On 2/27/07, Oleg Kobchenko <[EMAIL PROTECTED]> wrote:
Below is a quick sketch of vertical arrays
constructed from DataReader.

Hmm... that is simpler than I thought.

That said, while some of the complexity when compared
to my earlier implementation is eliminated by a better match
between .Net types and J  types, much of the simplicity of
this approach comes  because of issues that it ignores:

Column names are ignored.
Nulls are ignored.
Passing the data into J is ignored.
Some (perhaps irrelevant?) data types are ignored.

Hypothetically speaking, it's often (but not always) a bad
practice to ignore column names.

Hypothetically speaking, it's probably better practice to
not use Nulls (they're not relational), but that's not always
an option.

And, it seems to me that you need to manually code up
the mechanism to pass the data into J, for each different
table you're importing, if you're going to use SetM.

This isn't necessarily a bad thing, but it does suggest that
SetM (n different variables) does not  have a clear advantage
over SetB (with result intended for 3!:2).
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to