SPI_execute_with_args call

On 03/05/13 21:19, Yuriy Rusinov wrote:

Hello, colleagues !

I have to write random number generator state into database table
Table structure is
table rand_state
{
id serial not null primary key,
state_rand bytea
};

In C-function I do
size_t nr_ins = strlen ("insert into rand_state (state_rand) values
($1);");
char * r_sql = (char *) palloc (nr_ins + 1);
strncpy (r_sql, "insert into rand_state (state_rand) values ($1);",
nr_ins);

Oid * oids = (Oid *)palloc (sizeof (Oid));
Datum * val = PointerGetDatum (randBuf);
*oids = BYTEAOID;
const char * nulls = "NULL";
int rins = SPI_execute_with_args (r_sql, 1, oids, val, nulls, false, 1);

randBuf is a void * pointer that contains random number generator state,
when I try to execute SPI_execute_with_args (r_sql, 1, oids, val,
nulls, false, 1); I receive error
The connection to the server was lost. Attempting reset: Failed.
Could you give some work examples for SPI_execute_with_args because I
didn't find them in documentation.

Thanks a lot.

--
Best regards,
Sincerely yours,
Yuriy Rusinov.

I can't answer your question.

However, I can say that PRIMARY KEY implies NOT NULL (also an UNIQUE
index), so you don't need to explicitly add NOT NULL when you are
specifying PRIMARY KEY!

Cheers,
Gavin

I'm sorry !

But if I commented SPI_execute_with_args call, then all others works
without bugs.

On Fri, May 3, 2013 at 2:31 PM, Gavin Flower
<GavinFlower@archidevsys.co.nz>wrote:

On 03/05/13 21:19, Yuriy Rusinov wrote:

Hello, colleagues !

I have to write random number generator state into database table
Table structure is
table rand_state
{
id serial not null primary key,
state_rand bytea
};

In C-function I do
size_t nr_ins = strlen ("insert into rand_state (state_rand) values
($1);");
char * r_sql = (char *) palloc (nr_ins + 1);
strncpy (r_sql, "insert into rand_state (state_rand) values ($1);",
nr_ins);

Oid * oids = (Oid *)palloc (sizeof (Oid));
Datum * val = PointerGetDatum (randBuf);
*oids = BYTEAOID;
const char * nulls = "NULL";
int rins = SPI_execute_with_args (r_sql, 1, oids, val, nulls, false, 1);

randBuf is a void * pointer that contains random number generator state,
when I try to execute SPI_execute_with_args (r_sql, 1, oids, val, nulls,
false, 1); I receive error
The connection to the server was lost. Attempting reset: Failed.
Could you give some work examples for SPI_execute_with_args because I
didn't find them in documentation.

Thanks a lot.

--
Best regards,
Sincerely yours,
Yuriy Rusinov.

I can't answer your question.

However, I can say that PRIMARY KEY implies NOT NULL (also an UNIQUE
index), so you don't need to explicitly add NOT NULL when you are specifying
PRIMARY KEY!

Cheers,
Gavin

--
Best regards,
Sincerely yours,
Yuriy Rusinov.

Yuriy Rusinov <yrusinov@gmail.com> writes:

In C-function I do
size_t nr_ins = strlen ("insert into rand_state (state_rand) values ($1);");
char * r_sql = (char *) palloc (nr_ins + 1);
strncpy (r_sql, "insert into rand_state (state_rand) values ($1);", nr_ins);

This is a hard, error-prone, and ultimately incorrect way to do
pstrdup() --- you're not ensuring that the new string is
null-terminated.

Datum * val = PointerGetDatum (randBuf);

Didn't your compiler give you a warning about that? PointerGetDatum
produces a Datum, not a pointer to a Datum. You'd need something more
like

Datum val[1];
val[0] = PointerGetDatum (randBuf);

This is assuming that randBuf is even of the right format to be a bytea
value, which is unclear from your extract.

const char * nulls = "NULL";

And that's just wrong. Personally I'd just pass NULL to
SPI_execute_with_args since you don't have any null values to pass, but
if you don't want to do that you'd need something more like

char nulls[1];
nulls[0] = ' ';

(hmm, it looks like the SPI documentation leaves something to be desired
here --- the SPI_execute_with_args page, at least, isn't explaining the
convention for elements of the nulls[] array)

Could you give some work examples for SPI_execute_with_args because I
didn't find them in documentation.

A quick grep says there's a usage in src/pl/plpgsql/src/pl_exec.c

regards, tom lane

--
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

Thanks a lot, I have corrected and bug was fixed.

On Fri, May 3, 2013 at 6:42 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Yuriy Rusinov <yrusinov@gmail.com> writes:

In C-function I do
size_t nr_ins = strlen ("insert into rand_state (state_rand) values

($1);");

char * r_sql = (char *) palloc (nr_ins + 1);
strncpy (r_sql, "insert into rand_state (state_rand) values ($1);",

nr_ins);

This is a hard, error-prone, and ultimately incorrect way to do
pstrdup() --- you're not ensuring that the new string is
null-terminated.

Datum * val = PointerGetDatum (randBuf);

Didn't your compiler give you a warning about that? PointerGetDatum
produces a Datum, not a pointer to a Datum. You'd need something more
like

Datum val[1];
val[0] = PointerGetDatum (randBuf);

This is assuming that randBuf is even of the right format to be a bytea
value, which is unclear from your extract.

const char * nulls = "NULL";

And that's just wrong. Personally I'd just pass NULL to
SPI_execute_with_args since you don't have any null values to pass, but
if you don't want to do that you'd need something more like

char nulls[1];
nulls[0] = ' ';

(hmm, it looks like the SPI documentation leaves something to be desired
here --- the SPI_execute_with_args page, at least, isn't explaining the
convention for elements of the nulls[] array)

Could you give some work examples for SPI_execute_with_args because I
didn't find them in documentation.

A quick grep says there's a usage in src/pl/plpgsql/src/pl_exec.c

regards, tom lane

--
Best regards,
Sincerely yours,
Yuriy Rusinov.

Hello !

There is one's more question, which way I have to determine that void *
randBuf acceptable or not for bytea type ?

Thanks in advance.

On Fri, May 3, 2013 at 9:44 PM, Yuriy Rusinov <yrusinov@gmail.com> wrote:

Thanks a lot, I have corrected and bug was fixed.

On Fri, May 3, 2013 at 6:42 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Yuriy Rusinov <yrusinov@gmail.com> writes:

In C-function I do
size_t nr_ins = strlen ("insert into rand_state (state_rand) values

($1);");

char * r_sql = (char *) palloc (nr_ins + 1);
strncpy (r_sql, "insert into rand_state (state_rand) values ($1);",

nr_ins);

This is a hard, error-prone, and ultimately incorrect way to do
pstrdup() --- you're not ensuring that the new string is
null-terminated.

Datum * val = PointerGetDatum (randBuf);

Didn't your compiler give you a warning about that? PointerGetDatum
produces a Datum, not a pointer to a Datum. You'd need something more
like

Datum val[1];
val[0] = PointerGetDatum (randBuf);

This is assuming that randBuf is even of the right format to be a bytea
value, which is unclear from your extract.

const char * nulls = "NULL";

And that's just wrong. Personally I'd just pass NULL to
SPI_execute_with_args since you don't have any null values to pass, but
if you don't want to do that you'd need something more like

char nulls[1];
nulls[0] = ' ';

(hmm, it looks like the SPI documentation leaves something to be desired
here --- the SPI_execute_with_args page, at least, isn't explaining the
convention for elements of the nulls[] array)

Could you give some work examples for SPI_execute_with_args because I
didn't find them in documentation.

A quick grep says there's a usage in src/pl/plpgsql/src/pl_exec.c

regards, tom lane

--
Best regards,
Sincerely yours,
Yuriy Rusinov.

--
Best regards,
Sincerely yours,
Yuriy Rusinov.