Unify best practices for countDocuments() and estimatedDocumentCount()

[vkarpov15]

Re: discussion from #6713, leaving a note to think about whether we should add a method that switches between countDocuments() and estimatedDocumentCount() based on whether filter is empty or not.

[ILoveYouDrZaius]

I agree with this issue, I don't understand why I have to switch between two functions.

[vkarpov15]

There's some more detail about why you would use one vs the other in the API docs:

• https://mongoosejs.com/docs/api.html#model_Model.estimatedDocumentCount
• https://mongoosejs.com/docs/api.html#model_Model.countDocuments

TLDR; estimatedDocumentCount() is faster when you're looking to count how many documents are in the entire collection, but might not be 100% accurate in a sharded cluster. If you're using a standalone MongoDB instance and have less than 100k documents per collection, it doesn't make much difference whether you use estimatedDocumentCount() or countDocuments(). It can make a big difference with large collections though.

[xxyyzz2050]

use count(0) instead of count({filter}) or countDocument({filter}) to run estimatedDocumentCount()

[libook]

Is it the best practice to use estimatedDocumentCount() when there is no query parameter?
If it is, it will be very helpful to switch between estimatedDocumentCount() and countDocuments() with a single method.

[vkarpov15]

@libook yes, you should use estimatedDocumentCount when there's no filter param. There may be some reasons to use countDocumenta() if you're using sharding, but I'm not certain.

[shuwenhe]

CountDocuments