Style/FormatString cop breaks code

[michiomochi]

Rubocop auto-corrector breaks code by Style/FormatString cop.

Before code is

def hoge
  "%08x-%04x-%04x-%04x-%04x%08x" % ary
end

Execute auto-correct

> bundle exec rubocop -a

After code is

def hoge
  format("%08x-%04x-%04x-%04x-%04x%08x", ary)
end

Expected behavior

No effort.

Actual behavior

Fix to unexpected code.

RuboCop version

> bundle exec rubocop -v
0.71.0

[pocke]

String#% can receive an array, but format cannot receive it.

"%02x" % 1 # => "01"
"%02x" % [1] # => "01"
"%02x-%02x" % [1, 2] # => "01-02"

format("%02x", 1)# => "01"
format("%02x", [1])# => TypeError
format("%02x-%02x", [1, 2]) # => TypeError
format("%02x-%02x", 1, 2) # => "01-02"

I think we should mark the auto-correction as unsafe.

[tejasbubane]

@pocke Is is a good idea to not register offense when second argument to String#% is a variable?

[pocke]

@pocke Is is a good idea to not register offense when second argument to String#% is a variable?

It makes sense, but I prefer changing auto-correct safety.
Because if the cop does not register offense, user has no way to know format method.

---

By the way, I got other ideas.

First, ignore only auto-correction when the argument is a variable.
The cop registers an offesne for String% with a variable but does not apply auto-correction.

Second, I guess we can use splat operator for a variable, like format("str", *variable).
I think it is a better way, but I'm not sure splat operator does not change meaning anytime.

[tejasbubane]

I guess we can use splat operator for a variable, like format("str", *variable)

This will be an issue if variable is a hash. First approach looks better.