/*
* Used by js_split to find the next split point in target,
* starting at offset ip and looking either for the given
* separator substring, or for the next re match. ip and
* matchlen must be reference variables (assumed to be arrays of
* length 1) so they can be updated in the leading whitespace or
* re case.
*
* Return -1 on end of string, >= 0 for a valid index of the next
* separator occurrence if found, or the string length if no
* separator is found.
*/
private static int find_split(
Context cx,
Scriptable scope,
String target,
String separator,
int version,
RegExpProxy reProxy,
Scriptable re,
int[] ip,
int[] matchlen,
boolean[] matched,
String[][] parensp) {
int i = ip[0];
int length = target.length();
/*
* Perl4 special case for str.split(' '), only if the user has selected
* JavaScript1.2 explicitly. Split on whitespace, and skip leading w/s.
* Strange but true, apparently modeled after awk.
*/
if (version == Context.VERSION_1_2
&& re == null
&& separator.length() == 1
&& separator.charAt(0) == ' ') {
/* Skip leading whitespace if at front of str. */
if (i == 0) {
while (i < length && Character.isWhitespace(target.charAt(i))) i++;
ip[0] = i;
}
/* Don't delimit whitespace at end of string. */
if (i == length) return -1;
/* Skip over the non-whitespace chars. */
while (i < length && !Character.isWhitespace(target.charAt(i))) i++;
/* Now skip the next run of whitespace. */
int j = i;
while (j < length && Character.isWhitespace(target.charAt(j))) j++;
/* Update matchlen to count delimiter chars. */
matchlen[0] = j - i;
return i;
}
/*
* Stop if past end of string. If at end of string, we will
* return target length, so that
*
* "ab,".split(',') => new Array("ab", "")
*
* and the resulting array converts back to the string "ab,"
* for symmetry. NB: This differs from perl, which drops the
* trailing empty substring if the LIMIT argument is omitted.
*/
if (i > length) return -1;
/*
* Match a regular expression against the separator at or
* above index i. Return -1 at end of string instead of
* trying for a match, so we don't get stuck in a loop.
*/
if (re != null) {
return reProxy.find_split(cx, scope, target, separator, re, ip, matchlen, matched, parensp);
}
/*
* Deviate from ECMA by never splitting an empty string by any separator
* string into a non-empty array (an array of length 1 that contains the
* empty string).
*/
if (version != Context.VERSION_DEFAULT && version < Context.VERSION_1_3 && length == 0)
return -1;
/*
* Special case: if sep is the empty string, split str into
* one character substrings. Let our caller worry about
* whether to split once at end of string into an empty
* substring.
*
* For 1.2 compatibility, at the end of the string, we return the length as
* the result, and set the separator length to 1 -- this allows the caller
* to include an additional null string at the end of the substring list.
*/
if (separator.length() == 0) {
if (version == Context.VERSION_1_2) {
if (i == length) {
matchlen[0] = 1;
return i;
}
return i + 1;
}
return (i == length) ? -1 : i + 1;
}
/* Punt to j.l.s.indexOf; return target length if separator is
* not found.
*/
if (ip[0] >= length) return length;
i = target.indexOf(separator, ip[0]);
return (i != -1) ? i : length;
}
