Rockingcool/kleingrep
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fixed bug in character class implementation

Browse Source
This commit is contained in:
Aadhavan Srinivasan
2025-01-24 19:48:53 -05:00
parent 5a085907cf
commit d7c9c181e1
1 changed files with 33 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

51
compile.go
View File

@@ -123,9 +123,12 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
} else if c == '\\' && i < len(re_runes_orig)-1 && re_runes_orig[i+1] == '\\' { // Escaped backslash
re_runes = append(re_runes, ESC_BACKSLASH)
i++
} else if c == '[' && (i == 0 || re_runes_orig[i-1] != '\\')
} else if c == '[' && (i == 0 || re_runes[len(re_runes)-1] != '\\') {
re_runes = append(re_runes, LBRACKET)
continue
} else if c == ']' && (i == 0 || re_runes[len(re_runes)-1] != '\\') {
re_runes = append(re_runes, RBRACKET)
continue
} else {
re_runes = append(re_runes, c)
}
@@ -145,31 +148,28 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
i := 0
for i < len(re_runes) {
re_postfix = append(re_postfix, re_runes[i])
if re_runes[i] == '[' && (i == 0 || re_runes[i-1] != '\\') { // We do not touch things inside brackets, unless they are escaped. Inside this block, the only task is to expand character ranges into their constituent characters.
re_postfix[len(re_postfix)-1] = LBRACKET // Replace the '[' character with LBRACKET. This allows for easier parsing of all characters (including opening and closing brackets) within the character class
toAppend := make([]rune, 0) // Holds all the runes in the current character class
if i < len(re_runes)-1 && re_runes[i+1] == '^' { // Inverting class - match everything NOT in brackets
re_postfix = append(re_postfix, '^')
i++ // Skip opening bracket and caret
if re_runes[i] == LBRACKET && (i == 0 || re_runes[i-1] != '\\') { // We do not touch things inside brackets, unless they are escaped.
toAppend := make([]rune, 0) // Holds all the runes in the current character class
i++ // Skip past LBRACKET, because it was already added
if i >= len(re_runes) { // Sanity check before we start
return nil, fmt.Errorf("Opening bracket without closing bracket.")
}
if i < len(re_runes)-1 && re_runes[i+1] == ']' { // Nothing inside brackets - panic.
return nil, fmt.Errorf("Empty character class.")
}
for re_runes[i] != ']' || i == 0 || re_runes[i-1] == '\\' {
i++ // Skip all characters inside _unescaped_ brackets (we are _not_ at a closing bracket, or if we are, the previous character is a backslash)
for re_runes[i] != RBRACKET || i == 0 || re_runes[i-1] == '\\' { // Skip all characters inside _unescaped_ brackets (we are _not_ at a closing bracket, or if we are, the previous character is a backslash)
// Make sure we haven't exceeded the length of the string. If we did, then the regex doesn't actually have a closing bracket and we should throw an error.
if i >= len(re_runes) {
return nil, fmt.Errorf("Opening bracket without closing bracket.")
}
if re_runes[i] == '-' && (i > 0 && re_runes[i-1] != '\\') && (i < len(re_runes)-1 && re_runes[i+1] != ']') { // Unescaped hyphen, that has some character (not a RBRACKET) after it - This represents a character range, so we replace with CHAR_RANGE. This metacharacter will be used later on to construct the range
if re_runes[i] == '-' && (i > 0 && re_runes[i-1] != '\\') && (i < len(re_runes)-1 && re_runes[i+1] != RBRACKET) { // Unescaped hyphen, that has some character (not a RBRACKET) after it - This represents a character range, so we replace with CHAR_RANGE. This metacharacter will be used later on to construct the range
re_runes[i] = CHAR_RANGE
}
toAppend = append(toAppend, re_runes[i])
i++
}
// Replace the last character (which should have been ']', with RBRACKET
toAppend[len(toAppend)-1] = RBRACKET
// Add in the RBRACKET
toAppend = append(toAppend, RBRACKET)
re_postfix = append(re_postfix, toAppend...)
}
if i < len(re_runes) && re_runes[i] == '{' && (i > 0 && re_runes[i-1] != '\\') { // We don't touch things inside braces, either
@@ -284,6 +284,17 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
}
continue
}
// Since every unescaped bracket is replaced by a LBRACKET / RBRACKET, there may
// have been false positives. For example, the regex ']' has a closing bracket, but it
// isn't denoting a character class; it's just a regular character. Since it's not escaped,
// though, I would have converted this into an RBRACKET.
// To deal with this, I make the following assertion:
// If at any point I see an RBRACKET 'in the wild' (not in a character class), then it must be
// a regular character, with no special significance.
if c == RBRACKET {
outQueue = append(outQueue, newPostfixCharNode(']'))
continue
}
if c == '\\' { // Escape character - invert special and non-special characters eg. \( is treated as a literal parentheses, \b is treated as word boundary
if i == len(re_postfix)-1 { // End of string - panic, because backslash is an escape character (something needs to come after it)
@@ -419,6 +430,10 @@ func shuntingYard(re string, flags ...ReFlag) ([]postfixNode, error) {
firstCharAdded := false // A character class must have at least 1 character. This flag checks if the first character has been added.
endOfRange := false // Set to 'true' when we encounter a CHAR_RANGE metacharacter
i++ // Step forward so we can look at the character class
// Oops, there's nothing there to look at
if i >= len(re_postfix) {
return nil, fmt.Errorf("Opening bracket with no closing bracket.")
}
var invertMatch bool
if re_postfix[i] == '^' {
invertMatch = true