The Go Programming Language

Source file src/pkg/tabwriter/tabwriter.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 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	package tabwriter
    12	
    13	import (
    14		"bytes"
    15		"io"
    16		"os"
    17		"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	//
    28	type cell struct {
    29		size  int  // cell size in bytes
    30		width int  // cell width in runes
    31		htab  bool // true if the cell is terminated by an htab ('\t')
    32	}
    33	
    34	// A Writer is a filter that inserts padding around tab-delimited
    35	// columns in its input to align them in the output.
    36	//
    37	// The Writer treats incoming bytes as UTF-8 encoded text consisting
    38	// of cells terminated by (horizontal or vertical) tabs or line
    39	// breaks (newline or formfeed characters). Cells in adjacent lines
    40	// constitute a column. The Writer inserts padding as needed to
    41	// make all cells in a column have the same width, effectively
    42	// aligning the columns. It assumes that all characters have the
    43	// same width except for tabs for which a tabwidth must be specified.
    44	// Note that cells are tab-terminated, not tab-separated: trailing
    45	// non-tab text at the end of a line does not form a column cell.
    46	//
    47	// The Writer assumes that all Unicode code points have the same width;
    48	// this may not be true in some fonts.
    49	//
    50	// If DiscardEmptyColumns is set, empty columns that are terminated
    51	// entirely by vertical (or "soft") tabs are discarded. Columns
    52	// terminated by horizontal (or "hard") tabs are not affected by
    53	// this flag.
    54	//
    55	// If a Writer is configured to filter HTML, HTML tags and entities
    56	// are simply passed through. The widths of tags and entities are
    57	// assumed to be zero (tags) and one (entities) for formatting purposes.
    58	//
    59	// A segment of text may be escaped by bracketing it with Escape
    60	// characters. The tabwriter passes escaped text segments through
    61	// unchanged. In particular, it does not interpret any tabs or line
    62	// breaks within the segment. If the StripEscape flag is set, the
    63	// Escape characters are stripped from the output; otherwise they
    64	// are passed through as well. For the purpose of formatting, the
    65	// width of the escaped text is always computed excluding the Escape
    66	// characters.
    67	//
    68	// The formfeed character ('\f') acts like a newline but it also
    69	// terminates all columns in the current line (effectively calling
    70	// Flush). Cells in the next line start new columns. Unless found
    71	// inside an HTML tag or inside an escaped text segment, formfeed
    72	// characters appear as newlines in the output.
    73	//
    74	// The Writer must buffer input internally, because proper spacing
    75	// of one line may depend on the cells in future lines. Clients must
    76	// call Flush when done calling Write.
    77	//
    78	type Writer struct {
    79		// configuration
    80		output   io.Writer
    81		minwidth int
    82		tabwidth int
    83		padding  int
    84		padbytes [8]byte
    85		flags    uint
    86	
    87		// current state
    88		buf     bytes.Buffer // collected text excluding tabs or line breaks
    89		pos     int          // buffer position up to which cell.width of incomplete cell has been computed
    90		cell    cell         // current incomplete cell; cell.width is up to buf[pos] excluding ignored sections
    91		endChar byte         // terminating char of escaped sequence (Escape for escapes, '>', ';' for HTML tags/entities, or 0)
    92		lines   [][]cell     // list of lines; each line is a list of cells
    93		widths  []int        // list of column widths in runes - re-used during formatting
    94	}
    95	
    96	func (b *Writer) addLine() { b.lines = append(b.lines, []cell{}) }
    97	
    98	// Reset the current state.
    99	func (b *Writer) reset() {
   100		b.buf.Reset()
   101		b.pos = 0
   102		b.cell = cell{}
   103		b.endChar = 0
   104		b.lines = b.lines[0:0]
   105		b.widths = b.widths[0:0]
   106		b.addLine()
   107	}
   108	
   109	// Internal representation (current state):
   110	//
   111	// - all text written is appended to buf; tabs and line breaks are stripped away
   112	// - at any given time there is a (possibly empty) incomplete cell at the end
   113	//   (the cell starts after a tab or line break)
   114	// - cell.size is the number of bytes belonging to the cell so far
   115	// - cell.width is text width in runes of that cell from the start of the cell to
   116	//   position pos; html tags and entities are excluded from this width if html
   117	//   filtering is enabled
   118	// - the sizes and widths of processed text are kept in the lines list
   119	//   which contains a list of cells for each line
   120	// - the widths list is a temporary list with current widths used during
   121	//   formatting; it is kept in Writer because it's re-used
   122	//
   123	//                    |<---------- size ---------->|
   124	//                    |                            |
   125	//                    |<- width ->|<- ignored ->|  |
   126	//                    |           |             |  |
   127	// [---processed---tab------------<tag>...</tag>...]
   128	// ^                  ^                         ^
   129	// |                  |                         |
   130	// buf                start of incomplete cell  pos
   131	
   132	// Formatting can be controlled with these flags.
   133	const (
   134		// Ignore html tags and treat entities (starting with '&'
   135		// and ending in ';') as single characters (width = 1).
   136		FilterHTML uint = 1 << iota
   137	
   138		// Strip Escape characters bracketing escaped text segments
   139		// instead of passing them through unchanged with the text.
   140		StripEscape
   141	
   142		// Force right-alignment of cell content.
   143		// Default is left-alignment.
   144		AlignRight
   145	
   146		// Handle empty columns as if they were not present in
   147		// the input in the first place.
   148		DiscardEmptyColumns
   149	
   150		// Always use tabs for indentation columns (i.e., padding of
   151		// leading empty cells on the left) independent of padchar.
   152		TabIndent
   153	
   154		// Print a vertical bar ('|') between columns (after formatting).
   155		// Discarded columns appear as zero-width columns ("||").
   156		Debug
   157	)
   158	
   159	// A Writer must be initialized with a call to Init. The first parameter (output)
   160	// specifies the filter output. The remaining parameters control the formatting:
   161	//
   162	//	minwidth	minimal cell width including any padding
   163	//	tabwidth	width of tab characters (equivalent number of spaces)
   164	//	padding		padding added to a cell before computing its width
   165	//	padchar		ASCII char used for padding
   166	//			if padchar == '\t', the Writer will assume that the
   167	//			width of a '\t' in the formatted output is tabwidth,
   168	//			and cells are left-aligned independent of align_left
   169	//			(for correct-looking results, tabwidth must correspond
   170	//			to the tab width in the viewer displaying the result)
   171	//	flags		formatting control
   172	//
   173	// To format in tab-separated columns with a tab stop of 8:
   174	//	b.Init(w, 8, 1, 8, '\t', 0);
   175	//
   176	// To format in space-separated columns with at least 4 spaces between columns:
   177	//	b.Init(w, 0, 4, 8, ' ', 0);
   178	//
   179	func (b *Writer) Init(output io.Writer, minwidth, tabwidth, padding int, padchar byte, flags uint) *Writer {
   180		if minwidth < 0 || tabwidth < 0 || padding < 0 {
   181			panic("negative minwidth, tabwidth, or padding")
   182		}
   183		b.output = output
   184		b.minwidth = minwidth
   185		b.tabwidth = tabwidth
   186		b.padding = padding
   187		for i := range b.padbytes {
   188			b.padbytes[i] = padchar
   189		}
   190		if padchar == '\t' {
   191			// tab padding enforces left-alignment
   192			flags &^= AlignRight
   193		}
   194		b.flags = flags
   195	
   196		b.reset()
   197	
   198		return b
   199	}
   200	
   201	// debugging support (keep code around)
   202	func (b *Writer) dump() {
   203		pos := 0
   204		for i, line := range b.lines {
   205			print("(", i, ") ")
   206			for _, c := range line {
   207				print("[", string(b.buf.Bytes()[pos:pos+c.size]), "]")
   208				pos += c.size
   209			}
   210			print("\n")
   211		}
   212		print("\n")
   213	}
   214	
   215	// local error wrapper so we can distinguish os.Errors we want to return
   216	// as errors from genuine panics (which we don't want to return as errors)
   217	type osError struct {
   218		err os.Error
   219	}
   220	
   221	func (b *Writer) write0(buf []byte) {
   222		n, err := b.output.Write(buf)
   223		if n != len(buf) && err == nil {
   224			err = os.EIO
   225		}
   226		if err != nil {
   227			panic(osError{err})
   228		}
   229	}
   230	
   231	func (b *Writer) writeN(src []byte, n int) {
   232		for n > len(src) {
   233			b.write0(src)
   234			n -= len(src)
   235		}
   236		b.write0(src[0:n])
   237	}
   238	
   239	var (
   240		newline = []byte{'\n'}
   241		tabs    = []byte("\t\t\t\t\t\t\t\t")
   242	)
   243	
   244	func (b *Writer) writePadding(textw, cellw int, useTabs bool) {
   245		if b.padbytes[0] == '\t' || useTabs {
   246			// padding is done with tabs
   247			if b.tabwidth == 0 {
   248				return // tabs have no width - can't do any padding
   249			}
   250			// make cellw the smallest multiple of b.tabwidth
   251			cellw = (cellw + b.tabwidth - 1) / b.tabwidth * b.tabwidth
   252			n := cellw - textw // amount of padding
   253			if n < 0 {
   254				panic("internal error")
   255			}
   256			b.writeN(tabs, (n+b.tabwidth-1)/b.tabwidth)
   257			return
   258		}
   259	
   260		// padding is done with non-tab characters
   261		b.writeN(b.padbytes[0:], cellw-textw)
   262	}
   263	
   264	var vbar = []byte{'|'}
   265	
   266	func (b *Writer) writeLines(pos0 int, line0, line1 int) (pos int) {
   267		pos = pos0
   268		for i := line0; i < line1; i++ {
   269			line := b.lines[i]
   270	
   271			// if TabIndent is set, use tabs to pad leading empty cells
   272			useTabs := b.flags&TabIndent != 0
   273	
   274			for j, c := range line {
   275				if j > 0 && b.flags&Debug != 0 {
   276					// indicate column break
   277					b.write0(vbar)
   278				}
   279	
   280				if c.size == 0 {
   281					// empty cell
   282					if j < len(b.widths) {
   283						b.writePadding(c.width, b.widths[j], useTabs)
   284					}
   285				} else {
   286					// non-empty cell
   287					useTabs = false
   288					if b.flags&AlignRight == 0 { // align left
   289						b.write0(b.buf.Bytes()[pos : pos+c.size])
   290						pos += c.size
   291						if j < len(b.widths) {
   292							b.writePadding(c.width, b.widths[j], false)
   293						}
   294					} else { // align right
   295						if j < len(b.widths) {
   296							b.writePadding(c.width, b.widths[j], false)
   297						}
   298						b.write0(b.buf.Bytes()[pos : pos+c.size])
   299						pos += c.size
   300					}
   301				}
   302			}
   303	
   304			if i+1 == len(b.lines) {
   305				// last buffered line - we don't have a newline, so just write
   306				// any outstanding buffered data
   307				b.write0(b.buf.Bytes()[pos : pos+b.cell.size])
   308				pos += b.cell.size
   309			} else {
   310				// not the last line - write newline
   311				b.write0(newline)
   312			}
   313		}
   314		return
   315	}
   316	
   317	// Format the text between line0 and line1 (excluding line1); pos
   318	// is the buffer position corresponding to the beginning of line0.
   319	// Returns the buffer position corresponding to the beginning of
   320	// line1 and an error, if any.
   321	//
   322	func (b *Writer) format(pos0 int, line0, line1 int) (pos int) {
   323		pos = pos0
   324		column := len(b.widths)
   325		for this := line0; this < line1; this++ {
   326			line := b.lines[this]
   327	
   328			if column < len(line)-1 {
   329				// cell exists in this column => this line
   330				// has more cells than the previous line
   331				// (the last cell per line is ignored because cells are
   332				// tab-terminated; the last cell per line describes the
   333				// text before the newline/formfeed and does not belong
   334				// to a column)
   335	
   336				// print unprinted lines until beginning of block
   337				pos = b.writeLines(pos, line0, this)
   338				line0 = this
   339	
   340				// column block begin
   341				width := b.minwidth // minimal column width
   342				discardable := true // true if all cells in this column are empty and "soft"
   343				for ; this < line1; this++ {
   344					line = b.lines[this]
   345					if column < len(line)-1 {
   346						// cell exists in this column
   347						c := line[column]
   348						// update width
   349						if w := c.width + b.padding; w > width {
   350							width = w
   351						}
   352						// update discardable
   353						if c.width > 0 || c.htab {
   354							discardable = false
   355						}
   356					} else {
   357						break
   358					}
   359				}
   360				// column block end
   361	
   362				// discard empty columns if necessary
   363				if discardable && b.flags&DiscardEmptyColumns != 0 {
   364					width = 0
   365				}
   366	
   367				// format and print all columns to the right of this column
   368				// (we know the widths of this column and all columns to the left)
   369				b.widths = append(b.widths, width) // push width
   370				pos = b.format(pos, line0, this)
   371				b.widths = b.widths[0 : len(b.widths)-1] // pop width
   372				line0 = this
   373			}
   374		}
   375	
   376		// print unprinted lines until end
   377		return b.writeLines(pos, line0, line1)
   378	}
   379	
   380	// Append text to current cell.
   381	func (b *Writer) append(text []byte) {
   382		b.buf.Write(text)
   383		b.cell.size += len(text)
   384	}
   385	
   386	// Update the cell width.
   387	func (b *Writer) updateWidth() {
   388		b.cell.width += utf8.RuneCount(b.buf.Bytes()[b.pos:b.buf.Len()])
   389		b.pos = b.buf.Len()
   390	}
   391	
   392	// To escape a text segment, bracket it with Escape characters.
   393	// For instance, the tab in this string "Ignore this tab: \xff\t\xff"
   394	// does not terminate a cell and constitutes a single character of
   395	// width one for formatting purposes.
   396	//
   397	// The value 0xff was chosen because it cannot appear in a valid UTF-8 sequence.
   398	//
   399	const Escape = '\xff'
   400	
   401	// Start escaped mode.
   402	func (b *Writer) startEscape(ch byte) {
   403		switch ch {
   404		case Escape:
   405			b.endChar = Escape
   406		case '<':
   407			b.endChar = '>'
   408		case '&':
   409			b.endChar = ';'
   410		}
   411	}
   412	
   413	// Terminate escaped mode. If the escaped text was an HTML tag, its width
   414	// is assumed to be zero for formatting purposes; if it was an HTML entity,
   415	// its width is assumed to be one. In all other cases, the width is the
   416	// unicode width of the text.
   417	//
   418	func (b *Writer) endEscape() {
   419		switch b.endChar {
   420		case Escape:
   421			b.updateWidth()
   422			if b.flags&StripEscape == 0 {
   423				b.cell.width -= 2 // don't count the Escape chars
   424			}
   425		case '>': // tag of zero width
   426		case ';':
   427			b.cell.width++ // entity, count as one rune
   428		}
   429		b.pos = b.buf.Len()
   430		b.endChar = 0
   431	}
   432	
   433	// Terminate the current cell by adding it to the list of cells of the
   434	// current line. Returns the number of cells in that line.
   435	//
   436	func (b *Writer) terminateCell(htab bool) int {
   437		b.cell.htab = htab
   438		line := &b.lines[len(b.lines)-1]
   439		*line = append(*line, b.cell)
   440		b.cell = cell{}
   441		return len(*line)
   442	}
   443	
   444	func handlePanic(err *os.Error) {
   445		if e := recover(); e != nil {
   446			*err = e.(osError).err // re-panics if it's not a local osError
   447		}
   448	}
   449	
   450	// Flush should be called after the last call to Write to ensure
   451	// that any data buffered in the Writer is written to output. Any
   452	// incomplete escape sequence at the end is simply considered
   453	// complete for formatting purposes.
   454	//
   455	func (b *Writer) Flush() (err os.Error) {
   456		defer b.reset() // even in the presence of errors
   457		defer handlePanic(&err)
   458	
   459		// add current cell if not empty
   460		if b.cell.size > 0 {
   461			if b.endChar != 0 {
   462				// inside escape - terminate it even if incomplete
   463				b.endEscape()
   464			}
   465			b.terminateCell(false)
   466		}
   467	
   468		// format contents of buffer
   469		b.format(0, 0, len(b.lines))
   470	
   471		return
   472	}
   473	
   474	var hbar = []byte("---\n")
   475	
   476	// Write writes buf to the writer b.
   477	// The only errors returned are ones encountered
   478	// while writing to the underlying output stream.
   479	//
   480	func (b *Writer) Write(buf []byte) (n int, err os.Error) {
   481		defer handlePanic(&err)
   482	
   483		// split text into cells
   484		n = 0
   485		for i, ch := range buf {
   486			if b.endChar == 0 {
   487				// outside escape
   488				switch ch {
   489				case '\t', '\v', '\n', '\f':
   490					// end of cell
   491					b.append(buf[n:i])
   492					b.updateWidth()
   493					n = i + 1 // ch consumed
   494					ncells := b.terminateCell(ch == '\t')
   495					if ch == '\n' || ch == '\f' {
   496						// terminate line
   497						b.addLine()
   498						if ch == '\f' || ncells == 1 {
   499							// A '\f' always forces a flush. Otherwise, if the previous
   500							// line has only one cell which does not have an impact on
   501							// the formatting of the following lines (the last cell per
   502							// line is ignored by format()), thus we can flush the
   503							// Writer contents.
   504							if err = b.Flush(); err != nil {
   505								return
   506							}
   507							if ch == '\f' && b.flags&Debug != 0 {
   508								// indicate section break
   509								b.write0(hbar)
   510							}
   511						}
   512					}
   513	
   514				case Escape:
   515					// start of escaped sequence
   516					b.append(buf[n:i])
   517					b.updateWidth()
   518					n = i
   519					if b.flags&StripEscape != 0 {
   520						n++ // strip Escape
   521					}
   522					b.startEscape(Escape)
   523	
   524				case '<', '&':
   525					// possibly an html tag/entity
   526					if b.flags&FilterHTML != 0 {
   527						// begin of tag/entity
   528						b.append(buf[n:i])
   529						b.updateWidth()
   530						n = i
   531						b.startEscape(ch)
   532					}
   533				}
   534	
   535			} else {
   536				// inside escape
   537				if ch == b.endChar {
   538					// end of tag/entity
   539					j := i + 1
   540					if ch == Escape && b.flags&StripEscape != 0 {
   541						j = i // strip Escape
   542					}
   543					b.append(buf[n:j])
   544					n = i + 1 // ch consumed
   545					b.endEscape()
   546				}
   547			}
   548		}
   549	
   550		// append leftover text
   551		b.append(buf[n:])
   552		n = len(buf)
   553		return
   554	}
   555	
   556	// NewWriter allocates and initializes a new tabwriter.Writer.
   557	// The parameters are the same as for the the Init function.
   558	//
   559	func NewWriter(output io.Writer, minwidth, tabwidth, padding int, padchar byte, flags uint) *Writer {
   560		return new(Writer).Init(output, minwidth, tabwidth, padding, padchar, flags)
   561	}

release.r60.3. Except as noted, this content is licensed under a Creative Commons Attribution 3.0 License.