On Fri Oct 17 03:00:06 2014, ETHER wrote:
Show quoted text> On 2014-10-16 17:56:08, ETHER wrote:
> > See also
https://rt.cpan.org/Ticket/Display.html?id=99545
> >
> > When inflating a DateTime using a specific DateTime::Formatter::*
> > class, the formatter is not set on the resulting object - meaning if
> > it is stringified again, the format is not preserved from the
> > original. This leads to surprising results when copying column data
> > around in an application.
Can you explain the issue in more detail? The only thing I can think of is using data from Result Objects in search(), which is broken by design and should remain so without extra bandaids.
Show quoted text> >
> > I propose that DBIx::Class::InflateColumn::DateTime explicitly sets
> > the formatter when inflating, to overcome this. I've prepared a
> > preliminary patch in branch topic/preserve_datetime_formatter.
>
The patch in the branch is not acceptable for the same reason as outlined in
https://rt.cpan.org/Ticket/Display.html?id=99545#txn-1421646 - it is a very strong behavioral change.
Show quoted text> Alternatively, if this patch (or the intent thereof) is not
> acceptable, can we extend _post_inflate_datetime to also set the
> formatter field, if defined in $info? at the moment only the timezone
> and locale information can be set this way. I have been unable to
> find any other hooks in my application to add on extra inflation
> logic.
The general approach to augmenting IC::DT functionality is in fact to around() _post_inflate_datetime. There is a gray area of methods in DBIC which are private enough to never be called from the outside but not private enough to not be overloaded. I have not come across a good way of documenting this duality (see
https://metacpan.org/pod/DBIx::Class::Ordered#METHODS-FOR-EXTENDING-ORDERED for a clumsy attempt)
I am also open to add an extra column_info flag. The name is tricky because it *must* be able to accept a different formatter than the one used to talk to the database (the column info settings have to do with display, not with storage). I am thinking something like set_formatter => [class-name|1] where 1 means "use the storage formatter".
But first of all - explain your actual use case for this.
