Re: Define "flatten database" ?

> OK, I'd vote for both terms being incorrect. It is not denormalizing
> because it is starting with non-1NF data -- agreed? And it is not
> normalizing because, well, it clearly isn't (the key, the whole keyu and
> nothing but the key ...). However, the reason that "we" sometimes call it
> normalizing is that it takes a non-SQL92-compatible structure and turns it
> into something that can be used with ODBC, for example. By formal
> definitions (until perhaps recently) the source data are not in first
> normal form. So, informally, to get ODBC to work (for example) someone
> might suggest they are normalizing it. I'm one of the people who says
> "flatten" instead because normalizng is clearly wrong. However, it is
> still an informal way people might say it as in "we need to nornalize the
> data in order to pull it into Excel". (Wrong, but you can see how such
> language evolved).

I'll just add one other point that the reason that "flattening" from a non-1NF is sometimes referred to as normalizing is that there is a sense that the nested structures are pulled into separate "rows" thereby normalizing, and then a view created that ties them to their original table, but now as if both relations (the parent and the child) were normalized and then joined. Did that help clarify the langauge?

> You have it backwards and twisted.

This might still be true. --dawn Received on Fri Jan 28 2005 - 20:04:03 CET