...

Source file src/go/token/position.go

Documentation: go/token

     1  // Copyright 2010 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 token
     6  
     7  import (
     8  	"fmt"
     9  	"sort"
    10  	"strconv"
    11  	"sync"
    12  	"sync/atomic"
    13  )
    14  
    15  // -----------------------------------------------------------------------------
    16  // Positions
    17  
    18  // Position describes an arbitrary source position
    19  // including the file, line, and column location.
    20  // A Position is valid if the line number is > 0.
    21  type Position struct {
    22  	Filename string // filename, if any
    23  	Offset   int    // offset, starting at 0
    24  	Line     int    // line number, starting at 1
    25  	Column   int    // column number, starting at 1 (byte count)
    26  }
    27  
    28  // IsValid reports whether the position is valid.
    29  func (pos *Position) IsValid() bool { return pos.Line > 0 }
    30  
    31  // String returns a string in one of several forms:
    32  //
    33  //	file:line:column    valid position with file name
    34  //	file:line           valid position with file name but no column (column == 0)
    35  //	line:column         valid position without file name
    36  //	line                valid position without file name and no column (column == 0)
    37  //	file                invalid position with file name
    38  //	-                   invalid position without file name
    39  func (pos Position) String() string {
    40  	s := pos.Filename
    41  	if pos.IsValid() {
    42  		if s != "" {
    43  			s += ":"
    44  		}
    45  		s += strconv.Itoa(pos.Line)
    46  		if pos.Column != 0 {
    47  			s += fmt.Sprintf(":%d", pos.Column)
    48  		}
    49  	}
    50  	if s == "" {
    51  		s = "-"
    52  	}
    53  	return s
    54  }
    55  
    56  // Pos is a compact encoding of a source position within a file set.
    57  // It can be converted into a [Position] for a more convenient, but much
    58  // larger, representation.
    59  //
    60  // The Pos value for a given file is a number in the range [base, base+size],
    61  // where base and size are specified when a file is added to the file set.
    62  // The difference between a Pos value and the corresponding file base
    63  // corresponds to the byte offset of that position (represented by the Pos value)
    64  // from the beginning of the file. Thus, the file base offset is the Pos value
    65  // representing the first byte in the file.
    66  //
    67  // To create the Pos value for a specific source offset (measured in bytes),
    68  // first add the respective file to the current file set using [FileSet.AddFile]
    69  // and then call [File.Pos](offset) for that file. Given a Pos value p
    70  // for a specific file set fset, the corresponding [Position] value is
    71  // obtained by calling fset.Position(p).
    72  //
    73  // Pos values can be compared directly with the usual comparison operators:
    74  // If two Pos values p and q are in the same file, comparing p and q is
    75  // equivalent to comparing the respective source file offsets. If p and q
    76  // are in different files, p < q is true if the file implied by p was added
    77  // to the respective file set before the file implied by q.
    78  type Pos int
    79  
    80  // The zero value for [Pos] is NoPos; there is no file and line information
    81  // associated with it, and NoPos.IsValid() is false. NoPos is always
    82  // smaller than any other [Pos] value. The corresponding [Position] value
    83  // for NoPos is the zero value for [Position].
    84  const NoPos Pos = 0
    85  
    86  // IsValid reports whether the position is valid.
    87  func (p Pos) IsValid() bool {
    88  	return p != NoPos
    89  }
    90  
    91  // -----------------------------------------------------------------------------
    92  // File
    93  
    94  // A File is a handle for a file belonging to a [FileSet].
    95  // A File has a name, size, and line offset table.
    96  type File struct {
    97  	name string // file name as provided to AddFile
    98  	base int    // Pos value range for this file is [base...base+size]
    99  	size int    // file size as provided to AddFile
   100  
   101  	// lines and infos are protected by mutex
   102  	mutex sync.Mutex
   103  	lines []int // lines contains the offset of the first character for each line (the first entry is always 0)
   104  	infos []lineInfo
   105  }
   106  
   107  // Name returns the file name of file f as registered with AddFile.
   108  func (f *File) Name() string {
   109  	return f.name
   110  }
   111  
   112  // Base returns the base offset of file f as registered with AddFile.
   113  func (f *File) Base() int {
   114  	return f.base
   115  }
   116  
   117  // Size returns the size of file f as registered with AddFile.
   118  func (f *File) Size() int {
   119  	return f.size
   120  }
   121  
   122  // LineCount returns the number of lines in file f.
   123  func (f *File) LineCount() int {
   124  	f.mutex.Lock()
   125  	n := len(f.lines)
   126  	f.mutex.Unlock()
   127  	return n
   128  }
   129  
   130  // AddLine adds the line offset for a new line.
   131  // The line offset must be larger than the offset for the previous line
   132  // and smaller than the file size; otherwise the line offset is ignored.
   133  func (f *File) AddLine(offset int) {
   134  	f.mutex.Lock()
   135  	if i := len(f.lines); (i == 0 || f.lines[i-1] < offset) && offset < f.size {
   136  		f.lines = append(f.lines, offset)
   137  	}
   138  	f.mutex.Unlock()
   139  }
   140  
   141  // MergeLine merges a line with the following line. It is akin to replacing
   142  // the newline character at the end of the line with a space (to not change the
   143  // remaining offsets). To obtain the line number, consult e.g. [Position.Line].
   144  // MergeLine will panic if given an invalid line number.
   145  func (f *File) MergeLine(line int) {
   146  	if line < 1 {
   147  		panic(fmt.Sprintf("invalid line number %d (should be >= 1)", line))
   148  	}
   149  	f.mutex.Lock()
   150  	defer f.mutex.Unlock()
   151  	if line >= len(f.lines) {
   152  		panic(fmt.Sprintf("invalid line number %d (should be < %d)", line, len(f.lines)))
   153  	}
   154  	// To merge the line numbered <line> with the line numbered <line+1>,
   155  	// we need to remove the entry in lines corresponding to the line
   156  	// numbered <line+1>. The entry in lines corresponding to the line
   157  	// numbered <line+1> is located at index <line>, since indices in lines
   158  	// are 0-based and line numbers are 1-based.
   159  	copy(f.lines[line:], f.lines[line+1:])
   160  	f.lines = f.lines[:len(f.lines)-1]
   161  }
   162  
   163  // Lines returns the effective line offset table of the form described by [File.SetLines].
   164  // Callers must not mutate the result.
   165  func (f *File) Lines() []int {
   166  	f.mutex.Lock()
   167  	lines := f.lines
   168  	f.mutex.Unlock()
   169  	return lines
   170  }
   171  
   172  // SetLines sets the line offsets for a file and reports whether it succeeded.
   173  // The line offsets are the offsets of the first character of each line;
   174  // for instance for the content "ab\nc\n" the line offsets are {0, 3}.
   175  // An empty file has an empty line offset table.
   176  // Each line offset must be larger than the offset for the previous line
   177  // and smaller than the file size; otherwise SetLines fails and returns
   178  // false.
   179  // Callers must not mutate the provided slice after SetLines returns.
   180  func (f *File) SetLines(lines []int) bool {
   181  	// verify validity of lines table
   182  	size := f.size
   183  	for i, offset := range lines {
   184  		if i > 0 && offset <= lines[i-1] || size <= offset {
   185  			return false
   186  		}
   187  	}
   188  
   189  	// set lines table
   190  	f.mutex.Lock()
   191  	f.lines = lines
   192  	f.mutex.Unlock()
   193  	return true
   194  }
   195  
   196  // SetLinesForContent sets the line offsets for the given file content.
   197  // It ignores position-altering //line comments.
   198  func (f *File) SetLinesForContent(content []byte) {
   199  	var lines []int
   200  	line := 0
   201  	for offset, b := range content {
   202  		if line >= 0 {
   203  			lines = append(lines, line)
   204  		}
   205  		line = -1
   206  		if b == '\n' {
   207  			line = offset + 1
   208  		}
   209  	}
   210  
   211  	// set lines table
   212  	f.mutex.Lock()
   213  	f.lines = lines
   214  	f.mutex.Unlock()
   215  }
   216  
   217  // LineStart returns the [Pos] value of the start of the specified line.
   218  // It ignores any alternative positions set using [File.AddLineColumnInfo].
   219  // LineStart panics if the 1-based line number is invalid.
   220  func (f *File) LineStart(line int) Pos {
   221  	if line < 1 {
   222  		panic(fmt.Sprintf("invalid line number %d (should be >= 1)", line))
   223  	}
   224  	f.mutex.Lock()
   225  	defer f.mutex.Unlock()
   226  	if line > len(f.lines) {
   227  		panic(fmt.Sprintf("invalid line number %d (should be < %d)", line, len(f.lines)))
   228  	}
   229  	return Pos(f.base + f.lines[line-1])
   230  }
   231  
   232  // A lineInfo object describes alternative file, line, and column
   233  // number information (such as provided via a //line directive)
   234  // for a given file offset.
   235  type lineInfo struct {
   236  	// fields are exported to make them accessible to gob
   237  	Offset       int
   238  	Filename     string
   239  	Line, Column int
   240  }
   241  
   242  // AddLineInfo is like [File.AddLineColumnInfo] with a column = 1 argument.
   243  // It is here for backward-compatibility for code prior to Go 1.11.
   244  func (f *File) AddLineInfo(offset int, filename string, line int) {
   245  	f.AddLineColumnInfo(offset, filename, line, 1)
   246  }
   247  
   248  // AddLineColumnInfo adds alternative file, line, and column number
   249  // information for a given file offset. The offset must be larger
   250  // than the offset for the previously added alternative line info
   251  // and smaller than the file size; otherwise the information is
   252  // ignored.
   253  //
   254  // AddLineColumnInfo is typically used to register alternative position
   255  // information for line directives such as //line filename:line:column.
   256  func (f *File) AddLineColumnInfo(offset int, filename string, line, column int) {
   257  	f.mutex.Lock()
   258  	if i := len(f.infos); (i == 0 || f.infos[i-1].Offset < offset) && offset < f.size {
   259  		f.infos = append(f.infos, lineInfo{offset, filename, line, column})
   260  	}
   261  	f.mutex.Unlock()
   262  }
   263  
   264  // Pos returns the Pos value for the given file offset;
   265  // the offset must be <= f.Size().
   266  // f.Pos(f.Offset(p)) == p.
   267  func (f *File) Pos(offset int) Pos {
   268  	if offset > f.size {
   269  		panic(fmt.Sprintf("invalid file offset %d (should be <= %d)", offset, f.size))
   270  	}
   271  	return Pos(f.base + offset)
   272  }
   273  
   274  // Offset returns the offset for the given file position p;
   275  // p must be a valid [Pos] value in that file.
   276  // f.Offset(f.Pos(offset)) == offset.
   277  func (f *File) Offset(p Pos) int {
   278  	if int(p) < f.base || int(p) > f.base+f.size {
   279  		panic(fmt.Sprintf("invalid Pos value %d (should be in [%d, %d])", p, f.base, f.base+f.size))
   280  	}
   281  	return int(p) - f.base
   282  }
   283  
   284  // Line returns the line number for the given file position p;
   285  // p must be a [Pos] value in that file or [NoPos].
   286  func (f *File) Line(p Pos) int {
   287  	return f.Position(p).Line
   288  }
   289  
   290  func searchLineInfos(a []lineInfo, x int) int {
   291  	return sort.Search(len(a), func(i int) bool { return a[i].Offset > x }) - 1
   292  }
   293  
   294  // unpack returns the filename and line and column number for a file offset.
   295  // If adjusted is set, unpack will return the filename and line information
   296  // possibly adjusted by //line comments; otherwise those comments are ignored.
   297  func (f *File) unpack(offset int, adjusted bool) (filename string, line, column int) {
   298  	f.mutex.Lock()
   299  	filename = f.name
   300  	if i := searchInts(f.lines, offset); i >= 0 {
   301  		line, column = i+1, offset-f.lines[i]+1
   302  	}
   303  	if adjusted && len(f.infos) > 0 {
   304  		// few files have extra line infos
   305  		if i := searchLineInfos(f.infos, offset); i >= 0 {
   306  			alt := &f.infos[i]
   307  			filename = alt.Filename
   308  			if i := searchInts(f.lines, alt.Offset); i >= 0 {
   309  				// i+1 is the line at which the alternative position was recorded
   310  				d := line - (i + 1) // line distance from alternative position base
   311  				line = alt.Line + d
   312  				if alt.Column == 0 {
   313  					// alternative column is unknown => relative column is unknown
   314  					// (the current specification for line directives requires
   315  					// this to apply until the next PosBase/line directive,
   316  					// not just until the new newline)
   317  					column = 0
   318  				} else if d == 0 {
   319  					// the alternative position base is on the current line
   320  					// => column is relative to alternative column
   321  					column = alt.Column + (offset - alt.Offset)
   322  				}
   323  			}
   324  		}
   325  	}
   326  	// TODO(mvdan): move Unlock back under Lock with a defer statement once
   327  	// https://go.dev/issue/38471 is fixed to remove the performance penalty.
   328  	f.mutex.Unlock()
   329  	return
   330  }
   331  
   332  func (f *File) position(p Pos, adjusted bool) (pos Position) {
   333  	offset := int(p) - f.base
   334  	pos.Offset = offset
   335  	pos.Filename, pos.Line, pos.Column = f.unpack(offset, adjusted)
   336  	return
   337  }
   338  
   339  // PositionFor returns the Position value for the given file position p.
   340  // If adjusted is set, the position may be adjusted by position-altering
   341  // //line comments; otherwise those comments are ignored.
   342  // p must be a Pos value in f or NoPos.
   343  func (f *File) PositionFor(p Pos, adjusted bool) (pos Position) {
   344  	if p != NoPos {
   345  		if int(p) < f.base || int(p) > f.base+f.size {
   346  			panic(fmt.Sprintf("invalid Pos value %d (should be in [%d, %d])", p, f.base, f.base+f.size))
   347  		}
   348  		pos = f.position(p, adjusted)
   349  	}
   350  	return
   351  }
   352  
   353  // Position returns the Position value for the given file position p.
   354  // Calling f.Position(p) is equivalent to calling f.PositionFor(p, true).
   355  func (f *File) Position(p Pos) (pos Position) {
   356  	return f.PositionFor(p, true)
   357  }
   358  
   359  // -----------------------------------------------------------------------------
   360  // FileSet
   361  
   362  // A FileSet represents a set of source files.
   363  // Methods of file sets are synchronized; multiple goroutines
   364  // may invoke them concurrently.
   365  //
   366  // The byte offsets for each file in a file set are mapped into
   367  // distinct (integer) intervals, one interval [base, base+size]
   368  // per file. [FileSet.Base] represents the first byte in the file, and size
   369  // is the corresponding file size. A [Pos] value is a value in such
   370  // an interval. By determining the interval a [Pos] value belongs
   371  // to, the file, its file base, and thus the byte offset (position)
   372  // the [Pos] value is representing can be computed.
   373  //
   374  // When adding a new file, a file base must be provided. That can
   375  // be any integer value that is past the end of any interval of any
   376  // file already in the file set. For convenience, [FileSet.Base] provides
   377  // such a value, which is simply the end of the Pos interval of the most
   378  // recently added file, plus one. Unless there is a need to extend an
   379  // interval later, using the [FileSet.Base] should be used as argument
   380  // for [FileSet.AddFile].
   381  //
   382  // A [File] may be removed from a FileSet when it is no longer needed.
   383  // This may reduce memory usage in a long-running application.
   384  type FileSet struct {
   385  	mutex sync.RWMutex         // protects the file set
   386  	base  int                  // base offset for the next file
   387  	files []*File              // list of files in the order added to the set
   388  	last  atomic.Pointer[File] // cache of last file looked up
   389  }
   390  
   391  // NewFileSet creates a new file set.
   392  func NewFileSet() *FileSet {
   393  	return &FileSet{
   394  		base: 1, // 0 == NoPos
   395  	}
   396  }
   397  
   398  // Base returns the minimum base offset that must be provided to
   399  // [FileSet.AddFile] when adding the next file.
   400  func (s *FileSet) Base() int {
   401  	s.mutex.RLock()
   402  	b := s.base
   403  	s.mutex.RUnlock()
   404  	return b
   405  }
   406  
   407  // AddFile adds a new file with a given filename, base offset, and file size
   408  // to the file set s and returns the file. Multiple files may have the same
   409  // name. The base offset must not be smaller than the [FileSet.Base], and
   410  // size must not be negative. As a special case, if a negative base is provided,
   411  // the current value of the [FileSet.Base] is used instead.
   412  //
   413  // Adding the file will set the file set's [FileSet.Base] value to base + size + 1
   414  // as the minimum base value for the next file. The following relationship
   415  // exists between a [Pos] value p for a given file offset offs:
   416  //
   417  //	int(p) = base + offs
   418  //
   419  // with offs in the range [0, size] and thus p in the range [base, base+size].
   420  // For convenience, [File.Pos] may be used to create file-specific position
   421  // values from a file offset.
   422  func (s *FileSet) AddFile(filename string, base, size int) *File {
   423  	// Allocate f outside the critical section.
   424  	f := &File{name: filename, size: size, lines: []int{0}}
   425  
   426  	s.mutex.Lock()
   427  	defer s.mutex.Unlock()
   428  	if base < 0 {
   429  		base = s.base
   430  	}
   431  	if base < s.base {
   432  		panic(fmt.Sprintf("invalid base %d (should be >= %d)", base, s.base))
   433  	}
   434  	f.base = base
   435  	if size < 0 {
   436  		panic(fmt.Sprintf("invalid size %d (should be >= 0)", size))
   437  	}
   438  	// base >= s.base && size >= 0
   439  	base += size + 1 // +1 because EOF also has a position
   440  	if base < 0 {
   441  		panic("token.Pos offset overflow (> 2G of source code in file set)")
   442  	}
   443  	// add the file to the file set
   444  	s.base = base
   445  	s.files = append(s.files, f)
   446  	s.last.Store(f)
   447  	return f
   448  }
   449  
   450  // RemoveFile removes a file from the [FileSet] so that subsequent
   451  // queries for its [Pos] interval yield a negative result.
   452  // This reduces the memory usage of a long-lived [FileSet] that
   453  // encounters an unbounded stream of files.
   454  //
   455  // Removing a file that does not belong to the set has no effect.
   456  func (s *FileSet) RemoveFile(file *File) {
   457  	s.last.CompareAndSwap(file, nil) // clear last file cache
   458  
   459  	s.mutex.Lock()
   460  	defer s.mutex.Unlock()
   461  
   462  	if i := searchFiles(s.files, file.base); i >= 0 && s.files[i] == file {
   463  		last := &s.files[len(s.files)-1]
   464  		s.files = append(s.files[:i], s.files[i+1:]...)
   465  		*last = nil // don't prolong lifetime when popping last element
   466  	}
   467  }
   468  
   469  // Iterate calls f for the files in the file set in the order they were added
   470  // until f returns false.
   471  func (s *FileSet) Iterate(f func(*File) bool) {
   472  	for i := 0; ; i++ {
   473  		var file *File
   474  		s.mutex.RLock()
   475  		if i < len(s.files) {
   476  			file = s.files[i]
   477  		}
   478  		s.mutex.RUnlock()
   479  		if file == nil || !f(file) {
   480  			break
   481  		}
   482  	}
   483  }
   484  
   485  func searchFiles(a []*File, x int) int {
   486  	return sort.Search(len(a), func(i int) bool { return a[i].base > x }) - 1
   487  }
   488  
   489  func (s *FileSet) file(p Pos) *File {
   490  	// common case: p is in last file.
   491  	if f := s.last.Load(); f != nil && f.base <= int(p) && int(p) <= f.base+f.size {
   492  		return f
   493  	}
   494  
   495  	s.mutex.RLock()
   496  	defer s.mutex.RUnlock()
   497  
   498  	// p is not in last file - search all files
   499  	if i := searchFiles(s.files, int(p)); i >= 0 {
   500  		f := s.files[i]
   501  		// f.base <= int(p) by definition of searchFiles
   502  		if int(p) <= f.base+f.size {
   503  			// Update cache of last file. A race is ok,
   504  			// but an exclusive lock causes heavy contention.
   505  			s.last.Store(f)
   506  			return f
   507  		}
   508  	}
   509  	return nil
   510  }
   511  
   512  // File returns the file that contains the position p.
   513  // If no such file is found (for instance for p == [NoPos]),
   514  // the result is nil.
   515  func (s *FileSet) File(p Pos) (f *File) {
   516  	if p != NoPos {
   517  		f = s.file(p)
   518  	}
   519  	return
   520  }
   521  
   522  // PositionFor converts a [Pos] p in the fileset into a [Position] value.
   523  // If adjusted is set, the position may be adjusted by position-altering
   524  // //line comments; otherwise those comments are ignored.
   525  // p must be a [Pos] value in s or [NoPos].
   526  func (s *FileSet) PositionFor(p Pos, adjusted bool) (pos Position) {
   527  	if p != NoPos {
   528  		if f := s.file(p); f != nil {
   529  			return f.position(p, adjusted)
   530  		}
   531  	}
   532  	return
   533  }
   534  
   535  // Position converts a [Pos] p in the fileset into a Position value.
   536  // Calling s.Position(p) is equivalent to calling s.PositionFor(p, true).
   537  func (s *FileSet) Position(p Pos) (pos Position) {
   538  	return s.PositionFor(p, true)
   539  }
   540  
   541  // -----------------------------------------------------------------------------
   542  // Helper functions
   543  
   544  func searchInts(a []int, x int) int {
   545  	// This function body is a manually inlined version of:
   546  	//
   547  	//   return sort.Search(len(a), func(i int) bool { return a[i] > x }) - 1
   548  	//
   549  	// With better compiler optimizations, this may not be needed in the
   550  	// future, but at the moment this change improves the go/printer
   551  	// benchmark performance by ~30%. This has a direct impact on the
   552  	// speed of gofmt and thus seems worthwhile (2011-04-29).
   553  	// TODO(gri): Remove this when compilers have caught up.
   554  	i, j := 0, len(a)
   555  	for i < j {
   556  		h := int(uint(i+j) >> 1) // avoid overflow when computing h
   557  		// i ≤ h < j
   558  		if a[h] <= x {
   559  			i = h + 1
   560  		} else {
   561  			j = h
   562  		}
   563  	}
   564  	return i - 1
   565  }
   566  

View as plain text