Question

To check a user input english name is valid, I would usually match the input against regular expression such as [A-Za-z]. But how can I do this if multi-language(like Chinese, Japanese etc.) support is required with utf8 encoding?

Answer 1

You can approximate the Unicode derived property \p{Alphabetic} pretty succintly with [\pL\pM\p{Nl}] if your language doensn’t support a proper Alphabetic property directly.

Don’t use Java’s \p{Alpha}, because that’s ASCII-only.

But then you’ll notice that you’ve failed to account for dashes (\p{Pd} or DashPunctuation works, but that does not include most of the hyphens!), apostrophes (usually but not always one of U+27, U+2BC, U+2019, or U+FF07), comma, or full stop/period.

You probably had better include \p{Pc} ConnectorPunctuation, just in case.

If you have the Unicode derived property \p{Diacritic}, you should use that, too, because it includes things like the mid-dot needed for geminated L’s in Catalan and the non-combining forms of diacritic marks which people sometimes use.

But then you’ll find people who use ordinal numbers in their names in ways that \p{Nl} (LetterNumber) doesn’t accomodate, so you throw \p{Nd} (DecimalNumber) or even all of \pN (Number) into the mix.

Then you realize that Asian names often require the use of ZWJ or ZWNJ to be written correctly in their scripts, so then you have to add U+200D and U+200C to the mix, which are both \p{Cf} (Format) characters and indeed also JoinControl ones.

By the time you’re done looking up the various Unicode properties for the various and many exotic characters that keep cropping up — or when you think you’re done, rather — you’re almost certain to conclude that you would do a much better job at this if you simply allowed them to use whatever Unicode characters for their name that they wish, as the link Tim cites advises. Yes, you’ll get a few jokers putting in things like “əɯɐuʇƨɐ⅂ əɯɐuʇƨɹᴉℲ”, but that just goes with the territory, and you can’t preclude silly names in any reasonable way.