In UNICODE, there are 60 or more sequences, each containing ten code points, where unicode.IsDigit() returns true, but the numeric conversions in package strconv only work on digits in the ASCII range. The udigit package provides some analogous routines, most notably Atoi(), that work with all UNICODE digits, and utilty routines for transforming digit sequences embedded in UTF-8 strings. It also provides a version of unicode.IsDigit() that will be faster on some mixes of input, at the expense of some manual maintenance required when support for newer versions of UNICODE are incorporated into the standard Go packages.

go get github.com/tom-harvey/udigit

Only ASCII '+' and '-' characters are supported in udigits.Atoi().

Floating point conversions are not supported because even I think that is getting a little ridiculous.

This project was undertaken mainly for my educational purposes, with only a cursory amount of research into the problem space. If there is another facility for doing what these routines purport to do, or some reason why these routines are utterly unneeded, I'd be interested in hearing.

The source files are full of magic numbers which should be generated from UNICODE reference documents, but aren't. On the other hand, I believe the provided test routines are exhaustive enough to demonstrate when the code is completely consistent with the behavior of unicode.IsDigit(), and when it needs some manual maintenance to regain compatibility.