Skip to main content
Announcements
Live today at 11 AM ET. Get your questions about Qlik Connect answered, or just listen in. SIGN UP NOW
cancel
Showing results for 
Search instead for 
Did you mean: 
LasithaR
Contributor
Contributor

Does CreateSearchIndexOnReload only apply to smart Search feature?

Hi everyone,

Does the search index created by CreateSearchIndexOnReload only make searches done through smart search faster? or does it also apply to searches done in objects such as a filter pane?

I have a filter pane with a field of names with approx. 24 mill values of high cardinality. Searching in the filter pane  currently takes about 15 -20 seconds to return results. I guess this time would also be impacted by the specifications of our server.

in addition to the above question, can anyone else provide some other advice on what can be done to improve performance?


Kind Regards,
Lasitha

1 Solution

Accepted Solutions
Rodj
Luminary Alumni
Luminary Alumni

I'm almost certain that CreateSearchIndexOnReload (along with Search Include, Search Exclude) only affect the indexing done for smart search. The individual searches done on a single field in a filter pane are pattern matching using what is effectively the default index of unique values for that field. 24 million unique values is certainly getting up there for a field, and other than making your hardware faster there's no way that I"m aware of to increase performance other than to break down the values into other fields. For example, if it is a date/time you might separate the Date and time and then break up the time into hour/minute/second fields. That way each one has a much lower cardinality. Basically you need to create some kind of grouping or categorisation to the values so that you have smaller indexes to search.

Cheers,

Rod

View solution in original post

2 Replies
Rodj
Luminary Alumni
Luminary Alumni

I'm almost certain that CreateSearchIndexOnReload (along with Search Include, Search Exclude) only affect the indexing done for smart search. The individual searches done on a single field in a filter pane are pattern matching using what is effectively the default index of unique values for that field. 24 million unique values is certainly getting up there for a field, and other than making your hardware faster there's no way that I"m aware of to increase performance other than to break down the values into other fields. For example, if it is a date/time you might separate the Date and time and then break up the time into hour/minute/second fields. That way each one has a much lower cardinality. Basically you need to create some kind of grouping or categorisation to the values so that you have smaller indexes to search.

Cheers,

Rod

Vegar
MVP
MVP

The search index will be generated when your first user tries to do a search. I usually keep the search index on in production apps, but sometimes turn it of during the development phase to reduce load times.

If you do have a pure extract ot transform application that sole purpose is to create qvd files then you don't need the index either. 

https://help.qlik.com/en-US/sense/November2019/Subsystems/Hub/Content/Sense_Hub/Scripting/SystemVari...