On 27 Oct 2012, at 10:08 PM, Massimo Di Pierro <[email protected]> wrote: > There are two issues here. One is that validators were originally intended > for form input and form input never submit None. They submit '' or missing > value but never None. I agree with you that you may want to call validators > and pass a None. In that case there is no need to ignore it. One problem > remains, will this break backward compatibility. Perhaps no. Not sure.
Joe makes a pretty persuasive argument. Call it a bug fix. > > Can you show your modifications to dal.py? > > On Saturday, 27 October 2012 17:48:54 UTC-5, Joe Barnhart wrote: > But Massimo -- why do an early exit from "Field.formatter" if the value is > None? Why not give the formatter object a chance to format None values? I > just don't understand this decision. It is expedient for a small number of > programs where the user doesn't want to format None, but at the expense of > preventing anyone from ever formatting None. > > Your fix only papers over the problem by storing "NT" in the table instead of > None. What if I want to show it as "N/T" later, or "no time"? I have to do > a bulk change on the database just to accomplish a formatting task. There > could be many reasons for wanting to format None into some other value -- all > ignored because you chose to do an early exit in the formatter routine > instead of passing the real value along. > > It is also not explained anywhere in the documentation that None will never > be passed into the formatter, so it can't be a backwards-compatibility issue. > > -- Joe B. > > P.S. I'm perfectly happy to modify my version of dal.py forever, but I'm > really interested in your reasoning for the early exit on None and if this > pattern is carried out in other parts of web2py. > > On Saturday, October 27, 2012 12:44:21 PM UTC-7, Massimo Di Pierro wrote: > You can do > > db.table.filter_in = lambda value: 'NT' is value is None else value > > > On Saturday, 27 October 2012 02:54:53 UTC-5, Joe Barnhart wrote: > But Paolo -- you forgot the third option... > > FIX the problem in dal.py so that Validator classes work properly. It has no > deleterious effects on other validators (at least so far in my testing) and > it permits validators to format None values. > > -- Joe B. > > > > -- > > > --
