what is the better approach

[ali_hijazi]

Hello

I got 2 fact tables

one for sales and another for stock

we had to concatenate these two table to avoid synthetic keys and have star schema

my question is the following:

is it good to keep tables as are or unify column names?

in fact now the table looks something like this:

A    B      C             D      E      F           Date

sales sales sales    null    null   null        some_date

sales sales sales    null    null   null        some_date

sales sales sales    null    null   null        some_date

null    null   null      stock  stock  stock   some_date

null    null   null      stock stock   stock   some_date

can we keep it like this or rename D, E, and F columns as A,B, and C and add a flag

which is the better approach?

Please advise

[marcus_sommer]

It is a very interesting question and I think it couldn't be simply said that one is better then the nother - it will depend from your entire data and how they should be visualized which approach is more suitable.

I use such asynchronous concatenated tables quite often (mostly with sales and forecast data) and it worked very well and fast. It's simple and easy to create and the load run-times are less - this are beside hard-factors like the app-size or the calculation times from the biggest and/or most used objects also important points by the creation of a datamodel.

I wouldn't change it without serious performance issues. If you want to check both approaches against eachother you should use the mem-files to compare both and to try one or another optimization: Recipe for a Memory Statistics analysis.

- Marcus