On Fri, 8 Jun 2018 at 13:47, Geoff Winkless <pgsqladmin@geoff.dj> wrote:
> Answering my own question, looks like
And just in case anyone googling the question comes across this, this
example code works.
#include "postgres.h"
#include <string.h>
#include "fmgr.h"
#include "utils/geo_decls.h"
#include "funcapi.h"
#include "utils/array.h"
#include "utils/builtins.h"
#include "utils/numeric.h"
#include "catalog/pg_type.h"
PG_MODULE_MAGIC;
PG_FUNCTION_INFO_V1(pgnumeric_x10);
Datum
pgnumeric_x10(PG_FUNCTION_ARGS) {
Numeric v;
char *r;
char mybuff[1000];
double f;
v=PG_GETARG_NUMERIC(0);
r=numeric_normalize(v);
f=atof(r)*10;
sprintf(mybuff, "%f", f);
v = DatumGetNumeric(DirectFunctionCall3(numeric_in,
CStringGetDatum(mybuff), 0, -1));
pfree(r);
PG_RETURN_NUMERIC(v);
}
Example of it running:
=# CREATE OR REPLACE FUNCTION pgnumeric_x10(NUMERIC) RETURNS NUMERIC
AS 'testpgnumchange.so', 'pgnumeric_x10' LANGUAGE C STRICT IMMUTABLE;
CREATE FUNCTION
Time: 0.811 ms
=# select pgnumeric_x10(132387.4823487::NUMERIC);
pgnumeric_x10
----------------
1323874.823487
(1 row)
Time: 0.593 ms
=#
For obvious reasons I wouldn't suggest using atof on a numeric, we
have our own functions for manipulating _Decimal128 which is what I'll
actually be using in the end version, but this is easier to compile as
an example :)
Geoff
