Make docs of String to number funs a little more clear about digits

[mahozad]

Java Integer::parseInt and Kotlin String.toInt and other String to non-float functions accept non-ASCII (aka non-English aka non-Western-Arabic) digits in a String too. For example, Eastern Arabic digits:

println("12۳".toByte())        // 123
println("12۳۴".toShort())      // 1234
println("12۳۴۵".toIntOrNull()) // 12345
println("12۳۴۵6".toUInt())     // 123456
println("12۳۴۵67".toLong())    // 1234567

This is a good and expected behavior in my opinion. I don't know if Kotlin implementation of these functions for non-JVM targets also behave the same way.

Alos, Java Character::isDigit and Kotlin Char.isDigit return true for non-ASCII numerals too. Same for "\\p{javaDigit}" or "\\p{Nd}" regexes in both Java and Kotlin. See this code snippet. They all match a few hundred types of digit chars in JDK 24.

I have updated the docs to hopefully make it just a bit more clear that 0-9 are just one set of chars that are valid as digits. I don't know whether all those functions updated in this commit (especially non-JVM or expect ones) accept non-English digits too. If not, then my update for docs of those should be reverted and I hope their implementations is updated to make them behave the same way as on JVM implementation.

Another thing is, it seems most of those functions lack unit tests for non-ASCII digits to clarify their behavior. Found a weird bug:

println("123۴۵۶.789۳۲۱".toBigDecimal())       // 123456.789321
// Exactly the same string as above
println("123۴۵۶.789۳۲۱".toBigDecimalOrNull()) // null

And, BigDecimal allows non-ASCII digits in the string, but String.toDouble and variants throw exception.

Related:

• Convert String to double in Java
• String.toDouble does not parse non-English digits

[fzhinkin]

With toBigDecimal, it's clearly a bug.
Filed https://youtrack.jetbrains.com/issue/KT-81043