Avoid another (cached) query for next link

[lewispb]

Page#records already returns the records for the current page, so #next_param can use that to determine the last record for the next param, used in the Link header.

This is a performance optimization.

This avoids another query, albeit one that usually hits the Active Record Query Cache. The Query Cache is not always the most performant option.

cc @basecamp/sip

[djmb]

Do we need this check?

[lewispb]

The scope argument can be a full set of records, or now just a portion and the method supports both.

This check supports the full set of records. Although we now never call this method with a full set of records internally to Geared Pagination, there's a test for it and I wasn't sure about how much of this should be considered a public API?

geared_pagination/test/portion_at_cursor_test.rb

Lines 55 to 68 in bfce103

	test "next param" do
	create_recordings
	param = GearedPagination::PortionAtCursor.new(ordered_by: ORDERS).next_param(Recording.all)
	assert_equal GearedPagination::Cursor.encode(page_number: 2, values: { number: 106, id: 106 }), param
	param = GearedPagination::PortionAtCursor.new(
	cursor: GearedPagination::Cursor.decode(param), ordered_by: ORDERS).next_param(Recording.all)
	assert_equal GearedPagination::Cursor.encode(page_number: 3, values: { number: 76, id: 76 }), param
	param = GearedPagination::PortionAtCursor.new(
	cursor: GearedPagination::Cursor.decode(param), ordered_by: ORDERS).next_param(Recording.all)
	assert_equal GearedPagination::Cursor.encode(page_number: 4, values: { number: 26, id: 26 }), param
	end

[djmb]

Makes sense thanks! Could extract a method for the check?

[lewispb]

Pushed a3f9008 to avoid the check, by explicitly passing a portion in a kwarg without breaking the existing API.

[lewispb]

Pushed 22e140b as per your idea @djmb to memoize instead. Thanks!