Permalink
Browse files

remove redundant tests for tab characters in parsing

  • Loading branch information...
1 parent a99c922 commit d4353d99fa43a10b0c11a146b66151c795e2f129 Russ Ross committed Jul 1, 2011
Showing with 23 additions and 35 deletions.
  1. +19 −31 block.go
  2. +4 −4 inline.go
View
@@ -168,7 +168,7 @@ func (parser *Parser) isPrefixHeader(data []byte) bool {
for level < 6 && data[level] == '#' {
level++
}
- if data[level] != ' ' && data[level] != '\t' {
+ if data[level] != ' ' {
return false
}
}
@@ -181,15 +181,15 @@ func (parser *Parser) blockPrefixHeader(out *bytes.Buffer, data []byte) int {
level++
}
i, end := 0, 0
- for i = level; data[i] == ' ' || data[i] == '\t'; i++ {
+ for i = level; data[i] == ' '; i++ {
}
for end = i; data[end] != '\n'; end++ {
}
skip := end
for end > 0 && data[end-1] == '#' {
end--
}
- for end > 0 && (data[end-1] == ' ' || data[end-1] == '\t') {
+ for end > 0 && data[end-1] == ' ' {
end--
}
if end > i {
@@ -209,7 +209,7 @@ func (parser *Parser) isUnderlinedHeader(data []byte) int {
for data[i] == '=' {
i++
}
- for data[i] == ' ' || data[i] == '\t' {
+ for data[i] == ' ' {
i++
}
if data[i] == '\n' {
@@ -225,7 +225,7 @@ func (parser *Parser) isUnderlinedHeader(data []byte) int {
for data[i] == '-' {
i++
}
- for data[i] == ' ' || data[i] == '\t' {
+ for data[i] == ' ' {
i++
}
if data[i] == '\n' {
@@ -377,7 +377,7 @@ func (parser *Parser) blockHtmlHr(out *bytes.Buffer, data []byte, doRender bool)
if data[0] != '<' || (data[1] != 'h' && data[1] != 'H') || (data[2] != 'r' && data[2] != 'R') {
return 0
}
- if data[3] != ' ' && data[3] != '\t' && data[3] != '/' && data[3] != '>' {
+ if data[3] != ' ' && data[3] != '/' && data[3] != '>' {
// not an <hr> tag after all; at least not a valid one
return 0
}
@@ -459,7 +459,7 @@ func (parser *Parser) isEmpty(data []byte) int {
var i int
for i = 0; data[i] != '\n'; i++ {
- if data[i] != ' ' && data[i] != '\t' {
+ if data[i] != ' ' {
return 0
}
}
@@ -486,7 +486,7 @@ func (parser *Parser) isHRule(data []byte) bool {
switch {
case data[i] == c:
n++
- case data[i] != ' ' && data[i] != '\t':
+ case data[i] != ' ':
return false
}
i++
@@ -531,7 +531,7 @@ func (parser *Parser) isFencedCode(data []byte, syntax **string, oldmarker strin
if syntax != nil {
syn := 0
- for data[i] == ' ' || data[i] == '\t' {
+ for data[i] == ' ' {
i++
}
@@ -705,7 +705,7 @@ func (parser *Parser) blockTableHeader(out *bytes.Buffer, data []byte) (size int
for ; col < columns && i < underEnd; col++ {
dashes := 0
- for i < underEnd && (data[i] == ' ' || data[i] == '\t') {
+ for i < underEnd && data[i] == ' ' {
i++
}
@@ -726,7 +726,7 @@ func (parser *Parser) blockTableHeader(out *bytes.Buffer, data []byte) (size int
dashes++
}
- for i < underEnd && (data[i] == ' ' || data[i] == '\t') {
+ for i < underEnd && data[i] == ' ' {
i++
}
@@ -806,7 +806,7 @@ func (parser *Parser) blockQuotePrefix(data []byte) int {
i++
}
if data[i] == '>' {
- if data[i+1] == ' ' || data[i+1] == '\t' {
+ if data[i+1] == ' ' {
return i + 2
}
return i + 1
@@ -848,9 +848,6 @@ func (parser *Parser) blockQuote(out *bytes.Buffer, data []byte) int {
// returns prefix length for block code
func (parser *Parser) blockCodePrefix(data []byte) int {
- if len(data) > 0 && data[0] == '\t' {
- return 1
- }
if len(data) > 3 && data[0] == ' ' && data[1] == ' ' && data[2] == ' ' && data[3] == ' ' {
return 4
}
@@ -914,7 +911,7 @@ func (parser *Parser) blockUliPrefix(data []byte) int {
// need a *, +, or - followed by a space/tab
if i+1 >= len(data) ||
(data[i] != '*' && data[i] != '+' && data[i] != '-') ||
- (data[i+1] != ' ' && data[i+1] != '\t') {
+ data[i+1] != ' ' {
return 0
}
return i + 2
@@ -936,8 +933,7 @@ func (parser *Parser) blockOliPrefix(data []byte) int {
}
// we need >= 1 digits followed by a dot and a space/tab
- if start == i || data[i] != '.' || i+1 >= len(data) ||
- (data[i+1] != ' ' && data[i+1] != '\t') {
+ if start == i || data[i] != '.' || i+1 >= len(data) || data[i+1] != ' ' {
return 0
}
return i + 2
@@ -982,7 +978,7 @@ func (parser *Parser) blockListItem(out *bytes.Buffer, data []byte, flags *int)
}
// skip leading whitespace on first line
- for beg < len(data) && (data[beg] == ' ' || data[beg] == '\t') {
+ for beg < len(data) && data[beg] == ' ' {
beg++
}
@@ -1023,14 +1019,6 @@ func (parser *Parser) blockListItem(out *bytes.Buffer, data []byte, flags *int)
}
pre = i
- if data[beg] == '\t' {
- i = 1
- pre = TAB_SIZE_DEFAULT
- if parser.flags&EXTENSION_TAB_SIZE_EIGHT != 0 {
- pre = TAB_SIZE_EIGHT
- }
- }
-
chunk := data[beg+i : end]
// check for a nested list item
@@ -1052,14 +1040,14 @@ func (parser *Parser) blockListItem(out *bytes.Buffer, data []byte, flags *int)
// how about a nested prefix header?
if parser.isPrefixHeader(chunk) {
// only nest headers that are indented
- if containsBlankLine && i < 4 && data[beg] != '\t' {
+ if containsBlankLine && i < 4 {
*flags |= LIST_ITEM_END_OF_LIST
break
}
containsBlock = true
} else {
// only join stuff after empty lines when indented
- if containsBlankLine && i < 4 && data[beg] != '\t' {
+ if containsBlankLine && i < 4 {
*flags |= LIST_ITEM_END_OF_LIST
break
} else {
@@ -1164,10 +1152,10 @@ func (parser *Parser) blockParagraph(out *bytes.Buffer, data []byte) int {
// ignore leading and trailing whitespace
eol := i - 1
- for prev < eol && (data[prev] == ' ' || data[prev] == '\t') {
+ for prev < eol && data[prev] == ' ' {
prev++
}
- for eol > prev && (data[eol-1] == ' ' || data[eol-1] == '\t') {
+ for eol > prev && data[eol-1] == ' ' {
eol--
}
View
@@ -129,12 +129,12 @@ func inlineCodeSpan(parser *Parser, out *bytes.Buffer, data []byte, offset int)
// trim outside whitespace
fBegin := nb
- for fBegin < end && (data[fBegin] == ' ' || data[fBegin] == '\t') {
+ for fBegin < end && data[fBegin] == ' ' {
fBegin++
}
fEnd := end - nb
- for fEnd > fBegin && (data[fEnd-1] == ' ' || data[fEnd-1] == '\t') {
+ for fEnd > fBegin && data[fEnd-1] == ' ' {
fEnd--
}
@@ -154,7 +154,7 @@ func inlineLineBreak(parser *Parser, out *bytes.Buffer, data []byte, offset int)
outBytes := out.Bytes()
end := len(outBytes)
eol := end
- for eol > 0 && (outBytes[eol-1] == ' ' || outBytes[eol-1] == '\t') {
+ for eol > 0 && outBytes[eol-1] == ' ' {
eol--
}
out.Truncate(eol)
@@ -799,7 +799,7 @@ func inlineHelperFindEmphChar(data []byte, c byte) int {
i++
}
i++
- for i < len(data) && (data[i] == ' ' || data[i] == '\t' || data[i] == '\n') {
+ for i < len(data) && (data[i] == ' ' || data[i] == '\n') {
i++
}
if i >= len(data) {

0 comments on commit d4353d9

Please sign in to comment.