Re: random() (was Re: New GUC to sample log queries) - Mailing list pgsql-hackers

I wrote:
> On further reflection, it seems likely that in most installations a lot
> of processes never invoke drandom()/setseed() at all, making such work
> in InitProcessGlobals a waste of cycles.  Probably a better idea is to
> have drandom() initialize the seed on first use, if it wasn't already
> set by setseed().

Here's a proposed patch for that.

It occurs to me that there's still another good reason to decouple
drandom from everything else: the point of setseed(), if it has any
at all, is to allow a repeatable sequence of drandom values to be
generated.  Without this patch, no such guarantee exists because of
the possibility of the backend making additional internal calls of
libc random().

            regards, tom lane

diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
index 3786099..1df4356 100644
*** a/doc/src/sgml/func.sgml
--- b/doc/src/sgml/func.sgml
***************
*** 1136,1146 ****
     </table>

    <para>
!    The characteristics of the values returned by
!    <literal><function>random()</function></literal> depend
!    on the system implementation. It is not suitable for cryptographic
!    applications; see <xref linkend="pgcrypto"/> module for an alternative.
!    </para>

    <para>
     Finally, <xref linkend="functions-math-trig-table"/> shows the
--- 1136,1150 ----
     </table>

    <para>
!    The <function>random()</function> function uses a simple linear
!    congruential algorithm.  It is fast but not suitable for cryptographic
!    applications; see the <xref linkend="pgcrypto"/> module for an
!    alternative.
!    If <function>setseed()</function> is called, the results of
!    subsequent <function>random()</function> calls in the current session are
!    repeatable by re-issuing <function>setseed()</function> with the same
!    argument.
!   </para>

    <para>
     Finally, <xref linkend="functions-math-trig-table"/> shows the
diff --git a/src/backend/utils/adt/float.c b/src/backend/utils/adt/float.c
index cf9327f..add099e 100644
*** a/src/backend/utils/adt/float.c
--- b/src/backend/utils/adt/float.c
***************
*** 22,31 ****
--- 22,34 ----
  #include "catalog/pg_type.h"
  #include "common/int.h"
  #include "libpq/pqformat.h"
+ #include "miscadmin.h"
  #include "utils/array.h"
+ #include "utils/backend_random.h"
  #include "utils/float.h"
  #include "utils/fmgrprotos.h"
  #include "utils/sortsupport.h"
+ #include "utils/timestamp.h"


  /* Configurable GUC parameter */
*************** float8        degree_c_sixty = 60.0;
*** 53,58 ****
--- 56,65 ----
  float8        degree_c_one_half = 0.5;
  float8        degree_c_one = 1.0;

+ /* State for drandom() and setseed() */
+ static bool drandom_seed_set = false;
+ static unsigned short drandom_seed[3] = {0, 0, 0};
+
  /* Local function prototypes */
  static double sind_q1(double x);
  static double cosd_q1(double x);
*************** drandom(PG_FUNCTION_ARGS)
*** 2378,2385 ****
  {
      float8        result;

!     /* result [0.0 - 1.0) */
!     result = (double) random() / ((double) MAX_RANDOM_VALUE + 1);

      PG_RETURN_FLOAT8(result);
  }
--- 2385,2414 ----
  {
      float8        result;

!     /* Initialize random seed, if not done yet in this process */
!     if (unlikely(!drandom_seed_set))
!     {
!         /*
!          * If possible, initialize the seed using high-quality random bits.
!          * Should that fail for some reason, we fall back on a lower-quality
!          * seed based on current time and PID.
!          */
!         if (!pg_backend_random((char *) drandom_seed, sizeof(drandom_seed)))
!         {
!             TimestampTz now = GetCurrentTimestamp();
!             uint64        iseed;
!
!             /* Mix the PID with the most predictable bits of the timestamp */
!             iseed = (uint64) now ^ ((uint64) MyProcPid << 32);
!             drandom_seed[0] = (unsigned short) iseed;
!             drandom_seed[1] = (unsigned short) (iseed >> 16);
!             drandom_seed[2] = (unsigned short) (iseed >> 32);
!         }
!         drandom_seed_set = true;
!     }
!
!     /* pg_erand48 produces desired result range [0.0 - 1.0) */
!     result = pg_erand48(drandom_seed);

      PG_RETURN_FLOAT8(result);
  }
*************** Datum
*** 2392,2404 ****
  setseed(PG_FUNCTION_ARGS)
  {
      float8        seed = PG_GETARG_FLOAT8(0);
!     int            iseed;

!     if (seed < -1 || seed > 1)
!         elog(ERROR, "setseed parameter %f out of range [-1,1]", seed);

!     iseed = (int) (seed * MAX_RANDOM_VALUE);
!     srandom((unsigned int) iseed);

      PG_RETURN_VOID();
  }
--- 2421,2440 ----
  setseed(PG_FUNCTION_ARGS)
  {
      float8        seed = PG_GETARG_FLOAT8(0);
!     uint64        iseed;

!     if (seed < -1 || seed > 1 || isnan(seed))
!         ereport(ERROR,
!                 (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
!                  errmsg("setseed parameter %g is out of allowed range [-1,1]",
!                         seed)));

!     /* Use sign bit + 47 fractional bits to fill drandom_seed[] */
!     iseed = (int64) (seed * (float8) UINT64CONST(0x7FFFFFFFFFFF));
!     drandom_seed[0] = (unsigned short) iseed;
!     drandom_seed[1] = (unsigned short) (iseed >> 16);
!     drandom_seed[2] = (unsigned short) (iseed >> 32);
!     drandom_seed_set = true;

      PG_RETURN_VOID();
  }