AutoFetch doesn't work beyond a depth of 2 with (inversed) hasOne relationships

[entropitor]

My data-structure looks like this:

models.Team.hasOne('match',models.Match,{reverse: "teams"});
models.Player.hasOne('team',models.Team, {reverse: "players"});

now when I do

req.models.Match.get(id,{cache: false, autoFetchLimit: 2, autoFetch: true},function(error, match) {res.jsonp(match);});

I don't get my players into the match object :(

[entropitor]

I'm not sure if I should make a new issue for this, but I'm also noticing that autoFetch doesn't use JOIN queries to get its data, is there a reason for this?

[dresende]

autoFetch is just a shortcut for .find()+.get*(). I don't know if it could be changed to take advantage of JOINs.

About the bug, I think I saw a line that could be doing this, but I'm not sure, I need to investigate.

[idekterev]

I have also observed random behaviour using autoFetch: I have exactly the same set-up in terms of inverse associations described by caske33.
In some parts of my code the autoFetch works, and yet exactly the same code elsewhere sometimes fails to return any associations, and sometimes succeeds with 1 level, sometimes with 2 levels. It is truly 'random' as far as I can tell.
I had to stop using autoFetch, and do chained lookups - it seemed like the only way to fix the problem.

[dresende]

DIsable cache and your code will probably go away (see #241, #339).

// before connecting..
orm.settings.set("instance.cache", false);

[entropitor]

This still doesn't work btw. Nor with the cache disabled globally nor locally (like I did in the initial post).

[toddpi314]

FYI, disabling cache and turning all models into autoFetch (at the schema level) fixed for me.

The findChain syntax is ideal, where we can disable cache and force a deep auto-fetch on a query-by-query basis.

The current situation is disabling cache at a global level really slams the DB connection.

[toddpi314]

Is this a problem with the autoFetch state relative to caller and the cache?

Historically, when something like this comes up in an ORM, it is because in some scenarios an entity has a state which may or may not contain deep fetch elements based on the context/behavior of the caller. To avoid this, you may consider extending your cache key fields to contain meta-data relative to the auto-fetch history.

For instance, if entity is stored by "id" and keyed for hit/miss, you can extend the key to "id_propertyA_propertyB" to identify that the instance is associated to an auto fetch with those fields.

Hope this helps.