Set Analysis Performance-numeric vs. Text Filter

[yunus_emre]

Hi Experts,

I was wondering something. Are numeric filters faster than text filters in set analysis?

I have about 150 m rows fact table and four lines inline table associated with this fact table. I am using columns which is in this inline table to filter in set analysis. These columns are string values.

Should i convert my columns to numeric values? I have tried this way however i didn't get faster output.

What do you think about that?

Thanks

[marcus_sommer · Accepted solution]

From a theoretically point of view could be numbers faster evaluated as string-values. But it will depend from the specific use-case if any significantly difference within the performance could be noticed or not.

If I understand your question right you mean the difference between:

sum({< F = {'a'}>} value) and sum({< F = {1}>} value)

In this case should be no real difference because the content of {...} will be evaluated as string regardless if it's a number or a string. Further this evaluation happens against the system-tables to set the selection states to TRUE or FALSE before the virtual table which is behind the chart is created and on which the the aggregations are performed. The selection-part on the overall calculation time will be probably lesser as 1% and therefore  I doubt that it's noticeable within the most scenarios if numbers or strings are selected/defined.

- Marcus

[yunus_emre]

Yes, your sample set analysis is right. But in my case string vaIue is not single letter. A word which has about 10 characters like 'supermarket'. Nevertheless, i observe that's not significantly difference