RFC: Invalid Unicode Handling

[kbknapp]

This issue is to discuss how invalid unicode should be handled in 2.x and 1.x

Currently

• Consumers have the option to panic! on invalid unicode (the default)
• Consumers can get an Err on invalid unicode using the *_safe() methods
• Consumers can get a lossy value where invalid unicode is replaced with U+FFFD using *_lossy() methods

Problems

• Users cannot get values with invalid unicode (which is allowed on Unix systems for file names, paths, etc.)
• All the lossy, safe, and regular version of get_matches is somewhat messy. Granted in practice it's not an issue because you only use one version of that method...but for API space it looks a mess.

Future

One idea is to store the values internally in the ArgMatches struct as an OsString and by default give a &str, but allows the users the option to get an OsStr instead. This is the opt-in to invalid unicode. I am a firm believer the default should be strict unicode, but we should allows users to handle invalid unicode if they so choose.

Questions

How does this affect Windows? Or does it even affect Windows? If yes, should we use a #[cfg(not(windows))], or similar?

---

Question, comments, suggestions?

[sru]

👍

[kbknapp]

Closed with 2x

[remram44]

This is awesome! The ability to accept any filename is great, and the API is cool, and I love you all.