Sounds good, Daan. We're probably all looking forward to seeing your code.

On Mon, 5 Nov 2001, Daan Oosterveld wrote:

> A small example how I've implemented the datamodel so far in C++.
> The attachment is an PDF document of a class hierachy..

Sorry, but it's too little to give any real clue.

> In this model the values are fixed sizes.
> With exeption to complex or variable size values.

Figured as much. Did you comprehend enough of my "variable-size values"
ramblings to have an opinion about the issue?

Btw, there is a solution to the problem of how to throw away values with
block (which is automatic for contained values, but obviously not for
referenced ones). Just let value tables be local to containers. That's it.
Perhaps this was alreay obvious to others. :-/

I should probably try implementing these reference containers, when you
have released your code.

As for performance of reference blocks: While we lose some access speed
vs value blocks, we might gain some when manipulating (e.g sorting) them.
Also, reference blocks doesn't require as large continuous memory areas,
but on the other hand you need 4 extra ones to store the values in.

Reference blocks should be a win for storing small values (32 - 64 bits)
vs fixed value (128bit) blocks, but a loss for large values. It's hard to
predict which the general case is.

Marcus
------------------------------------
If you find that life spits on you
calm down and pretend it's raining