Avoid caching sorted id's, just use relation.key.

[jhirn]

...ting

Hi. Because ID's are cached when the model is loaded, a reload is required after pushing objects onto the model. The ID's are already cached in the model as relation.key, so it's possible to just use that. Observe this behavior:

1.9.2p290> a.b << b2
0.0060 => [BSON::ObjectId('4fbe68488451661ad800000d')]
1.9.2p290> a.b << b
0.0035 => [BSON::ObjectId('4fbe68468451661ad8000009')]
1.9.2p290> a.b.map(&:id)
0.0001 => [BSON::ObjectId('4fbe68468451661ad8000009'), BSON::ObjectId('4fbe68488451661ad800000d')]
1.9.2p290> a.sorted_b.map(&:id)
0.0004 => [BSON::ObjectId('4fbe68468451661ad8000009'), BSON::ObjectId('4fbe68488451661ad800000d')]
1.9.2p290> a.reload 
1.9.2p290> a.sorted_b.map(&:id)
0.0019 => [BSON::ObjectId('4fbe68488451661ad800000d'), BSON::ObjectId('4fbe68468451661ad8000009')]

This commit fixes that, but you still have issue of the documents being cached after calling a.sorted_b. So this still is an issue:

1.9.2p290> a.sorted_b.map(&:id)
0.0002 => [BSON::ObjectId('4fbe68488451661ad800000d'), BSON::ObjectId('4fbe68468451661ad8000009')]
1.9.2p290> a.sorted_b << o3
0.0021 => [BSON::ObjectId('4fbe6a158451661ad8000011')]
1.9.2p290> a.sorted_b.map(&:id)
0.0001 => [BSON::ObjectId('4fbe68488451661ad800000d'), BSON::ObjectId('4fbe68468451661ad8000009')]

I understand a desire for caching here and would like to discuss the need for it vs making it optional, but this pull request can be merged in regardless to eliminate the unnecessary cache object. I can open a separate issue for the latter.

[srosa]

Hi,

First of all, thanks for the fix and the test!

But before I merge let me explain why the cache is there (apart of the performance gain). On our current version of Mongoid (2.0.2 we cannot upgrade yet until 3.0 is out), if we call the relation without the sorted_ prefix, mongoid will set the array of ids back to the "natural order". See the example below :

b = Books.first
# => #<Book _id: 4fb16167942179a62000007d, author_ids: [BSON::ObjectId('4fa2b9ab9421791024000010'), BSON::ObjectId('4f9161829421797635000142'), BSON::ObjectId('4f916182942179763500012c')]> 
b.authors.count
# => 10 
b.author_ids
# => [BSON::ObjectId('4f916182942179763500012c'), BSON::ObjectId('4f9161829421797635000142'), BSON::ObjectId('4fa2b9ab9421791024000010')]

Worst, if the model is saved afterward, the order is completely lost. The cache is an attempt to protect us from this behavior. I don't know yet if this is fix in future release of Mongoid but for now, it's useful to us.

I think your suggestion of making it optional is great and we will surely implement that in a near future.

I'll open a new issue right away ...

[jhirn]

Ah I see. We are using 2.4.10 of mongoid and it seems to behave a bit better.

Well thanks for this gem. I literally needed it the day after you published it. I will keep an eye on this and help out where I can.