*/ REBINT Compare_String_Vals(REBVAL *v1, REBVAL *v2, REBOOL uncase)
/*
** Compare two string values. Either can be byte or unicode wide.
**
** Uncase: compare is case-insensitive.
**
** Used for: general string comparions (various places)
**
***********************************************************************/
{
REBCNT l1 = VAL_LEN(v1);
REBCNT l2 = VAL_LEN(v2);
REBCNT len = MIN(l1, l2);
REBINT n;
if (IS_BINARY(v1) || IS_BINARY(v2)) uncase = FALSE;
if (VAL_BYTE_SIZE(v1)) { // v1 is 8
if (VAL_BYTE_SIZE(v2))
n = Compare_Bytes(VAL_BIN_DATA(v1), VAL_BIN_DATA(v2), len, uncase);
else
n = -Compare_Uni_Byte(VAL_UNI_DATA(v2), VAL_BIN_DATA(v1), len, uncase);
}
else { // v1 is 16
if (VAL_BYTE_SIZE(v2))
n = Compare_Uni_Byte(VAL_UNI_DATA(v1), VAL_BIN_DATA(v2), len, uncase);
else
n = Compare_Uni_Str(VAL_UNI_DATA(v1), VAL_UNI_DATA(v2), len, uncase);
}
if (n != 0) return n;
return l1 - l2;
}
*/ REBINT PD_String(REBPVS *pvs)
/*
***********************************************************************/
{
REBVAL *data = pvs->value;
REBVAL *val = pvs->setval;
REBINT n = 0;
REBCNT i;
REBINT c;
REBSER *ser = VAL_SERIES(data);
if (IS_INTEGER(pvs->select)) {
n = Int32(pvs->select) + VAL_INDEX(data) - 1;
}
else return PE_BAD_SELECT;
if (val == 0) {
if (n < 0 || (REBCNT)n >= SERIES_TAIL(ser)) return PE_NONE;
if (IS_BINARY(data)) {
SET_INTEGER(pvs->store, *BIN_SKIP(ser, n));
} else {
SET_CHAR(pvs->store, GET_ANY_CHAR(ser, n));
}
return PE_USE;
}
if (n < 0 || (REBCNT)n >= SERIES_TAIL(ser)) return PE_BAD_RANGE;
if (IS_CHAR(val)) {
c = VAL_CHAR(val);
if (c > MAX_CHAR) return PE_BAD_SET;
}
else if (IS_INTEGER(val)) {
c = Int32(val);
if (c > MAX_CHAR || c < 0) return PE_BAD_SET;
if (IS_BINARY(data)) { // special case for binary
if (c > 0xff) Trap_Range(val);
BIN_HEAD(ser)[n] = (REBYTE)c;
return PE_OK;
}
}
else if (ANY_BINSTR(val)) {
i = VAL_INDEX(val);
if (i >= VAL_TAIL(val)) return PE_BAD_SET;
c = GET_ANY_CHAR(VAL_SERIES(val), i);
}
else
return PE_BAD_SELECT;
TRAP_PROTECT(ser);
if (BYTE_SIZE(ser) && c > 0xff) Widen_String(ser);
SET_ANY_CHAR(ser, n, c);
return PE_OK;
}
*/ REBSER *Temp_Bin_Str_Managed(REBVAL *val, REBCNT *index, REBCNT *length)
/*
** Determines if UTF8 conversion is needed for a series before it
** is used with a byte-oriented function.
**
** If conversion is needed, a UTF8 series will be created. Otherwise,
** the source series is returned as-is.
**
** Note: This routine should only be used to generate a value used
** for temporary purposes, because it has a "surprising variance"
** regarding its input. If the value's series can be reused, it is--
** and this depends on an implementation detail of internal encoding
** that the user should not be aware of (they need not know if the
** internal representation of an ASCII string uses 1, 2, or however
** many bytes). But copying vs. non-copying means the resulting
** data might or might not have previous values available to step
** back into from the originating series!
**
** !!! Should performance dictate it, the callsites could be
** adapted to know whether this produced a new series or not, and
** instead of managing a created result they could be responsible
** for freeing it if so.
**
***********************************************************************/
{
REBCNT len = (length && *length) ? *length : VAL_LEN(val);
REBSER *series;
assert(IS_BINARY(val) || ANY_STR(val));
if (len == 0 || IS_BINARY(val) || VAL_STR_IS_ASCII(val)) {
// If it's zero length, BINARY!, or an ANY-STRING! whose bytes are
// all values less than 128, we reuse the series.
series = VAL_SERIES(val);
ASSERT_SERIES_MANAGED(series);
if (index) *index = VAL_INDEX(val);
if (length) *length = len;
}
else {
// UTF-8 conversion is required, and we manage the result.
series = Make_UTF8_From_Any_String(val, len, OPT_ENC_CRLF_MAYBE);
MANAGE_SERIES(series);
if (index) *index = 0;
if (length) *length = SERIES_TAIL(series);
}
return series;
}
static REBCNT find_string(REBSER *series, REBCNT index, REBCNT end, REBVAL *target, REBCNT len, REBCNT flags, REBINT skip)
{
REBCNT start = index;
if (flags & (AM_FIND_REVERSE | AM_FIND_LAST)) {
skip = -1;
start = 0;
if (flags & AM_FIND_LAST) index = end - len;
else index--;
}
if (ANY_BINSTR(target)) {
// Do the optimal search or the general search?
if (BYTE_SIZE(series) && VAL_BYTE_SIZE(target) && !(flags & ~(AM_FIND_CASE|AM_FIND_MATCH)))
return Find_Byte_Str(series, start, VAL_BIN_DATA(target), len, !GET_FLAG(flags, ARG_FIND_CASE-1), GET_FLAG(flags, ARG_FIND_MATCH-1));
else
return Find_Str_Str(series, start, index, end, skip, VAL_SERIES(target), VAL_INDEX(target), len, flags & (AM_FIND_MATCH|AM_FIND_CASE));
}
else if (IS_BINARY(target)) {
return Find_Byte_Str(series, start, VAL_BIN_DATA(target), len, 0, GET_FLAG(flags, ARG_FIND_MATCH-1));
}
else if (IS_CHAR(target)) {
return Find_Str_Char(series, start, index, end, skip, VAL_CHAR(target), flags);
}
else if (IS_INTEGER(target)) {
return Find_Str_Char(series, start, index, end, skip, (REBUNI)VAL_INT32(target), flags);
}
else if (IS_BITSET(target)) {
return Find_Str_Bitset(series, start, index, end, skip, VAL_SERIES(target), flags);
}
return NOT_FOUND;
}
//
// Find_Key: C
//
// Returns hash index (either the match or the new one).
// A return of zero is valid (as a hash index);
//
// Wide: width of record (normally 2, a key and a value).
//
// Modes:
// 0 - search, return hash if found or not
// 1 - search, return hash, else return -1 if not
// 2 - search, return hash, else append value and return -1
//
REBINT Find_Key(REBSER *series, REBSER *hser, const REBVAL *key, REBINT wide, REBCNT cased, REBYTE mode)
{
REBCNT *hashes;
REBCNT skip;
REBCNT hash;
REBCNT len;
REBCNT n;
REBVAL *val;
// Compute hash for value:
len = hser->tail;
hash = Hash_Value(key, len);
if (!hash) fail (Error_Has_Bad_Type(key));
// Determine skip and first index:
skip = (len == 0) ? 0 : (hash & 0x0000FFFF) % len;
if (skip == 0) skip = 1;
hash = (len == 0) ? 0 : (hash & 0x00FFFF00) % len;
// Scan hash table for match:
hashes = (REBCNT*)hser->data;
if (ANY_WORD(key)) {
while ((n = hashes[hash])) {
val = BLK_SKIP(series, (n-1) * wide);
if (
ANY_WORD(val) &&
(VAL_WORD_SYM(key) == VAL_WORD_SYM(val) ||
(!cased && VAL_WORD_CANON(key) == VAL_WORD_CANON(val)))
) return hash;
hash += skip;
if (hash >= len) hash -= len;
}
}
else if (ANY_BINSTR(key)) {
while ((n = hashes[hash])) {
val = BLK_SKIP(series, (n-1) * wide);
if (
VAL_TYPE(val) == VAL_TYPE(key)
&& 0 == Compare_String_Vals(key, val, (REBOOL)(!IS_BINARY(key) && !cased))
) return hash;
hash += skip;
if (hash >= len) hash -= len;
}
} else {
while ((n = hashes[hash])) {
val = BLK_SKIP(series, (n-1) * wide);
if (VAL_TYPE(val) == VAL_TYPE(key) && 0 == Cmp_Value(key, val, !cased)) return hash;
hash += skip;
if (hash >= len) hash -= len;
}
}
// Append new value the target series:
if (mode > 1) {
hashes[hash] = SERIES_TAIL(series) + 1;
Append_Values_Len(series, key, wide);
}
return (mode > 0) ? NOT_FOUND : hash;
}
*/ void *Use_Port_State(REBSER *port, REBCNT device, REBCNT size)
/*
** Use private state area in a port. Create if necessary.
** The size is that of a binary structure used by
** the port for storing internal information.
**
***********************************************************************/
{
REBVAL *state = BLK_SKIP(port, STD_PORT_STATE);
// If state is not a binary structure, create it:
if (!IS_BINARY(state)) {
REBSER *data = Make_Binary(size);
REBREQ *req = (REBREQ*)STR_HEAD(data);
req->clen = size;
CLEAR(STR_HEAD(data), size);
//data->tail = size; // makes it easier for ACCEPT to clone the port
SET_FLAG(req->flags, RRF_ALLOC); // not on stack
req->port = port;
req->device = device;
Val_Init_Binary(state, data);
}
return (void *)VAL_BIN(state);
}
//
// MAKE_String: C
//
void MAKE_String(REBVAL *out, enum Reb_Kind kind, const REBVAL *def) {
REBSER *ser; // goto would cross initialization
if (IS_INTEGER(def)) {
//
// !!! R3-Alpha tolerated decimal, e.g. `make string! 3.14`, which
// is semantically nebulous (round up, down?) and generally bad.
//
ser = Make_Binary(Int32s(def, 0));
Val_Init_Series(out, kind, ser);
return;
}
else if (IS_BLOCK(def)) {
//
// The construction syntax for making strings or binaries that are
// preloaded with an offset into the data is #[binary [#{0001} 2]].
// In R3-Alpha make definitions didn't have to be a single value
// (they are for compatibility between construction syntax and MAKE
// in Ren-C). So the positional syntax was #[binary! #{0001} 2]...
// while #[binary [#{0001} 2]] would join the pieces together in order
// to produce #{000102}. That behavior is not available in Ren-C.
if (VAL_ARRAY_LEN_AT(def) != 2)
goto bad_make;
RELVAL *any_binstr = VAL_ARRAY_AT(def);
if (!ANY_BINSTR(any_binstr))
goto bad_make;
if (IS_BINARY(any_binstr) != LOGICAL(kind == REB_BINARY))
goto bad_make;
RELVAL *index = VAL_ARRAY_AT(def) + 1;
if (!IS_INTEGER(index))
goto bad_make;
REBINT i = Int32(index) - 1 + VAL_INDEX(any_binstr);
if (i < 0 || i > cast(REBINT, VAL_LEN_AT(any_binstr)))
goto bad_make;
Val_Init_Series_Index(out, kind, VAL_SERIES(any_binstr), i);
return;
}
if (kind == REB_BINARY)
ser = make_binary(def, TRUE);
else
ser = MAKE_TO_String_Common(def);
if (!ser)
goto bad_make;
Val_Init_Series_Index(out, kind, ser, 0);
return;
bad_make:
fail (Error_Bad_Make(kind, def));
}
*/ REBFLG Is_Port_Open(REBSER *port)
/*
** Standard method for checking if port is open.
** A convention. Not all ports use this method.
**
***********************************************************************/
{
REBVAL *state = BLK_SKIP(port, STD_PORT_STATE);
if (!IS_BINARY(state)) return FALSE;
return IS_OPEN(VAL_BIN_DATA(state));
}
*/ static REBCNT Set_Parse_Series(REBPARSE *parse, REBVAL *item)
/*
** Change the series and return the new index.
**
***********************************************************************/
{
parse->series = VAL_SERIES(item);
parse->type = VAL_TYPE(item);
if (IS_BINARY(item) || (parse->flags & PF_CASED)) parse->flags |= PF_CASE;
else parse->flags &= ~PF_CASE;
return (VAL_INDEX(item) > VAL_TAIL(item)) ? VAL_TAIL(item) : VAL_INDEX(item);
}
*/ void Set_Port_Open(REBSER *port, REBFLG flag)
/*
** Standard method for setting a port open/closed.
** A convention. Not all ports use this method.
**
***********************************************************************/
{
REBVAL *state = BLK_SKIP(port, STD_PORT_STATE);
if (IS_BINARY(state)) {
if (flag) SET_OPEN(VAL_BIN_DATA(state));
else SET_CLOSED(VAL_BIN_DATA(state));
}
}
Handle<Value> CreateExternalArrayBuffer(const Arguments& args) {
if (args.Length() == 0) {
return ThrowException(
String::New("ArrayBuffer constructor must have one parameter."));
}
if (args[0]->IsObject() && IS_BINARY(args[0])) {
return CreateExternalArrayBuffer(args[0]);
}
TryCatch try_catch;
int32_t length = convertToUint(args[0], &try_catch);
if (try_catch.HasCaught()) return try_catch.Exception();
return CreateExternalArrayBuffer(length);
}
static REBSER *make_string(REBVAL *arg, REBOOL make)
{
REBSER *ser = 0;
// MAKE <type> 123
if (make && (IS_INTEGER(arg) || IS_DECIMAL(arg))) {
ser = Make_Binary(Int32s(arg, 0));
}
// MAKE/TO <type> <binary!>
else if (IS_BINARY(arg)) {
REBYTE *bp = VAL_BIN_DATA(arg);
REBCNT len = VAL_LEN(arg);
switch (What_UTF(bp, len)) {
case 0:
break;
case 8: // UTF-8 encoded
bp += 3;
len -= 3;
break;
default:
Trap0(RE_BAD_DECODE);
}
ser = Decode_UTF_String(bp, len, 8); // UTF-8
}
// MAKE/TO <type> <any-string>
else if (ANY_BINSTR(arg)) {
ser = Copy_String(VAL_SERIES(arg), VAL_INDEX(arg), VAL_LEN(arg));
}
// MAKE/TO <type> <any-word>
else if (ANY_WORD(arg)) {
ser = Copy_Mold_Value(arg, TRUE);
//ser = Append_UTF8(0, Get_Word_Name(arg), -1);
}
// MAKE/TO <type> #"A"
else if (IS_CHAR(arg)) {
ser = (VAL_CHAR(arg) > 0xff) ? Make_Unicode(2) : Make_Binary(2);
Append_Byte(ser, VAL_CHAR(arg));
}
// MAKE/TO <type> <any-value>
// else if (IS_NONE(arg)) {
// ser = Make_Binary(0);
// }
else
ser = Copy_Form_Value(arg, 1<<MOPT_TIGHT);
return ser;
}
//
// Bin_To_Money_May_Fail: C
//
// Will successfully convert or fail (longjmp) with an error.
//
void Bin_To_Money_May_Fail(REBVAL *result, REBVAL *val)
{
REBCNT len;
REBYTE buf[MAX_HEX_LEN+4] = {0}; // binary to convert
if (IS_BINARY(val)) {
len = VAL_LEN_AT(val);
if (len > 12) len = 12;
memcpy(buf, VAL_BIN_AT(val), len);
}
else
fail (Error_Invalid_Arg(val));
memcpy(buf + 12 - len, buf, len); // shift to right side
memset(buf, 0, 12 - len);
VAL_MONEY_AMOUNT(result) = binary_to_deci(buf);
}
STOID Mold_All_String(REBVAL *value, REB_MOLD *mold)
{
// The string that is molded for /all option:
REBVAL val;
//// ???? move to above Mold_String_Series function????
Pre_Mold(value, mold); // #[file! part
val = *value;
VAL_INDEX(&val) = 0;
if (IS_BINARY(value)) Mold_Binary(&val, mold);
else {
VAL_SET(&val, REB_STRING);
Mold_String_Series(&val, mold);
}
Post_Mold(value, mold);
}
*/ REBFLG Pending_Port(REBVAL *port)
/*
** Return TRUE if port value is pending a signal.
** Not valid for all ports - requires request struct!!!
**
***********************************************************************/
{
REBVAL *state;
REBREQ *req;
if (IS_PORT(port)) {
state = BLK_SKIP(VAL_PORT(port), STD_PORT_STATE);
if (IS_BINARY(state)) {
req = (REBREQ*)VAL_BIN(state);
if (!GET_FLAG(req->flags, RRF_PENDING)) return FALSE;
}
}
return TRUE;
}
//
// MT_Bitset: C
//
REBOOL MT_Bitset(REBVAL *out, REBVAL *data, enum Reb_Kind type)
{
REBOOL is_not = FALSE;
if (IS_BLOCK(data)) {
REBINT len = Find_Max_Bit(data);
REBSER *ser;
if (len < 0 || len > 0xFFFFFF) fail (Error_Invalid_Arg(data));
ser = Make_Bitset(len);
Set_Bits(ser, data, TRUE);
Val_Init_Bitset(out, ser);
return TRUE;
}
if (!IS_BINARY(data)) return FALSE;
Val_Init_Bitset(out, Copy_Sequence_At_Position(data));
BITS_NOT(VAL_SERIES(out)) = FALSE;
return TRUE;
}
*/ REBFLG MT_Bitset(REBVAL *out, REBVAL *data, REBCNT type)
/*
***********************************************************************/
{
REBFLG is_not = 0;
if (IS_BLOCK(data)) {
REBINT len = Find_Max_Bit(data);
REBSER *ser;
if (len < 0 || len > 0xFFFFFF) raise Error_Invalid_Arg(data);
ser = Make_Bitset(len);
Set_Bits(ser, data, TRUE);
Val_Init_Bitset(out, ser);
return TRUE;
}
if (!IS_BINARY(data)) return FALSE;
Val_Init_Bitset(out, Copy_Sequence_At_Position(data));
BITS_NOT(VAL_SERIES(out)) = 0;
return TRUE;
}
*/ REBINT Bin_To_Money(REBVAL *result, REBVAL *val)
/*
***********************************************************************/
{
REBCNT len;
REBYTE buf[MAX_HEX_LEN+4] = {0}; // binary to convert
if (IS_BINARY(val)) {
len = VAL_LEN(val);
if (len > 12) len = 12;
memcpy(buf, VAL_BIN_DATA(val), len);
}
#ifdef removed
else if (IS_ISSUE(val)) {
//if (!(len = Scan_Hex_Bytes(val, 24, buf))) return FALSE;
REBYTE *ap = Get_Word_Name(val);
REBYTE *bp = &buf[0];
REBCNT alen;
REBUNI c;
len = LEN_BYTES(ap); // UTF-8 len
if (len & 1) return FALSE; // must have even # of chars
len /= 2;
if (len > 12) return FALSE; // valid even for UTF-8
for (alen = 0; alen < len; alen++) {
if (!Scan_Hex2(ap, &c, 0)) return FALSE;
*bp++ = (REBYTE)c;
ap += 2;
}
}
#endif
else
raise Error_Invalid_Arg(val);
memcpy(buf + 12 - len, buf, len); // shift to right side
memset(buf, 0, 12 - len);
VAL_MONEY_AMOUNT(result) = binary_to_deci(buf);
return TRUE;
}
static REBSER *MAKE_TO_String_Common(const REBVAL *arg)
{
REBSER *ser = 0;
// MAKE/TO <type> <binary!>
if (IS_BINARY(arg)) {
REBYTE *bp = VAL_BIN_AT(arg);
REBCNT len = VAL_LEN_AT(arg);
switch (What_UTF(bp, len)) {
case 0:
break;
case 8: // UTF-8 encoded
bp += 3;
len -= 3;
break;
default:
fail (Error(RE_BAD_UTF8));
}
ser = Decode_UTF_String(bp, len, 8); // UTF-8
}
// MAKE/TO <type> <any-string>
else if (ANY_BINSTR(arg)) {
ser = Copy_String_Slimming(VAL_SERIES(arg), VAL_INDEX(arg), VAL_LEN_AT(arg));
}
// MAKE/TO <type> <any-word>
else if (ANY_WORD(arg)) {
ser = Copy_Mold_Value(arg, 0 /* opts... MOPT_0? */);
}
// MAKE/TO <type> #"A"
else if (IS_CHAR(arg)) {
ser = (VAL_CHAR(arg) > 0xff) ? Make_Unicode(2) : Make_Binary(2);
Append_Codepoint_Raw(ser, VAL_CHAR(arg));
}
else
ser = Copy_Form_Value(arg, 1 << MOPT_TIGHT);
return ser;
}
/* We get here after reading the header+body into an item.
*/
void cproxy_process_upstream_binary_nread(conn *c) {
cb_assert(c != NULL);
cb_assert(c->cmd >= 0);
cb_assert(c->next == NULL);
cb_assert(c->cmd_start == NULL);
cb_assert(IS_BINARY(c->protocol));
cb_assert(IS_PROXY(c->protocol));
protocol_binary_request_header *header =
(protocol_binary_request_header *) &c->binary_header;
int extlen = header->request.extlen;
int keylen = header->request.keylen;
uint32_t bodylen = header->request.bodylen;
if (settings.verbose > 2) {
moxi_log_write("<%d cproxy_process_upstream_binary_nread %x %d %d %u\n",
c->sfd, c->cmd, extlen, keylen, bodylen);
}
/* pthread_mutex_lock(&c->thread->stats.mutex); */
/* c->thread->stats.slab_stats[it->slabs_clsid].set_cmds++; */
/* pthread_mutex_unlock(&c->thread->stats.mutex); */
proxy_td *ptd = c->extra;
cb_assert(ptd != NULL);
if (header->request.opcode == PROTOCOL_BINARY_CMD_SASL_AUTH) {
item *it = c->item;
cb_assert(it);
cproxy_sasl_plain_auth(c, (char *) ITEM_data(it));
return;
}
if (header->request.opcode == PROTOCOL_BINARY_CMD_SASL_STEP) {
write_bin_error(c, PROTOCOL_BINARY_RESPONSE_AUTH_ERROR, 0);
return;
}
if (c->binary_header.request.opcode == PROTOCOL_BINARY_CMD_STAT) {
char *subcommand = binary_get_key(c);
size_t nkey = c->binary_header.request.keylen;
if (nkey == 13 && memcmp(subcommand, "proxy buckets", 13) == 0) {
process_bin_proxy_stats(c);
return;
}
}
if (c->noreply) {
if (settings.verbose > 2) {
moxi_log_write("<%d cproxy_process_upstream_binary_nread "
"corking quiet command %x %d\n",
c->sfd, c->cmd, (c->corked != NULL));
}
/* TODO: We currently don't support binary FLUSHQ. */
/* Rather than having the downstream connections get */
/* into a wonky state, prevent it. */
if (header->request.opcode == PROTOCOL_BINARY_CMD_FLUSHQ) {
/* Note: don't use cproxy_close_conn(c), as it goes */
/* through the drive_machine() loop again. */
/* cproxy_close_conn(c); */
conn_set_state(c, conn_closing);
return;
}
/* Hold onto or 'cork' all the binary quiet commands */
/* until there's a later non-quiet command. */
if (cproxy_binary_cork_cmd(c)) {
conn_set_state(c, conn_new_cmd);
} else {
ptd->stats.stats.err_oom++;
cproxy_close_conn(c);
}
return;
}
cb_assert(c->item == NULL || ((item *) c->item)->refcount == 1);
cproxy_pause_upstream_for_downstream(ptd, c);
}
void cproxy_process_upstream_binary(conn *c) {
cb_assert(c != NULL);
cb_assert(c->cmd >= 0);
cb_assert(c->next == NULL);
cb_assert(c->item == NULL);
cb_assert(IS_BINARY(c->protocol));
cb_assert(IS_PROXY(c->protocol));
proxy_td *ptd = c->extra;
cb_assert(ptd != NULL);
if (!cproxy_prep_conn_for_write(c)) {
ptd->stats.stats.err_upstream_write_prep++;
conn_set_state(c, conn_closing);
return;
}
c->cmd_curr = -1;
c->cmd_start = NULL;
c->cmd_start_time = msec_current_time;
c->cmd_retries = 0;
int extlen = c->binary_header.request.extlen;
int keylen = c->binary_header.request.keylen;
uint32_t bodylen = c->binary_header.request.bodylen;
cb_assert(bodylen >= (uint32_t) keylen + extlen);
if (settings.verbose > 2) {
moxi_log_write("<%d cproxy_process_upstream_binary %x %d %d %u\n",
c->sfd, c->cmd, extlen, keylen, bodylen);
}
process_bin_noreply(c); /* Map quiet c->cmd values into non-quiet. */
if (c->cmd == PROTOCOL_BINARY_CMD_VERSION ||
c->cmd == PROTOCOL_BINARY_CMD_QUIT) {
dispatch_bin_command(c);
return;
}
/* Alloc an item and continue with an rest-of-body nread if */
/* necessary. The item will hold the entire request message */
/* (the header + body). */
char *ikey = "u";
int ikeylen = 1;
c->item = item_alloc(ikey, ikeylen, 0, 0,
sizeof(c->binary_header) + bodylen);
if (c->item != NULL) {
item *it = c->item;
void *rb = c->rcurr;
cb_assert(it->refcount == 1);
memcpy(ITEM_data(it), rb, sizeof(c->binary_header));
if (bodylen > 0) {
c->ritem = ITEM_data(it) + sizeof(c->binary_header);
c->rlbytes = bodylen;
c->substate = bin_read_set_value;
conn_set_state(c, conn_nread);
} else {
/* Since we have no body bytes, we can go immediately to */
/* the nread completed processing step. */
if (c->binary_header.request.opcode == PROTOCOL_BINARY_CMD_SASL_LIST_MECHS) {
/* TODO: One day handle more than just PLAIN sasl auth. */
write_bin_response(c, "PLAIN", 0, 0, strlen("PLAIN"));
return;
}
cproxy_pause_upstream_for_downstream(ptd, c);
}
} else {
if (settings.verbose > 2) {
moxi_log_write("<%d cproxy_process_upstream_binary OOM\n",
c->sfd);
}
ptd->stats.stats.err_oom++;
cproxy_close_conn(c);
}
}
/* We reach here after nread'ing a header+body into an item.
*/
void cproxy_process_b2b_downstream_nread(conn *c) {
conn *uc;
item *it;
downstream *d;
protocol_binary_response_header *header;
int extlen;
int keylen;
uint32_t bodylen;
int status;
int opcode;
cb_assert(c != NULL);
cb_assert(c->cmd >= 0);
cb_assert(c->next == NULL);
cb_assert(c->cmd_start == NULL);
cb_assert(IS_BINARY(c->protocol));
cb_assert(IS_PROXY(c->protocol));
header = (protocol_binary_response_header *) &c->binary_header;
extlen = header->response.extlen;
keylen = header->response.keylen;
bodylen = header->response.bodylen;
status = ntohs(header->response.status);
opcode = header->response.opcode;
if (settings.verbose > 2) {
moxi_log_write("<%d cproxy_process_b2b_downstream_nread %x %x %d %d %u %d %x\n",
c->sfd, c->cmd, opcode, extlen, keylen, bodylen, c->noreply, status);
}
d = c->extra;
cb_assert(d != NULL);
cb_assert(d->ptd != NULL);
cb_assert(d->ptd->proxy != NULL);
/* TODO: Need to handle quiet binary command error response, */
/* in the right order. */
/* TODO: Need to handle not-my-vbucket error during a quiet cmd. */
uc = d->upstream_conn;
it = c->item;
/* Clear c->item because we either move it to the upstream or */
/* item_remove() it on error. */
c->item = NULL;
cb_assert(it != NULL);
cb_assert(it->refcount == 1);
if (cproxy_binary_ignore_reply(c, header, it)) {
return;
}
if (c->noreply) {
conn_set_state(c, conn_new_cmd);
} else {
conn_set_state(c, conn_pause);
if (opcode == PROTOCOL_BINARY_CMD_NOOP ||
opcode == PROTOCOL_BINARY_CMD_FLUSH) {
goto done;
}
if (opcode == PROTOCOL_BINARY_CMD_STAT) {
if (status == PROTOCOL_BINARY_RESPONSE_SUCCESS) {
if (keylen > 0) {
if (d->merger != NULL) {
char *key = (ITEM_data(it)) + sizeof(*header) + extlen;
char *val = key + keylen;
protocol_stats_merge_name_val(d->merger, "STAT", 4,
key, keylen,
val, bodylen - keylen - extlen);
}
conn_set_state(c, conn_new_cmd); /* Get next STATS response. */
}
}
goto done;
}
/* If the client is still there, we should handle */
/* a not-my-vbucket error with a possible retry. */
if (uc != NULL &&
status == PROTOCOL_BINARY_RESPONSE_NOT_MY_VBUCKET) {
int max_retries;
protocol_binary_request_header *req;
int vbucket;
int sindex;
if (settings.verbose > 2) {
moxi_log_write("<%d cproxy_process_b2b_downstream_nread not-my-vbucket, "
"cmd: %x %d\n",
c->sfd, header->response.opcode, uc->item != NULL);
}
cb_assert(uc->item != NULL);
req = (protocol_binary_request_header *)ITEM_data((item*)uc->item);
vbucket = ntohs(req->request.reserved);
sindex = downstream_conn_index(d, c);
if (settings.verbose > 2) {
moxi_log_write("<%d cproxy_process_b2b_downstream_nread not-my-vbucket, "
"cmd: %x not multi-key get, sindex %d, vbucket %d, retries %d\n",
c->sfd, header->response.opcode,
sindex, vbucket, uc->cmd_retries);
}
mcs_server_invalid_vbucket(&d->mst, sindex, vbucket);
/* As long as the upstream is still open and we haven't */
/* retried too many times already. */
max_retries = cproxy_max_retries(d);
if (uc->cmd_retries < max_retries) {
uc->cmd_retries++;
d->upstream_retry++;
d->ptd->stats.stats.tot_retry_vbucket++;
goto done;
}
if (settings.verbose > 2) {
moxi_log_write("%d: cproxy_process_b2b_downstream_nread not-my-vbucket, "
"cmd: %x skipping retry %d >= %d\n",
c->sfd, header->response.opcode, uc->cmd_retries,
max_retries);
}
}
}
/* Write the response to the upstream connection. */
if (uc != NULL) {
if (settings.verbose > 2) {
moxi_log_write("<%d cproxy_process_b2b_downstream_nread got %u\n",
c->sfd, it->nbytes);
cproxy_dump_header(c->sfd, ITEM_data(it));
}
if (add_conn_item(uc, it) == true) {
it->refcount++;
if (add_iov(uc, ITEM_data(it), it->nbytes) == 0) {
/* If we got a quiet response, however, don't change the */
/* upstream connection's state (should be in paused state), */
/* as we expect the downstream server to provide a */
/* verbal/non-quiet response that moves the downstream */
/* conn through the conn_pause countdown codepath. */
if (c->noreply == false) {
cproxy_update_event_write(d, uc);
conn_set_state(uc, conn_mwrite);
}
goto done;
}
}
d->ptd->stats.stats.err_oom++;
cproxy_close_conn(uc);
}
done:
if (it != NULL) {
item_remove(it);
}
}
//
// Modify_String: C
//
// Returns new dst_idx.
//
REBCNT Modify_String(
REBCNT action, // INSERT, APPEND, CHANGE
REBSER *dst_ser, // target
REBCNT dst_idx, // position
const REBVAL *src_val, // source
REBFLGS flags, // AN_PART
REBINT dst_len, // length to remove
REBINT dups // dup count
) {
REBSER *src_ser = 0;
REBCNT src_idx = 0;
REBCNT src_len;
REBCNT tail = SER_LEN(dst_ser);
REBINT size; // total to insert
REBOOL needs_free;
REBINT limit;
// For INSERT/PART and APPEND/PART
if (action != SYM_CHANGE && GET_FLAG(flags, AN_PART))
limit = dst_len; // should be non-negative
else
limit = -1;
if (limit == 0 || dups < 0) return (action == SYM_APPEND) ? 0 : dst_idx;
if (action == SYM_APPEND || dst_idx > tail) dst_idx = tail;
// If the src_val is not a string, then we need to create a string:
if (GET_FLAG(flags, AN_SERIES)) { // used to indicate a BINARY series
if (IS_INTEGER(src_val)) {
src_ser = Make_Series_Codepoint(Int8u(src_val));
needs_free = TRUE;
limit = -1;
}
else if (IS_BLOCK(src_val)) {
src_ser = Join_Binary(src_val, limit); // NOTE: it's the shared FORM buffer!
needs_free = FALSE;
limit = -1;
}
else if (IS_CHAR(src_val)) {
//
// "UTF-8 was originally specified to allow codepoints with up to
// 31 bits (or 6 bytes). But with RFC3629, this was reduced to 4
// bytes max. to be more compatible to UTF-16." So depending on
// which RFC you consider "the UTF-8", max size is either 4 or 6.
//
src_ser = Make_Binary(6);
SET_SERIES_LEN(
src_ser,
Encode_UTF8_Char(BIN_HEAD(src_ser), VAL_CHAR(src_val))
);
needs_free = TRUE;
limit = -1;
}
else if (ANY_STRING(src_val)) {
src_len = VAL_LEN_AT(src_val);
if (limit >= 0 && src_len > cast(REBCNT, limit))
src_len = limit;
src_ser = Make_UTF8_From_Any_String(src_val, src_len, 0);
needs_free = TRUE;
limit = -1;
}
else if (!IS_BINARY(src_val))
fail (Error_Invalid_Arg(src_val));
}
else if (IS_CHAR(src_val)) {
src_ser = Make_Series_Codepoint(VAL_CHAR(src_val));
needs_free = TRUE;
}
else if (IS_BLOCK(src_val)) {
src_ser = Form_Tight_Block(src_val);
needs_free = TRUE;
}
else if (!ANY_STRING(src_val) || IS_TAG(src_val)) {
src_ser = Copy_Form_Value(src_val, 0);
needs_free = TRUE;
}
// Use either new src or the one that was passed:
if (src_ser) {
src_len = SER_LEN(src_ser);
}
else {
src_ser = VAL_SERIES(src_val);
src_idx = VAL_INDEX(src_val);
src_len = VAL_LEN_AT(src_val);
needs_free = FALSE;
}
if (limit >= 0) src_len = limit;
// If Source == Destination we need to prevent possible conflicts.
// Clone the argument just to be safe.
// (Note: It may be possible to optimize special cases like append !!)
if (dst_ser == src_ser) {
assert(!needs_free);
src_ser = Copy_Sequence_At_Len(src_ser, src_idx, src_len);
needs_free = TRUE;
src_idx = 0;
}
// Total to insert:
size = dups * src_len;
if (action != SYM_CHANGE) {
// Always expand dst_ser for INSERT and APPEND actions:
Expand_Series(dst_ser, dst_idx, size);
} else {
if (size > dst_len)
Expand_Series(dst_ser, dst_idx, size - dst_len);
else if (size < dst_len && GET_FLAG(flags, AN_PART))
Remove_Series(dst_ser, dst_idx, dst_len - size);
else if (size + dst_idx > tail) {
EXPAND_SERIES_TAIL(dst_ser, size - (tail - dst_idx));
}
}
// For dup count:
for (; dups > 0; dups--) {
Insert_String(dst_ser, dst_idx, src_ser, src_idx, src_len, TRUE);
dst_idx += src_len;
}
TERM_SEQUENCE(dst_ser);
if (needs_free) {
// If we did not use the series that was passed in, but rather
// created an internal temporary one, we need to free it.
Free_Series(src_ser);
}
return (action == SYM_APPEND) ? 0 : dst_idx;
}
/* Do the actual work of forwarding the command from an
* upstream binary conn to its assigned binary downstream.
*/
bool cproxy_forward_b2b_downstream(downstream *d) {
int nc;
int server_index;
conn *uc;
cb_assert(d != NULL);
cb_assert(d->ptd != NULL);
cb_assert(d->ptd->proxy != NULL);
cb_assert(d->downstream_conns != NULL);
cb_assert(d->downstream_used == 0);
cb_assert(d->multiget == NULL);
cb_assert(d->merger == NULL);
d->downstream_used_start = 0;
uc = d->upstream_conn;
if (settings.verbose > 2) {
moxi_log_write("%d: cproxy_forward_b2b_downstream %x\n",
uc->sfd, uc->cmd);
}
cb_assert(uc != NULL);
cb_assert(uc->state == conn_pause);
cb_assert(uc->cmd >= 0);
cb_assert(uc->cmd_start == NULL);
cb_assert(uc->thread != NULL);
cb_assert(uc->thread->base != NULL);
cb_assert(uc->noreply == false);
cb_assert(IS_BINARY(uc->protocol));
cb_assert(IS_PROXY(uc->protocol));
server_index = -1;
if (cproxy_is_broadcast_cmd(uc->cmd) == false && uc->corked == NULL) {
item *it = uc->item;
protocol_binary_request_header *req;
char *key;
int key_len;
cb_assert(it != NULL);
req = (protocol_binary_request_header *) ITEM_data(it);
key = ((char *) req) + sizeof(*req) + req->request.extlen;
key_len = ntohs(req->request.keylen);
if (key_len > 0) {
server_index = cproxy_server_index(d, key, key_len, NULL);
if (server_index < 0) {
return false;
}
}
}
nc = cproxy_connect_downstream(d, uc->thread, server_index);
if (nc == -1) {
return true;
}
if (nc > 0) {
int i;
int nconns;
cb_assert(d->downstream_conns != NULL);
if (d->usec_start == 0 &&
d->ptd->behavior_pool.base.time_stats) {
d->usec_start = usec_now();
}
nconns = mcs_server_count(&d->mst);
for (i = 0; i < nconns; i++) {
conn *c = d->downstream_conns[i];
if (c != NULL &&
c != NULL_CONN) {
cb_assert(c->state == conn_pause);
cb_assert(c->item == NULL);
if (cproxy_prep_conn_for_write(c) == false) {
d->ptd->stats.stats.err_downstream_write_prep++;
cproxy_close_conn(c);
return false;
}
}
}
/* Uncork the saved-up quiet binary commands. */
cproxy_binary_uncork_cmds(d, uc);
if (uc->cmd == PROTOCOL_BINARY_CMD_FLUSH ||
uc->cmd == PROTOCOL_BINARY_CMD_NOOP ||
uc->cmd == PROTOCOL_BINARY_CMD_STAT) {
return cproxy_broadcast_b2b_downstream(d, uc);
}
return cproxy_forward_b2b_simple_downstream(d, uc);
}
if (settings.verbose > 2) {
moxi_log_write("%d: cproxy_forward_b2b_downstream connect failed\n",
uc->sfd);
}
return false;
}
/* Called when we receive a binary response header from
* a downstream server, via try_read_command()/drive_machine().
*/
void cproxy_process_b2b_downstream(conn *c) {
char *ikey;
int ikeylen;
downstream *d;
int extlen;
int keylen;
uint32_t bodylen;
cb_assert(c != NULL);
cb_assert(c->cmd >= 0);
cb_assert(c->next == NULL);
cb_assert(c->item == NULL);
cb_assert(IS_BINARY(c->protocol));
cb_assert(IS_PROXY(c->protocol));
cb_assert(c->substate == bin_no_state);
d = c->extra;
cb_assert(d);
c->cmd_curr = -1;
c->cmd_start = NULL;
c->cmd_start_time = msec_current_time;
c->cmd_retries = 0;
extlen = c->binary_header.request.extlen;
keylen = c->binary_header.request.keylen;
bodylen = c->binary_header.request.bodylen;
if (settings.verbose > 2) {
moxi_log_write("<%d cproxy_process_b2b_downstream %x %d %d %u\n",
c->sfd, c->cmd, extlen, keylen, bodylen);
}
cb_assert(bodylen >= (uint32_t) keylen + extlen);
process_bin_noreply(c); /* Map quiet c->cmd values into non-quiet. */
/* Our approach is to read everything we can before */
/* getting into big switch/case statements for the */
/* actual processing. */
/* Alloc an item and continue with an rest-of-body nread if */
/* necessary. The item will hold the entire response message */
/* (the header + body). */
ikey = "q";
ikeylen = 1;
c->item = item_alloc(ikey, ikeylen, 0, 0,
sizeof(c->binary_header) + bodylen);
if (c->item != NULL) {
item *it = c->item;
void *rb = c->rcurr;
cb_assert(it->refcount == 1);
memcpy(ITEM_data(it), rb, sizeof(c->binary_header));
if (bodylen > 0) {
c->ritem = ITEM_data(it) + sizeof(c->binary_header);
c->rlbytes = bodylen;
c->substate = bin_read_set_value;
conn_set_state(c, conn_nread);
} else {
/* Since we have no body bytes, we can go immediately to */
/* the nread completed processing step. */
cproxy_process_b2b_downstream_nread(c);
}
} else {
d->ptd->stats.stats.err_oom++;
cproxy_close_conn(c);
}
}
*/ static REBFLG Set_GOB_Var(REBGOB *gob, REBVAL *word, REBVAL *val)
/*
***********************************************************************/
{
switch (VAL_WORD_CANON(word)) {
case SYM_OFFSET:
return Set_Pair(&(gob->offset), val);
case SYM_SIZE:
return Set_Pair(&gob->size, val);
case SYM_IMAGE:
CLR_GOB_OPAQUE(gob);
if (IS_IMAGE(val)) {
SET_GOB_TYPE(gob, GOBT_IMAGE);
GOB_W(gob) = (REBD32)VAL_IMAGE_WIDE(val);
GOB_H(gob) = (REBD32)VAL_IMAGE_HIGH(val);
GOB_CONTENT(gob) = VAL_SERIES(val);
// if (!VAL_IMAGE_TRANSP(val)) SET_GOB_OPAQUE(gob);
}
else if (IS_NONE(val)) SET_GOB_TYPE(gob, GOBT_NONE);
else return FALSE;
break;
case SYM_DRAW:
CLR_GOB_OPAQUE(gob);
if (IS_BLOCK(val)) {
SET_GOB_TYPE(gob, GOBT_DRAW);
GOB_CONTENT(gob) = VAL_SERIES(val);
}
else if (IS_NONE(val)) SET_GOB_TYPE(gob, GOBT_NONE);
else return FALSE;
break;
case SYM_TEXT:
CLR_GOB_OPAQUE(gob);
if (IS_BLOCK(val)) {
SET_GOB_TYPE(gob, GOBT_TEXT);
GOB_CONTENT(gob) = VAL_SERIES(val);
}
else if (IS_STRING(val)) {
SET_GOB_TYPE(gob, GOBT_STRING);
GOB_CONTENT(gob) = VAL_SERIES(val);
}
else if (IS_NONE(val)) SET_GOB_TYPE(gob, GOBT_NONE);
else return FALSE;
break;
case SYM_EFFECT:
CLR_GOB_OPAQUE(gob);
if (IS_BLOCK(val)) {
SET_GOB_TYPE(gob, GOBT_EFFECT);
GOB_CONTENT(gob) = VAL_SERIES(val);
}
else if (IS_NONE(val)) SET_GOB_TYPE(gob, GOBT_NONE);
else return FALSE;
break;
case SYM_COLOR:
CLR_GOB_OPAQUE(gob);
if (IS_TUPLE(val)) {
SET_GOB_TYPE(gob, GOBT_COLOR);
Set_Pixel_Tuple((REBYTE*)&GOB_CONTENT(gob), val);
if (VAL_TUPLE_LEN(val) < 4 || VAL_TUPLE(val)[3] == 0)
SET_GOB_OPAQUE(gob);
}
else if (IS_NONE(val)) SET_GOB_TYPE(gob, GOBT_NONE);
break;
case SYM_PANE:
if (GOB_PANE(gob)) Clear_Series(GOB_PANE(gob));
if (IS_BLOCK(val))
Insert_Gobs(gob, VAL_BLK_DATA(val), 0, VAL_BLK_LEN(val), 0);
else if (IS_GOB(val))
Insert_Gobs(gob, val, 0, 1, 0);
else if (IS_NONE(val))
gob->pane = 0;
else
return FALSE;
break;
case SYM_ALPHA:
GOB_ALPHA(gob) = Clip_Int(Int32(val), 0, 255);
break;
case SYM_DATA:
SET_GOB_DTYPE(gob, GOBD_NONE);
if (IS_OBJECT(val)) {
SET_GOB_DTYPE(gob, GOBD_OBJECT);
SET_GOB_DATA(gob, VAL_OBJ_FRAME(val));
}
else if (IS_BLOCK(val)) {
SET_GOB_DTYPE(gob, GOBD_BLOCK);
SET_GOB_DATA(gob, VAL_SERIES(val));
}
else if (IS_STRING(val)) {
SET_GOB_DTYPE(gob, GOBD_STRING);
SET_GOB_DATA(gob, VAL_SERIES(val));
}
else if (IS_BINARY(val)) {
SET_GOB_DTYPE(gob, GOBD_BINARY);
SET_GOB_DATA(gob, VAL_SERIES(val));
}
else if (IS_INTEGER(val)) {
SET_GOB_DTYPE(gob, GOBD_INTEGER);
SET_GOB_DATA(gob, (void*)(REBIPT)VAL_INT64(val));
}
else if (IS_NONE(val))
SET_GOB_TYPE(gob, GOBT_NONE);
else return FALSE;
break;
case SYM_FLAGS:
if (IS_WORD(val)) Set_Gob_Flag(gob, val);
else if (IS_BLOCK(val)) {
gob->flags = 0;
for (val = VAL_BLK(val); NOT_END(val); val++) {
if (IS_WORD(val)) Set_Gob_Flag(gob, val);
}
}
break;
case SYM_OWNER:
if (IS_GOB(val))
GOB_TMP_OWNER(gob) = VAL_GOB(val);
else
return FALSE;
break;
default:
return FALSE;
}
return TRUE;
}
x*/ void Modify_StringX(REBCNT action, REBVAL *string, REBVAL *arg)
/*
** Actions: INSERT, APPEND, CHANGE
**
** string [string!] {Series at point to insert}
** value [any-type!] {The value to insert}
** /part {Limits to a given length or position.}
** length [number! series! pair!]
** /only {Inserts a series as a series.}
** /dup {Duplicates the insert a specified number of times.}
** count [number! pair!]
**
***********************************************************************/
{
REBSER *series = VAL_SERIES(string);
REBCNT index = VAL_INDEX(string);
REBCNT tail = VAL_TAIL(string);
REBINT rlen; // length to be removed
REBINT ilen = 1; // length to be inserted
REBINT cnt = 1; // DUP count
REBINT size;
REBVAL *val;
REBSER *arg_ser = 0; // argument series
// Length of target (may modify index): (arg can be anything)
rlen = Partial1((action == A_CHANGE) ? string : arg, DS_ARG(AN_LENGTH));
index = VAL_INDEX(string);
if (action == A_APPEND || index > tail) index = tail;
// If the arg is not a string, then we need to create a string:
if (IS_BINARY(string)) {
if (IS_INTEGER(arg)) {
if (VAL_INT64(arg) > 255 || VAL_INT64(arg) < 0)
Trap_Range(arg);
arg_ser = Make_Binary(1);
Append_Byte(arg_ser, VAL_CHAR(arg)); // check for size!!!
}
else if (!ANY_BINSTR(arg)) Trap_Arg(arg);
}
else if (IS_BLOCK(arg)) {
// MOVE!
REB_MOLD mo = {0};
arg_ser = mo.series = Make_Unicode(VAL_BLK_LEN(arg) * 10); // GC!?
for (val = VAL_BLK_DATA(arg); NOT_END(val); val++)
Mold_Value(&mo, val, 0);
}
else if (IS_CHAR(arg)) {
// Optimize this case !!!
arg_ser = Make_Unicode(1);
Append_Byte(arg_ser, VAL_CHAR(arg));
}
else if (!ANY_STR(arg) || IS_TAG(arg)) {
arg_ser = Copy_Form_Value(arg, 0);
}
if (arg_ser) Set_String(arg, arg_ser);
else arg_ser = VAL_SERIES(arg);
// Length of insertion:
ilen = (action != A_CHANGE && DS_REF(AN_PART)) ? rlen : VAL_LEN(arg);
// If Source == Destination we need to prevent possible conflicts.
// Clone the argument just to be safe.
// (Note: It may be possible to optimize special cases like append !!)
if (series == VAL_SERIES(arg)) {
arg_ser = Copy_Series_Part(arg_ser, VAL_INDEX(arg), ilen); // GC!?
}
// Get /DUP count:
if (DS_REF(AN_DUP)) {
cnt = Int32(DS_ARG(AN_COUNT));
if (cnt <= 0) return; // no changes
}
// Total to insert:
size = cnt * ilen;
if (action != A_CHANGE) {
// Always expand series for INSERT and APPEND actions:
Expand_Series(series, index, size);
} else {
if (size > rlen)
Expand_Series(series, index, size-rlen);
else if (size < rlen && DS_REF(AN_PART))
Remove_Series(series, index, rlen-size);
else if (size + index > tail) {
EXPAND_SERIES_TAIL(series, size - (tail - index));
}
}
// For dup count:
for (; cnt > 0; cnt--) {
Insert_String(series, index, arg_ser, VAL_INDEX(arg), ilen, TRUE);
index += ilen;
}
TERM_SERIES(series);
VAL_INDEX(string) = (action == A_APPEND) ? 0 : index;
}
*/ static REBINT Do_Set_Operation(struct Reb_Call *call_, REBCNT flags)
/*
** Do set operations on a series.
**
***********************************************************************/
{
REBVAL *val;
REBVAL *val1;
REBVAL *val2 = 0;
REBSER *ser;
REBSER *hser = 0; // hash table for series
REBSER *retser; // return series
REBSER *hret; // hash table for return series
REBCNT i;
REBINT h = TRUE;
REBCNT skip = 1; // record size
REBCNT cased = 0; // case sensitive when TRUE
SET_NONE(D_OUT);
val1 = D_ARG(1);
i = 2;
// Check for second series argument:
if (flags != SET_OP_UNIQUE) {
val2 = D_ARG(i++);
if (VAL_TYPE(val1) != VAL_TYPE(val2))
raise Error_Unexpected_Type(VAL_TYPE(val1), VAL_TYPE(val2));
}
// Refinements /case and /skip N
cased = D_REF(i++); // cased
if (D_REF(i++)) skip = Int32s(D_ARG(i), 1);
switch (VAL_TYPE(val1)) {
case REB_BLOCK:
i = VAL_LEN(val1);
// Setup result block:
if (GET_FLAG(flags, SOP_BOTH)) i += VAL_LEN(val2);
retser = BUF_EMIT; // use preallocated shared block
Resize_Series(retser, 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 {
// Check what is in series1 but not in series2:
if (GET_FLAG(flags, SOP_CHECK))
hser = Hash_Block(val2, cased);
// Iterate over first series:
ser = VAL_SERIES(val1);
i = VAL_INDEX(val1);
for (; val = BLK_SKIP(ser, i), i < SERIES_TAIL(ser); i += skip) {
if (GET_FLAG(flags, SOP_CHECK)) {
h = Find_Key(VAL_SERIES(val2), hser, val, skip, cased, 1) >= 0;
if (GET_FLAG(flags, SOP_INVERT)) h = !h;
}
if (h) Find_Key(retser, hret, val, skip, cased, 2);
}
// Iterate over second series?
if ((i = GET_FLAG(flags, SOP_BOTH))) {
val = val1;
val1 = val2;
val2 = val;
CLR_FLAG(flags, SOP_BOTH);
}
if (GET_FLAG(flags, SOP_CHECK))
Free_Series(hser);
} while (i);
if (hret)
Free_Series(hret);
Val_Init_Block(D_OUT, Copy_Array_Shallow(retser));
RESET_TAIL(retser); // required - allow reuse
break;
case REB_BINARY:
cased = TRUE;
SET_TYPE(D_OUT, REB_BINARY);
case REB_STRING:
i = VAL_LEN(val1);
// Setup result block:
if (GET_FLAG(flags, SOP_BOTH)) i += VAL_LEN(val2);
retser = BUF_MOLD;
Reset_Buffer(retser, i);
RESET_TAIL(retser);
do {
REBUNI uc;
cased = cased ? AM_FIND_CASE : 0;
// Iterate over first series:
ser = VAL_SERIES(val1);
i = VAL_INDEX(val1);
for (; i < SERIES_TAIL(ser); i += skip) {
uc = GET_ANY_CHAR(ser, i);
if (GET_FLAG(flags, SOP_CHECK)) {
h = Find_Str_Char(VAL_SERIES(val2), 0, VAL_INDEX(val2), VAL_TAIL(val2), skip, uc, cased) != NOT_FOUND;
if (GET_FLAG(flags, SOP_INVERT)) h = !h;
}
if (h && (Find_Str_Char(retser, 0, 0, SERIES_TAIL(retser), skip, uc, cased) == NOT_FOUND)) {
Append_String(retser, ser, i, skip);
}
}
// Iterate over second series?
if ((i = GET_FLAG(flags, SOP_BOTH))) {
val = val1;
val1 = val2;
val2 = val;
CLR_FLAG(flags, SOP_BOTH);
}
} while (i);
ser = Copy_String(retser, 0, -1);
if (IS_BINARY(D_OUT))
Val_Init_Binary(D_OUT, ser);
else
Val_Init_String(D_OUT, ser);
break;
case REB_BITSET:
switch (flags) {
case SET_OP_UNIQUE:
return R_ARG1;
case SET_OP_UNION:
i = A_OR;
break;
case SET_OP_INTERSECT:
i = A_AND;
break;
case SET_OP_DIFFERENCE:
i = A_XOR;
break;
case SET_OP_EXCLUDE:
i = 0; // special case
break;
}
ser = Xandor_Binary(i, val1, val2);
Val_Init_Bitset(D_OUT, ser);
break;
case REB_TYPESET:
switch (flags) {
case SET_OP_UNIQUE:
break;
case SET_OP_UNION:
VAL_TYPESET(val1) |= VAL_TYPESET(val2);
break;
case SET_OP_INTERSECT:
VAL_TYPESET(val1) &= VAL_TYPESET(val2);
break;
case SET_OP_DIFFERENCE:
VAL_TYPESET(val1) ^= VAL_TYPESET(val2);
break;
case SET_OP_EXCLUDE:
VAL_TYPESET(val1) &= ~VAL_TYPESET(val2);
break;
}
return R_ARG1;
default:
raise Error_Invalid_Arg(val1);
}
return R_OUT;
}
//
// MAKE_Tuple: C
//
REB_R MAKE_Tuple(
REBVAL *out,
enum Reb_Kind kind,
const REBVAL *opt_parent,
const REBVAL *arg
){
assert(kind == REB_TUPLE);
if (opt_parent)
fail (Error_Bad_Make_Parent(kind, opt_parent));
if (IS_TUPLE(arg))
return Move_Value(out, arg);
RESET_CELL(out, REB_TUPLE, CELL_MASK_NONE);
REBYTE *vp = VAL_TUPLE(out);
// !!! Net lookup parses IP addresses out of `tcp://93.184.216.34` or
// similar URL!s. In Rebol3 these captures come back the same type
// as the input instead of as STRING!, which was a latent bug in the
// network code of the 12-Dec-2012 release:
//
// https://github.com/rebol/rebol/blob/master/src/mezz/sys-ports.r#L110
//
// All attempts to convert a URL!-flavored IP address failed. Taking
// URL! here fixes it, though there are still open questions.
//
if (IS_TEXT(arg) or IS_URL(arg)) {
REBSIZ size;
const REBYTE *bp
= Analyze_String_For_Scan(&size, arg, MAX_SCAN_TUPLE);
if (Scan_Tuple(out, bp, size) == nullptr)
fail (arg);
return out;
}
if (ANY_ARRAY(arg)) {
REBCNT len = 0;
REBINT n;
RELVAL *item = VAL_ARRAY_AT(arg);
for (; NOT_END(item); ++item, ++vp, ++len) {
if (len >= MAX_TUPLE)
goto bad_make;
if (IS_INTEGER(item)) {
n = Int32(item);
}
else if (IS_CHAR(item)) {
n = VAL_CHAR(item);
}
else
goto bad_make;
if (n > 255 || n < 0)
goto bad_make;
*vp = n;
}
VAL_TUPLE_LEN(out) = len;
for (; len < MAX_TUPLE; len++) *vp++ = 0;
return out;
}
REBCNT alen;
if (IS_ISSUE(arg)) {
REBSTR *spelling = VAL_STRING(arg);
const REBYTE *ap = STR_HEAD(spelling);
size_t size = STR_SIZE(spelling); // UTF-8 len
if (size & 1)
fail (arg); // must have even # of chars
size /= 2;
if (size > MAX_TUPLE)
fail (arg); // valid even for UTF-8
VAL_TUPLE_LEN(out) = size;
for (alen = 0; alen < size; alen++) {
REBYTE decoded;
if ((ap = Scan_Hex2(&decoded, ap)) == NULL)
fail (arg);
*vp++ = decoded;
}
}
else if (IS_BINARY(arg)) {
REBYTE *ap = VAL_BIN_AT(arg);
REBCNT len = VAL_LEN_AT(arg);
if (len > MAX_TUPLE) len = MAX_TUPLE;
VAL_TUPLE_LEN(out) = len;
for (alen = 0; alen < len; alen++) *vp++ = *ap++;
}
else
fail (arg);
for (; alen < MAX_TUPLE; alen++) *vp++ = 0;
return out;
bad_make:
fail (Error_Bad_Make(REB_TUPLE, arg));
}
//
// 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;
}
