is Count still used in DataProvider ?

I was told setItems no longer use count but I’m seeing it in the log. I’m using @Transactional(readOnly = true) public Page<CustomerEntity> list(Pageable pageable, Specification<CustomerEntity> filter) { return repository.findAll(filter, pageable); }. But still the logs says Hibernate: select c1_0.id, c1_0.address, c1_0.city, c1_0.contact_name, c1_0.country, c1_0.created_date, c1_0.customer_name, c1_0.last_modified_date, c1_0.postal_code from customer_entity c1_0 offset ? rows fetch first ? rows only Hibernate: select count(c1_0.id) from customer_entity c1_0

Spring is doing the count query, you can see the discussion here: Don't use findAll(Pageable pageable) · Issue #103 · vaadin/starters · GitHub

Also, the count query is optional, but it’s better to provide it if you don’t have a performance reason not to

what is the use of this count query ?

If you provide a count the scroll bar will have the correct size. Without the scroll bar might shrink while scrolling

so we’ve to live with this

when there are thousands of records, there is a huge impact on the performance

the progres bar keeps runing for about 1 min

Sounds like you have done no database optimization like reduced columns, added indexes or reduced joins / queries

For your initial question, the count is used to fill the Page information, if you return a List or a Stream then you don’t need the count query.

( as it’s written in the link). Of course you can also try to optimize your database or query. A thousand records is nothing for a database, so there must be something else.

We have millions of records without an issue. How does your grid look like and how long takes your query?

yes it is returning stream. Should I try to create custom repo as given in the link ?

it is a simple table having 7 columns and it takes 1-2 minutes

Do you have Joins or complex where condition?

no. simple SELECT with two WHERE conditions

it checks for a matching STRING value

Lower / upper casing, like, binary field or something else that “intensive” to compute?

yes lower with like operator

Create a proper index for that