Stephen Frost wrote:
>
> I'm honestly not a big fan of the "columnlist" approach that's been
> taken with the options. While I understand the desire to seperate the
> parsing from the typing, making the users essentially do that association
> for us by way of making them specify how to handle each column explicitly
> is worse than just accepting that different types may need to be handled
> in different ways.
>
Whether or not you like it, the fact is it's there. I think any solution
should be consistent with what is done.
When CSV was first discussed we looked at doing type-specific behaviour.
The end of the long debate was that we simply couldn't do that safely,
and the only recourse was to require the user to specify the behaviour
required if it differed from the default. You might be inclined to want
to revisit that, but I am not.
cheers
andrew