Best way to fetch the continuous list with PostgreSQL in web

Assuming it's only the ordering of the results that fluctuates and not the data in the rows, Fredrik's answer makes sense. However, I'd suggest the following additions:

• store the id list in a postgresql table using the array type rather than in memory. Doing it in memory, unless you carefully use something like redis with auto expiry and memory limits, is setting yourself up for a DOS memory consumption attack. I imagine it would look something like this:

  create table foo_paging_cursor (
    cursor_token ..., -- probably a uuid is best or timestamp (see below)
    result_ids integer[], -- or text[] if you have non-integer ids
    expiry_time TIMESTAMP
  );
  

• You need to decide if the cursor_token and result_ids can be shared between users to reduce your storage needs and the time needed to run the initial query per user. If they can be shared, chose a cache window, say 1 or 5 minute(s), and then upon a new request create the cache_token for that time period and then check to see if the results ids have already been calculated for that token. If not, add a new row for that token. You should probably add a lock around the check/insert code to handle concurrent requests for a new token.

• Have a scheduled background job that purges old tokens/results and make sure your client code can handle any errors related to expired/invalid tokens.

Don't even consider using real db cursors for this.

Keeping the result ids in Redis lists is another way to handle this (see the LRANGE command), but be careful with expiry and memory usage if you go down that path. Your Redis key would be the cursor_token and the ids would be the members of the list.