[Tarantool-patches] [PATCH v5 45/52] sql: introduce mem_get_int()
Mergen Imeev
imeevma at tarantool.org
Wed Apr 14 03:28:32 MSK 2021
Thank you for the review! My answers, diff and new patch below.
On Wed, Apr 14, 2021 at 01:01:02AM +0200, Vladislav Shpilevoy wrote:
> Nice fixes!
>
> On 09.04.2021 22:53, Mergen Imeev via Tarantool-patches wrote:
> > Thank you for the review! My answers and new patch below.
> >
> >
> > On 30.03.2021 02:08, Vladislav Shpilevoy wrote:
> >> Thanks for the patch!
> >>
> >>> diff --git a/src/box/sql/func.c b/src/box/sql/func.c
> >>> index b644c39d8..0fa0f6ac7 100644
> >>> --- a/src/box/sql/func.c
> >>> +++ b/src/box/sql/func.c
> >>> @@ -1532,10 +1543,11 @@ hexFunc(sql_context * context, int argc, sql_value ** argv)
> >>> static void
> >>> zeroblobFunc(sql_context * context, int argc, sql_value ** argv)
> >>> {
> >>> - i64 n;
> >>> + int64_t n;
> >>> assert(argc == 1);
> >>> UNUSED_PARAMETER(argc);
> >>> - n = sql_value_int64(argv[0]);
> >>> + bool unused;
> >>> + mem_get_integer(argv[0], &n, &unused);
> >>
> >> The flag is never used anywhere except one assertion where you can
> >> check the integer value instead. I think you can drop this out
> >> parameter. In future we could add mem_get_int_with_sign() or something
> >> like that if necessary.
> > I think the problem here mostly because most of built-in functions and bitwise
> > operations cannot work with our INTEGER. They can only work with int64. I
> > believe, if we fix this problem, there will be no problems with having this
> > flag.
>
> My complaint is about the flag. The third argument which is almost never
> used. It makes the code ugly, and does not give a clue it is broken in fact.
> When uint64_t is > INT64_MAX and is returned as int64_t and the flag is
> ignored.
>
> What about mem_get_int_unsafe()? It would return int64_t truncated like
> before. Return as 'return', not out parameter. Because we also never check
> for fail as I see. And no 'unused' flag. But we would clearly see that these
> places are broken and need attention.
I agree that function with such name would be a good indicator that something
may go wrong here. I created new function, mem_get_int_unfase().
Diff:
diff --git a/src/box/sql/func.c b/src/box/sql/func.c
index 701e77d49..0db698174 100644
--- a/src/box/sql/func.c
+++ b/src/box/sql/func.c
@@ -205,9 +205,7 @@ absFunc(sql_context * context, int argc, sql_value ** argv)
break;
}
case MP_INT: {
- bool unused;
- int64_t value;
- mem_get_int(argv[0], &value, &unused);
+ int64_t value = mem_get_int_unsafe(argv[0]);
assert(value < 0);
sql_result_uint(context, -value);
break;
@@ -435,8 +433,7 @@ substrFunc(sql_context * context, int argc, sql_value ** argv)
if (mem_is_null(argv[1]) || (argc == 3 && mem_is_null(argv[2])))
return;
p0type = sql_value_type(argv[0]);
- bool unused;
- mem_get_int(argv[1], &p1, &unused);
+ p1 = mem_get_int_unsafe(argv[1]);
if (p0type == MP_BIN) {
len = sql_value_bytes(argv[0]);
z = sql_value_blob(argv[0]);
@@ -452,7 +449,7 @@ substrFunc(sql_context * context, int argc, sql_value ** argv)
len = sql_utf8_char_count(z, sql_value_bytes(argv[0]));
}
if (argc == 3) {
- mem_get_int(argv[2], &p2, &unused);
+ p2 = mem_get_int_unsafe(argv[2]);
if (p2 < 0) {
p2 = -p2;
negP2 = 1;
@@ -534,8 +531,7 @@ roundFunc(sql_context * context, int argc, sql_value ** argv)
if (argc == 2) {
if (mem_is_null(argv[1]))
return;
- bool unused;
- mem_get_int(argv[1], &n, &unused);
+ n = mem_get_int_unsafe(argv[1]);
if (n < 0)
n = 0;
}
@@ -688,8 +684,7 @@ randomBlob(sql_context * context, int argc, sql_value ** argv)
context->is_aborted = true;
return;
}
- bool unused;
- mem_get_int(argv[0], &n, &unused);
+ n = mem_get_int_unsafe(argv[0]);
if (n < 1)
return;
p = contextMalloc(context, n);
@@ -1233,8 +1228,7 @@ zeroblobFunc(sql_context * context, int argc, sql_value ** argv)
int64_t n;
assert(argc == 1);
UNUSED_PARAMETER(argc);
- bool unused;
- mem_get_int(argv[0], &n, &unused);
+ n = mem_get_int_unsafe(argv[0]);
if (n < 0)
n = 0;
if (sql_result_zeroblob64(context, n) != 0) {
@@ -1478,9 +1472,9 @@ trim_func_two_args(struct sql_context *context, sql_value *arg1,
int input_str_sz = sql_value_bytes(arg2);
if (sql_value_type(arg1) == MP_INT || sql_value_type(arg1) == MP_UINT) {
uint8_t len_one = 1;
- uint64_t n = sql_value_uint64(arg1);
- trim_procedure(context, n, (const unsigned char *) " ",
- &len_one, 1, input_str, input_str_sz);
+ trim_procedure(context, mem_get_int_unsafe(arg1),
+ (const unsigned char *) " ", &len_one, 1,
+ input_str, input_str_sz);
} else if ((trim_set = sql_value_text(arg1)) != NULL) {
int trim_set_sz = sql_value_bytes(arg1);
uint8_t *char_len;
@@ -1518,8 +1512,7 @@ trim_func_three_args(struct sql_context *context, sql_value *arg1,
&char_len);
if (char_cnt == -1)
return;
- uint64_t n = sql_value_uint64(arg1);
- trim_procedure(context, n, trim_set, char_len,
+ trim_procedure(context, mem_get_int_unsafe(arg1), trim_set, char_len,
char_cnt, input_str, input_str_sz);
sql_free(char_len);
}
@@ -1658,9 +1651,7 @@ sum_step(struct sql_context *context, int argc, sql_value **argv)
}
p->cnt++;
if (type == MP_INT || type == MP_UINT) {
- bool unused;
- int64_t v;
- mem_get_int(argv[0], &v, &unused);
+ int64_t v = mem_get_int_unsafe(argv[0]);
if (type == MP_INT)
p->rSum += v;
else
diff --git a/src/box/sql/mem.h b/src/box/sql/mem.h
index 313ca0ab2..f3d9043e5 100644
--- a/src/box/sql/mem.h
+++ b/src/box/sql/mem.h
@@ -754,6 +754,20 @@ mem_cast_implicit_old(struct Mem *mem, enum field_type type);
int
mem_get_int(const struct Mem *mem, int64_t *i, bool *is_neg);
+/**
+ * Return value of MEM converted to int64_t. This function is not safe, since it
+ * works incorrectly with integer values that are more than INT64_MAX. Also, its
+ * behaviour is undefined if mem_get_int() returned an error.
+ */
+static inline int64_t
+mem_get_int_unsafe(const struct Mem *mem)
+{
+ int64_t i;
+ bool is_neg;
+ mem_get_int(mem, &i, &is_neg);
+ return i;
+}
+
/**
* Simple type to str convertor. It is used to simplify
* error reporting.
diff --git a/src/box/sql/printf.c b/src/box/sql/printf.c
index 09da39e81..eb8413f9c 100644
--- a/src/box/sql/printf.c
+++ b/src/box/sql/printf.c
@@ -144,10 +144,7 @@ getIntArg(PrintfArguments * p)
{
if (p->nArg <= p->nUsed)
return 0;
- int64_t i;
- bool unused;
- mem_get_int(p->apArg[p->nUsed++], &i, &unused);
- return (sql_int64)i;
+ return mem_get_int_unsafe(p->apArg[p->nUsed++]);
}
static double
New patch:
commit 9472eeb564bd0caa039fe49c8fef4c1a7775ce80
Author: Mergen Imeev <imeevma at gmail.com>
Date: Wed Mar 17 13:20:37 2021 +0300
sql: introduce mem_get_int()
This patch introduces mem_get_int() function. This function is used to
receive integer value from MEM. If value of MEM is not integer, it is
converted to integer if possible. MEM is not changed.
Part of #5818
diff --git a/src/box/sql/func.c b/src/box/sql/func.c
index 0282aec74..0db698174 100644
--- a/src/box/sql/func.c
+++ b/src/box/sql/func.c
@@ -205,7 +205,7 @@ absFunc(sql_context * context, int argc, sql_value ** argv)
break;
}
case MP_INT: {
- int64_t value = sql_value_int64(argv[0]);
+ int64_t value = mem_get_int_unsafe(argv[0]);
assert(value < 0);
sql_result_uint(context, -value);
break;
@@ -421,7 +421,7 @@ substrFunc(sql_context * context, int argc, sql_value ** argv)
const unsigned char *z2;
int len;
int p0type;
- i64 p1, p2;
+ int64_t p1, p2;
int negP2 = 0;
if (argc != 2 && argc != 3) {
@@ -433,7 +433,7 @@ substrFunc(sql_context * context, int argc, sql_value ** argv)
if (mem_is_null(argv[1]) || (argc == 3 && mem_is_null(argv[2])))
return;
p0type = sql_value_type(argv[0]);
- p1 = sql_value_int(argv[1]);
+ p1 = mem_get_int_unsafe(argv[1]);
if (p0type == MP_BIN) {
len = sql_value_bytes(argv[0]);
z = sql_value_blob(argv[0]);
@@ -449,7 +449,7 @@ substrFunc(sql_context * context, int argc, sql_value ** argv)
len = sql_utf8_char_count(z, sql_value_bytes(argv[0]));
}
if (argc == 3) {
- p2 = sql_value_int(argv[2]);
+ p2 = mem_get_int_unsafe(argv[2]);
if (p2 < 0) {
p2 = -p2;
negP2 = 1;
@@ -520,7 +520,7 @@ substrFunc(sql_context * context, int argc, sql_value ** argv)
static void
roundFunc(sql_context * context, int argc, sql_value ** argv)
{
- int n = 0;
+ int64_t n = 0;
double r;
if (argc != 1 && argc != 2) {
diag_set(ClientError, ER_FUNC_WRONG_ARG_COUNT, "ROUND",
@@ -531,7 +531,7 @@ roundFunc(sql_context * context, int argc, sql_value ** argv)
if (argc == 2) {
if (mem_is_null(argv[1]))
return;
- n = sql_value_int(argv[1]);
+ n = mem_get_int_unsafe(argv[1]);
if (n < 0)
n = 0;
}
@@ -674,7 +674,7 @@ randomFunc(sql_context * context, int NotUsed, sql_value ** NotUsed2)
static void
randomBlob(sql_context * context, int argc, sql_value ** argv)
{
- int n;
+ int64_t n;
unsigned char *p;
assert(argc == 1);
UNUSED_PARAMETER(argc);
@@ -684,7 +684,7 @@ randomBlob(sql_context * context, int argc, sql_value ** argv)
context->is_aborted = true;
return;
}
- n = sql_value_int(argv[0]);
+ n = mem_get_int_unsafe(argv[0]);
if (n < 1)
return;
p = contextMalloc(context, n);
@@ -1225,10 +1225,10 @@ hexFunc(sql_context * context, int argc, sql_value ** argv)
static void
zeroblobFunc(sql_context * context, int argc, sql_value ** argv)
{
- i64 n;
+ int64_t n;
assert(argc == 1);
UNUSED_PARAMETER(argc);
- n = sql_value_int64(argv[0]);
+ n = mem_get_int_unsafe(argv[0]);
if (n < 0)
n = 0;
if (sql_result_zeroblob64(context, n) != 0) {
@@ -1472,7 +1472,7 @@ trim_func_two_args(struct sql_context *context, sql_value *arg1,
int input_str_sz = sql_value_bytes(arg2);
if (sql_value_type(arg1) == MP_INT || sql_value_type(arg1) == MP_UINT) {
uint8_t len_one = 1;
- trim_procedure(context, sql_value_int(arg1),
+ trim_procedure(context, mem_get_int_unsafe(arg1),
(const unsigned char *) " ", &len_one, 1,
input_str, input_str_sz);
} else if ((trim_set = sql_value_text(arg1)) != NULL) {
@@ -1512,7 +1512,7 @@ trim_func_three_args(struct sql_context *context, sql_value *arg1,
&char_len);
if (char_cnt == -1)
return;
- trim_procedure(context, sql_value_int(arg1), trim_set, char_len,
+ trim_procedure(context, mem_get_int_unsafe(arg1), trim_set, char_len,
char_cnt, input_str, input_str_sz);
sql_free(char_len);
}
@@ -1651,7 +1651,7 @@ sum_step(struct sql_context *context, int argc, sql_value **argv)
}
p->cnt++;
if (type == MP_INT || type == MP_UINT) {
- int64_t v = sql_value_int64(argv[0]);
+ int64_t v = mem_get_int_unsafe(argv[0]);
if (type == MP_INT)
p->rSum += v;
else
diff --git a/src/box/sql/mem.c b/src/box/sql/mem.c
index 537288c14..adf5e236b 100644
--- a/src/box/sql/mem.c
+++ b/src/box/sql/mem.c
@@ -1042,6 +1042,38 @@ mem_cast_implicit_old(struct Mem *mem, enum field_type type)
return -1;
}
+int
+mem_get_int(const struct Mem *mem, int64_t *i, bool *is_neg)
+{
+ if ((mem->flags & MEM_Int) != 0) {
+ *i = mem->u.i;
+ *is_neg = true;
+ return 0;
+ }
+ if ((mem->flags & MEM_UInt) != 0) {
+ *i = mem->u.i;
+ *is_neg = false;
+ return 0;
+ }
+ if ((mem->flags & (MEM_Str | MEM_Blob)) != 0)
+ return sql_atoi64(mem->z, i, is_neg, mem->n);
+ if ((mem->flags & MEM_Real) != 0) {
+ double d = mem->u.r;
+ if (d < 0 && d >= (double)INT64_MIN) {
+ *i = (int64_t)d;
+ *is_neg = true;
+ return 0;
+ }
+ if (d >= 0 && d < (double)UINT64_MAX) {
+ *i = (int64_t)(uint64_t)d;
+ *is_neg = false;
+ return 0;
+ }
+ return -1;
+ }
+ return -1;
+}
+
int
mem_copy(struct Mem *to, const struct Mem *from)
{
@@ -1400,12 +1432,12 @@ bitwise_prepare(const struct Mem *left, const struct Mem *right,
int64_t *a, int64_t *b)
{
bool unused;
- if (sqlVdbeIntValue(left, a, &unused) != 0) {
+ if (mem_get_int(left, a, &unused) != 0) {
diag_set(ClientError, ER_SQL_TYPE_MISMATCH, mem_str(left),
"integer");
return -1;
}
- if (sqlVdbeIntValue(right, b, &unused) != 0) {
+ if (mem_get_int(right, b, &unused) != 0) {
diag_set(ClientError, ER_SQL_TYPE_MISMATCH, mem_str(right),
"integer");
return -1;
@@ -1494,7 +1526,7 @@ mem_bit_not(const struct Mem *mem, struct Mem *result)
return 0;
int64_t i;
bool unused;
- if (sqlVdbeIntValue(mem, &i, &unused) != 0) {
+ if (mem_get_int(mem, &i, &unused) != 0) {
diag_set(ClientError, ER_SQL_TYPE_MISMATCH, mem_str(mem),
"integer");
return -1;
@@ -1708,35 +1740,6 @@ valueToText(sql_value * pVal)
return pVal->z;
}
-/*
- * Convert a 64-bit IEEE double into a 64-bit signed integer.
- * If the double is out of range of a 64-bit signed integer then
- * return the closest available 64-bit signed integer.
- */
-static int
-doubleToInt64(double r, int64_t *i)
-{
- /*
- * Many compilers we encounter do not define constants for the
- * minimum and maximum 64-bit integers, or they define them
- * inconsistently. And many do not understand the "LL" notation.
- * So we define our own static constants here using nothing
- * larger than a 32-bit integer constant.
- */
- static const int64_t maxInt = LARGEST_INT64;
- static const int64_t minInt = SMALLEST_INT64;
- if (r <= (double)minInt) {
- *i = minInt;
- return -1;
- } else if (r >= (double)maxInt) {
- *i = maxInt;
- return -1;
- } else {
- *i = (int64_t) r;
- return *i != r;
- }
-}
-
/*
* It is already known that pMem contains an unterminated string.
* Add the zero terminator.
@@ -2210,42 +2213,6 @@ mem_value_bool(const struct Mem *mem, bool *b)
return -1;
}
-/*
- * Return some kind of integer value which is the best we can do
- * at representing the value that *pMem describes as an integer.
- * If pMem is an integer, then the value is exact. If pMem is
- * a floating-point then the value returned is the integer part.
- * If pMem is a string or blob, then we make an attempt to convert
- * it into an integer and return that. If pMem represents an
- * an SQL-NULL value, return 0.
- *
- * If pMem represents a string value, its encoding might be changed.
- */
-int
-sqlVdbeIntValue(const struct Mem *pMem, int64_t *i, bool *is_neg)
-{
- int flags;
- assert(EIGHT_BYTE_ALIGNMENT(pMem));
- flags = pMem->flags;
- if (flags & MEM_Int) {
- *i = pMem->u.i;
- *is_neg = true;
- return 0;
- } else if (flags & MEM_UInt) {
- *i = pMem->u.u;
- *is_neg = false;
- return 0;
- } else if (flags & MEM_Real) {
- *is_neg = pMem->u.r < 0;
- return doubleToInt64(pMem->u.r, i);
- } else if (flags & (MEM_Str)) {
- assert(pMem->z || pMem->n == 0);
- if (sql_atoi64(pMem->z, i, is_neg, pMem->n) == 0)
- return 0;
- }
- return -1;
-}
-
/*
* Return the best representation of pMem that we can get into a
* double. If pMem is already a double or an integer, return its
@@ -2316,30 +2283,12 @@ sql_value_boolean(sql_value *val)
return b;
}
-int
-sql_value_int(sql_value * pVal)
-{
- int64_t i = 0;
- bool is_neg;
- sqlVdbeIntValue((Mem *) pVal, &i, &is_neg);
- return (int)i;
-}
-
-sql_int64
-sql_value_int64(sql_value * pVal)
-{
- int64_t i = 0;
- bool unused;
- sqlVdbeIntValue((Mem *) pVal, &i, &unused);
- return i;
-}
-
uint64_t
sql_value_uint64(sql_value *val)
{
int64_t i = 0;
bool is_neg;
- sqlVdbeIntValue((struct Mem *) val, &i, &is_neg);
+ mem_get_int((struct Mem *) val, &i, &is_neg);
assert(!is_neg);
return i;
}
diff --git a/src/box/sql/mem.h b/src/box/sql/mem.h
index 91c1c464f..f3d9043e5 100644
--- a/src/box/sql/mem.h
+++ b/src/box/sql/mem.h
@@ -746,6 +746,28 @@ mem_cast_implicit(struct Mem *mem, enum field_type type);
int
mem_cast_implicit_old(struct Mem *mem, enum field_type type);
+/**
+ * Return value for MEM of INTEGER type. For MEM of all other types convert
+ * value of the MEM to INTEGER if possible and return converted value. Original
+ * MEM is not changed.
+ */
+int
+mem_get_int(const struct Mem *mem, int64_t *i, bool *is_neg);
+
+/**
+ * Return value of MEM converted to int64_t. This function is not safe, since it
+ * works incorrectly with integer values that are more than INT64_MAX. Also, its
+ * behaviour is undefined if mem_get_int() returned an error.
+ */
+static inline int64_t
+mem_get_int_unsafe(const struct Mem *mem)
+{
+ int64_t i;
+ bool is_neg;
+ mem_get_int(mem, &i, &is_neg);
+ return i;
+}
+
/**
* Simple type to str convertor. It is used to simplify
* error reporting.
@@ -800,7 +822,6 @@ releaseMemArray(Mem * p, int N);
int
mem_value_bool(const struct Mem *mem, bool *b);
-int sqlVdbeIntValue(const struct Mem *, int64_t *, bool *is_neg);
int sqlVdbeRealValue(struct Mem *, double *);
const void *
sql_value_blob(struct Mem *);
@@ -814,12 +835,6 @@ sql_value_double(struct Mem *);
bool
sql_value_boolean(struct Mem *val);
-int
-sql_value_int(struct Mem *);
-
-sql_int64
-sql_value_int64(struct Mem *);
-
uint64_t
sql_value_uint64(struct Mem *val);
diff --git a/src/box/sql/printf.c b/src/box/sql/printf.c
index cf32ba3f3..eb8413f9c 100644
--- a/src/box/sql/printf.c
+++ b/src/box/sql/printf.c
@@ -144,7 +144,7 @@ getIntArg(PrintfArguments * p)
{
if (p->nArg <= p->nUsed)
return 0;
- return sql_value_int64(p->apArg[p->nUsed++]);
+ return mem_get_int_unsafe(p->apArg[p->nUsed++]);
}
static double
diff --git a/src/box/sql/sqlInt.h b/src/box/sql/sqlInt.h
index 7a026d21b..0af247ebf 100644
--- a/src/box/sql/sqlInt.h
+++ b/src/box/sql/sqlInt.h
@@ -445,15 +445,9 @@ sql_column_bytes16(sql_stmt *, int iCol);
double
sql_column_double(sql_stmt *, int iCol);
-int
-sql_column_int(sql_stmt *, int iCol);
-
bool
sql_column_boolean(struct sql_stmt *stmt, int column);
-sql_int64
-sql_column_int64(sql_stmt *, int iCol);
-
uint64_t
sql_column_uint64(struct sql_stmt *stmt, int column);
diff --git a/src/box/sql/vdbeapi.c b/src/box/sql/vdbeapi.c
index af1174d0a..5e5957496 100644
--- a/src/box/sql/vdbeapi.c
+++ b/src/box/sql/vdbeapi.c
@@ -470,24 +470,12 @@ sql_column_double(sql_stmt * pStmt, int i)
return sql_value_double(columnMem(pStmt, i));
}
-int
-sql_column_int(sql_stmt * pStmt, int i)
-{
- return sql_value_int(columnMem(pStmt, i));
-}
-
bool
sql_column_boolean(struct sql_stmt *stmt, int i)
{
return sql_value_boolean(columnMem(stmt, i));
}
-sql_int64
-sql_column_int64(sql_stmt * pStmt, int i)
-{
- return sql_value_int64(columnMem(pStmt, i));
-}
-
uint64_t
sql_column_uint64(sql_stmt * pStmt, int i)
{
More information about the Tarantool-patches
mailing list