Thread: Type implemented in plpythonu crashes backend

I'd like to implement new postgres types in a language other than C for
prototyping speed.  plpgsql won't let me create a function which takes a
cstring, and neither will pltcl.  Happily, plpythonu will, and it seems
to work fine.

However, the code below causes the backend to crash...

I'm guessing that plpython is trying to format the return value using
the appropriate registered function, which just happens to be itself.


My goal is to prototype new types, operators and gist indexes without
getting bogged down in with the guts of posgresql in C, so if it's
unreasonable to expect the below code to work, do you have any other
suggestions?


Thanks.




create or replace function pycomplex_in(cstring)
returns opaque as '

import struct
import re

p = re.compile(" *\\( *([0-9]+(\\.[0-9]+)?) *, *([0-9]+(\\.[0-9]+)?)*\\)*")
m = p.match(args[0])

if m:   (x, y) = m.group(1, 3)
else:   (x, y) = (0, 0)   #plpy.error("incorrect format for pycomplex: %s" % args[0])

return struct.pack("dd", float(x), float(y))

' language plpythonu immutable strict;



create or replace function pycomplex_out(opaque)
returns cstring as '

import struct

(x, y) = struct.unpack("dd", args[0])
return "(%f, %f)", (x, y)

' language plpythonu immutable strict;


create type pycomplex (  internallength = 16,  input          = pycomplex_in,  output         = pycomplex_out,
alignment     = double
 
);


select pycomplex_out(pycomplex_in('(1,2)'));

pgsql@groks.org writes:
> I'd like to implement new postgres types in a language other than C for
> prototyping speed.  plpgsql won't let me create a function which takes a
> cstring, and neither will pltcl.  Happily, plpythonu will, and it seems
> to work fine.

I'd say this is an oversight in the plpython code --- it should be
rejecting pseudotype arguments, since as you've discovered it cannot
really cope.  It should *certainly* reject "opaque" ... might be okay to
allow cstring but I'm not sure.
        regards, tom lane

Added to open items list:
* have plpython reject pseudotype arguments because it crashes

---------------------------------------------------------------------------

Tom Lane wrote:
> pgsql@groks.org writes:
> > I'd like to implement new postgres types in a language other than C for
> > prototyping speed.  plpgsql won't let me create a function which takes a
> > cstring, and neither will pltcl.  Happily, plpythonu will, and it seems
> > to work fine.
> 
> I'd say this is an oversight in the plpython code --- it should be
> rejecting pseudotype arguments, since as you've discovered it cannot
> really cope.  It should *certainly* reject "opaque" ... might be okay to
> allow cstring but I'm not sure.
> 
>             regards, tom lane
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 8: explain analyze is your friend
> 

--  Bruce Momjian                        |  http://candle.pha.pa.us pgman@candle.pha.pa.us               |  (610)
359-1001+  If your life is a hard drive,     |  13 Roberts Road +  Christ can be your backup.        |  Newtown Square,
Pennsylvania19073