Improved Number.prototype.toString for radix other than 10

[chqrlie]

• fix the conversions for integers and exact fractions
• approximate approach for other cases.
• bypass floating point conversions for JS_TAG_INT values
• avoid divisions for base 10 integer conversions
• fixes Improve number->string conversion #242

[chqrlie]

I purposely did not remove the radix argument from js_dtoa and js_dtoa1 because we should change the implementation to avoid relying on the C library snprintf which has some issues on Windows.

[bnoordhuis]

You will want to update v8.txt in the top-level directory but otherwise nice work.

[chqrlie]

I updated v8.txt. Why does v8.sh tee the output to the console? Just showing the diff would suffice and avoid verbose listings.

[bnoordhuis]

Why does v8.sh tee the output to the console?

It made it easier for me to debug things.

[bnoordhuis]

For my understanding, this is sort of incorrect when radix > 10 but works okay in practice? I tried to deduce if you could get an accidental match in e.g. base-18 or base-36 (because then nine is 'A' or 'S') but I'm not sure.

I suppose it could also lead to somewhat unexpected results on EBCDIC systems but 🤷

[chqrlie]

This is incorrect indeed for bases larger than 10. I shall post a fix. The code should be

    char nine = digits[radix - 1];

I am using an array for the digit characters, so EBCDIC should work fine :)

[bnoordhuis]

LGTM. Nice work.

[chqrlie]

LGTM. Nice work.

Thank you. For completeness, this implementation is partial, it is not fully precise for the last fractional digit or two because of precision issues. I will post a multi-precision version when I get it done.