Support bag and duplicate bag

[fcevado]

issue #25

[fcevado]

I didn't update the docs and readme, firsts I wanted to know if there isn't anything missing. I'll do that in this same pr, as soon as it is approved.

[sasa1977]

This looks good and mostly consistent with behaviour of raw ets tables.

The only difference I see is that retrieval operations in ConCache return a single value or a list of values, whereas ets functions always return a list of values. I'm not 100% sure if ConCache should be changed, and it would be a breaking change anyway, so let's leave that as it is for now.

I think the following is still missing:

• Since we agreed update_* won't be supported for bags, I think we should raise in these functions if ets table is of the wrong type.
• get_or_store function should also be tested for bag tables
• Return types in typespecs should be adapted. Currently most of functions return value, but now the correct type would be value | [value]
• Docs must be updated, and differences between set/bag types should be explained for every function. From what I can tell, expiry will remove all items under the given key. That's fine, but it should also be explicitly mentioned in docs.

Thanks for doing this!

[fcevado]

The only difference I see is that retrieval operations in ConCache return a single value or a list of values, whereas ets functions always return a list of values. I'm not 100% sure if ConCache should be changed, and it would be a breaking change anyway, so let's leave that as it is for now.

I totally agree about that, I think a change like that to fit more a change to a 1.0 version, since it would break applications already using ConCache.

About the necessary changes, I'll try to do that during the week, I dont have too much free time but I'll do it.

[fcevado]

@sasa1977 I started to apply the changes, and I saw that insert_new uses update to execute. If I validate the ets type on update or dirty_update, insert_new wont be supported anymore. Should I change insert_new or include it on non-supported functions for bag types?

You said that I should raise on wrong ets type. I thought about return only the tuple {:error, :ets_type_not_valid}. Should I return the error or raise?

[sasa1977]

Looking at :ets.insert_new I see that it works for bag types, which makes sense to me as well. So we should support insert_new for bags.

There is in fact no need for insert_new to rely on update, since there's ETS native version of insert_new. Luckily, I've actually made that change in a 0.12.0 branch, but I still didn't decide on merging that branch. I think if you cherry pick this commit into your own branch, the problem would be solved.

You said that I should raise on wrong ets type. I thought about return only the tuple {:error, :ets_type_not_valid}. Should I return the error or raise?

Error tuples are usually returned if the caller can do something meaningful with them. I believe this is not the case here. Moreover, :ets raises in such situations, so I think that would be consistent behaviour. I'd say raising an ArgumentError with a helpful message would be a good approach here.

[fcevado]

I think that's it. I read the docs a few times and I think that were it needs to be mentioned, I put disclaimers about bag and duplicate_bag.

Do you want that I up the version and update changelog?

[sasa1977]

Thanks for contributing!