Re: Database design issue: seems really bad, yet I'm having trouble coming up with anything better

First thoughts on this is that with a conventional approach, maintenance of this file could become a HUGE nightmare over time.

I believe I would at least look at a self-building menu process where you build some sort of sparse 3 dimensional matrix (protein, starch, vegetable) and then as people make decisions, those decisions become stored and ranked as "good", "ok", and "never again" and also by count. Then, as each choice is recorded you will have a self-building menu model that users can use to build the menu choices for you. Also, giving the users an opportunity to score the choice, and using the counts of usage, you can, over time, develop a dynamic menu based upon popularity, and/or by score,

If that process seems to easy you could add in search-ability by adding the ability for users to "tag" menu choices with words like "Holiday", "cold weather," "Snacks," "brunch," "spicy" and so forth.

This looks like a process where user feedback is required to make the process really fly and to keep downstream maintenance under control.




On 12/10/2014 11:35 AM, James H. H. Lampert wrote:

Consider this scenario:

We have three sets of items. To avoid discussing the real-world situation, I'll use a food analogy, and call them protein, starch, and vegetable.

We have perhaps as many as 300 distinct protein items (to which others may be added in the future), some 60 starch items (to which others may be added in the future), and some 80 vegetable items (once again, to which others may be added in the future). A complete meal includes a protein, an arbitrary number of starches, and an arbitrary number of vegetables.

We always select the protein first. Each protein has its own subset of compatible starches and vegetables, e.g., if the protein is prime rib, 10 oz cut, then the available starches would include Yorkshire pudding, mashed potatoes, and baked potato, and the available vegetables would include creamed spinach and creamed corn, while if the protein is roast turkey, 8 oz, then the starches might be dressing and mashed potatoes, and the vegetables might be creamed corn, Brussels sprouts, and carrots.

Now, the solution that's been presented to me is for the protein file to have a record for every protein, the starch file to have a record for every valid combination of one protein and one starch, and the vegetable file to have a record for every valid combination of one protein and one vegetable.

Something about this seems wrong. Mainly because it would produce so many starch and vegetable records, repeating any given starch or vegetable for every protein with which it's a valid combination.

Yet I'm having trouble coming up with anything better. Any suggestions?

--
JHHL