replace.go

     1  // Copyright 2011 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 strings
     6  
     7  import (
     8  	"io"
     9  	"sync"
    10  )
    11  
    12  // Replacer replaces a list of strings with replacements.
    13  // It is safe for concurrent use by multiple goroutines.
    14  type Replacer struct {
    15  	once   sync.Once // guards buildOnce method
    16  	r      replacer
    17  	oldnew []string
    18  }
    19  
    20  // replacer is the interface that a replacement algorithm needs to implement.
    21  type replacer interface {
    22  	Replace(s string) string
    23  	WriteString(w io.Writer, s string) (n int, err error)
    24  }
    25  
    26  // NewReplacer returns a new Replacer from a list of old, new string
    27  // pairs. Replacements are performed in the order they appear in the
    28  // target string, without overlapping matches. The old string
    29  // comparisons are done in argument order.
    30  //
    31  // NewReplacer panics if given an odd number of arguments.
    32  func NewReplacer(oldnew ...string) *Replacer {
    33  	if len(oldnew)%2 == 1 {
    34  		panic("strings.NewReplacer: odd argument count")
    35  	}
    36  	return &Replacer{oldnew: append([]string(nil), oldnew...)}
    37  }
    38  
    39  func (r *Replacer) buildOnce() {
    40  	r.r = r.build()
    41  	r.oldnew = nil
    42  }
    43  
    44  func (b *Replacer) build() replacer {
    45  	oldnew := b.oldnew
    46  	if len(oldnew) == 2 && len(oldnew[0]) > 1 {
    47  		return makeSingleStringReplacer(oldnew[0], oldnew[1])
    48  	}
    49  
    50  	allNewBytes := true
    51  	for i := 0; i < len(oldnew); i += 2 {
    52  		if len(oldnew[i]) != 1 {
    53  			return makeGenericReplacer(oldnew)
    54  		}
    55  		if len(oldnew[i+1]) != 1 {
    56  			allNewBytes = false
    57  		}
    58  	}
    59  
    60  	if allNewBytes {
    61  		r := byteReplacer{}
    62  		for i := range r {
    63  			r[i] = byte(i)
    64  		}
    65  		// The first occurrence of old->new map takes precedence
    66  		// over the others with the same old string.
    67  		for i := len(oldnew) - 2; i >= 0; i -= 2 {
    68  			o := oldnew[i][0]
    69  			n := oldnew[i+1][0]
    70  			r[o] = n
    71  		}
    72  		return &r
    73  	}
    74  
    75  	r := byteStringReplacer{toReplace: make([]string, 0, len(oldnew)/2)}
    76  	// The first occurrence of old->new map takes precedence
    77  	// over the others with the same old string.
    78  	for i := len(oldnew) - 2; i >= 0; i -= 2 {
    79  		o := oldnew[i][0]
    80  		n := oldnew[i+1]
    81  		// To avoid counting repetitions multiple times.
    82  		if r.replacements[o] == nil {
    83  			// We need to use string([]byte{o}) instead of string(o),
    84  			// to avoid utf8 encoding of o.
    85  			// E. g. byte(150) produces string of length 2.
    86  			r.toReplace = append(r.toReplace, string([]byte{o}))
    87  		}
    88  		r.replacements[o] = []byte(n)
    89  
    90  	}
    91  	return &r
    92  }
    93  
    94  // Replace returns a copy of s with all replacements performed.
    95  func (r *Replacer) Replace(s string) string {
    96  	r.once.Do(r.buildOnce)
    97  	return r.r.Replace(s)
    98  }
    99  
   100  // WriteString writes s to w with all replacements performed.
   101  func (r *Replacer) WriteString(w io.Writer, s string) (n int, err error) {
   102  	r.once.Do(r.buildOnce)
   103  	return r.r.WriteString(w, s)
   104  }
   105  
   106  // trieNode is a node in a lookup trie for prioritized key/value pairs. Keys
   107  // and values may be empty. For example, the trie containing keys "ax", "ay",
   108  // "bcbc", "x" and "xy" could have eight nodes:
   109  //
   110  //  n0  -
   111  //  n1  a-
   112  //  n2  .x+
   113  //  n3  .y+
   114  //  n4  b-
   115  //  n5  .cbc+
   116  //  n6  x+
   117  //  n7  .y+
   118  //
   119  // n0 is the root node, and its children are n1, n4 and n6; n1's children are
   120  // n2 and n3; n4's child is n5; n6's child is n7. Nodes n0, n1 and n4 (marked
   121  // with a trailing "-") are partial keys, and nodes n2, n3, n5, n6 and n7
   122  // (marked with a trailing "+") are complete keys.
   123  type trieNode struct {
   124  	// value is the value of the trie node's key/value pair. It is empty if
   125  	// this node is not a complete key.
   126  	value string
   127  	// priority is the priority (higher is more important) of the trie node's
   128  	// key/value pair; keys are not necessarily matched shortest- or longest-
   129  	// first. Priority is positive if this node is a complete key, and zero
   130  	// otherwise. In the example above, positive/zero priorities are marked
   131  	// with a trailing "+" or "-".
   132  	priority int
   133  
   134  	// A trie node may have zero, one or more child nodes:
   135  	//  * if the remaining fields are zero, there are no children.
   136  	//  * if prefix and next are non-zero, there is one child in next.
   137  	//  * if table is non-zero, it defines all the children.
   138  	//
   139  	// Prefixes are preferred over tables when there is one child, but the
   140  	// root node always uses a table for lookup efficiency.
   141  
   142  	// prefix is the difference in keys between this trie node and the next.
   143  	// In the example above, node n4 has prefix "cbc" and n4's next node is n5.
   144  	// Node n5 has no children and so has zero prefix, next and table fields.
   145  	prefix string
   146  	next   *trieNode
   147  
   148  	// table is a lookup table indexed by the next byte in the key, after
   149  	// remapping that byte through genericReplacer.mapping to create a dense
   150  	// index. In the example above, the keys only use 'a', 'b', 'c', 'x' and
   151  	// 'y', which remap to 0, 1, 2, 3 and 4. All other bytes remap to 5, and
   152  	// genericReplacer.tableSize will be 5. Node n0's table will be
   153  	// []*trieNode{ 0:n1, 1:n4, 3:n6 }, where the 0, 1 and 3 are the remapped
   154  	// 'a', 'b' and 'x'.
   155  	table []*trieNode
   156  }
   157  
   158  func (t *trieNode) add(key, val string, priority int, r *genericReplacer) {
   159  	if key == "" {
   160  		if t.priority == 0 {
   161  			t.value = val
   162  			t.priority = priority
   163  		}
   164  		return
   165  	}
   166  
   167  	if t.prefix != "" {
   168  		// Need to split the prefix among multiple nodes.
   169  		var n int // length of the longest common prefix
   170  		for ; n < len(t.prefix) && n < len(key); n++ {
   171  			if t.prefix[n] != key[n] {
   172  				break
   173  			}
   174  		}
   175  		if n == len(t.prefix) {
   176  			t.next.add(key[n:], val, priority, r)
   177  		} else if n == 0 {
   178  			// First byte differs, start a new lookup table here. Looking up
   179  			// what is currently t.prefix[0] will lead to prefixNode, and
   180  			// looking up key[0] will lead to keyNode.
   181  			var prefixNode *trieNode
   182  			if len(t.prefix) == 1 {
   183  				prefixNode = t.next
   184  			} else {
   185  				prefixNode = &trieNode{
   186  					prefix: t.prefix[1:],
   187  					next:   t.next,
   188  				}
   189  			}
   190  			keyNode := new(trieNode)
   191  			t.table = make([]*trieNode, r.tableSize)
   192  			t.table[r.mapping[t.prefix[0]]] = prefixNode
   193  			t.table[r.mapping[key[0]]] = keyNode
   194  			t.prefix = ""
   195  			t.next = nil
   196  			keyNode.add(key[1:], val, priority, r)
   197  		} else {
   198  			// Insert new node after the common section of the prefix.
   199  			next := &trieNode{
   200  				prefix: t.prefix[n:],
   201  				next:   t.next,
   202  			}
   203  			t.prefix = t.prefix[:n]
   204  			t.next = next
   205  			next.add(key[n:], val, priority, r)
   206  		}
   207  	} else if t.table != nil {
   208  		// Insert into existing table.
   209  		m := r.mapping[key[0]]
   210  		if t.table[m] == nil {
   211  			t.table[m] = new(trieNode)
   212  		}
   213  		t.table[m].add(key[1:], val, priority, r)
   214  	} else {
   215  		t.prefix = key
   216  		t.next = new(trieNode)
   217  		t.next.add("", val, priority, r)
   218  	}
   219  }
   220  
   221  func (r *genericReplacer) lookup(s string, ignoreRoot bool) (val string, keylen int, found bool) {
   222  	// Iterate down the trie to the end, and grab the value and keylen with
   223  	// the highest priority.
   224  	bestPriority := 0
   225  	node := &r.root
   226  	n := 0
   227  	for node != nil {
   228  		if node.priority > bestPriority && !(ignoreRoot && node == &r.root) {
   229  			bestPriority = node.priority
   230  			val = node.value
   231  			keylen = n
   232  			found = true
   233  		}
   234  
   235  		if s == "" {
   236  			break
   237  		}
   238  		if node.table != nil {
   239  			index := r.mapping[s[0]]
   240  			if int(index) == r.tableSize {
   241  				break
   242  			}
   243  			node = node.table[index]
   244  			s = s[1:]
   245  			n++
   246  		} else if node.prefix != "" && HasPrefix(s, node.prefix) {
   247  			n += len(node.prefix)
   248  			s = s[len(node.prefix):]
   249  			node = node.next
   250  		} else {
   251  			break
   252  		}
   253  	}
   254  	return
   255  }
   256  
   257  // genericReplacer is the fully generic algorithm.
   258  // It's used as a fallback when nothing faster can be used.
   259  type genericReplacer struct {
   260  	root trieNode
   261  	// tableSize is the size of a trie node's lookup table. It is the number
   262  	// of unique key bytes.
   263  	tableSize int
   264  	// mapping maps from key bytes to a dense index for trieNode.table.
   265  	mapping [256]byte
   266  }
   267  
   268  func makeGenericReplacer(oldnew []string) *genericReplacer {
   269  	r := new(genericReplacer)
   270  	// Find each byte used, then assign them each an index.
   271  	for i := 0; i < len(oldnew); i += 2 {
   272  		key := oldnew[i]
   273  		for j := 0; j < len(key); j++ {
   274  			r.mapping[key[j]] = 1
   275  		}
   276  	}
   277  
   278  	for _, b := range r.mapping {
   279  		r.tableSize += int(b)
   280  	}
   281  
   282  	var index byte
   283  	for i, b := range r.mapping {
   284  		if b == 0 {
   285  			r.mapping[i] = byte(r.tableSize)
   286  		} else {
   287  			r.mapping[i] = index
   288  			index++
   289  		}
   290  	}
   291  	// Ensure root node uses a lookup table (for performance).
   292  	r.root.table = make([]*trieNode, r.tableSize)
   293  
   294  	for i := 0; i < len(oldnew); i += 2 {
   295  		r.root.add(oldnew[i], oldnew[i+1], len(oldnew)-i, r)
   296  	}
   297  	return r
   298  }
   299  
   300  type appendSliceWriter []byte
   301  
   302  // Write writes to the buffer to satisfy io.Writer.
   303  func (w *appendSliceWriter) Write(p []byte) (int, error) {
   304  	*w = append(*w, p...)
   305  	return len(p), nil
   306  }
   307  
   308  // WriteString writes to the buffer without string->[]byte->string allocations.
   309  func (w *appendSliceWriter) WriteString(s string) (int, error) {
   310  	*w = append(*w, s...)
   311  	return len(s), nil
   312  }
   313  
   314  type stringWriter struct {
   315  	w io.Writer
   316  }
   317  
   318  func (w stringWriter) WriteString(s string) (int, error) {
   319  	return w.w.Write([]byte(s))
   320  }
   321  
   322  func getStringWriter(w io.Writer) io.StringWriter {
   323  	sw, ok := w.(io.StringWriter)
   324  	if !ok {
   325  		sw = stringWriter{w}
   326  	}
   327  	return sw
   328  }
   329  
   330  func (r *genericReplacer) Replace(s string) string {
   331  	buf := make(appendSliceWriter, 0, len(s))
   332  	r.WriteString(&buf, s)
   333  	return string(buf)
   334  }
   335  
   336  func (r *genericReplacer) WriteString(w io.Writer, s string) (n int, err error) {
   337  	sw := getStringWriter(w)
   338  	var last, wn int
   339  	var prevMatchEmpty bool
   340  	for i := 0; i <= len(s); {
   341  		// Fast path: s[i] is not a prefix of any pattern.
   342  		if i != len(s) && r.root.priority == 0 {
   343  			index := int(r.mapping[s[i]])
   344  			if index == r.tableSize || r.root.table[index] == nil {
   345  				i++
   346  				continue
   347  			}
   348  		}
   349  
   350  		// Ignore the empty match iff the previous loop found the empty match.
   351  		val, keylen, match := r.lookup(s[i:], prevMatchEmpty)
   352  		prevMatchEmpty = match && keylen == 0
   353  		if match {
   354  			wn, err = sw.WriteString(s[last:i])
   355  			n += wn
   356  			if err != nil {
   357  				return
   358  			}
   359  			wn, err = sw.WriteString(val)
   360  			n += wn
   361  			if err != nil {
   362  				return
   363  			}
   364  			i += keylen
   365  			last = i
   366  			continue
   367  		}
   368  		i++
   369  	}
   370  	if last != len(s) {
   371  		wn, err = sw.WriteString(s[last:])
   372  		n += wn
   373  	}
   374  	return
   375  }
   376  
   377  // singleStringReplacer is the implementation that's used when there is only
   378  // one string to replace (and that string has more than one byte).
   379  type singleStringReplacer struct {
   380  	finder *stringFinder
   381  	// value is the new string that replaces that pattern when it's found.
   382  	value string
   383  }
   384  
   385  func makeSingleStringReplacer(pattern string, value string) *singleStringReplacer {
   386  	return &singleStringReplacer{finder: makeStringFinder(pattern), value: value}
   387  }
   388  
   389  func (r *singleStringReplacer) Replace(s string) string {
   390  	var buf Builder
   391  	i, matched := 0, false
   392  	for {
   393  		match := r.finder.next(s[i:])
   394  		if match == -1 {
   395  			break
   396  		}
   397  		matched = true
   398  		buf.Grow(match + len(r.value))
   399  		buf.WriteString(s[i : i+match])
   400  		buf.WriteString(r.value)
   401  		i += match + len(r.finder.pattern)
   402  	}
   403  	if !matched {
   404  		return s
   405  	}
   406  	buf.WriteString(s[i:])
   407  	return buf.String()
   408  }
   409  
   410  func (r *singleStringReplacer) WriteString(w io.Writer, s string) (n int, err error) {
   411  	sw := getStringWriter(w)
   412  	var i, wn int
   413  	for {
   414  		match := r.finder.next(s[i:])
   415  		if match == -1 {
   416  			break
   417  		}
   418  		wn, err = sw.WriteString(s[i : i+match])
   419  		n += wn
   420  		if err != nil {
   421  			return
   422  		}
   423  		wn, err = sw.WriteString(r.value)
   424  		n += wn
   425  		if err != nil {
   426  			return
   427  		}
   428  		i += match + len(r.finder.pattern)
   429  	}
   430  	wn, err = sw.WriteString(s[i:])
   431  	n += wn
   432  	return
   433  }
   434  
   435  // byteReplacer is the implementation that's used when all the "old"
   436  // and "new" values are single ASCII bytes.
   437  // The array contains replacement bytes indexed by old byte.
   438  type byteReplacer [256]byte
   439  
   440  func (r *byteReplacer) Replace(s string) string {
   441  	var buf []byte // lazily allocated
   442  	for i := 0; i < len(s); i++ {
   443  		b := s[i]
   444  		if r[b] != b {
   445  			if buf == nil {
   446  				buf = []byte(s)
   447  			}
   448  			buf[i] = r[b]
   449  		}
   450  	}
   451  	if buf == nil {
   452  		return s
   453  	}
   454  	return string(buf)
   455  }
   456  
   457  func (r *byteReplacer) WriteString(w io.Writer, s string) (n int, err error) {
   458  	// TODO(bradfitz): use io.WriteString with slices of s, avoiding allocation.
   459  	bufsize := 32 << 10
   460  	if len(s) < bufsize {
   461  		bufsize = len(s)
   462  	}
   463  	buf := make([]byte, bufsize)
   464  
   465  	for len(s) > 0 {
   466  		ncopy := copy(buf, s)
   467  		s = s[ncopy:]
   468  		for i, b := range buf[:ncopy] {
   469  			buf[i] = r[b]
   470  		}
   471  		wn, err := w.Write(buf[:ncopy])
   472  		n += wn
   473  		if err != nil {
   474  			return n, err
   475  		}
   476  	}
   477  	return n, nil
   478  }
   479  
   480  // byteStringReplacer is the implementation that's used when all the
   481  // "old" values are single ASCII bytes but the "new" values vary in size.
   482  type byteStringReplacer struct {
   483  	// replacements contains replacement byte slices indexed by old byte.
   484  	// A nil []byte means that the old byte should not be replaced.
   485  	replacements [256][]byte
   486  	// toReplace keeps a list of bytes to replace. Depending on length of toReplace
   487  	// and length of target string it may be faster to use Count, or a plain loop.
   488  	// We store single byte as a string, because Count takes a string.
   489  	toReplace []string
   490  }
   491  
   492  // countCutOff controls the ratio of a string length to a number of replacements
   493  // at which (*byteStringReplacer).Replace switches algorithms.
   494  // For strings with higher ration of length to replacements than that value,
   495  // we call Count, for each replacement from toReplace.
   496  // For strings, with a lower ratio we use simple loop, because of Count overhead.
   497  // countCutOff is an empirically determined overhead multiplier.
   498  // TODO(tocarip) revisit once we have register-based abi/mid-stack inlining.
   499  const countCutOff = 8
   500  
   501  func (r *byteStringReplacer) Replace(s string) string {
   502  	newSize := len(s)
   503  	anyChanges := false
   504  	// Is it faster to use Count?
   505  	if len(r.toReplace)*countCutOff <= len(s) {
   506  		for _, x := range r.toReplace {
   507  			if c := Count(s, x); c != 0 {
   508  				// The -1 is because we are replacing 1 byte with len(replacements[b]) bytes.
   509  				newSize += c * (len(r.replacements[x[0]]) - 1)
   510  				anyChanges = true
   511  			}
   512  
   513  		}
   514  	} else {
   515  		for i := 0; i < len(s); i++ {
   516  			b := s[i]
   517  			if r.replacements[b] != nil {
   518  				// See above for explanation of -1
   519  				newSize += len(r.replacements[b]) - 1
   520  				anyChanges = true
   521  			}
   522  		}
   523  	}
   524  	if !anyChanges {
   525  		return s
   526  	}
   527  	buf := make([]byte, newSize)
   528  	j := 0
   529  	for i := 0; i < len(s); i++ {
   530  		b := s[i]
   531  		if r.replacements[b] != nil {
   532  			j += copy(buf[j:], r.replacements[b])
   533  		} else {
   534  			buf[j] = b
   535  			j++
   536  		}
   537  	}
   538  	return string(buf)
   539  }
   540  
   541  func (r *byteStringReplacer) WriteString(w io.Writer, s string) (n int, err error) {
   542  	sw := getStringWriter(w)
   543  	last := 0
   544  	for i := 0; i < len(s); i++ {
   545  		b := s[i]
   546  		if r.replacements[b] == nil {
   547  			continue
   548  		}
   549  		if last != i {
   550  			nw, err := sw.WriteString(s[last:i])
   551  			n += nw
   552  			if err != nil {
   553  				return n, err
   554  			}
   555  		}
   556  		last = i + 1
   557  		nw, err := w.Write(r.replacements[b])
   558  		n += nw
   559  		if err != nil {
   560  			return n, err
   561  		}
   562  	}
   563  	if last != len(s) {
   564  		var nw int
   565  		nw, err = sw.WriteString(s[last:])
   566  		n += nw
   567  	}
   568  	return
   569  }
   570
View as plain text