fix: handle big numbers when calculating nice range/numbers

[bernardobelchior]

Problem

Currently, graphic throws an error when calculating the step in linearNiceRange when using big numbers (e.g., 999999999999999999999999999999).

This happens because the step is being stored in an int which overflows and becomes negative.

In _tickIncrement we calculate the dart_math.log(step), which returns NaN because step is negative, which crashes the algorithm.

A similar thing happens in linearNiceNumbers.

Solution

By using a double instead of an int for storing the step, we are sure it never overflows. It may lose some precision, which I'd argue is ok, since this is a visualization tool and eyes can't see that precisely, but it doesn't throw an exception due to an overflow.

I'm open to other solutions as well, but this one seems to fix the aforementioned issue and does minimal impact.

Let me know what you think 😄

[bernardobelchior]

Hey, @entronad! Let me know what you think about this 😄

[entronad]

Sorry missed this PR.

I believe your solution is proper, because this algorithm is from a JavaScript lib (AntV). we know that all numbers in JavaScript are doubles. I haven't noticed that 10 in Dart will limit pow to int scope.

Only one thing, I suggest, that 10.0 is better than 10.toDouble() to indicate a double. Const literals seem more concise and efficient.

[bernardobelchior]

Updated 😄

[bernardobelchior]

Thanks, @entronad! 😄 Could we create a new release with this change?

[entronad]

I would wait a few days to see if there are other changes to collect in one release.

[entronad]

Published in v0.10.5