Calling len() on a QuerySet object mid-iteration stops iteration at current hundred documents

[eli-b]

Please see PR #1072 for a test demonstrating this behavior.

[eli-b]

Anybody?

[touilleMan]

Hi, sorry for the delay of the response...

Concerning your issue, it's obviously a cache trouble: the QuerySet fetches the items lazily.
This way the first iteration builds the cache at the same time and returns when the cache is done.
If the first iteration (building the cache) another call on the object ask for the items, the 2nd items iteration will build the cache and the first iteration next statement will returns just like the iteration was over.

Fixing this issue seems a bit complicated given we should have a lazy cache that could be partially populated (chunk by chunk) by any iterator (each keeping it own current state)

I think it could be easier to add some information about this behavior in the documentation (and to raise an exception if the reentrant call is done on the object) and to redirect the user on the QuerySetNoCache object which don't have this limitation.

[eli-b]

I'd be happy with an exception being raised - that's a lot better than not knowing you're getting incomplete results.

[gordol]

I made a pass at fixing this, and it seems to be working now, but I'm not sure whether this will have further ramifications... I'm not familiar enough with how these QuerySets are being implemented to know this won't break anything... but... your test passes now, and I'm using this change in production with no noticeable issues yet... knock on wood

[gordol]

Also, perhaps related? When I set no_cache() I get the following error when trying to check the length of the QuerySet:

TypeError: object of type 'QuerySetNoCache' has no len()

This happening is what prompted me to try to fix this.

[wojcikstefan]

I just proposed a proper fix for this in #1427

[eli-b]

Thanks for the fix @wojcikstefan!