search

radzenhq / radzen-blazor

Radzen Blazor is a set of 90+ free native Blazor UI components packed with DataGrid, Scheduler, Charts and robust theming including Material design and FluentUI.
https://www.radzen.com
MIT License
3.52k stars 785 forks source link

RadzenNumeric with TValue set to double does not correctly increment with the step up or step down buttons #1667

Closed Daeymon closed 1 month ago

Daeymon commented 1 month ago

Step up or step down on RadzenNumeric is adding additional tiny decimals to the number on top of the designated step amount when TValue is set to "double".

To reproduce add the following to a default Radzen enabled Blazor project and click the step up or step down button a few times. The additional decimal doesn't get added on every step.

<RadzenNumeric TValue=double? Value=0.1 Step="0.1" />

image ...press step up twice and then ..... image

This started happened after NuGet version 4.32.9 and is still happening in 5.1.2.

enchev commented 1 month ago

Looks like it's caused by this pull request: https://github.com/radzenhq/radzen-blazor/commit/98967410cb38d115e9a4bb0e1db4ac5aa22ae151 I'll check how to fix it.

UPDATE: Before this change the Numeric component calculated new value with step only using decimal type exactly to avoid such problems. For example the result of (double)0.2 + (double)0.1 will be 0.30000000000000004, now the Numeric component will use TValue to perform the calculation and as a result we have floating-point value loss of precision. More info can be found here: https://stackoverflow.com/questions/5903003/addition-of-double-values-inconsistent

Cosmatevs commented 1 month ago

I think that there were two solutions to this problem:

  1. Use decimal instead of double or float if you need decimal precision. double and float types are meant to be approximations.
  2. Use the Format attribute to show only some decimal places, and round the value when needed.

@enchev Converting floats and doubles to decimals will certainly work in most cases. However, decimal is not able to represent the full range of float and double types. For this reason, you may lose precision for numbers really close to zero and get conversion errors for large numbers. I'm not sure if this is intended – I was just concerned after seeing your commit.

enchev commented 1 month ago

Feel free to propose better solution.

Cosmatevs commented 1 month ago

@enchev I've already suggested two user-side solutions 😅 I think that it wasn't a problem with the Radzen component, it was just unrealistic expectations about the double and float types. Therefore, in my opinion, the responsibility for using appropriate types or solutions should lie with the developers using Radzen components, not with the developers of the Radzen. But that's just my opinion – perhaps there are better reasons to actually convert floating types to decimal.