tabwriter.mx raw

   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 tabwriter implements a write filter (tabwriter.Writer) that
   6  // translates tabbed columns in input into properly aligned text.
   7  //
   8  // The package is using the Elastic Tabstops algorithm described at
   9  // http://nickgravgaard.com/elastictabstops/index.html.
  10  //
  11  // The text/tabwriter package is frozen and is not accepting new features.
  12  package tabwriter
  13  
  14  import (
  15  	"fmt"
  16  	"io"
  17  	"unicode/utf8"
  18  )
  19  
  20  // ----------------------------------------------------------------------------
  21  // Filter implementation
  22  
  23  // A cell represents a segment of text terminated by tabs or line breaks.
  24  // The text itself is stored in a separate buffer; cell only describes the
  25  // segment's size in bytes, its width in runes, and whether it's an htab
  26  // ('\t') terminated cell.
  27  type cell struct {
  28  	size  int  // cell size in bytes
  29  	width int  // cell width in runes
  30  	htab  bool // true if the cell is terminated by an htab ('\t')
  31  }
  32  
  33  // A Writer is a filter that inserts padding around tab-delimited
  34  // columns in its input to align them in the output.
  35  //
  36  // The Writer treats incoming bytes as UTF-8-encoded text consisting
  37  // of cells terminated by horizontal ('\t') or vertical ('\v') tabs,
  38  // and newline ('\n') or formfeed ('\f') characters; both newline and
  39  // formfeed act as line breaks.
  40  //
  41  // Tab-terminated cells in contiguous lines constitute a column. The
  42  // Writer inserts padding as needed to make all cells in a column have
  43  // the same width, effectively aligning the columns. It assumes that
  44  // all characters have the same width, except for tabs for which a
  45  // tabwidth must be specified. Column cells must be tab-terminated, not
  46  // tab-separated: non-tab terminated trailing text at the end of a line
  47  // forms a cell but that cell is not part of an aligned column.
  48  // For instance, in this example (where | stands for a horizontal tab):
  49  //
  50  //	aaaa|bbb|d
  51  //	aa  |b  |dd
  52  //	a   |
  53  //	aa  |cccc|eee
  54  //
  55  // the b and c are in distinct columns (the b column is not contiguous
  56  // all the way). The d and e are not in a column at all (there's no
  57  // terminating tab, nor would the column be contiguous).
  58  //
  59  // The Writer assumes that all Unicode code points have the same width;
  60  // this may not be true in some fonts or if the string contains combining
  61  // characters.
  62  //
  63  // If [DiscardEmptyColumns] is set, empty columns that are terminated
  64  // entirely by vertical (or "soft") tabs are discarded. Columns
  65  // terminated by horizontal (or "hard") tabs are not affected by
  66  // this flag.
  67  //
  68  // If a Writer is configured to filter HTML, HTML tags and entities
  69  // are passed through. The widths of tags and entities are
  70  // assumed to be zero (tags) and one (entities) for formatting purposes.
  71  //
  72  // A segment of text may be escaped by bracketing it with [Escape]
  73  // characters. The tabwriter passes escaped text segments through
  74  // unchanged. In particular, it does not interpret any tabs or line
  75  // breaks within the segment. If the [StripEscape] flag is set, the
  76  // Escape characters are stripped from the output; otherwise they
  77  // are passed through as well. For the purpose of formatting, the
  78  // width of the escaped text is always computed excluding the Escape
  79  // characters.
  80  //
  81  // The formfeed character acts like a newline but it also terminates
  82  // all columns in the current line (effectively calling [Writer.Flush]). Tab-
  83  // terminated cells in the next line start new columns. Unless found
  84  // inside an HTML tag or inside an escaped text segment, formfeed
  85  // characters appear as newlines in the output.
  86  //
  87  // The Writer must buffer input internally, because proper spacing
  88  // of one line may depend on the cells in future lines. Clients must
  89  // call Flush when done calling [Writer.Write].
  90  type Writer struct {
  91  	// configuration
  92  	output   io.Writer
  93  	minwidth int
  94  	tabwidth int
  95  	padding  int
  96  	padbytes [8]byte
  97  	flags    uint
  98  
  99  	// current state
 100  	buf     []byte   // collected text excluding tabs or line breaks
 101  	pos     int      // buffer position up to which cell.width of incomplete cell has been computed
 102  	cell    cell     // current incomplete cell; cell.width is up to buf[pos] excluding ignored sections
 103  	endChar byte     // terminating char of escaped sequence (Escape for escapes, '>', ';' for HTML tags/entities, or 0)
 104  	lines   [][]cell // list of lines; each line is a list of cells
 105  	widths  []int    // list of column widths in runes - re-used during formatting
 106  }
 107  
 108  // addLine adds a new line.
 109  // flushed is a hint indicating whether the underlying writer was just flushed.
 110  // If so, the previous line is not likely to be a good indicator of the new line's cells.
 111  func (b *Writer) addLine(flushed bool) {
 112  	// Grow slice instead of appending,
 113  	// as that gives us an opportunity
 114  	// to re-use an existing []cell.
 115  	if n := len(b.lines) + 1; n <= cap(b.lines) {
 116  		b.lines = b.lines[:n]
 117  		b.lines[n-1] = b.lines[n-1][:0]
 118  	} else {
 119  		b.lines = append(b.lines, nil)
 120  	}
 121  
 122  	if !flushed {
 123  		// The previous line is probably a good indicator
 124  		// of how many cells the current line will have.
 125  		// If the current line's capacity is smaller than that,
 126  		// abandon it and make a new one.
 127  		if n := len(b.lines); n >= 2 {
 128  			if prev := len(b.lines[n-2]); prev > cap(b.lines[n-1]) {
 129  				b.lines[n-1] = []cell{:0:prev}
 130  			}
 131  		}
 132  	}
 133  }
 134  
 135  // Reset the current state.
 136  func (b *Writer) reset() {
 137  	b.buf = b.buf[:0]
 138  	b.pos = 0
 139  	b.cell = cell{}
 140  	b.endChar = 0
 141  	b.lines = b.lines[0:0]
 142  	b.widths = b.widths[0:0]
 143  	b.addLine(true)
 144  }
 145  
 146  // Internal representation (current state):
 147  //
 148  // - all text written is appended to buf; tabs and line breaks are stripped away
 149  // - at any given time there is a (possibly empty) incomplete cell at the end
 150  //   (the cell starts after a tab or line break)
 151  // - cell.size is the number of bytes belonging to the cell so far
 152  // - cell.width is text width in runes of that cell from the start of the cell to
 153  //   position pos; html tags and entities are excluded from this width if html
 154  //   filtering is enabled
 155  // - the sizes and widths of processed text are kept in the lines list
 156  //   which contains a list of cells for each line
 157  // - the widths list is a temporary list with current widths used during
 158  //   formatting; it is kept in Writer because it's re-used
 159  //
 160  //                    |<---------- size ---------->|
 161  //                    |                            |
 162  //                    |<- width ->|<- ignored ->|  |
 163  //                    |           |             |  |
 164  // [---processed---tab------------<tag>...</tag>...]
 165  // ^                  ^                         ^
 166  // |                  |                         |
 167  // buf                start of incomplete cell  pos
 168  
 169  // Formatting can be controlled with these flags.
 170  const (
 171  	// Ignore html tags and treat entities (starting with '&'
 172  	// and ending in ';') as single characters (width = 1).
 173  	FilterHTML uint = 1 << iota
 174  
 175  	// Strip Escape characters bracketing escaped text segments
 176  	// instead of passing them through unchanged with the text.
 177  	StripEscape
 178  
 179  	// Force right-alignment of cell content.
 180  	// Default is left-alignment.
 181  	AlignRight
 182  
 183  	// Handle empty columns as if they were not present in
 184  	// the input in the first place.
 185  	DiscardEmptyColumns
 186  
 187  	// Always use tabs for indentation columns (i.e., padding of
 188  	// leading empty cells on the left) independent of padchar.
 189  	TabIndent
 190  
 191  	// Print a vertical bar ('|') between columns (after formatting).
 192  	// Discarded columns appear as zero-width columns ("||").
 193  	Debug
 194  )
 195  
 196  // A [Writer] must be initialized with a call to Init. The first parameter (output)
 197  // specifies the filter output. The remaining parameters control the formatting:
 198  //
 199  //	minwidth	minimal cell width including any padding
 200  //	tabwidth	width of tab characters (equivalent number of spaces)
 201  //	padding		padding added to a cell before computing its width
 202  //	padchar		ASCII char used for padding
 203  //			if padchar == '\t', the Writer will assume that the
 204  //			width of a '\t' in the formatted output is tabwidth,
 205  //			and cells are left-aligned independent of align_left
 206  //			(for correct-looking results, tabwidth must correspond
 207  //			to the tab width in the viewer displaying the result)
 208  //	flags		formatting control
 209  func (b *Writer) Init(output io.Writer, minwidth, tabwidth, padding int, padchar byte, flags uint) *Writer {
 210  	if minwidth < 0 || tabwidth < 0 || padding < 0 {
 211  		panic("negative minwidth, tabwidth, or padding")
 212  	}
 213  	b.output = output
 214  	b.minwidth = minwidth
 215  	b.tabwidth = tabwidth
 216  	b.padding = padding
 217  	for i := range b.padbytes {
 218  		b.padbytes[i] = padchar
 219  	}
 220  	if padchar == '\t' {
 221  		// tab padding enforces left-alignment
 222  		flags &^= AlignRight
 223  	}
 224  	b.flags = flags
 225  
 226  	b.reset()
 227  
 228  	return b
 229  }
 230  
 231  // debugging support (keep code around)
 232  func (b *Writer) dump() {
 233  	pos := 0
 234  	for i, line := range b.lines {
 235  		print("(", i, ") ")
 236  		for _, c := range line {
 237  			print("[", string(b.buf[pos:pos+c.size]), "]")
 238  			pos += c.size
 239  		}
 240  		print("\n")
 241  	}
 242  	print("\n")
 243  }
 244  
 245  // local error wrapper so we can distinguish errors we want to return
 246  // as errors from genuine panics (which we don't want to return as errors)
 247  type osError struct {
 248  	err error
 249  }
 250  
 251  func (b *Writer) write0(buf []byte) {
 252  	n, err := b.output.Write(buf)
 253  	if n != len(buf) && err == nil {
 254  		err = io.ErrShortWrite
 255  	}
 256  	if err != nil {
 257  		panic(osError{err})
 258  	}
 259  }
 260  
 261  func (b *Writer) writeN(src []byte, n int) {
 262  	for n > len(src) {
 263  		b.write0(src)
 264  		n -= len(src)
 265  	}
 266  	b.write0(src[0:n])
 267  }
 268  
 269  var (
 270  	newline = []byte{'\n'}
 271  	tabs    = []byte("\t\t\t\t\t\t\t\t")
 272  )
 273  
 274  func (b *Writer) writePadding(textw, cellw int, useTabs bool) {
 275  	if b.padbytes[0] == '\t' || useTabs {
 276  		// padding is done with tabs
 277  		if b.tabwidth == 0 {
 278  			return // tabs have no width - can't do any padding
 279  		}
 280  		// make cellw the smallest multiple of b.tabwidth
 281  		cellw = (cellw + b.tabwidth - 1) / b.tabwidth * b.tabwidth
 282  		n := cellw - textw // amount of padding
 283  		if n < 0 {
 284  			panic("internal error")
 285  		}
 286  		b.writeN(tabs, (n+b.tabwidth-1)/b.tabwidth)
 287  		return
 288  	}
 289  
 290  	// padding is done with non-tab characters
 291  	b.writeN(b.padbytes[0:], cellw-textw)
 292  }
 293  
 294  var vbar = []byte{'|'}
 295  
 296  func (b *Writer) writeLines(pos0 int, line0, line1 int) (pos int) {
 297  	pos = pos0
 298  	for i := line0; i < line1; i++ {
 299  		line := b.lines[i]
 300  
 301  		// if TabIndent is set, use tabs to pad leading empty cells
 302  		useTabs := b.flags&TabIndent != 0
 303  
 304  		for j, c := range line {
 305  			if j > 0 && b.flags&Debug != 0 {
 306  				// indicate column break
 307  				b.write0(vbar)
 308  			}
 309  
 310  			if c.size == 0 {
 311  				// empty cell
 312  				if j < len(b.widths) {
 313  					b.writePadding(c.width, b.widths[j], useTabs)
 314  				}
 315  			} else {
 316  				// non-empty cell
 317  				useTabs = false
 318  				if b.flags&AlignRight == 0 { // align left
 319  					b.write0(b.buf[pos : pos+c.size])
 320  					pos += c.size
 321  					if j < len(b.widths) {
 322  						b.writePadding(c.width, b.widths[j], false)
 323  					}
 324  				} else { // align right
 325  					if j < len(b.widths) {
 326  						b.writePadding(c.width, b.widths[j], false)
 327  					}
 328  					b.write0(b.buf[pos : pos+c.size])
 329  					pos += c.size
 330  				}
 331  			}
 332  		}
 333  
 334  		if i+1 == len(b.lines) {
 335  			// last buffered line - we don't have a newline, so just write
 336  			// any outstanding buffered data
 337  			b.write0(b.buf[pos : pos+b.cell.size])
 338  			pos += b.cell.size
 339  		} else {
 340  			// not the last line - write newline
 341  			b.write0(newline)
 342  		}
 343  	}
 344  	return
 345  }
 346  
 347  // Format the text between line0 and line1 (excluding line1); pos
 348  // is the buffer position corresponding to the beginning of line0.
 349  // Returns the buffer position corresponding to the beginning of
 350  // line1 and an error, if any.
 351  func (b *Writer) format(pos0 int, line0, line1 int) (pos int) {
 352  	pos = pos0
 353  	column := len(b.widths)
 354  	for this := line0; this < line1; this++ {
 355  		line := b.lines[this]
 356  
 357  		if column >= len(line)-1 {
 358  			continue
 359  		}
 360  		// cell exists in this column => this line
 361  		// has more cells than the previous line
 362  		// (the last cell per line is ignored because cells are
 363  		// tab-terminated; the last cell per line describes the
 364  		// text before the newline/formfeed and does not belong
 365  		// to a column)
 366  
 367  		// print unprinted lines until beginning of block
 368  		pos = b.writeLines(pos, line0, this)
 369  		line0 = this
 370  
 371  		// column block begin
 372  		width := b.minwidth // minimal column width
 373  		discardable := true // true if all cells in this column are empty and "soft"
 374  		for ; this < line1; this++ {
 375  			line = b.lines[this]
 376  			if column >= len(line)-1 {
 377  				break
 378  			}
 379  			// cell exists in this column
 380  			c := line[column]
 381  			// update width
 382  			if w := c.width + b.padding; w > width {
 383  				width = w
 384  			}
 385  			// update discardable
 386  			if c.width > 0 || c.htab {
 387  				discardable = false
 388  			}
 389  		}
 390  		// column block end
 391  
 392  		// discard empty columns if necessary
 393  		if discardable && b.flags&DiscardEmptyColumns != 0 {
 394  			width = 0
 395  		}
 396  
 397  		// format and print all columns to the right of this column
 398  		// (we know the widths of this column and all columns to the left)
 399  		b.widths = append(b.widths, width) // push width
 400  		pos = b.format(pos, line0, this)
 401  		b.widths = b.widths[0 : len(b.widths)-1] // pop width
 402  		line0 = this
 403  	}
 404  
 405  	// print unprinted lines until end
 406  	return b.writeLines(pos, line0, line1)
 407  }
 408  
 409  // Append text to current cell.
 410  func (b *Writer) append(text []byte) {
 411  	b.buf = append(b.buf, text...)
 412  	b.cell.size += len(text)
 413  }
 414  
 415  // Update the cell width.
 416  func (b *Writer) updateWidth() {
 417  	b.cell.width += utf8.RuneCount(b.buf[b.pos:])
 418  	b.pos = len(b.buf)
 419  }
 420  
 421  // To escape a text segment, bracket it with Escape characters.
 422  // For instance, the tab in this string "Ignore this tab: \xff\t\xff"
 423  // does not terminate a cell and constitutes a single character of
 424  // width one for formatting purposes.
 425  //
 426  // The value 0xff was chosen because it cannot appear in a valid UTF-8 sequence.
 427  const Escape = '\xff'
 428  
 429  // Start escaped mode.
 430  func (b *Writer) startEscape(ch byte) {
 431  	switch ch {
 432  	case Escape:
 433  		b.endChar = Escape
 434  	case '<':
 435  		b.endChar = '>'
 436  	case '&':
 437  		b.endChar = ';'
 438  	}
 439  }
 440  
 441  // Terminate escaped mode. If the escaped text was an HTML tag, its width
 442  // is assumed to be zero for formatting purposes; if it was an HTML entity,
 443  // its width is assumed to be one. In all other cases, the width is the
 444  // unicode width of the text.
 445  func (b *Writer) endEscape() {
 446  	switch b.endChar {
 447  	case Escape:
 448  		b.updateWidth()
 449  		if b.flags&StripEscape == 0 {
 450  			b.cell.width -= 2 // don't count the Escape chars
 451  		}
 452  	case '>': // tag of zero width
 453  	case ';':
 454  		b.cell.width++ // entity, count as one rune
 455  	}
 456  	b.pos = len(b.buf)
 457  	b.endChar = 0
 458  }
 459  
 460  // Terminate the current cell by adding it to the list of cells of the
 461  // current line. Returns the number of cells in that line.
 462  func (b *Writer) terminateCell(htab bool) int {
 463  	b.cell.htab = htab
 464  	line := &b.lines[len(b.lines)-1]
 465  	*line = append(*line, b.cell)
 466  	b.cell = cell{}
 467  	return len(*line)
 468  }
 469  
 470  func (b *Writer) handlePanic(err *error, op string) {
 471  	if e := recover(); e != nil {
 472  		if op == "Flush" {
 473  			// If Flush ran into a panic, we still need to reset.
 474  			b.reset()
 475  		}
 476  		if nerr, ok := e.(osError); ok {
 477  			*err = nerr.err
 478  			return
 479  		}
 480  		panic(fmt.Sprintf("tabwriter: panic during %s (%v)", op, e))
 481  	}
 482  }
 483  
 484  // Flush should be called after the last call to [Writer.Write] to ensure
 485  // that any data buffered in the [Writer] is written to output. Any
 486  // incomplete escape sequence at the end is considered
 487  // complete for formatting purposes.
 488  func (b *Writer) Flush() error {
 489  	return b.flush()
 490  }
 491  
 492  // flush is the internal version of Flush, with a named return value which we
 493  // don't want to expose.
 494  func (b *Writer) flush() (err error) {
 495  	defer b.handlePanic(&err, "Flush")
 496  	b.flushNoDefers()
 497  	return nil
 498  }
 499  
 500  // flushNoDefers is like flush, but without a deferred handlePanic call. This
 501  // can be called from other methods which already have their own deferred
 502  // handlePanic calls, such as Write, and avoid the extra defer work.
 503  func (b *Writer) flushNoDefers() {
 504  	// add current cell if not empty
 505  	if b.cell.size > 0 {
 506  		if b.endChar != 0 {
 507  			// inside escape - terminate it even if incomplete
 508  			b.endEscape()
 509  		}
 510  		b.terminateCell(false)
 511  	}
 512  
 513  	// format contents of buffer
 514  	b.format(0, 0, len(b.lines))
 515  	b.reset()
 516  }
 517  
 518  var hbar = []byte("---\n")
 519  
 520  // Write writes buf to the writer b.
 521  // The only errors returned are ones encountered
 522  // while writing to the underlying output stream.
 523  func (b *Writer) Write(buf []byte) (n int, err error) {
 524  	defer b.handlePanic(&err, "Write")
 525  
 526  	// split text into cells
 527  	n = 0
 528  	for i, ch := range buf {
 529  		if b.endChar == 0 {
 530  			// outside escape
 531  			switch ch {
 532  			case '\t', '\v', '\n', '\f':
 533  				// end of cell
 534  				b.append(buf[n:i])
 535  				b.updateWidth()
 536  				n = i + 1 // ch consumed
 537  				ncells := b.terminateCell(ch == '\t')
 538  				if ch == '\n' || ch == '\f' {
 539  					// terminate line
 540  					b.addLine(ch == '\f')
 541  					if ch == '\f' || ncells == 1 {
 542  						// A '\f' always forces a flush. Otherwise, if the previous
 543  						// line has only one cell which does not have an impact on
 544  						// the formatting of the following lines (the last cell per
 545  						// line is ignored by format()), thus we can flush the
 546  						// Writer contents.
 547  						b.flushNoDefers()
 548  						if ch == '\f' && b.flags&Debug != 0 {
 549  							// indicate section break
 550  							b.write0(hbar)
 551  						}
 552  					}
 553  				}
 554  
 555  			case Escape:
 556  				// start of escaped sequence
 557  				b.append(buf[n:i])
 558  				b.updateWidth()
 559  				n = i
 560  				if b.flags&StripEscape != 0 {
 561  					n++ // strip Escape
 562  				}
 563  				b.startEscape(Escape)
 564  
 565  			case '<', '&':
 566  				// possibly an html tag/entity
 567  				if b.flags&FilterHTML != 0 {
 568  					// begin of tag/entity
 569  					b.append(buf[n:i])
 570  					b.updateWidth()
 571  					n = i
 572  					b.startEscape(ch)
 573  				}
 574  			}
 575  
 576  		} else {
 577  			// inside escape
 578  			if ch == b.endChar {
 579  				// end of tag/entity
 580  				j := i + 1
 581  				if ch == Escape && b.flags&StripEscape != 0 {
 582  					j = i // strip Escape
 583  				}
 584  				b.append(buf[n:j])
 585  				n = i + 1 // ch consumed
 586  				b.endEscape()
 587  			}
 588  		}
 589  	}
 590  
 591  	// append leftover text
 592  	b.append(buf[n:])
 593  	n = len(buf)
 594  	return
 595  }
 596  
 597  // NewWriter allocates and initializes a new [Writer].
 598  // The parameters are the same as for the Init function.
 599  func NewWriter(output io.Writer, minwidth, tabwidth, padding int, padchar byte, flags uint) *Writer {
 600  	return (&Writer{}).Init(output, minwidth, tabwidth, padding, padchar, flags)
 601  }
 602