.ToString("0.000000") fails

[IvoTops]

Hi Takuma,

Testing your library at the moment because I needed a higher precision float in c# which is faster than the decimal. Looking good so far, but somehow I get a System.FormatException doing

ddouble x = 1;
var s= x.ToString("0.0");

while

ddouble x = 1;
var s= x.ToString();

works.

Replacing all the decimal math with ddouble worked fine, but all the logging broke ;-)

If you have no time just let me know and I will look into it myself.

Kind regards,

Ivo Tops

[tk-yoshimura]

IvoTops, I am glad you are using my library.

DoubleDouble is designed with compatibility with MultiPrecision in mind.
Therefore, the implementation priority of formats other than exponential notation is low.

DoubleDouble/DoubleDouble/DDouble/DDouble_tostr.cs

Lines 57 to 66 in 7df7478

	public string ToString([AllowNull] string format, [AllowNull] IFormatProvider formatProvider) {
	if (string.IsNullOrWhiteSpace(format)) {
	return ToString();
	}
	format = format.Trim();
	if (format.Length < 2 || (format[0] != 'e' && format[0] != 'E')) {
	throw new FormatException(format);
	}

Especially with fixed-point numbers, the string expressing a huge number becomes very long, so it is not implemented from the standpoint of practicality.

double val = 1e+300;
string s = val.ToString("0.0");

Not all formats supported by Double are supported. Sorry.

[IvoTops]

I understand the priority is low, but interested people might become hesitant if stuff that did work no longer works after changing from decimal or double to ddouble. And my own usecase is keeping the lower digits correct and rouding errors away while displaying the first 10 or so decimals for logging all over the place ;-) If I implemented support for this format and create a pull request will you merge it?

…

On Mon, 26 May 2025 at 13:48, T.Yoshimura ***@***.***> wrote: *tk-yoshimura* left a comment (tk-yoshimura/DoubleDouble#17) <#17 (comment)> IvoTops, I am glad you are using my library. DoubleDouble is designed with compatibility with MultiPrecision <https://github.com/tk-yoshimura/MultiPrecision> in mind. Therefore, the implementation priority of formats other than exponential notation is low. https://github.com/tk-yoshimura/DoubleDouble/blob/7df74786ae1afce5c2165083ad138a62e1e062f2/DoubleDouble/DDouble/DDouble_tostr.cs#L57-L66 Especially with fixed-point numbers, the string expressing a huge number becomes very long, so it is not implemented from the standpoint of practicality. double val = 1e+300;string s = val.ToString("0.0"); Not all formats supported by Double are supported. Sorry. — Reply to this email directly, view it on GitHub <#17 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA5YHTSWZDNPT624N2JLKL33AL5RTAVCNFSM6AAAAAB547BRGOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDSMBZGQ2TCOJRGA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[tk-yoshimura]

Thank you very much!
Once I confirm that the exhaustive test cases for multiple fixed-point digits for infinity, huge numbers, minute numbers, NaN, and denormalized numbers work as well as the Double's format, I will reflect your request!

https://learn.microsoft.com/en-us/dotnet/standard/base-types/standard-numeric-format-strings

May I modify and publish the code you provided for this issue in the future without your permission?
This is the first time I have received a pull request for a private work, so I am unfamiliar with the appropriate way to respond...

[IvoTops]

Ok. Will put some time in it and create a pull request when finished. And yes, anything you pull from me you can use, re-publish etc. without my permission. I have done this a few times with other repo's and that's how I see it. I use your library, which is nice and if I can attribute a little while fixing some issue I have with it that feels as the appropriate way to do that. Probably covered in the opensource license you chose somewhere? No idea ;-)

…

On Tue, 27 May 2025 at 16:35, T.Yoshimura ***@***.***> wrote: *tk-yoshimura* left a comment (tk-yoshimura/DoubleDouble#17) <#17 (comment)> Thank you very much! Once I confirm that the exhaustive test cases for multiple fixed-point digits for infinity, huge numbers, minute numbers, NaN, and denormalized numbers work as well as the Double's format, I will reflect your request! https://learn.microsoft.com/en-us/dotnet/standard/base-types/standard-numeric-format-strings May I modify and publish the code you provided for this issue in the future without your permission? This is the first time I have received a pull request for a private work, so I am unfamiliar with the appropriate way to respond... — Reply to this email directly, view it on GitHub <#17 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA5YHTUP6F34DZ7W7YUMEHD3ARZ4DAVCNFSM6AAAAAB547BRGOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDSMJSG43DAMJWHA> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[IvoTops]

Instead of trying to replicate a lot of tricky formats we could also simply cast to double and use the original formatting code. That would mean that the first two formatexceptions would be replaced like below. That will keep all the special formats that work for the large numbers and still support all of the 'regular' formats, just on a lower precision....

public string ToString([AllowNull] string format, [AllowNull] IFormatProvider formatProvider) {
if (string.IsNullOrWhiteSpace(format)) {
return ToString();
}

  format = format.Trim();

  if (format.Length < 2 || (format[0] != 'e' && format[0] != 'E')) {
      return ((double)this).ToString(format,formatProvider);
  }

  if (!(int.TryParse(format[1..], NumberStyles.Integer, CultureInfo.InvariantCulture, out int digits)) || digits <= 1) {
     return ((double)this).ToString(format, formatProvider);
  }

[tk-yoshimura]

Thank you for your ideas.

Yours implementation is contrary to the concept of a high precision calculation library.
It should be explicitly cast to double and ToString at the user's end.

[IvoTops]

I agree with you. Wanted to check and therefore did not make it a pull request just yet. Will see if I can come up with a good way to have parity with all the format masks that double and decimal support. Ivo

…

On Tue, 3 Jun 2025, 06:34 T.Yoshimura, ***@***.***> wrote: *tk-yoshimura* left a comment (tk-yoshimura/DoubleDouble#17) <#17 (comment)> Thank you for your ideas. Yours implementation is contrary to the concept of a high precision calculation library. It should be explicitly cast to double and ToString at the user's end. — Reply to this email directly, view it on GitHub <#17 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA5YHTRARMNDOQWMDLJG6LD3BUQXBAVCNFSM6AAAAAB547BRGOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDSMZTGM3TKOJZG4> . You are receiving this because you authored the thread.Message ID: ***@***.***>