fix negative-number engineering formatting #33
Conversation
|
@timholy I have no idea how to split the test set gracefully for both Ryu and Grisu versions. The problem is that Grisu code was not without its issues hence these series of PRs. Perhaps it should just be declared obsolete and closed for further modification and testing. If new features to be added it should all go to ryu branch? In case of complaints from package mainteners, they could be asked to pin their dependency on the older Grisu only version of the code? |
| @@ -36,7 +36,7 @@ end | |||
| @test Showoff.format_fixed_scientific(0.012345678, 4, true) == "12.346×10⁻³" | |||
| @test Showoff.format_fixed_scientific(0.012345678, 4, false) == "1.2346×10⁻²" | |||
| @test Showoff.format_fixed_scientific(-10.0, 4, false) == "-1.0000×10¹" | |||
| @test Showoff.format_fixed_scientific(-10.0, 4, false) == "-1.0000×10¹" | |||
| @test Showoff.format_fixed_scientific(-10.0, 4, true) == "-10.000×10⁰" | |||
There was a problem hiding this comment.
It was same with line 38, thats why I just modified this line in accordance with the bug you found
| @@ -103,11 +103,12 @@ function get_engineering_string(x::AbstractFloat, precision::Integer) | |||
| end | |||
|
|
|||
| buf = IOBuffer() | |||
| negative_base_compensation = base_digits[1] == '-' ? 1 : 0 | |||
There was a problem hiding this comment.
I could not find a better solution to compensate for an extra index in case of negative numbers,
|
Awesome. I agree with splitting the tests since, as you point out, some of the Grisu results just seem wrong (which makes it weird to enforce them with a test). Thanks for making this more self-consistent! |
Tests passing now. The problem was with putting the floating point in case of negative numbers