Float-to-string conversion while controlling decimal place count

[Enyium]

It's a basic UI necessity to display floats in a controlled format. Currently, you can only let your float be displayed with inappropriate accuracy (like 5.9993034 in stead of 6.0) or implement it yourself. Trying to round yourself in the float domain, then accepting implicit conversion to string can lead to base-2-related problems like 3.299999999999 or omission of an intended .0 suffix. Slint should provide such functionality. It could copy JavaScript's toFixed() and maybe also toPrecision(), which return strings.

Ideally, Slint would be aware of the current locale (from the translation context?) and use the appropriate decimal separator - also with regular auto-conversions.

[ogoffart]

Thanks for filing an issue.

• I think indeed there is value in having number.to-fixed(xx) and number.to-precision(xx)
• In addition to that we probably do need formatting ability in the @tr macro, but that is a different issue.

As a workaround, toPrecision can be emulated with something like round(foo * 100) / 100 but it's just a workaround. As another workaround, it is possible to implement in native code in a global callback.

[ogoffart]

Added as good first issue now that the number.sqrt() and similar are merged, it can just be done similarly.

What needs to be done is:

• Add "to-fixed" and "to-precision" in the lookup for NumberExpression https://github.com/slint-ui/slint/blob/8a0df2b3a5b6e74e4c52f8177d20d33c7ab5aa54/internal/compiler/lookup.rs#L1124
• Add a values to the BuiltinFunction enum (BuiltinFunction::ToFixed and BuiltinFunction::ToPrecision) https://github.com/slint-ui/slint/blob/8a0df2b3a5b6e74e4c52f8177d20d33c7ab5aa54/internal/compiler/expression_tree.rs#L21
• Do the runtime for each of them (use BuiltinFunction::StringToFloat or others as examples)
• documentation/tests

[marcothaller]

I was also looking for a convenient solution to convert floats to a string on the UI with a fixed number of decimals and potentially a minimum width.

I looked into the slint code Olivier pointed to and attempted to implement a solution utilizing std::fmt. std::fmt's concept of precision is this:

For floating-point types, this indicates how many digits after the decimal point should be printed.

So specifying the digits to appear after the decimal points (like with toFixed()) can easily be done with std::fmt but specifying the number of significant digits (like with toPrecision()) is not supported out-of-the-box with std::fmt IMO.

My main interest was in providing a proposal for setting the number of digits after the decimal point and for setting the width. The code can be found here

[ogoffart]

@marcothaller Thanks. I added a few comment to the commit.

On a unrelated note, i wonder if the function should follow the locale or not.