hi-
i'm looking to convert a pro*c application to something
postgresql-based. i've just started to experiment with ecpg, and have
been pleasantly surprised that large parts of out existing code will be
useable without too much alteration.
the problems that i've run into all involve variable declarations. we
have a large number of complex structures that map in-house
datatypes. with oracle, we could do something like this:
--------------------------------------------------------
typedef struct { int id; char name[30]; char dsc[1000];
} rec_t;
rec_t r;
EXEC SQL SELECT id, name, dsc INTO :r.id, :r.name, :r.dsc
FROM table WHERE id = :r.id;
--------------------------------------------------------
to make this code work with ecpg, i've made the following changes:
--------------------------------------------------------
EXEC SQL BEGIN DECLARE SECTION;
typedef struct { int id; char name[30]; char dsc[1000];
} rec_t;
rec_t r;
EXEC SQL END DECLARE SECTION;
--------------------------------------------------------
which yields the following error from ecpg:
postgres.pgc:15: ERROR: invalid datatype 'typedef'
is there are facility that i might be missing that would allow me to use
our existing structures as host variables w/ postgres embedded sql? or am
i better off just skipping ecpg and coding directly with libpq?
thanks in advance for any insight you can offer.
dan
