//
// Split_Lines: C
//
// Given a string series, split lines on CR-LF. Give back array of strings.
//
// Note: The definition of "line" in POSIX is a sequence of characters that
// end with a newline. Hence, the last line of a file should have a newline
// marker, or it's not a "line")
//
// https://stackoverflow.com/a/729795
//
// This routine does not require it.
//
// !!! CR support is likely to be removed...and CR will be handled as a normal
// character, with special code needed to process it.
//
REBARR *Split_Lines(const REBVAL *str)
{
REBDSP dsp_orig = DSP;
REBCNT len = VAL_LEN_AT(str);
REBCNT i = VAL_INDEX(str);
if (i == len)
return Make_Array(0);
DECLARE_MOLD (mo);
Push_Mold(mo);
REBCHR(const*) cp = VAL_STRING_AT(str);
REBUNI c;
cp = NEXT_CHR(&c, cp);
for (; i < len; ++i, cp = NEXT_CHR(&c, cp)) {
if (c != LF && c != CR) {
Append_Codepoint(mo->series, c);
continue;
}
Init_Text(DS_PUSH(), Pop_Molded_String(mo));
SET_CELL_FLAG(DS_TOP, NEWLINE_BEFORE);
Push_Mold(mo);
if (c == CR) {
REBCHR(const*) tp = NEXT_CHR(&c, cp);
if (c == LF) {
++i;
cp = tp; // treat CR LF as LF, lone CR as LF
}
}
}
// If there's any remainder we pushed in the buffer, consider the end of
// string to be an implicit line-break
if (STR_SIZE(mo->series) == mo->offset)
Drop_Mold(mo);
else {
Init_Text(DS_PUSH(), Pop_Molded_String(mo));
SET_CELL_FLAG(DS_TOP, NEWLINE_BEFORE);
}
return Pop_Stack_Values_Core(dsp_orig, ARRAY_FLAG_NEWLINE_AT_TAIL);
}
//
// Entab_Unicode: C
//
// Entab a string and return a new series.
//
REBSER *Entab_Unicode(REBUNI *bp, REBCNT index, REBCNT len, REBINT tabsize)
{
REBINT n = 0;
REBUNI *dp;
REBUNI c;
REB_MOLD mo;
CLEARS(&mo);
mo.opts = MOPT_RESERVE;
mo.reserve = len;
Push_Mold(&mo);
dp = UNI_AT(mo.series, mo.start);
for (; index < len; index++) {
c = bp[index];
// Count leading spaces, insert TAB for each tabsize:
if (c == ' ') {
if (++n >= tabsize) {
*dp++ = '\t';
n = 0;
}
continue;
}
// Hitting a leading TAB resets space counter:
if (c == '\t') {
*dp++ = (REBYTE)c;
n = 0;
}
else {
// Incomplete tab space, pad with spaces:
for (; n > 0; n--) *dp++ = ' ';
// Copy chars thru end-of-line (or end of buffer):
while (index < len) {
if ((*dp++ = bp[index++]) == '\n') break;
}
}
}
SET_SERIES_LEN(mo.series, mo.start + cast(REBCNT, dp - UNI_AT(mo.series, mo.start)));
UNI_TERM(mo.series);
return Pop_Molded_String(&mo);
}
//
// Detab_Unicode: C
//
// Detab a unicode string and return a new series.
//
REBSER *Detab_Unicode(REBUNI *bp, REBCNT index, REBCNT len, REBINT tabsize)
{
REBCNT cnt = 0;
REBCNT n;
REBUNI *dp;
REBUNI c;
REB_MOLD mo;
CLEARS(&mo);
// Estimate new length based on tab expansion:
for (n = index; n < len; n++)
if (bp[n] == TAB) cnt++;
mo.opts = MOPT_RESERVE;
mo.reserve = len + (cnt * (tabsize - 1));
Push_Mold(&mo);
dp = UNI_AT(mo.series, mo.start);
n = 0;
while (index < len) {
c = bp[index++];
if (c == '\t') {
*dp++ = ' ';
n++;
for (; n % tabsize != 0; n++) *dp++ = ' ';
continue;
}
if (c == '\n') n = 0;
else n++;
*dp++ = c;
}
SET_SERIES_LEN(mo.series, mo.start + cast(REBCNT, dp - UNI_AT(mo.series, mo.start)));
UNI_TERM(mo.series);
return Pop_Molded_String(&mo);
}
//
// Make_Set_Operation_Series: C
//
// Do set operations on a series. Case-sensitive if `cased` is TRUE.
// `skip` is the record size.
//
static REBSER *Make_Set_Operation_Series(
const REBVAL *val1,
const REBVAL *val2,
REBFLGS flags,
REBOOL cased,
REBCNT skip
) {
REBCNT i;
REBINT h = 1; // used for both logic true/false and hash check
REBOOL first_pass = TRUE; // are we in the first pass over the series?
REBSER *out_ser;
assert(ANY_SERIES(val1));
if (val2) {
assert(ANY_SERIES(val2));
if (ANY_ARRAY(val1)) {
if (!ANY_ARRAY(val2))
fail (Error_Unexpected_Type(VAL_TYPE(val1), VAL_TYPE(val2)));
// As long as they're both arrays, we're willing to do:
//
// >> union quote (a b c) 'b/d/e
// (a b c d e)
//
// The type of the result will match the first value.
}
else if (!IS_BINARY(val1)) {
// We will similarly do any two ANY-STRING! types:
//
// >> union <abc> "bde"
// <abcde>
if (IS_BINARY(val2))
fail (Error_Unexpected_Type(VAL_TYPE(val1), VAL_TYPE(val2)));
}
else {
// Binaries only operate with other binaries
if (!IS_BINARY(val2))
fail (Error_Unexpected_Type(VAL_TYPE(val1), VAL_TYPE(val2)));
}
}
// Calculate `i` as maximum length of result block. The temporary buffer
// will be allocated at this size, but copied out at the exact size of
// the actual result.
//
i = VAL_LEN_AT(val1);
if (flags & SOP_FLAG_BOTH) i += VAL_LEN_AT(val2);
if (ANY_ARRAY(val1)) {
REBSER *hser = 0; // hash table for series
REBSER *hret; // hash table for return series
// The buffer used for building the return series. Currently it
// reuses BUF_EMIT, because that buffer is not likely to be in
// use (emit doesn't call set operations, nor vice versa). However,
// other routines may get the same idea and start recursing so it
// may be better to use something more similar to the mold stack
// approach of marking off successive ranges in the array.
//
REBSER *buffer = ARR_SERIES(BUF_EMIT);
Resize_Series(buffer, i);
hret = Make_Hash_Sequence(i); // allocated
// Optimization note: !!
// This code could be optimized for small blocks by not hashing them
// and extending Find_Key to FIND on the value itself w/o the hash.
do {
REBARR *array1 = VAL_ARRAY(val1); // val1 and val2 swapped 2nd pass!
// Check what is in series1 but not in series2
//
if (flags & SOP_FLAG_CHECK)
hser = Hash_Block(val2, skip, cased);
// Iterate over first series
//
i = VAL_INDEX(val1);
for (; i < ARR_LEN(array1); i += skip) {
RELVAL *item = ARR_AT(array1, i);
if (flags & SOP_FLAG_CHECK) {
h = Find_Key_Hashed(
VAL_ARRAY(val2),
hser,
item,
VAL_SPECIFIER(val1),
skip,
cased,
1
);
h = (h >= 0);
if (flags & SOP_FLAG_INVERT) h = !h;
}
if (h) {
Find_Key_Hashed(
AS_ARRAY(buffer),
hret,
item,
VAL_SPECIFIER(val1),
skip,
cased,
2
);
}
}
if (i != ARR_LEN(array1)) {
//
// In the current philosophy, the semantics of what to do
// with things like `intersect/skip [1 2 3] [7] 2` is too
// shaky to deal with, so an error is reported if it does
// not work out evenly to the skip size.
//
fail (Error(RE_BLOCK_SKIP_WRONG));
}
if (flags & SOP_FLAG_CHECK)
Free_Series(hser);
if (!first_pass) break;
first_pass = FALSE;
// Iterate over second series?
//
if ((i = ((flags & SOP_FLAG_BOTH) != 0))) {
const REBVAL *temp = val1;
val1 = val2;
val2 = temp;
}
} while (i);
if (hret)
Free_Series(hret);
out_ser = ARR_SERIES(Copy_Array_Shallow(AS_ARRAY(buffer), SPECIFIED));
SET_SERIES_LEN(buffer, 0); // required - allow reuse
}
else {
REB_MOLD mo;
CLEARS(&mo);
if (IS_BINARY(val1)) {
//
// All binaries use "case-sensitive" comparison (e.g. each byte
// is treated distinctly)
//
cased = TRUE;
}
// ask mo.series to have at least `i` capacity beyond mo.start
//
mo.opts = MOPT_RESERVE;
mo.reserve = i;
Push_Mold(&mo);
do {
REBSER *ser = VAL_SERIES(val1); // val1 and val2 swapped 2nd pass!
REBUNI uc;
// Iterate over first series
//
i = VAL_INDEX(val1);
for (; i < SER_LEN(ser); i += skip) {
uc = GET_ANY_CHAR(ser, i);
if (flags & SOP_FLAG_CHECK) {
h = (NOT_FOUND != Find_Str_Char(
uc,
VAL_SERIES(val2),
0,
VAL_INDEX(val2),
VAL_LEN_HEAD(val2),
skip,
cased ? AM_FIND_CASE : 0
));
if (flags & SOP_FLAG_INVERT) h = !h;
}
if (!h) continue;
if (
NOT_FOUND == Find_Str_Char(
uc, // c2 (the character to find)
mo.series, // ser
mo.start, // head
mo.start, // index
SER_LEN(mo.series), // tail
skip, // skip
cased ? AM_FIND_CASE : 0 // flags
)
) {
Append_String(mo.series, ser, i, skip);
}
}
if (!first_pass) break;
first_pass = FALSE;
// Iterate over second series?
//
if ((i = ((flags & SOP_FLAG_BOTH) != 0))) {
const REBVAL *temp = val1;
val1 = val2;
val2 = temp;
}
} while (i);
out_ser = Pop_Molded_String(&mo);
}
return out_ser;
}