Add a cursor for the current page.

[GCorbel]

I added current_cursor to Mongoid::Scroll::Scrollable::Iterator.

current_cursor = nil

cursor = cursor_type.from_record Feed::Item.find_by(name: '7'), field_name: field_name, field_type: field_type

Feed::Item.asc(field_name).limit(2).scroll(cursor) do |_, iterator|
  current_cursor = iterator.current_cursor
end

Feed::Item.asc(field_name).limit(2).scroll(current_cursor) do |record|
  # loop over the same records
end

This allows to store the cursor and have the same result.

We use it to have this kind of payload for an API response :

GET {{baseUrl}}/attachments?pageToken=eyJ2YWx1ZSI6bnVsbCwiZmllbGRfdHlwZSI6IlRpbWUiLCJmaWVsZF9uYW1lIjoiY3JlYXRlZF9hdCIsImRpcmVjdGlvbiI6MSwiaW5jbHVkZV9jdXJyZW50IjpmYWxzZSwidGllYnJlYWtfaWQiOm51bGwsInR5cGUiOiJuZXh0In0=
{
    "records": [
        // ....
    ],
    "paging": {
        "pageToken": "eyJ2YWx1ZSI6MTM0OTM2MTI4MS4wMjcsImZpZWxkX3R5cGUiOiJUaW1lIiwiZmllbGRfbmFtZSI6ImNyZWF0ZWRfYXQiLCJkaXJlY3Rpb24iOjEsImluY2x1ZGVfY3VycmVudCI6dHJ1ZSwidGllYnJlYWtfaWQiOiI1MDZkOWU4MTdhYTU4ZDEyNTkwMDBmMTIiLCJ0eXBlIjoibmV4dCJ9",
        "perPage": 1,
        "firstPageToken": "eyJ2YWx1ZSI6bnVsbCwiZmllbGRfdHlwZSI6IlRpbWUiLCJmaWVsZF9uYW1lIjoiY3JlYXRlZF9hdCIsImRpcmVjdGlvbiI6MSwiaW5jbHVkZV9jdXJyZW50IjpmYWxzZSwidGllYnJlYWtfaWQiOm51bGwsInR5cGUiOiJuZXh0In0=",
        "nextPageToken": "eyJ2YWx1ZSI6MTM0OTM2MTI4MS4wMjcsImZpZWxkX3R5cGUiOiJUaW1lIiwiZmllbGRfbmFtZSI6ImNyZWF0ZWRfYXQiLCJkaXJlY3Rpb24iOjEsImluY2x1ZGVfY3VycmVudCI6ZmFsc2UsInRpZWJyZWFrX2lkIjoiNTA2ZDllODE3YWE1OGQxMjU5MDAwZjEyIiwidHlwZSI6Im5leHQifQ=="
    }
}

Given that, we can find the same records if we do a call with the token related to the current cursor.

[dblock]

Add to README.

Not for this PR, but I wonder whether we can make some of the cursor construction lazy to avoid paying the cost of the cursors being generated without being used.

[dblock]

Some rubocop will be needed here.

[GCorbel]

I changed, thanks.

Rubocop show me 138 offenses with the config in .rubocop.yml 😥

[dblock]

Looks like CI failed for something legit.

[GCorbel]

Not for this PR, but I wonder whether we can make some of the cursor construction lazy to avoid paying the cost of the cursors being generated without being used.

The problem is that we have to build cursors with the first and the last record. We don't know how many records there are without an extra query.

We can do :

records = records.to_a

iterator = Mongoid::Criteria::Scrollable::Iterator.new(
  previous_cursor: cursor_type.from_record(records.first, cursor_options.merge(type: :previous)),
  next_cursor: cursor_type.from_record(records.last, cursor_options),
  current_cursor: cursor_type.from_record(records.first, cursor_options.merge(include_current: true))
)

records.each do |record|
  yield record, iterator
end

But it means we converts records to an array, so we fetch documents first and I guess we will have other errors.

[GCorbel]

Actually, we can have data about the records with records.view.cursor.entries.

[GCorbel]

Looks like CI failed for something legit.

I don't know why but I fixed it.

[GCorbel]

#42 have to be merged first so I will change the Changelog and Readme in this PR.

[dblock]

Thanks! Rebase?

[dblock]

Something's odd in the README.

[dblock]

FYI, 2.0 has been released, https://twitter.com/dblockdotorg/status/1832420865847259391/