Mongodb - unique Ids with count()

Question

I build a site with referral links to products and working with mongodb. We all know that the long _id field looks ugly in Urls e.g. http://website.com/p/532600d9a715ea980a00002b/u/532af835a715eae43f00002c.

I have read some posts here and googled a while and wondered why people don't generate this easy unique id:

1. Count current documents in collection
2. add 1 to the number
3. append a field to document called e.g. 'unique'

So each document got a unique number starting with 1 and urls would look like this then http://website.com/p/1/u/2.

I won't delete any document.

Where is the problem with this solution?

Answer 1

The sane variant of this approach, where you increment a counter and assign that number, is part of the official documentation.

The problem with your approach is that counting is a relatively expensive operation. Moreover, it's not even collision safe because two threads could count simultaneously and get the same number, so the second insert will fail.

Generally speaking, sequences create a bottleneck because you need a single point of coordination. For a write-heavy system, the number of operations is doubled (an additional $inc for every insert). To make matters worse, sequences leak a lot of information to the outside (i.e. the number of entries in the database).

If you're running a web shop, blog, etc. you probably want your URLs to be a lot prettier, descriptive and SEO-friendly like /products/displays/27-inch/iiyama-2734FW.

If you're building a web application where the URLs aren't visible to search engines anyway, I believe that neither /users/433 nor /users/532af835a715eae43f00002c will be a reason for your customer to choose or not choose your application. Maybe a wise customer would, but then they'd reject the former option...