mirror of
https://github.com/strongdm/comply
synced 2024-11-05 07:25:26 +00:00
588 lines
17 KiB
Go
588 lines
17 KiB
Go
// Copyright 2009 The Go Authors. All rights reserved.
|
|
// Use of this source code is governed by a BSD-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
// This file is mostly a copy of the go standard library text/tabwriter. With
|
|
// the additional stripping of ansi control characters for width calculations.
|
|
|
|
// Package tabwriter implements a write filter (tabwriter.Writer) that
|
|
// translates tabbed columns in input into properly aligned text.
|
|
//
|
|
// The package is using the Elastic Tabstops algorithm described at
|
|
// http://nickgravgaard.com/elastictabstops/index.html.
|
|
//
|
|
package tabwriter
|
|
|
|
import (
|
|
"bytes"
|
|
"io"
|
|
"unicode/utf8"
|
|
|
|
"github.com/lunixbochs/vtclean"
|
|
)
|
|
|
|
// ----------------------------------------------------------------------------
|
|
// Filter implementation
|
|
|
|
// A cell represents a segment of text terminated by tabs or line breaks.
|
|
// The text itself is stored in a separate buffer; cell only describes the
|
|
// segment's size in bytes, its width in runes, and whether it's an htab
|
|
// ('\t') terminated cell.
|
|
//
|
|
type cell struct {
|
|
size int // cell size in bytes
|
|
width int // cell width in runes
|
|
htab bool // true if the cell is terminated by an htab ('\t')
|
|
}
|
|
|
|
// A Writer is a filter that inserts padding around tab-delimited
|
|
// columns in its input to align them in the output.
|
|
//
|
|
// The Writer treats incoming bytes as UTF-8 encoded text consisting
|
|
// of cells terminated by (horizontal or vertical) tabs or line
|
|
// breaks (newline or formfeed characters). Cells in adjacent lines
|
|
// constitute a column. The Writer inserts padding as needed to
|
|
// make all cells in a column have the same width, effectively
|
|
// aligning the columns. It assumes that all characters have the
|
|
// same width except for tabs for which a tabwidth must be specified.
|
|
// Note that cells are tab-terminated, not tab-separated: trailing
|
|
// non-tab text at the end of a line does not form a column cell.
|
|
//
|
|
// The Writer assumes that all Unicode code points have the same width;
|
|
// this may not be true in some fonts.
|
|
//
|
|
// If DiscardEmptyColumns is set, empty columns that are terminated
|
|
// entirely by vertical (or "soft") tabs are discarded. Columns
|
|
// terminated by horizontal (or "hard") tabs are not affected by
|
|
// this flag.
|
|
//
|
|
// If a Writer is configured to filter HTML, HTML tags and entities
|
|
// are passed through. The widths of tags and entities are
|
|
// assumed to be zero (tags) and one (entities) for formatting purposes.
|
|
//
|
|
// A segment of text may be escaped by bracketing it with Escape
|
|
// characters. The tabwriter passes escaped text segments through
|
|
// unchanged. In particular, it does not interpret any tabs or line
|
|
// breaks within the segment. If the StripEscape flag is set, the
|
|
// Escape characters are stripped from the output; otherwise they
|
|
// are passed through as well. For the purpose of formatting, the
|
|
// width of the escaped text is always computed excluding the Escape
|
|
// characters.
|
|
//
|
|
// The formfeed character ('\f') acts like a newline but it also
|
|
// terminates all columns in the current line (effectively calling
|
|
// Flush). Cells in the next line start new columns. Unless found
|
|
// inside an HTML tag or inside an escaped text segment, formfeed
|
|
// characters appear as newlines in the output.
|
|
//
|
|
// The Writer must buffer input internally, because proper spacing
|
|
// of one line may depend on the cells in future lines. Clients must
|
|
// call Flush when done calling Write.
|
|
//
|
|
type Writer struct {
|
|
// configuration
|
|
output io.Writer
|
|
minwidth int
|
|
tabwidth int
|
|
padding int
|
|
padbytes [8]byte
|
|
flags uint
|
|
|
|
// current state
|
|
buf bytes.Buffer // collected text excluding tabs or line breaks
|
|
pos int // buffer position up to which cell.width of incomplete cell has been computed
|
|
cell cell // current incomplete cell; cell.width is up to buf[pos] excluding ignored sections
|
|
endChar byte // terminating char of escaped sequence (Escape for escapes, '>', ';' for HTML tags/entities, or 0)
|
|
lines [][]cell // list of lines; each line is a list of cells
|
|
widths []int // list of column widths in runes - re-used during formatting
|
|
alignment map[int]uint // column alignment
|
|
}
|
|
|
|
func (b *Writer) addLine() { b.lines = append(b.lines, []cell{}) }
|
|
|
|
// Reset the current state.
|
|
func (b *Writer) reset() {
|
|
b.buf.Reset()
|
|
b.pos = 0
|
|
b.cell = cell{}
|
|
b.endChar = 0
|
|
b.lines = b.lines[0:0]
|
|
b.widths = b.widths[0:0]
|
|
b.alignment = make(map[int]uint)
|
|
b.addLine()
|
|
}
|
|
|
|
// Internal representation (current state):
|
|
//
|
|
// - all text written is appended to buf; tabs and line breaks are stripped away
|
|
// - at any given time there is a (possibly empty) incomplete cell at the end
|
|
// (the cell starts after a tab or line break)
|
|
// - cell.size is the number of bytes belonging to the cell so far
|
|
// - cell.width is text width in runes of that cell from the start of the cell to
|
|
// position pos; html tags and entities are excluded from this width if html
|
|
// filtering is enabled
|
|
// - the sizes and widths of processed text are kept in the lines list
|
|
// which contains a list of cells for each line
|
|
// - the widths list is a temporary list with current widths used during
|
|
// formatting; it is kept in Writer because it's re-used
|
|
//
|
|
// |<---------- size ---------->|
|
|
// | |
|
|
// |<- width ->|<- ignored ->| |
|
|
// | | | |
|
|
// [---processed---tab------------<tag>...</tag>...]
|
|
// ^ ^ ^
|
|
// | | |
|
|
// buf start of incomplete cell pos
|
|
|
|
// Formatting can be controlled with these flags.
|
|
const (
|
|
// Ignore html tags and treat entities (starting with '&'
|
|
// and ending in ';') as single characters (width = 1).
|
|
FilterHTML uint = 1 << iota
|
|
|
|
// Strip Escape characters bracketing escaped text segments
|
|
// instead of passing them through unchanged with the text.
|
|
StripEscape
|
|
|
|
// Force right-alignment of cell content.
|
|
// Default is left-alignment.
|
|
AlignRight
|
|
|
|
// Handle empty columns as if they were not present in
|
|
// the input in the first place.
|
|
DiscardEmptyColumns
|
|
|
|
// Always use tabs for indentation columns (i.e., padding of
|
|
// leading empty cells on the left) independent of padchar.
|
|
TabIndent
|
|
|
|
// Print a vertical bar ('|') between columns (after formatting).
|
|
// Discarded columns appear as zero-width columns ("||").
|
|
Debug
|
|
)
|
|
|
|
// A Writer must be initialized with a call to Init. The first parameter (output)
|
|
// specifies the filter output. The remaining parameters control the formatting:
|
|
//
|
|
// minwidth minimal cell width including any padding
|
|
// tabwidth width of tab characters (equivalent number of spaces)
|
|
// padding padding added to a cell before computing its width
|
|
// padchar ASCII char used for padding
|
|
// if padchar == '\t', the Writer will assume that the
|
|
// width of a '\t' in the formatted output is tabwidth,
|
|
// and cells are left-aligned independent of align_left
|
|
// (for correct-looking results, tabwidth must correspond
|
|
// to the tab width in the viewer displaying the result)
|
|
// flags formatting control
|
|
//
|
|
func (b *Writer) Init(output io.Writer, minwidth, tabwidth, padding int, padchar byte, flags uint) *Writer {
|
|
if minwidth < 0 || tabwidth < 0 || padding < 0 {
|
|
panic("negative minwidth, tabwidth, or padding")
|
|
}
|
|
b.output = output
|
|
b.minwidth = minwidth
|
|
b.tabwidth = tabwidth
|
|
b.padding = padding
|
|
for i := range b.padbytes {
|
|
b.padbytes[i] = padchar
|
|
}
|
|
if padchar == '\t' {
|
|
// tab padding enforces left-alignment
|
|
flags &^= AlignRight
|
|
}
|
|
b.flags = flags
|
|
|
|
b.reset()
|
|
|
|
return b
|
|
}
|
|
|
|
// debugging support (keep code around)
|
|
func (b *Writer) dump() {
|
|
pos := 0
|
|
for i, line := range b.lines {
|
|
print("(", i, ") ")
|
|
for _, c := range line {
|
|
print("[", string(b.buf.Bytes()[pos:pos+c.size]), "]")
|
|
pos += c.size
|
|
}
|
|
print("\n")
|
|
}
|
|
print("\n")
|
|
}
|
|
|
|
// local error wrapper so we can distinguish errors we want to return
|
|
// as errors from genuine panics (which we don't want to return as errors)
|
|
type osError struct {
|
|
err error
|
|
}
|
|
|
|
func (b *Writer) write0(buf []byte) {
|
|
n, err := b.output.Write(buf)
|
|
if n != len(buf) && err == nil {
|
|
err = io.ErrShortWrite
|
|
}
|
|
if err != nil {
|
|
panic(osError{err})
|
|
}
|
|
}
|
|
|
|
func (b *Writer) writeN(src []byte, n int) {
|
|
for n > len(src) {
|
|
b.write0(src)
|
|
n -= len(src)
|
|
}
|
|
b.write0(src[0:n])
|
|
}
|
|
|
|
var (
|
|
newline = []byte{'\n'}
|
|
tabs = []byte("\t\t\t\t\t\t\t\t")
|
|
)
|
|
|
|
func (b *Writer) writePadding(textw, cellw int, useTabs bool) {
|
|
if b.padbytes[0] == '\t' || useTabs {
|
|
// padding is done with tabs
|
|
if b.tabwidth == 0 {
|
|
return // tabs have no width - can't do any padding
|
|
}
|
|
// make cellw the smallest multiple of b.tabwidth
|
|
cellw = (cellw + b.tabwidth - 1) / b.tabwidth * b.tabwidth
|
|
n := cellw - textw // amount of padding
|
|
if n < 0 {
|
|
panic("internal error")
|
|
}
|
|
b.writeN(tabs, (n+b.tabwidth-1)/b.tabwidth)
|
|
return
|
|
}
|
|
|
|
// padding is done with non-tab characters
|
|
b.writeN(b.padbytes[0:], cellw-textw)
|
|
}
|
|
|
|
var vbar = []byte{'|'}
|
|
|
|
func (b *Writer) writeLines(pos0 int, line0, line1 int) (pos int) {
|
|
pos = pos0
|
|
for i := line0; i < line1; i++ {
|
|
line := b.lines[i]
|
|
|
|
// if TabIndent is set, use tabs to pad leading empty cells
|
|
useTabs := b.flags&TabIndent != 0
|
|
|
|
for j, c := range line {
|
|
if j > 0 && b.flags&Debug != 0 {
|
|
// indicate column break
|
|
b.write0(vbar)
|
|
}
|
|
|
|
if c.size == 0 {
|
|
// empty cell
|
|
if j < len(b.widths) {
|
|
b.writePadding(c.width, b.widths[j], useTabs)
|
|
}
|
|
} else {
|
|
// non-empty cell
|
|
useTabs = false
|
|
alignColumnRight := b.alignment[j] == AlignRight
|
|
if (b.flags&AlignRight == 0) && !alignColumnRight { // align left
|
|
b.write0(b.buf.Bytes()[pos : pos+c.size])
|
|
pos += c.size
|
|
if j < len(b.widths) {
|
|
b.writePadding(c.width, b.widths[j], false)
|
|
}
|
|
} else if alignColumnRight && j < len(b.widths) {
|
|
// just this column
|
|
internalSize := b.widths[j] - b.padding
|
|
if j < len(b.widths) {
|
|
b.writePadding(c.width, internalSize, false)
|
|
}
|
|
b.write0(b.buf.Bytes()[pos : pos+c.size])
|
|
if b.padding > 0 {
|
|
b.writePadding(0, b.padding, false)
|
|
}
|
|
pos += c.size
|
|
} else { // align right
|
|
if j < len(b.widths) {
|
|
b.writePadding(c.width, b.widths[j], false)
|
|
}
|
|
b.write0(b.buf.Bytes()[pos : pos+c.size])
|
|
pos += c.size
|
|
}
|
|
}
|
|
}
|
|
|
|
if i+1 == len(b.lines) {
|
|
// last buffered line - we don't have a newline, so just write
|
|
// any outstanding buffered data
|
|
b.write0(b.buf.Bytes()[pos : pos+b.cell.size])
|
|
pos += b.cell.size
|
|
} else {
|
|
// not the last line - write newline
|
|
b.write0(newline)
|
|
}
|
|
}
|
|
return
|
|
}
|
|
|
|
// Format the text between line0 and line1 (excluding line1); pos
|
|
// is the buffer position corresponding to the beginning of line0.
|
|
// Returns the buffer position corresponding to the beginning of
|
|
// line1 and an error, if any.
|
|
//
|
|
func (b *Writer) format(pos0 int, line0, line1 int) (pos int) {
|
|
pos = pos0
|
|
column := len(b.widths)
|
|
for this := line0; this < line1; this++ {
|
|
line := b.lines[this]
|
|
|
|
if column < len(line)-1 {
|
|
// cell exists in this column => this line
|
|
// has more cells than the previous line
|
|
// (the last cell per line is ignored because cells are
|
|
// tab-terminated; the last cell per line describes the
|
|
// text before the newline/formfeed and does not belong
|
|
// to a column)
|
|
|
|
// print unprinted lines until beginning of block
|
|
pos = b.writeLines(pos, line0, this)
|
|
line0 = this
|
|
|
|
// column block begin
|
|
width := b.minwidth // minimal column width
|
|
discardable := true // true if all cells in this column are empty and "soft"
|
|
for ; this < line1; this++ {
|
|
line = b.lines[this]
|
|
if column < len(line)-1 {
|
|
// cell exists in this column
|
|
c := line[column]
|
|
// update width
|
|
if w := c.width + b.padding; w > width {
|
|
width = w
|
|
}
|
|
// update discardable
|
|
if c.width > 0 || c.htab {
|
|
discardable = false
|
|
}
|
|
} else {
|
|
break
|
|
}
|
|
}
|
|
// column block end
|
|
|
|
// discard empty columns if necessary
|
|
if discardable && b.flags&DiscardEmptyColumns != 0 {
|
|
width = 0
|
|
}
|
|
|
|
// format and print all columns to the right of this column
|
|
// (we know the widths of this column and all columns to the left)
|
|
b.widths = append(b.widths, width) // push width
|
|
pos = b.format(pos, line0, this)
|
|
b.widths = b.widths[0 : len(b.widths)-1] // pop width
|
|
line0 = this
|
|
}
|
|
}
|
|
|
|
// print unprinted lines until end
|
|
return b.writeLines(pos, line0, line1)
|
|
}
|
|
|
|
// Append text to current cell.
|
|
func (b *Writer) append(text []byte) {
|
|
b.buf.Write(text)
|
|
b.cell.size += len(text)
|
|
}
|
|
|
|
// Update the cell width.
|
|
func (b *Writer) updateWidth() {
|
|
// ---- Changes here -----
|
|
newChars := b.buf.Bytes()[b.pos:b.buf.Len()]
|
|
cleaned := vtclean.Clean(string(newChars), false) // false to strip colors
|
|
b.cell.width += utf8.RuneCount([]byte(cleaned))
|
|
// --- end of changes ----
|
|
b.pos = b.buf.Len()
|
|
}
|
|
|
|
// To escape a text segment, bracket it with Escape characters.
|
|
// For instance, the tab in this string "Ignore this tab: \xff\t\xff"
|
|
// does not terminate a cell and constitutes a single character of
|
|
// width one for formatting purposes.
|
|
//
|
|
// The value 0xff was chosen because it cannot appear in a valid UTF-8 sequence.
|
|
//
|
|
const Escape = '\xff'
|
|
|
|
// Start escaped mode.
|
|
func (b *Writer) startEscape(ch byte) {
|
|
switch ch {
|
|
case Escape:
|
|
b.endChar = Escape
|
|
case '<':
|
|
b.endChar = '>'
|
|
case '&':
|
|
b.endChar = ';'
|
|
}
|
|
}
|
|
|
|
// Terminate escaped mode. If the escaped text was an HTML tag, its width
|
|
// is assumed to be zero for formatting purposes; if it was an HTML entity,
|
|
// its width is assumed to be one. In all other cases, the width is the
|
|
// unicode width of the text.
|
|
//
|
|
func (b *Writer) endEscape() {
|
|
switch b.endChar {
|
|
case Escape:
|
|
b.updateWidth()
|
|
if b.flags&StripEscape == 0 {
|
|
b.cell.width -= 2 // don't count the Escape chars
|
|
}
|
|
case '>': // tag of zero width
|
|
case ';':
|
|
b.cell.width++ // entity, count as one rune
|
|
}
|
|
b.pos = b.buf.Len()
|
|
b.endChar = 0
|
|
}
|
|
|
|
// Terminate the current cell by adding it to the list of cells of the
|
|
// current line. Returns the number of cells in that line.
|
|
//
|
|
func (b *Writer) terminateCell(htab bool) int {
|
|
b.cell.htab = htab
|
|
line := &b.lines[len(b.lines)-1]
|
|
*line = append(*line, b.cell)
|
|
b.cell = cell{}
|
|
return len(*line)
|
|
}
|
|
|
|
func handlePanic(err *error, op string) {
|
|
if e := recover(); e != nil {
|
|
if nerr, ok := e.(osError); ok {
|
|
*err = nerr.err
|
|
return
|
|
}
|
|
panic("tabwriter: panic during " + op)
|
|
}
|
|
}
|
|
|
|
// Flush should be called after the last call to Write to ensure
|
|
// that any data buffered in the Writer is written to output. Any
|
|
// incomplete escape sequence at the end is considered
|
|
// complete for formatting purposes.
|
|
//
|
|
func (b *Writer) Flush() (err error) {
|
|
defer b.reset() // even in the presence of errors
|
|
defer handlePanic(&err, "Flush")
|
|
|
|
// add current cell if not empty
|
|
if b.cell.size > 0 {
|
|
if b.endChar != 0 {
|
|
// inside escape - terminate it even if incomplete
|
|
b.endEscape()
|
|
}
|
|
b.terminateCell(false)
|
|
}
|
|
|
|
// format contents of buffer
|
|
b.format(0, 0, len(b.lines))
|
|
|
|
return
|
|
}
|
|
|
|
var hbar = []byte("---\n")
|
|
|
|
// SetColumnAlignRight will mark a particular column as align right.
|
|
// This is reset on the next flush.
|
|
func (b *Writer) SetColumnAlignRight(column int) {
|
|
b.alignment[column] = AlignRight
|
|
}
|
|
|
|
// Write writes buf to the writer b.
|
|
// The only errors returned are ones encountered
|
|
// while writing to the underlying output stream.
|
|
//
|
|
func (b *Writer) Write(buf []byte) (n int, err error) {
|
|
defer handlePanic(&err, "Write")
|
|
|
|
// split text into cells
|
|
n = 0
|
|
for i, ch := range buf {
|
|
if b.endChar == 0 {
|
|
// outside escape
|
|
switch ch {
|
|
case '\t', '\v', '\n', '\f':
|
|
// end of cell
|
|
b.append(buf[n:i])
|
|
b.updateWidth()
|
|
n = i + 1 // ch consumed
|
|
ncells := b.terminateCell(ch == '\t')
|
|
if ch == '\n' || ch == '\f' {
|
|
// terminate line
|
|
b.addLine()
|
|
if ch == '\f' || ncells == 1 {
|
|
// A '\f' always forces a flush. Otherwise, if the previous
|
|
// line has only one cell which does not have an impact on
|
|
// the formatting of the following lines (the last cell per
|
|
// line is ignored by format()), thus we can flush the
|
|
// Writer contents.
|
|
if err = b.Flush(); err != nil {
|
|
return
|
|
}
|
|
if ch == '\f' && b.flags&Debug != 0 {
|
|
// indicate section break
|
|
b.write0(hbar)
|
|
}
|
|
}
|
|
}
|
|
|
|
case Escape:
|
|
// start of escaped sequence
|
|
b.append(buf[n:i])
|
|
b.updateWidth()
|
|
n = i
|
|
if b.flags&StripEscape != 0 {
|
|
n++ // strip Escape
|
|
}
|
|
b.startEscape(Escape)
|
|
|
|
case '<', '&':
|
|
// possibly an html tag/entity
|
|
if b.flags&FilterHTML != 0 {
|
|
// begin of tag/entity
|
|
b.append(buf[n:i])
|
|
b.updateWidth()
|
|
n = i
|
|
b.startEscape(ch)
|
|
}
|
|
}
|
|
|
|
} else {
|
|
// inside escape
|
|
if ch == b.endChar {
|
|
// end of tag/entity
|
|
j := i + 1
|
|
if ch == Escape && b.flags&StripEscape != 0 {
|
|
j = i // strip Escape
|
|
}
|
|
b.append(buf[n:j])
|
|
n = i + 1 // ch consumed
|
|
b.endEscape()
|
|
}
|
|
}
|
|
}
|
|
|
|
// append leftover text
|
|
b.append(buf[n:])
|
|
n = len(buf)
|
|
return
|
|
}
|
|
|
|
// NewWriter allocates and initializes a new tabwriter.Writer.
|
|
// The parameters are the same as for the Init function.
|
|
//
|
|
func NewWriter(output io.Writer, minwidth, tabwidth, padding int, padchar byte, flags uint) *Writer {
|
|
return new(Writer).Init(output, minwidth, tabwidth, padding, padchar, flags)
|
|
}
|