Fixes has many through counter-cache bug

[matthewrobertson]

This commit fixes reported issue #7630 in which counter caches were not being updated properly when replacing has_many_through relationships.

eg if posts have many tags through taggings

and I did:

post = Post.create!
tag = post.tags.create!
tag.posts = []

then it was: post.reload.taggings_count == 1

this pull makes it: post.reload.taggings_count == 0

[senny]

I receive some errors when running all the tests but this could be because the master build is red at the moment (all errors are related to "deprecated_dynamic_methods_test.rb".

Just some cosmetic corrections:

• Could you add a CHANGELOG entry, which explains the change?
• Could you trim the commit message subject to <= 72 characters (you can use the body to add more details to the commit message)?
• You can find more information related to contributions in our guide

[matthewrobertson]

@senny I trimmed the message, and amended the CHANGELOG.

I am also getting 3 failed tests, but those fail on master with and without this change. So I assume they are unrelated.

[carlosantoniodasilva]

I think you should use has_cached_counter? for this check.

[matthewrobertson]

@carlosantoniodasilva I thought so too, but that method just takes a wild stab at determining if a has_many association has a counter cache. In this case we are operating on the belongs_to side, so Reflection#counter_cache_column will be set if there is a counter_cache.

I think there might actually be problems with the way has_cached_counter? works but if there is I don't think it is in the scope of this pull.

[carlosantoniodasilva]

Alright, I'll check that method later.

[tobidelius]

This is really awesome. I've been waiting for this fix a long time.

[jeremy]

👍

[rafaelfranca]

@matthewrobertson this needs a rebase.

@carlosantoniodasilva

[matthewrobertson]

@rafaelfranca I rebased, should be good to go now.

[carlosantoniodasilva]

@matthewrobertson thanks!

[carlosantoniodasilva]

This added the AR gem to the repo, and I committed it without seeing through the github interface (that doesn't make it very apparent) =/. It has been reverted already.

[matthewrobertson]

@carlosantoniodasilva sorry about that. I also didn't notice that it was in there. I don't know why it didn't show up on Github...

[carlosantoniodasilva]

@matthewrobertson no problem. It did show up, but since there's no diff, only the file, I missed while scanning the changes, my bad. It'd be way easier to see if the diff states were shown by default as previously.