Re: [Sqlalchemy-tickets] [sqlalchemy] #2916: possible numeric regression on 0.9.1

[sqlalchemy <mi...@zz...>]

#2916: possible numeric regression on 0.9.1
-----------------------------------+-------------------------------
      Reporter:  zzzeek            |      Owner:
          Type:  defect            |     Status:  closed
      Priority:  highest           |  Milestone:  0.9.xx
     Component:  mssql             |   Severity:  major - 1-3 hours
    Resolution:  worksforme        |   Keywords:
Progress State:  completed/closed  |
-----------------------------------+-------------------------------
Changes (by zzzeek):

 * status:  new => closed
 * resolution:   => worksforme
 * status_field:  in queue => completed/closed


Comment:

 this is due to the column being actually DECIMAL on the DB side.  the
 MSSQL dialect in 0.9 specifically has a pyodbc float type and it does not
 check for a type handler.  We can check for pyodbc sending us the
 "Decimal" class here in the result processor; some dialects, like that of
 psycopg2, are smart enough to look at the incoming "coltype" to determine
 the correct handling, while others (most?) are not (e.g. MySQLdb sends us
 integer codes but we don't have a lookup for them).

 So in this case the behavior of using a Float to receive Numeric/DECIMAL
 values is undefined; it will vary by backend.   In MSSQL's case it's just
 changed.  I'd rather not get into trying to support mismatches like this
 consistently in all cases, it's better to just use the correct type.

 here's a test just for record:

 {{{
 #!python
 import sys
 import cdecimal
 sys.modules["decimal"] = cdecimal

 from sqlalchemy import *

 e = create_engine("mssql://scott:tiger@ms_2005", echo='debug')

 conn = e.connect()
 trans = conn.begin()

 t = Table('t', MetaData(), Column('data', Numeric(10, 5)))
 t.create(conn)
 conn.execute(t.insert().values(data="45.17"))

 t2 = Table('t', MetaData(), Column('data', Float()))

 print(conn.execute(t2.select()).fetchall())

 t3 = Table('t', MetaData(), Column('data', Numeric(asdecimal=False)))

 print(conn.execute(t3.select()).fetchall())
 }}}

--
Ticket URL: <http://www.sqlalchemy.org/trac/ticket/2916#comment:2>
sqlalchemy <http://www.sqlalchemy.org/>
The Database Toolkit for Python