors returned by Scanner.
    63  var (
    64  	ErrTooLong         = errors.New("bufio.Scanner: token too long")
    65  	ErrNegativeAdvance = errors.New("bufio.Scanner: SplitFunc returns negative advance count")
    66  	ErrAdvanceTooFar   = errors.New("bufio.Scanner: SplitFunc returns advance count beyond input")
    67  )
    68  
    69  const (
    70  	// MaxScanTokenSize is the maximum size used to buffer a token
    71  	// unless the user provides an explicit buffer with Scan.Buffer.
    72  	// The actual maximum token size may be smaller as the buffer
    73  	// may need to include, for instance, a newline.
    74  	MaxScanTokenSize = 64 * 1024
    75  
    76  	startBufSize = 4096 // Size of initial allocation for buffer.
    77  )
    78  
    79  // NewScanner returns a new Scanner to read from r.
    80  // The split function defaults to ScanLines.
    81  func NewScanner(r io.Reader) *Scanner {
    82  	return &Scanner{
    83  		r:            r,
    84  		split:        ScanLines,
    85  		maxTokenSize: MaxScanTokenSize,
    86  	}
    87  }
    88  
    89  // Err returns the first non-EOF error that was encountered by the Scanner.
    90  func (s *Scanner) Err() error {
    91  	if s.err == io.EOF {
    92  		return nil
    93  	}
    94  	return s.err
    95  }
    96  
    97  // Bytes returns the most recent token generated by a call to Scan.
    98  // The underlying array may point to data that will be overwritten
    99  // by a subsequent call to Scan. It does no allocation.
   100  func (s *Scanner) Bytes() []byte {
   101  	return s.token
   102  }
   103  
   104  // Text returns the most recent token generated by a call to Scan
   105  // as a newly allocated string holding its bytes.
   106  func (s *Scanner) Text() string {
   107  	return string(s.token)
   108  }
   109  
   110  // ErrFinalToken is a special sentinel error value. It is intended to be
   111  // returned by a Split function to indicate that the token being delivered
   112  // with the error is the last token and scanning should stop after this one.
   113  // After ErrFinalToken is received by Scan, scanning stops with no error.
   114  // The value is useful to stop processing early or when it is necessary to
   115  // deliver a final empty token. One could achieve the same behavior
   116  // with a custom error value but providing one here is tidier.
   117  // See the emptyFinalToken example for a use of this value.
   118  var ErrFinalToken = errors.New("final token")
   119  
   120  // Scan advances the Scanner to the next token, which will then be
   121  // available through the Bytes or Text method. It returns false when the
   122  // scan stops, either by reaching the end of the input or an error.
   123  // After Scan returns false, the Err method will return any error that
   124  // occurred during scanning, except that if it was io.EOF, Err
   125  // will return nil.
   126  // Scan panics if the split function returns too many empty
   127  // tokens without advancing the input. This is a common error mode for
   128  // scanners.
   129  func (s *Scanner) Scan() bool {
   130  	if s.done {
   131  		return false
   132  	}
   133  	s.scanCalled = true
   134  	// Loop until we have a token.
   135  	for {
   136  		// See if we can get a token with what we already have.
   137  		// If we've run out of data but have an error, give the split function
   138  		// a chance to recover any remaining, possibly empty token.
   139  		if s.end > s.start || s.err != nil {
   140  			advance, token, err := s.split(s.buf[s.start:s.end], s.err != nil)
   141  			if err != nil {
   142  				if err == ErrFinalToken {
   143  					s.token = token
   144  					s.done = true
   145  					return true
   146  				}
   147  				s.setErr(err)
   148  				return false
   149  			}
   150  			if !s.advance(advance) {
   151  				return false
   152  			}
   153  			s.token = token
   154  			if token != nil {
   155  				if s.err == nil || advance > 0 {
   156  					s.empties = 0
   157  				} else {
   158  					// Returning tokens not advancing input at EOF.
   159  					s.empties++
   160  					if s.empties > maxConsecutiveEmptyReads {
   161  						panic("bufio.Scan: too many empty tokens without progressing")
   162  					}
   163  				}
   164  				return true
   165  			}
   166  		}
   167  		// We cannot generate a token with what we are holding.
   168  		// If we've already hit EOF or an I/O error, we are done.
   169  		if s.err != nil {
   170  			// Shut it down.
   171  			s.start = 0
   172  			s.end = 0
   173  			return false
   174  		}
   175  		// Must read more data.
   176  		// First, shift data to beginning of buffer if there's lots of empty space
   177  		// or space is needed.
   178  		if s.start > 0 && (s.end == len(s.buf) || s.start > len(s.buf)/2) {
   179  			copy(s.buf, s.buf[s.start:s.end])
   180  			s.end -= s.start
   181  			s.start = 0
   182  		}
   183  		// Is the buffer full? If so, resize.
   184  		if s.end == len(s.buf) {
   185  			// Guarantee no overflow in the multiplication below.
   186  			const maxInt = int(^uint(0) >> 1)
   187  			if len(s.buf) >= s.maxTokenSize || len(s.buf) > maxInt/2 {
   188  				s.setErr(ErrTooLong)
   189  				return false
   190  			}
   191  			newSize := len(s.buf) * 2
   192  			if newSize == 0 {
   193  				newSize = startBufSize
   194  			}
   195  			if newSize > s.maxTokenSize {
   196  				newSize = s.maxTokenSize
   197  			}
   198  			newBuf := make([]byte, newSize)
   199  			copy(newBuf, s.buf[s.start:s.end])
   200  			s.buf = newBuf
   201  			s.end -= s.start
   202  			s.start = 0
   203  		}
   204  		// Finally we can read some input. Make sure we don't get stuck with
   205  		// a misbehaving Reader. Officially we don't need to do this, but let's
   206  		// be extra careful: Scanner is for safe, simple jobs.
   207  		for loop := 0; ; {
   208  			n, err := s.r.Read(s.buf[s.end:len(s.buf)])
   209  			s.end += n
   210  			if err != nil {
   211  				s.setErr(err)
   212  				break
   213  			}
   214  			if n > 0 {
   215  				s.empties = 0
   216  				break
   217  			}
   218  			loop++
   219  			if loop > maxConsecutiveEmptyReads {
   220  				s.setErr(io.ErrNoProgress)
   221  				break
   222  			}
   223  		}
   224  	}
   225  }
   226  
   227  // advance consumes n bytes of the buffer. It reports whether the advance was legal.
   228  func (s *Scanner) advance(n int) bool {
   229  	if n < 0 {
   230  		s.setErr(ErrNegativeAdvance)
   231  		return false
   232  	}
   233  	if n > s.end-s.start {
   234  		s.setErr(ErrAdvanceTooFar)
   235  		return false
   236  	}
   237  	s.start += n
   238  	return true
   239  }
   240  
   241  // setErr records the first error encountered.
   242  func (s *Scanner) setErr(err error) {
   243  	if s.err == nil || s.err == io.EOF {
   244  		s.err = err
   245  	}
   246  }
   247  
   248  // Buffer sets the initial buffer to use when scanning and the maximum
   249  // size of buffer that may be allocated during scanning. The maximum
   250  // token size is the larger of max and cap(buf). If max <= cap(buf),
   251  // Scan will use this buffer only and do no allocation.
   252  //
   253  // By default, Scan uses an internal buffer and sets the
   254  // maximum token size to MaxScanTokenSize.
   255  //
   256  // Buffer panics if it is called after scanning has started.
   257  func (s *Scanner) Buffer(buf []byte, max int) {
   258  	if s.scanCalled {
   259  		panic("Buffer called after Scan")
   260  	}
   261  	s.buf = buf[0:cap(buf)]
   262  	s.maxTokenSize = max
   263  }
   264  
   265  // Split sets the split function for the Scanner.
   266  // The default split function is ScanLines.
   267  //
   268  // Split panics if it is called after scanning has started.
   269  func (s *Scanner) Split(split SplitFunc) {
   270  	if s.scanCalled {
   271  		panic("Split called after Scan")
   272  	}
   273  	s.split = split
   274  }
   275  
   276  // Split functions
   277  
   278  // ScanBytes is a split function for a Scanner that returns each byte as a token.
   279  func ScanBytes(data []byte, atEOF bool) (advance int, token []byte, err error) {
   280  	if atEOF && len(data) == 0 {
   281  		return 0, nil, nil
   282  	}
   283  	return 1, data[0:1], nil
   284  }
   285  
   286  var errorRune = []byte(string(utf8.RuneError))
   287  
   288  // ScanRunes is a split function for a Scanner that returns each
   289  // UTF-8-encoded rune as a token. The sequence of runes returned is
   290  // equivalent to that from a range loop over the input as a string, which
   291  // means that erroneous UTF-8 encodings translate to U+FFFD = "\xef\xbf\xbd".
   292  // Because of the Scan interface, this makes it impossible for the client to
   293  // distinguish correctly encoded replacement runes from encoding errors.
   294  func ScanRunes(data []byte, atEOF bool) (advance int, token []byte, err error) {
   295  	if atEOF && len(data) == 0 {
   296  		return 0, nil, nil
   297  	}
   298  
   299  	// Fast path 1: ASCII.
   300  	if data[0] < utf8.RuneSelf {
   301  		return 1, data[0:1], nil
   302  	}
   303  
   304  	// Fast path 2: Correct UTF-8 decode without error.
   305  	_, width := utf8.DecodeRune(data)
   306  	if width > 1 {
   307  		// It's a valid encoding. Width cannot be one for a correctly encoded
   308  		// non-ASCII rune.
   309  		return width, data[0:width], nil
   310  	}
   311  
   312  	// We know it's an error: we have width==1 and implicitly r==utf8.RuneError.
   313  	// Is the error because there wasn't a full rune to be decoded?
   314  	// FullRune distinguishes correctly between erroneous and incomplete encodings.
   315  	if !atEOF && !utf8.FullRune(data) {
   316  		// Incomplete; get more bytes.
   317  		return 0, nil, nil
   318  	}
   319  
   320  	// We have a real UTF-8 encoding error. Return a properly encoded error rune
   321  	// but advance only one byte. This matches the behavior of a range loop over
   322  	// an incorrectly encoded string.
   323  	return 1, errorRune, nil
   324  }
   325  
   326  // dropCR drops a terminal \r from the data.
   327  func dropCR(data []byte) []byte {
   328  	if len(data) > 0 && data[len(data)-1] == '\r' {
   329  		return data[0 : len(data)-1]
   330  	}
   331  	return data
   332  }
   333  
   334  // ScanLines is a split function for a Scanner that returns each line of
   335  // text, stripped of any trailing end-of-line marker. The returned line may
   336  // be empty. The end-of-line marker is one optional carriage return followed
   337  // by one mandatory newline. In regular expression notation, it is `\r?\n`.
   338  // The last non-empty line of input will be returned even if it has no
   339  // newline.
   340  func ScanLines(data []byte, atEOF bool) (advance int, token []byte, err error) {
   341  	if atEOF && len(data) == 0 {
   342  		return 0, nil, nil
   343  	}
   344  	if i := bytes.IndexByte(data, '\n'); i >= 0 {
   345  		// We have a full newline-terminated line.
   346  		return i + 1, dropCR(data[0:i]), nil
   347  	}
   348  	// If we're at EOF, we have a final, non-terminated line. Return it.
   349  	if atEOF {
   350  		return len(data), dropCR(data), nil
   351  	}
   352  	// Request more data.
   353  	return 0, nil, nil
   354  }
   355  
   356  // isSpace reports whether the character is a Unicode white space character.
   357  // We avoid dependency on the unicode package, but check validity of the implementation
   358  // in the tests.
   359  func isSpace(r rune) bool {
   360  	if r <= '\u00FF' {
   361  		// Obvious ASCII ones: \t through \r plus space. Plus two Latin-1 oddballs.
   362  		switch r {
   363  		case ' ', '\t', '\n', '\v', '\f', '\r':
   364  			return true
   365  		case '\u0085', '\u00A0':
   366  			return true
   367  		}
   368  		return false
   369  	}
   370  	// High-valued ones.
   371  	if '\u2000' <= r && r <= '\u200a' {
   372  		return true
   373  	}
   374  	switch r {
   375  	case '\u1680', '\u2028', '\u2029', '\u202f', '\u205f', '\u3000':
   376  		return true
   377  	}
   378  	return false
   379  }
   380  
   381  // ScanWords is a split function for a Scanner that returns each
   382  // space-separated word of text, with surrounding spaces deleted. It will
   383  // never return an empty string. The definition of space is set by
   384  // unicode.IsSpace.
   385  func ScanWords(data []byte, atEOF bool) (advance int, token []byte, err error) {
   386  	// Skip leading spaces.
   387  	start := 0
   388  	for width := 0; start < len(data); start += width {
   389  		var r rune
   390  		r, width = utf8.DecodeRune(data[start:])
   391  		if !isSpace(r) {
   392  			break
   393  		}
   394  	}
   395  	// Scan until space, marking end of word.
   396  	for width, i := 0, start; i < len(data); i += width {
   397  		var r rune
   398  		r, width = utf8.DecodeRune(data[i:])
   399  		if isSpace(r) {
   400  			return i + width, data[start:i], nil
   401  		}
   402  	}
   403  	// If we're at EOF, we have a final, non-empty, non-terminated word. Return it.
   404  	if atEOF && len(data) > start {
   405  		return len(data), data[start:], nil
   406  	}
   407  	// Request more data.
   408  	return start, nil, nil
   409  }
   410  

View as plain text