[Proposal] Map.take!/2

[Aaron Tinio]

Map.take!(map, keys)

Same as Map.take/2 but raises a KeyError when one of the keys doesn't exist.

[Tallys Martins]

Hi!

I could not find a final reply for this thread, here nor at Github, so I am here to ask and argue why this would be a useful function.

I am writing some tests where I pattern match just a few keys of my map objects using Map.take. Developers could type invalid keys and the tests would still pass.
If you are open to receive an implementation for this I can work on it and give my first piece of contribution.

Sorry for reviving something deep in the past if its already solved.

Warm Regards,

Tallys Martins

[Peter Hamilton]

Thanks for searching for an old thread rather than making a new one!

> I pattern match just a few keys of my map objects

Would just doing a direct pattern match work here?

%{foo: foo, bar: bar} = my_map

That would fail if foo or bar was missing in my_map.

--
You received this message because you are subscribed to the Google Groups "elixir-lang-core" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elixir-lang-co...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elixir-lang-core/6c898b72-6626-4738-8331-cdee13cff251%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Tallys Martins]

Well, this does not suites for my cases because I actually have object maps like %Measurement{}. The problem comes when I need to pattern match its attributes like id and others. Let me give you some context, but maybe I am going through the wrong solution, though.

This code:

assert expected_measurement = measurement 

would fail because association attributes in measurement are not loaded Ecto.Association.NotLoaded (and they are supposed to)

So I selected the ones that I care, not simple the id, but to resume:

assert %Measurement{id: ^expected_measurement.id} = measurement

And this raise an error: cannot invoke remote function expected_measurement.id/0 inside match 

So I decided to use Map.take/2 on both maps checking a list of attributes that I think that matters the most. The problem is that if one comes with an attribute that does not exists, the tests will equally pass, so I thought it would be good to have a Map.take!/2 function to help on that.

To view this discussion on the web visit https://groups.google.com/d/msgid/elixir-lang-core/CAOMhEnwKOfH4scFO7JKK3nBqZUKj_CP9mV0km6%3D2gw%3D3UiNGQg%40mail.gmail.com.

[Martin Svalin]

So essentially you are concerned with this scenario?

```elixir
actual = Map.take(DateTime.utc_now, [:yaer, :month]) # note misspelling
expected = Map.take(%{year: 2017, month: 6, day: 1}, [:yaer, :month]) # same misspelling
assert actual == expected  # passes even though we're in 2018
```

I would write that as 

```elixir
assert %{yaer: 2018, month: 6} = DateTime.utc_now
```

Which would fail for the correct reason: the misspelling. Fixing this would reveal the test failure that 5 != 6.

You'd have similar failures if your tests were correctly spelled, but your data was misspelled or missing keys.

I'm not opposed to the idea of a `Map.take!` that raises KeyError. I don't have a use case handy, though.

- Martin

To view this discussion on the web visit https://groups.google.com/d/msgid/elixir-lang-core/CAHtu5s8pHEqQEnYcNAXbvSnRRXeciXrQSwmk1OoXZ%2Bn1PyYnUA%40mail.gmail.com.

[Martin Svalin]

(edited to have the year fail instead of the month, but didn't edit enough… month: 5, 2017 != 2018… you get the gist. Sorry)

[Tallys Martins]

Thanks for the replies!

Em sex, 25 de mai de 2018 às 06:00, Martin Svalin <martin...@gmail.com> escreveu:

So essentially you are concerned with this scenario?

```elixir
actual = Map.take(DateTime.utc_now, [:yaer, :month]) # note misspelling
expected = Map.take(%{year: 2017, month: 6, day: 1}, [:yaer, :month]) # same misspelling
assert actual == expected  # passes even though we're in 2018
```

Yes, something like this!

I would write that as 

```elixir
assert %{yaer: 2018, month: 6} = DateTime.utc_now
```

You are right, this is a better approach. No more need for Map.Take/2. I realized that my scenario was similar to testing Repo.one, I took a look at ecto's codebase and this is exactly what they did in their tests. Thanks!

Which would fail for the correct reason: the misspelling. Fixing this would reveal the test failure that 5 != 6.

You'd have similar failures if your tests were correctly spelled, but your data was misspelled or missing keys.

I'm not opposed to the idea of a `Map.take!` that raises KeyError. I don't have a use case handy, though.

We can wait someone bring real a use case. Thanks for your reply.

Tallys

To view this discussion on the web visit https://groups.google.com/d/msgid/elixir-lang-core/CAAHw6CJkR75u2DVJcj%2BHn47Sx0PG6Acz-KkbvpkxFE9DSATPdg%40mail.gmail.com.