Looking for MNESIA and AMNESIA learning resources

Learning Questions & Discussions Resources Chat/Questions
,
1

As a novice dev I’m trying to keep the curriculum as lean as possible.

My requirements are modest: build simple CRUD apps with Phoenix liveview, a key value store and auth. Scalability is a non concern.

MNESIA and AMNESIA seem like viable alternatives without the added learning overhead of SQL, Postgres, and Ecto.

But, I’m struggling to find practical learning resources, outside of these few:

I’d appreciate any other recommendations, the more project based/practical the better.

2
5 Likes
3

Thanks, this is exactly what I was looking for. Especially, “Erlang for Great Good!” - engaging, in depth and applied to a practical context that makes it easier to grok.

1 Like
4

I’m the author of Memento, an Elixir wrapper around :mnesia. I’ve written various use-cases and examples in the Memento docs, which might be somewhat useful.

6 Likes
5

Thank you, this looks a lot more intuitive than vanilla Mnesia and I appreciate the effort that went into the docs.

1 Like
6

Just had a look at Memento.

I only did very basic stuff with ets/mnesia, but what bugged me were the match-specs.

Please correct if I’m wrong, but as I understand it …

if you got a table like this

Mnesia.create_table(Person, [attributes: [:id, :name, :job]])

and you want to query all persons with an :id > 3 you got to tell the match-spec about the structure you want to query by giving the positions of all attributes, these you can then in turn use in an expression.

Mnesia.dirty_select(Person, [{{Person, :"$1", :"$2", :"$3"}, [{:>, :"$1", 3}], [:"$$"]}])

This becomes annoying at the latest when you want to change the table, eg if you want to add an attribute you’d have to change all match-specs (add "$4" etc).

Helpers like :ets.fun2ms

iex> :ets.fun2ms(fn {id, name, job} when id > 3 -> {id, name, job} end) 
[{{:"$1", :"$2", :"$3"}, [{:>, :"$1", 3}], [{{:"$1", :"$2", :"$3"}}]}]

or GitHub - ericmj/ex2ms: :ets.fun2ms for Elixir, translate functions to match specifications do not really help, because while they make it easier to write specs, you’ll still have to update when you change the table.

Memento keeps track of the positions of the table’s attributes so you can use keys in your queries:

Memento.Query.select(Person, [{:>, :id, 3}])

:+1:

1 Like
7

Just be very (very) careful and with both eyes (all four in my case) WIDE open about the pitfalls of deploying mnesia in production environments due to the split-brain/resolution issues.

Great for learning, I liked it a lot, but operationally it became a nightmare/tech-debt item.

Cheers,
-mox

8

what’s that? can you elaborate?

9

Mnesia is a simple solution designed for a small number of nodes (maybe a couple of dozens max) all reside in the same data center. If you stretch it beyond that you will get burned. Most people are perfectly happy with the limit though.

1 Like