Fix Bug#13833 String#scanf("%a") incorrectly requires a sign on the (binary) exponent

[tarotaro0]

Issue

• String#scanf("%a") incorrectly requires a sign on the (binary) exponent.
  https://bugs.ruby-lang.org/issues/13833

Implementation

• When we use String#scanf("%a"), it doesn't require a sign on the (binary) exponent similar to Float class.

Before :

>> "0x1p+9".scanf("%a")
=> [512.0]
>> "0x1p9".scanf("%a")
=> [0.0]  

After :

>> "0x1p+9".scanf("%a")
=> [512.0]
>> "0x1p9".scanf("%a")
=> [512.0]  

[rhenium]

It looks like (decimal-)exponent-part shares the same bug. Would you mind fixing this as well?

$ ruby -rscanf -e'p "1.e+10".scanf("%a")'
[10000000000.0]
$ ruby -rscanf -e'p "1.e10".scanf("%a")'
[1.0]

[tarotaro0]

I would fix (decimal-) exponent-part.

$ ruby -rscanf -e'p "1.0e+10".scanf("%a")'
[10000000000.0]
$ ruby -rscanf -e'p "1.0e10".scanf("%a")'
[10000000000.0]

but it failed this test :

TestIOScanf#test_"12.e23 45 string"("%f %*d %s") [/ruby/test/scanf/test_scanf.rb:303]:
<[12.0, "string"]> expected but was
<[1.2e+24, "string"]>.

This failure happens in /test/scanf/test_scanf.rb:225 :

[ "%f %*d %s", "12.e23 45 string", ["12.e23".to_f, "string"] ],

I think this cause is String#to_f :

irb(main):001:0> "12.e23".to_f
=> 12.0
irb(main):002:0> "12.0e23".to_f
=> 1.2e+24

I think this behavior is not expected, so I fixed test_scanf.rb as below :

Before :

[ "%f %*d %s", "12.e23 45 string", ["12.e23".to_f, "string"] ]

After :

[ "%f %*d %s", "12.0e23 45 string", ["12.0e23".to_f, "string"] ]

[tarotaro0]

I discussed with @matz and ruby committers directly about this problem.

This includes two problems.

1. String#scanf("%a") (decimal-) exponent-part incorrectly requires a sign on the (binary) exponent.

like below :

$ ruby -rscanf -e'p "1.e+10".scanf("%a")'
[10000000000.0]
$ ruby -rscanf -e'p "1.e10".scanf("%a")'
[1.0]

2. String#to_f parses .e~~ as a method.

like below :

irb(main):001:0> "12.e23".to_f
=> 12.0
irb(main):002:0> "12.0e23".to_f
=> 1.2e+24

I think this problem should be created as new issues.

After the second problem is fixed, then the first problem should be fixed I think.

So I revert the change about only (decimal-) exponent-part, and I'll create new issues after merge this PR.