Once in a while, you might have to catch an error. Of course, this is also possible in Python. The following example shows how this works:

CREATE OR REPLACE FUNCTION trial_error()  
  RETURNS text  AS 
$$ 
try: 
  rv = plpy.execute("SELECT surely_a_syntax_error") 
except plpy.SPIError: 
  return "we caught the error" else: 
else: 
  return "all fine" 
$$ LANGUAGE 'plpythonu'; 

You can use a normal try/except block and access plpy to treat the error you want to catch. The function can then return normally without destroying your transaction, as follows:

test=# SELECT trial_error(); 
 trial_error 
---------------------  
 we caught the error 
(1 row) 

Remember, PL/Python has full access to the internals of PostgreSQL. Therefore, it can also expose all kinds of errors to your procedure. Here is an example:

except spiexceptions.DivisionByZero: 
  return "found a division by zero"  
except spiexceptions.UniqueViolation:  
  return "found a unique violation" 
except plpy.SPIError, e: 
  return "other error, SQLSTATE %s" % e.sqlstate 

Catching errors in Python is really easy and can help prevent your functions from failing.