""" Using a tuple as a bound variable in "SELECT ... IN (...)" clauses in PostgreSQL using psycopg 2 Some time ago someone asked on the psycopg mailing list how to have a bound variable expand to the right SQL for an SELECT IN clause: SELECT * FROM atable WHERE afield IN (value1, value2, value3) with the values to be used in the IN clause to be passed to the cursor .execute() method in a tuple as a bound variable, i.e.: in_values = ("value1", "value2", "value3") curs.execute("SELECT ... IN %s", (in_values,)) psycopg 1 does support typecasting from Python to PostgreSQL (and back) only for simple types and this problem has no elegant solution (short or writing a wrapper class returning the pre-quoted text in an __str__ method. But psycopg 2 offers a simple and elegant solution by partially implementing the Object Adaptation from PEP 246. psycopg 2 (still in beta and currently labeled as 1.99.9) moves the type-casting logic into external adapters and a somehow broken adapt() function. While the original adapt() takes 3 arguments, psycopg's one only takes 1: the bound variable to be adapted. The result is an object supporting a not-yet well defined protocol that we can call IPsycopgSQLQuote: class IPsycopgSQLQuote: def getquoted(self): "Returns a quoted string representing the bound variable." def getbinary(self): "Returns a binary quoted string representing the bound variable." def getbuffer(self): "Returns the wrapped object itself." __str__ = getquoted Then one of the functions (usually .getquoted()) is called by psycopg at the right time to obtain the right, sql-quoted representation for the corresponding bound variable. The nice part is that the default, built-in adapters, derived from psycopg 1 tyecasting code can be overridden by the programmer, simply replacing them in the psycopg.extensions.adapters dictionary. Then the solution to the original problem is now obvious: write an adapter that adapts tuple objects into the right SQL string, by calling recursively adapt() on each element. Note: psycopg 2 adapter code is still very young and will probably move to a more 'standard' (3 arguments) implementation for the adapt() function; as long as that does not slow down too much query execution. Psycopg 2 development can be tracked on the psycopg mailing list: http://lists.initd.org/mailman/listinfo/psycopg and on the psycopg 2 wiki: http://wiki.initd.org/Projects/Psycopg2 """ import psycopg2 import psycopg2.extensions from psycopg2.extensions import adapt as psycoadapt from psycopg2.extensions import register_adapter class AsIs(object): """An adapter that just return the object 'as is'. psycopg 1.99.9 has some optimizations that make impossible to call adapt() without adding some basic adapters externally. This limitation will be lifted in a future release. """ def __init__(self, obj): self.__obj = obj def getquoted(self): return self.__obj class SQL_IN(object): """Adapt a tuple to an SQL quotable object.""" def __init__(self, seq): self._seq = seq def prepare(self, conn): pass def getquoted(self): # this is the important line: note how every object in the # list is adapted and then how getquoted() is called on it qobjs = [str(psycoadapt(o).getquoted()) for o in self._seq] return '(' + ', '.join(qobjs) + ')' __str__ = getquoted # add our new adapter class to psycopg list of adapters register_adapter(tuple, SQL_IN) register_adapter(float, AsIs) register_adapter(int, AsIs) # usually we would call: # # conn = psycopg.connect("...") # curs = conn.cursor() # curs.execute("SELECT ...", (("this", "is", "the", "tuple"),)) # # but we have no connection to a database right now, so we just check # the SQL_IN class by calling psycopg's adapt() directly: if __name__ == '__main__': print "Note how the string will be SQL-quoted, but the number will not:" print psycoadapt(("this is an 'sql quoted' str\\ing", 1, 2.0))