Cop idea: numbered variable names (user_1 vs user1)

[mockdeep]

We've got a lot of places in our code where we use a couple of records, something like:

user_1 = blah
user_2 = bloo

Sometimes people number them instead, user1 and user2. Would be nice to have a rule to enforce consistent numbering of variable names, so always with _ or always without.

[Drenmi]

Would it also be useful to have a configuration option to disallow numbered variables? I can see how some teams might want this to encourage better design (i.e. using a data structure like Array) and naming conventions.

[mockdeep]

I can't say that it would be something that we would use. We do a lot of testing around things like one-off scripts and database scopes to ensure that only specific records are returned or modified. So something like:

user_1 = create(:user, ...)
user_2 = create(:user, ...)
expect do
  SomeThing.call
end.to change { user_1.reload.some_attribute }
  .and not_change { user_2.reload.some_attribute }

[mikegee]

I agree with both of you. I think this new cop would have some value and disallowing numbered variables might be the best default config.

[Drenmi]

There is, of course, a small chance for false positives. Consider, for example, a method call base64 within a class, which would get flagged.

[mikegee]

What if the cop specifically looked for more than one local variable with incrementing suffix numbers? Perhaps starting at 0 or 1?

I think that would catch the problem and not miss any violations.

[mockdeep]

It almost sounds like there are multiple rules being discussed here:

1. (my suggestion) variable names with a number should all be consistently name_1 or name1
2. variable names should not be named with a number
3. variable names with numbers should be sequential starting from 1 (or 0)
4. (another I just thought of) variable names shouldn't have a number if there's only one of them

[Drenmi]

@mockdeep: These would almost certainly go in the same cop (with configuration options), which is why we're discussing them all in the same thread. 😀 Checking adjacent variables is a bit more involved, but might be possible if the rules are well defined.