Question

Let’s say I’ve got n properties (pairs key-value) that are either dates, numbers or strings. Which data model is the most efficient to store them?

Here is a solution I thought about:

Each row of the data table must be able to be linked to n properties of several types.

The advantage of this solution is that if I want to add a new type, I just have to add two new tables (label_xy and value_xy) instead of modifying the structure of an existing table. But is this solution really the most convenient one?

How would you have done this?

Thanks :)

Answer 1

I’ve used all of the scenarios:

1. the one you described (each data type in own table)
2. the one from Anthony’s answer – with value only string casted to ither datatypes on demand
3. one table with several columns with distinct datatypes, of them only one is used in each record (rest of them are sparse columns)

From those solutions: 1. was joining hell, 2. is good if your data is mainly strings, 3. although a bit of space-over-consuming is by far the best and most efficient.

As for 3. I must explain that the table looks like this:

EntityID int,
TypeID int,
LabelID int,
ValueInt int,
ValueDecimal decimal(12,4),
ValueString nvarchar(max),
ValueDate datetime

(Although LabelID defined TypeID I used both for join-less-usage purposes.)