*/ static REBVAL *Do_Parse_Path(REBVAL *item, REBPARSE *parse, REBCNT *index)
/*
** Handle a PATH, including get and set, that's found in a rule.
**
***********************************************************************/
{
REBVAL *path = item;
REBVAL tmp;
if (IS_PATH(item)) {
if (Do_Path(&path, 0)) return item; // found a function
item = DS_TOP;
}
else if (IS_SET_PATH(item)) {
Set_Series(parse->type, &tmp, parse->series);
VAL_INDEX(&tmp) = *index;
if (Do_Path(&path, &tmp)) return item; // found a function
return 0;
}
else if (IS_GET_PATH(item)) {
if (Do_Path(&path, 0)) return item; // found a function
item = DS_TOP;
// CureCode #1263 change
// if (parse->type != VAL_TYPE(item) || VAL_SERIES(item) != parse->series)
if (!ANY_SERIES(item)) Trap1(RE_PARSE_SERIES, path);
*index = Set_Parse_Series(parse, item);
return 0;
}
return item;
}
//
// Partial1: C
//
// Process the /part (or /skip) and other length modifying
// arguments.
//
REBINT Partial1(REBVAL *sval, REBVAL *lval)
{
REBI64 len;
REBINT maxlen;
REBINT is_ser = ANY_SERIES(sval);
// If lval is not set or is BAR!, use the current len of the target value:
if (IS_UNSET(lval) || IS_BAR(lval)) {
if (!is_ser) return 1;
if (VAL_INDEX(sval) >= VAL_LEN_HEAD(sval)) return 0;
return (VAL_LEN_HEAD(sval) - VAL_INDEX(sval));
}
if (IS_INTEGER(lval) || IS_DECIMAL(lval)) len = Int32(lval);
else {
if (is_ser && VAL_TYPE(sval) == VAL_TYPE(lval) && VAL_SERIES(sval) == VAL_SERIES(lval))
len = (REBINT)VAL_INDEX(lval) - (REBINT)VAL_INDEX(sval);
else
fail (Error(RE_INVALID_PART, lval));
}
if (is_ser) {
// Restrict length to the size available:
if (len >= 0) {
maxlen = (REBINT)VAL_LEN_AT(sval);
if (len > maxlen) len = maxlen;
} else {
len = -len;
if (len > (REBINT)VAL_INDEX(sval)) len = (REBINT)VAL_INDEX(sval);
VAL_INDEX(sval) -= (REBCNT)len;
}
}
return (REBINT)len;
}
*/ static REBFLG Get_Index_Var(REBVAL *item, REBSER *series, REBINT *index)
/*
** Get the series index from a word or path or integer.
**
** Returns: TRUE if value was a series. FALSE if integer.
**
***********************************************************************/
{
REBVAL *hold = item;
if (IS_END(item)) Trap1(RE_PARSE_END, item);
if (IS_WORD(item)) {
if (!VAL_CMD(item)) item = Get_Var(item);
}
else if (IS_PATH(item)) {
REBVAL *path = item;
Do_Path(&path, 0); //!!! function!
item = DS_TOP;
}
else if (!IS_INTEGER(item))
Trap1(RE_PARSE_VARIABLE, hold);
if (IS_INTEGER(item)) {
*index = Int32(item);
return FALSE;
}
if (!ANY_SERIES(item) || VAL_SERIES(item) != series)
Trap1(RE_PARSE_SERIES, hold);
*index = VAL_INDEX(item);
return TRUE;
}
//
// Partial: C
//
// Args:
// aval: target value
// bval: argument to modify target (optional)
// lval: length value (or none)
//
// Determine the length of a /PART value. It can be:
// 1. integer or decimal
// 2. relative to A value (bval is null)
// 3. relative to B value
//
// NOTE: Can modify the value's index!
// The result can be negative. ???
//
REBINT Partial(REBVAL *aval, REBVAL *bval, REBVAL *lval)
{
REBVAL *val;
REBINT len;
REBINT maxlen;
// If lval is unset, use the current len of the target value:
if (IS_UNSET(lval)) {
val = (bval && ANY_SERIES(bval)) ? bval : aval;
if (VAL_INDEX(val) >= VAL_LEN_HEAD(val)) return 0;
return (VAL_LEN_HEAD(val) - VAL_INDEX(val));
}
if (IS_INTEGER(lval) || IS_DECIMAL(lval)) {
len = Int32(lval);
val = bval;
}
else {
// So, lval must be relative to aval or bval series:
if (
VAL_TYPE(aval) == VAL_TYPE(lval)
&& VAL_SERIES(aval) == VAL_SERIES(lval)
) {
val = aval;
}
else if (
bval
&& VAL_TYPE(bval) == VAL_TYPE(lval)
&& VAL_SERIES(bval) == VAL_SERIES(lval)
) {
val = bval;
}
else
fail (Error(RE_INVALID_PART, lval));
len = cast(REBINT, VAL_INDEX(lval)) - cast(REBINT, VAL_INDEX(val));
}
if (!val) val = aval;
// Restrict length to the size available
//
if (len >= 0) {
maxlen = (REBINT)VAL_LEN_AT(val);
if (len > maxlen) len = maxlen;
}
else {
len = -len;
if (len > cast(REBINT, VAL_INDEX(val)))
len = cast(REBINT, VAL_INDEX(val));
VAL_INDEX(val) -= (REBCNT)len;
}
return len;
}
*/ REBINT Partial1(REBVAL *sval, REBVAL *lval)
/*
** Process the /part (or /skip) and other length modifying
** arguments.
**
***********************************************************************/
{
REBI64 len;
REBINT maxlen;
REBINT is_ser = ANY_SERIES(sval);
// If lval = NONE, use the current len of the target value:
if (IS_NONE(lval)) {
if (!is_ser) return 1;
if (VAL_INDEX(sval) >= VAL_TAIL(sval)) return 0;
return (VAL_TAIL(sval) - VAL_INDEX(sval));
}
if (IS_INTEGER(lval) || IS_DECIMAL(lval)) len = Int32(lval);
else {
if (is_ser && VAL_TYPE(sval) == VAL_TYPE(lval) && VAL_SERIES(sval) == VAL_SERIES(lval))
len = (REBINT)VAL_INDEX(lval) - (REBINT)VAL_INDEX(sval);
else
Trap1(RE_INVALID_PART, lval);
}
if (is_ser) {
// Restrict length to the size available:
if (len >= 0) {
maxlen = (REBINT)VAL_LEN(sval);
if (len > maxlen) len = maxlen;
} else {
len = -len;
if (len > (REBINT)VAL_INDEX(sval)) len = (REBINT)VAL_INDEX(sval);
VAL_INDEX(sval) -= (REBCNT)len;
}
}
return (REBINT)len;
}
//
// Uncolor: C
//
// Clear the recusion markers for series and object trees.
//
void Uncolor(RELVAL *v)
{
REBARR *array;
if (ANY_ARRAY_OR_PATH(v))
array = VAL_ARRAY(v);
else if (IS_MAP(v))
array = MAP_PAIRLIST(VAL_MAP(v));
else if (ANY_CONTEXT(v))
array = CTX_VARLIST(VAL_CONTEXT(v));
else {
// Shouldn't have marked recursively any non-array series (no need)
//
assert(
not ANY_SERIES(v)
or Is_Series_White(VAL_SERIES(v))
);
return;
}
Uncolor_Array(array);
}
//
// 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, REBCNT flags, REBCNT cased, REBCNT skip)
{
REBSER *buffer; // buffer for building the return series
REBCNT i;
REBINT h = TRUE;
REBFLG first_pass = TRUE; // are we in the first pass over the series?
REBSER *out_ser;
// This routine should only be called with SERIES! values
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 length of result block.
i = VAL_LEN(val1);
if (flags & SOP_FLAG_BOTH) i += VAL_LEN(val2);
if (ANY_ARRAY(val1)) {
REBSER *hser = 0; // hash table for series
REBSER *hret; // hash table for return series
buffer = BUF_EMIT; // use preallocated shared block
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 do a FIND on the value itself w/o the hash.
do {
REBSER *ser = VAL_SERIES(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, cased);
// Iterate over first series:
i = VAL_INDEX(val1);
for (; i < SERIES_TAIL(ser); i += skip) {
REBVAL *item = BLK_SKIP(ser, i);
if (flags & SOP_FLAG_CHECK) {
h = Find_Key(VAL_SERIES(val2), hser, item, skip, cased, 1);
h = (h >= 0);
if (flags & SOP_FLAG_INVERT) h = !h;
}
if (h) Find_Key(buffer, hret, item, skip, cased, 2);
}
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 = Copy_Array_Shallow(buffer);
RESET_TAIL(buffer); // required - allow reuse
}
else {
if (IS_BINARY(val1)) {
// All binaries use "case-sensitive" comparison (e.g. each byte
// is treated distinctly)
cased = TRUE;
}
buffer = BUF_MOLD;
Reset_Buffer(buffer, i);
RESET_TAIL(buffer);
do {
REBSER *ser = VAL_SERIES(val1); // val1 and val2 swapped 2nd pass!
REBUNI uc;
// Iterate over first series:
i = VAL_INDEX(val1);
for (; i < SERIES_TAIL(ser); i += skip) {
uc = GET_ANY_CHAR(ser, i);
if (flags & SOP_FLAG_CHECK) {
h = (NOT_FOUND != Find_Str_Char(
VAL_SERIES(val2),
0,
VAL_INDEX(val2),
VAL_TAIL(val2),
skip,
uc,
cased ? AM_FIND_CASE : 0
));
if (flags & SOP_FLAG_INVERT) h = !h;
}
if (!h) continue;
if (
NOT_FOUND == Find_Str_Char(
buffer,
0,
0,
SERIES_TAIL(buffer),
skip,
uc,
cased ? AM_FIND_CASE : 0
)
) {
Append_String(buffer, 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 = Copy_String(buffer, 0, -1);
}
return out_ser;
}
*/ REBINT Partial(REBVAL *aval, REBVAL *bval, REBVAL *lval, REBFLG flag)
/*
** Args:
** aval: target value
** bval: argument to modify target (optional)
** lval: length value (or none)
**
** Determine the length of a /PART value. It can be:
** 1. integer or decimal
** 2. relative to A value (bval is null)
** 3. relative to B value
**
** Flag: indicates special treatment for CHANGE. As in:
** CHANGE/part "abcde" "xy" 3 => "xyde"
**
** NOTE: Can modify the value's index!
** The result can be negative. ???
**
***********************************************************************/
{
REBVAL *val;
REBINT len;
REBINT maxlen;
// If lval = NONE, use the current len of the target value:
if (IS_NONE(lval)) {
val = (bval && ANY_SERIES(bval)) ? bval : aval;
if (VAL_INDEX(val) >= VAL_TAIL(val)) return 0;
return (VAL_TAIL(val) - VAL_INDEX(val));
}
if (IS_INTEGER(lval)) {
len = Int32(lval);
val = flag ? aval : bval;
}
else if (IS_DECIMAL(lval)) {
len = Int32(lval);
val = bval;
}
else {
// So, lval must be relative to aval or bval series:
if (VAL_TYPE(aval) == VAL_TYPE(lval) && VAL_SERIES(aval) == VAL_SERIES(lval))
val = aval;
else if (bval && VAL_TYPE(bval) == VAL_TYPE(lval) && VAL_SERIES(bval) == VAL_SERIES(lval))
val = bval;
else
Trap1(RE_INVALID_PART, lval);
len = (REBINT)VAL_INDEX(lval) - (REBINT)VAL_INDEX(val);
}
if (!val) val = aval;
// Restrict length to the size available:
if (len >= 0) {
maxlen = (REBINT)VAL_LEN(val);
if (len > maxlen) len = maxlen;
} else {
len = -len;
if (len > (REBINT)VAL_INDEX(val)) len = (REBINT)VAL_INDEX(val);
VAL_INDEX(val) -= (REBCNT)len;
// if ((-len) > (REBINT)VAL_INDEX(val)) len = -(REBINT)VAL_INDEX(val);
}
return len;
}
//
// Transport_Actor: C
//
static REB_R Transport_Actor(struct Reb_Call *call_, REBSER *port, REBCNT action, enum Transport_Types proto)
{
REBREQ *sock; // IO request
REBVAL *spec; // port spec
REBVAL *arg; // action argument value
REBVAL *val; // e.g. port number value
REBINT result; // IO result
REBCNT refs; // refinement argument flags
REBCNT len; // generic length
REBSER *ser; // simplifier
Validate_Port(port, action);
*D_OUT = *D_ARG(1);
arg = DS_ARGC > 1 ? D_ARG(2) : NULL;
sock = cast(REBREQ*, Use_Port_State(port, RDI_NET, sizeof(*sock)));
if (proto == TRANSPORT_UDP) {
SET_FLAG(sock->modes, RST_UDP);
}
//Debug_Fmt("Sock: %x", sock);
spec = OFV(port, STD_PORT_SPEC);
if (!IS_OBJECT(spec)) fail (Error(RE_INVALID_PORT));
// sock->timeout = 4000; // where does this go? !!!
// HOW TO PREVENT OVERWRITE DURING BUSY OPERATION!!!
// Should it just ignore it or cause an error?
// Actions for an unopened socket:
if (!IS_OPEN(sock)) {
switch (action) { // Ordered by frequency
case A_OPEN:
arg = Obj_Value(spec, STD_PORT_SPEC_NET_HOST);
val = Obj_Value(spec, STD_PORT_SPEC_NET_PORT_ID);
if (OS_DO_DEVICE(sock, RDC_OPEN))
fail (Error_On_Port(RE_CANNOT_OPEN, port, -12));
SET_OPEN(sock);
// Lookup host name (an extra TCP device step):
if (IS_STRING(arg)) {
sock->common.data = VAL_BIN(arg);
sock->special.net.remote_port = IS_INTEGER(val) ? VAL_INT32(val) : 80;
result = OS_DO_DEVICE(sock, RDC_LOOKUP); // sets remote_ip field
if (result < 0)
fail (Error_On_Port(RE_NO_CONNECT, port, sock->error));
return R_OUT;
}
// Host IP specified:
else if (IS_TUPLE(arg)) {
sock->special.net.remote_port = IS_INTEGER(val) ? VAL_INT32(val) : 80;
memcpy(&sock->special.net.remote_ip, VAL_TUPLE(arg), 4);
break;
}
// No host, must be a LISTEN socket:
else if (IS_NONE(arg)) {
SET_FLAG(sock->modes, RST_LISTEN);
sock->common.data = 0; // where ACCEPT requests are queued
sock->special.net.local_port = IS_INTEGER(val) ? VAL_INT32(val) : 8000;
break;
}
else
fail (Error_On_Port(RE_INVALID_SPEC, port, -10));
case A_CLOSE:
return R_OUT;
case A_OPENQ:
return R_FALSE;
case A_UPDATE: // allowed after a close
break;
default:
fail (Error_On_Port(RE_NOT_OPEN, port, -12));
}
}
// Actions for an open socket:
switch (action) { // Ordered by frequency
case A_UPDATE:
// Update the port object after a READ or WRITE operation.
// This is normally called by the WAKE-UP function.
arg = OFV(port, STD_PORT_DATA);
if (sock->command == RDC_READ) {
if (ANY_BINSTR(arg)) VAL_TAIL(arg) += sock->actual;
}
else if (sock->command == RDC_WRITE) {
SET_NONE(arg); // Write is done.
}
return R_NONE;
case A_READ:
// Read data into a buffer, expanding the buffer if needed.
// If no length is given, program must stop it at some point.
refs = Find_Refines(call_, ALL_READ_REFS);
if (
!GET_FLAG(sock->modes, RST_UDP)
&& !GET_FLAG(sock->state, RSM_CONNECT)
) {
fail (Error_On_Port(RE_NOT_CONNECTED, port, -15));
}
// Setup the read buffer (allocate a buffer if needed):
arg = OFV(port, STD_PORT_DATA);
if (!IS_STRING(arg) && !IS_BINARY(arg)) {
Val_Init_Binary(arg, Make_Binary(NET_BUF_SIZE));
}
ser = VAL_SERIES(arg);
sock->length = SERIES_AVAIL(ser); // space available
if (sock->length < NET_BUF_SIZE/2) Extend_Series(ser, NET_BUF_SIZE);
sock->length = SERIES_AVAIL(ser);
sock->common.data = STR_TAIL(ser); // write at tail
//if (SERIES_TAIL(ser) == 0)
sock->actual = 0; // Actual for THIS read, not for total.
//Print("(max read length %d)", sock->length);
result = OS_DO_DEVICE(sock, RDC_READ); // recv can happen immediately
if (result < 0) fail (Error_On_Port(RE_READ_ERROR, port, sock->error));
break;
case A_WRITE:
// Write the entire argument string to the network.
// The lower level write code continues until done.
refs = Find_Refines(call_, ALL_WRITE_REFS);
if (!GET_FLAG(sock->modes, RST_UDP)
&& !GET_FLAG(sock->state, RSM_CONNECT))
fail (Error_On_Port(RE_NOT_CONNECTED, port, -15));
// Determine length. Clip /PART to size of string if needed.
spec = D_ARG(2);
len = VAL_LEN(spec);
if (refs & AM_WRITE_PART) {
REBCNT n = Int32s(D_ARG(ARG_WRITE_LIMIT), 0);
if (n <= len) len = n;
}
// Setup the write:
*OFV(port, STD_PORT_DATA) = *spec; // keep it GC safe
sock->length = len;
sock->common.data = VAL_BIN_DATA(spec);
sock->actual = 0;
//Print("(write length %d)", len);
result = OS_DO_DEVICE(sock, RDC_WRITE); // send can happen immediately
if (result < 0) fail (Error_On_Port(RE_WRITE_ERROR, port, sock->error));
if (result == DR_DONE) SET_NONE(OFV(port, STD_PORT_DATA));
break;
case A_PICK:
// FIRST server-port returns new port connection.
len = Get_Num_Arg(arg); // Position
if (len == 1 && GET_FLAG(sock->modes, RST_LISTEN) && sock->common.data)
Accept_New_Port(D_OUT, port, sock); // sets D_OUT
else
fail (Error_Out_Of_Range(arg));
break;
case A_QUERY:
// Get specific information - the scheme's info object.
// Special notation allows just getting part of the info.
Ret_Query_Net(port, sock, D_OUT);
break;
case A_OPENQ:
// Connect for clients, bind for servers:
if (sock->state & ((1<<RSM_CONNECT) | (1<<RSM_BIND))) return R_TRUE;
return R_FALSE;
case A_CLOSE:
if (IS_OPEN(sock)) {
OS_DO_DEVICE(sock, RDC_CLOSE);
SET_CLOSED(sock);
}
break;
case A_LENGTH:
arg = OFV(port, STD_PORT_DATA);
len = ANY_SERIES(arg) ? VAL_TAIL(arg) : 0;
SET_INTEGER(D_OUT, len);
break;
case A_OPEN:
result = OS_DO_DEVICE(sock, RDC_CONNECT);
if (result < 0)
fail (Error_On_Port(RE_NO_CONNECT, port, sock->error));
break;
case A_DELETE: // Temporary to TEST error handler!
{
REBVAL *event = Append_Event(); // sets signal
VAL_SET(event, REB_EVENT); // (has more space, if we need it)
VAL_EVENT_TYPE(event) = EVT_ERROR;
VAL_EVENT_DATA(event) = 101;
VAL_EVENT_REQ(event) = sock;
}
break;
default:
fail (Error_Illegal_Action(REB_PORT, action));
}
return R_OUT;
}
*/ static int Loop_All(struct Reb_Call *call_, REBINT mode)
/*
** 0: forall
** 1: forskip
**
***********************************************************************/
{
REBVAL *var;
REBSER *body;
REBCNT bodi;
REBSER *dat;
REBINT idx;
REBINT inc = 1;
REBCNT type;
REBVAL *ds;
var = GET_MUTABLE_VAR(D_ARG(1));
if (IS_NONE(var)) return R_NONE;
// Save the starting var value:
*D_ARG(1) = *var;
SET_NONE(D_OUT);
if (mode == 1) inc = Int32(D_ARG(2));
type = VAL_TYPE(var);
body = VAL_SERIES(D_ARG(mode+2));
bodi = VAL_INDEX(D_ARG(mode+2));
// Starting location when past end with negative skip:
if (inc < 0 && VAL_INDEX(var) >= VAL_TAIL(var)) {
VAL_INDEX(var) = VAL_TAIL(var) + inc;
}
// NOTE: This math only works for index in positive ranges!
if (ANY_SERIES(var)) {
while (TRUE) {
dat = VAL_SERIES(var);
idx = VAL_INDEX(var);
if (idx < 0) break;
if (idx >= cast(REBINT, SERIES_TAIL(dat))) {
if (inc >= 0) break;
idx = SERIES_TAIL(dat) + inc; // negative
if (idx < 0) break;
VAL_INDEX(var) = idx;
}
if (!DO_BLOCK(D_OUT, body, bodi)) { // Break, throw, continue, error.
if (Check_Error(D_OUT) >= 0) {
break;
}
}
if (VAL_TYPE(var) != type) Trap_Arg_DEAD_END(var);
VAL_INDEX(var) += inc;
}
}
else Trap_Arg_DEAD_END(var);
// !!!!! ???? allowed to write VAR????
*var = *D_ARG(1);
return R_OUT;
}
//
// 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;
}
*/ static REBCNT Parse_Rules_Loop(REBPARSE *parse, REBCNT index, REBVAL *rules, REBCNT depth)
/*
***********************************************************************/
{
REBSER *series = parse->series;
REBVAL *item; // current rule item
REBVAL *word; // active word to be set
REBCNT start; // recovery restart point
REBCNT i; // temp index point
REBCNT begin; // point at beginning of match
REBINT count; // iterated pattern counter
REBINT mincount; // min pattern count
REBINT maxcount; // max pattern count
REBVAL *item_hold;
REBVAL *val; // spare
REBCNT rulen;
REBSER *ser;
REBFLG flags;
REBCNT cmd;
REBVAL *rule_head = rules;
CHECK_STACK(&flags);
//if (depth > MAX_PARSE_DEPTH) Trap_Word(RE_LIMIT_HIT, SYM_PARSE, 0);
flags = 0;
word = 0;
mincount = maxcount = 1;
start = begin = index;
// For each rule in the rule block:
while (NOT_END(rules)) {
//Print_Parse_Index(parse->type, rules, series, index);
if (--Eval_Count <= 0 || Eval_Signals) Do_Signals();
//--------------------------------------------------------------------
// Pre-Rule Processing Section
//
// For non-iterated rules, including setup for iterated rules.
// The input index is not advanced here, but may be changed by
// a GET-WORD variable.
//--------------------------------------------------------------------
item = rules++;
// If word, set-word, or get-word, process it:
if (VAL_TYPE(item) >= REB_WORD && VAL_TYPE(item) <= REB_GET_WORD) {
// Is it a command word?
if (cmd = VAL_CMD(item)) {
if (!IS_WORD(item)) Trap1(RE_PARSE_COMMAND, item); // SET or GET not allowed
if (cmd <= SYM_BREAK) { // optimization
switch (cmd) {
case SYM_OR_BAR:
return index; // reached it successfully
// Note: mincount = maxcount = 1 on entry
case SYM_WHILE:
SET_FLAG(flags, PF_WHILE);
case SYM_ANY:
mincount = 0;
case SYM_SOME:
maxcount = MAX_I32;
continue;
case SYM_OPT:
mincount = 0;
continue;
case SYM_COPY:
SET_FLAG(flags, PF_COPY);
case SYM_SET:
SET_FLAG(flags, PF_SET);
item = rules++;
if (!IS_WORD(item)) Trap1(RE_PARSE_VARIABLE, item);
if (VAL_CMD(item)) Trap1(RE_PARSE_COMMAND, item);
word = item;
continue;
case SYM_NOT:
SET_FLAG(flags, PF_NOT);
flags ^= (1<<PF_NOT2);
continue;
case SYM_AND:
SET_FLAG(flags, PF_AND);
continue;
case SYM_THEN:
SET_FLAG(flags, PF_THEN);
continue;
case SYM_REMOVE:
SET_FLAG(flags, PF_REMOVE);
continue;
case SYM_INSERT:
SET_FLAG(flags, PF_INSERT);
goto post;
case SYM_CHANGE:
SET_FLAG(flags, PF_CHANGE);
continue;
case SYM_RETURN:
if (IS_PAREN(rules)) {
item = Do_Block_Value_Throw(rules); // might GC
Throw_Return_Value(item);
}
SET_FLAG(flags, PF_RETURN);
continue;
case SYM_ACCEPT:
case SYM_BREAK:
parse->result = 1;
return index;
case SYM_REJECT:
parse->result = -1;
return index;
case SYM_FAIL:
index = NOT_FOUND;
goto post;
case SYM_IF:
item = rules++;
if (IS_END(item)) goto bad_end;
if (!IS_PAREN(item)) Trap1(RE_PARSE_RULE, item);
item = Do_Block_Value_Throw(item); // might GC
if (IS_TRUE(item)) continue;
else {
index = NOT_FOUND;
goto post;
}
case SYM_LIMIT:
Trap0(RE_NOT_DONE);
//val = Get_Parse_Value(rules++);
// if (IS_INTEGER(val)) limit = index + Int32(val);
// else if (ANY_SERIES(val)) limit = VAL_INDEX(val);
// else goto
//goto bad_rule;
// goto post;
case SYM_QQ:
Print_Parse_Index(parse->type, rules, series, index);
continue;
}
}
// Any other cmd must be a match command, so proceed...
} else { // It's not a PARSE command, get or set it:
// word: - set a variable to the series at current index
if (IS_SET_WORD(item)) {
Set_Var_Series(item, parse->type, series, index);
continue;
}
// :word - change the index for the series to a new position
if (IS_GET_WORD(item)) {
item = Get_Var(item);
// CureCode #1263 change
//if (parse->type != VAL_TYPE(item) || VAL_SERIES(item) != series)
// Trap1(RE_PARSE_SERIES, rules-1);
if (!ANY_SERIES(item)) Trap1(RE_PARSE_SERIES, rules-1);
index = Set_Parse_Series(parse, item);
series = parse->series;
continue;
}
// word - some other variable
if (IS_WORD(item)) {
item = Get_Var(item);
}
// item can still be 'word or /word
}
}
else if (ANY_PATH(item)) {
item = Do_Parse_Path(item, parse, &index); // index can be modified
if (index > series->tail) index = series->tail;
if (item == 0) continue; // for SET and GET cases
}
if (IS_PAREN(item)) {
Do_Block_Value_Throw(item); // might GC
if (index > series->tail) index = series->tail;
continue;
}
// Counter? 123
if (IS_INTEGER(item)) { // Specify count or range count
SET_FLAG(flags, PF_WHILE);
mincount = maxcount = Int32s(item, 0);
item = Get_Parse_Value(rules++);
if (IS_END(item)) Trap1(RE_PARSE_END, rules-2);
if (IS_INTEGER(item)) {
maxcount = Int32s(item, 0);
item = Get_Parse_Value(rules++);
if (IS_END(item)) Trap1(RE_PARSE_END, rules-2);
}
}
// else fall through on other values and words
//--------------------------------------------------------------------
// Iterated Rule Matching Section:
//
// Repeats the same rule N times or until the rule fails.
// The index is advanced and stored in a temp variable i until
// the entire rule has been satisfied.
//--------------------------------------------------------------------
item_hold = item; // a command or literal match value
if (VAL_TYPE(item) <= REB_UNSET || VAL_TYPE(item) >= REB_NATIVE) goto bad_rule;
begin = index; // input at beginning of match section
rulen = 0; // rules consumed (do not use rule++ below)
i = index;
//note: rules var already advanced
for (count = 0; count < maxcount;) {
item = item_hold;
if (IS_WORD(item)) {
switch (cmd = VAL_WORD_CANON(item)) {
case SYM_SKIP:
i = (index < series->tail) ? index+1 : NOT_FOUND;
break;
case SYM_END:
i = (index < series->tail) ? NOT_FOUND : series->tail;
break;
case SYM_TO:
case SYM_THRU:
if (IS_END(rules)) goto bad_end;
item = Get_Parse_Value(rules);
rulen = 1;
i = Parse_To(parse, index, item, cmd == SYM_THRU);
break;
case SYM_QUOTE:
if (IS_END(rules)) goto bad_end;
rulen = 1;
if (IS_PAREN(rules)) {
item = Do_Block_Value_Throw(rules); // might GC
}
else item = rules;
i = (0 == Cmp_Value(BLK_SKIP(series, index), item, parse->flags & AM_FIND_CASE)) ? index+1 : NOT_FOUND;
break;
case SYM_INTO:
if (IS_END(rules)) goto bad_end;
rulen = 1;
item = Get_Parse_Value(rules); // sub-rules
if (!IS_BLOCK(item)) goto bad_rule;
val = BLK_SKIP(series, index);
i = (
(ANY_BINSTR(val) || ANY_BLOCK(val))
&& (Parse_Series(val, VAL_BLK_DATA(item), parse->flags, depth+1) == VAL_TAIL(val))
) ? index+1 : NOT_FOUND;
break;
case SYM_DO:
if (!IS_BLOCK_INPUT(parse)) goto bad_rule;
i = Do_Eval_Rule(parse, index, &rules);
rulen = 1;
break;
default:
goto bad_rule;
}
}
else if (IS_BLOCK(item)) {
item = VAL_BLK_DATA(item);
//if (IS_END(rules) && item == rule_head) {
// rules = item;
// goto top;
//}
i = Parse_Rules_Loop(parse, index, item, depth+1);
if (parse->result) {
index = (parse->result > 0) ? i : NOT_FOUND;
parse->result = 0;
break;
}
}
// Parse according to datatype:
else {
if (IS_BLOCK_INPUT(parse))
i = Parse_Next_Block(parse, index, item, depth+1);
else
i = Parse_Next_String(parse, index, item, depth+1);
}
// Necessary for special cases like: some [to end]
// i: indicates new index or failure of the match, but
// that does not mean failure of the rule, because optional
// matches can still succeed, if if the last match failed.
if (i != NOT_FOUND) {
count++; // may overflow to negative
if (count < 0) count = MAX_I32; // the forever case
// If input did not advance:
if (i == index && !GET_FLAG(flags, PF_WHILE)) {
if (count < mincount) index = NOT_FOUND; // was not enough
break;
}
}
//if (i >= series->tail) { // OLD check: no more input
else {
if (count < mincount) index = NOT_FOUND; // was not enough
else if (i != NOT_FOUND) index = i;
// else keep index as is.
break;
}
index = i;
// A BREAK word stopped us:
//if (parse->result) {parse->result = 0; break;}
}
rules += rulen;
//if (index > series->tail && index != NOT_FOUND) index = series->tail;
if (index > series->tail) index = NOT_FOUND;
//--------------------------------------------------------------------
// Post Match Processing:
//--------------------------------------------------------------------
post:
// Process special flags:
if (flags) {
// NOT before all others:
if (GET_FLAG(flags, PF_NOT)) {
if (GET_FLAG(flags, PF_NOT2) && index != NOT_FOUND) index = NOT_FOUND;
else index = begin;
}
if (index == NOT_FOUND) { // Failure actions:
// not decided: if (word) Set_Var_Basic(word, REB_NONE);
if (GET_FLAG(flags, PF_THEN)) {
SKIP_TO_BAR(rules);
if (!IS_END(rules)) rules++;
}
}
else { // Success actions:
count = (begin > index) ? 0 : index - begin; // how much we advanced the input
if (GET_FLAG(flags, PF_COPY)) {
ser = (IS_BLOCK_INPUT(parse))
? Copy_Block_Len(series, begin, count)
: Copy_String(series, begin, count); // condenses
Set_Var_Series(word, parse->type, ser, 0);
}
else if (GET_FLAG(flags, PF_SET)) {
if (IS_BLOCK_INPUT(parse)) {
item = Get_Var_Safe(word);
if (count == 0) SET_NONE(item);
else *item = *BLK_SKIP(series, begin);
}
else {
item = Get_Var_Safe(word);
if (count == 0) SET_NONE(item);
else {
i = GET_ANY_CHAR(series, begin);
if (parse->type == REB_BINARY) {
SET_INTEGER(item, i);
} else {
SET_CHAR(item, i);
}
}
}
}
if (GET_FLAG(flags, PF_RETURN)) {
ser = (IS_BLOCK_INPUT(parse))
? Copy_Block_Len(series, begin, count)
: Copy_String(series, begin, count); // condenses
Throw_Return_Series(parse->type, ser);
}
if (GET_FLAG(flags, PF_REMOVE)) {
if (count) Remove_Series(series, begin, count);
index = begin;
}
if (flags & (1<<PF_INSERT | 1<<PF_CHANGE)) {
count = GET_FLAG(flags, PF_INSERT) ? 0 : count;
cmd = GET_FLAG(flags, PF_INSERT) ? 0 : (1<<AN_PART);
item = rules++;
if (IS_END(item)) goto bad_end;
// Check for ONLY flag:
if (IS_WORD(item) && NZ(cmd = VAL_CMD(item))) {
if (cmd != SYM_ONLY) goto bad_rule;
cmd |= (1<<AN_ONLY);
item = rules++;
}
// CHECK FOR QUOTE!!
item = Get_Parse_Value(item); // new value
if (IS_UNSET(item)) Trap1(RE_NO_VALUE, rules-1);
if (IS_END(item)) goto bad_end;
if (IS_BLOCK_INPUT(parse)) {
index = Modify_Block(GET_FLAG(flags, PF_CHANGE) ? A_CHANGE : A_INSERT,
series, begin, item, cmd, count, 1);
if (IS_LIT_WORD(item)) SET_TYPE(BLK_SKIP(series, index-1), REB_WORD);
}
else {
if (parse->type == REB_BINARY) cmd |= (1<<AN_SERIES); // special flag
index = Modify_String(GET_FLAG(flags, PF_CHANGE) ? A_CHANGE : A_INSERT,
series, begin, item, cmd, count, 1);
}
}
if (GET_FLAG(flags, PF_AND)) index = begin;
}
flags = 0;
word = 0;
}
// Goto alternate rule and reset input:
if (index == NOT_FOUND) {
SKIP_TO_BAR(rules);
if (IS_END(rules)) break;
rules++;
index = begin = start;
}
begin = index;
mincount = maxcount = 1;
}
return index;
bad_rule:
Trap1(RE_PARSE_RULE, rules-1);
bad_end:
Trap1(RE_PARSE_END, rules-1);
return 0;
}
*/ static int Loop_All(REBVAL *ds, REBINT mode)
/*
** 0: forall
** 1: forskip
**
***********************************************************************/
{
REBVAL *var;
REBSER *body;
REBCNT bodi;
REBSER *dat;
REBINT idx;
REBINT inc = 1;
REBCNT type;
var = Get_Var(D_ARG(1));
if (IS_NONE(var)) return R_NONE;
// Save the starting var value:
*D_ARG(1) = *var;
SET_NONE(D_RET);
if (mode == 1) inc = Int32(D_ARG(2));
type = VAL_TYPE(var);
body = VAL_SERIES(D_ARG(mode+2));
bodi = VAL_INDEX(D_ARG(mode+2));
// Starting location when past end with negative skip:
if (inc < 0 && VAL_INDEX(var) >= (REBINT)VAL_TAIL(var)) {
VAL_INDEX(var) = (REBINT)VAL_TAIL(var) + inc;
}
// NOTE: This math only works for index in positive ranges!
if (ANY_SERIES(var)) {
while (TRUE) {
dat = VAL_SERIES(var);
idx = (REBINT)VAL_INDEX(var);
if (idx < 0) break;
if (idx >= (REBINT)SERIES_TAIL(dat)) {
if (inc >= 0) break;
idx = (REBINT)SERIES_TAIL(dat) + inc; // negative
if (idx < 0) break;
VAL_INDEX(var) = idx;
}
ds = Do_Blk(body, bodi); // (may move stack)
if (THROWN(ds)) { // Break, throw, continue, error.
if (Check_Error(ds) >= 0) {
*DS_RETURN = *DS_NEXT;
break;
}
}
*DS_RETURN = *ds;
if (VAL_TYPE(var) != type) Trap_Arg(var);
VAL_INDEX(var) += inc;
}
}
else Trap_Arg(var);
// !!!!! ???? allowed to write VAR????
*var = *DS_ARG(1);
return R_RET;
}
*/ static int Loop_All(struct Reb_Call *call_, REBINT mode)
/*
** 0: forall
** 1: forskip
**
***********************************************************************/
{
REBVAL *var;
REBSER *body;
REBCNT bodi;
REBSER *dat;
REBINT idx;
REBINT inc = 1;
REBCNT type;
REBVAL *ds;
var = GET_MUTABLE_VAR(D_ARG(1));
if (IS_NONE(var)) return R_NONE;
// Save the starting var value:
*D_ARG(1) = *var;
SET_NONE(D_OUT);
if (mode == 1) inc = Int32(D_ARG(2));
type = VAL_TYPE(var);
body = VAL_SERIES(D_ARG(mode+2));
bodi = VAL_INDEX(D_ARG(mode+2));
// Starting location when past end with negative skip:
if (inc < 0 && VAL_INDEX(var) >= VAL_TAIL(var)) {
VAL_INDEX(var) = VAL_TAIL(var) + inc;
}
// NOTE: This math only works for index in positive ranges!
if (ANY_SERIES(var)) {
while (TRUE) {
dat = VAL_SERIES(var);
idx = VAL_INDEX(var);
if (idx < 0) break;
if (idx >= cast(REBINT, SERIES_TAIL(dat))) {
if (inc >= 0) break;
idx = SERIES_TAIL(dat) + inc; // negative
if (idx < 0) break;
VAL_INDEX(var) = idx;
}
if (Do_Block_Throws(D_OUT, body, bodi)) {
if (Loop_Throw_Should_Return(D_OUT)) {
// return value is set, but we still need to assign var
break;
}
}
if (VAL_TYPE(var) != type) raise Error_Invalid_Arg(var);
VAL_INDEX(var) += inc;
}
}
else
raise Error_Invalid_Arg(var);
// !!!!! ???? allowed to write VAR????
*var = *D_ARG(1);
return R_OUT;
}
