provide faster way to perform a count() on million documents
a couple of customer a shifting from MongoDB to other DBs like Redis or even PostreSQL as count is too slow on MongoDB when there are a lot of document to count even with use of indexes
Gabriel
shared this idea
-
Alec
commented
To build on this, we have really bad performance doing a `countDocuments()` call on a covered index. It takes >5s on a collection with:
```
totalKeysExamined: 769298,
totalDocsExamined: 0,
```I get that under the hood, the countDocuments() needs to scan the index, so will have O(n) time.
Since we want to call this count frequently (ideally every minute), it would be nice if this wasn't so bad. An acceptable workaround for us would be building a partial index with a filter that matches our countDocuments() filter, and then it would be really nice if we could get a estimatedCountDocuments() from the metadata of the index size, which is currently unavailable at the index level
-
Yiru
commented
More specifically, it is better having meta on index to get the key below to the index page. So you don't have to scan the key, but return the meta count.