> Perhaps … a third mode is useful? I know, it makes it more complicated, but I think we need it:
NO.
A UriBuilder.quote(String) static method can %-escape all unreserved chars, ensuring the input provides “content” – not “structure” – to a URI.
A mode that does not %-escape percent signs allows the caller to build every possible valid URI.
The remaining modes add too little to warrant any complications.
If a caller really wants to escape percent chars as well, just change:
UriBuilder.fromPath(s);
to
UriBuilder.fromPath(s.replaceAll(“%”, “%25”));
If a caller really wants an exception if non-uri chars are present:
if (!s.matches(“[-._~a-zA-Z0-9:/?#[\\]@!$&’()*+,;=%]*”))
throw new IllegalArgumentException();
Hence, if the API just supports mode 3, a caller can still implement modes 1 & 2 without much code – if they really need to, which I believe will be extremely rare.
James Manger
________________________________
From: Stephan.Koops_at_web.de [mailto:Stephan.Koops_at_web.de]
Sent: Thursday, 17 July 2008 7:20 PM
To: users_at_jsr311.dev.java.net
Hi James, hi Marc,
Perhaps we a third mode is useful? I know, it makes it more complicated, but I think we need it:
1. no encode, only check (same as current false)
2. encode everything, including '%' (what Marc intended for true)
3. encode everything excluding valid percent encoded values (suggestion of James and also my understanding of true)
It could be explicit documented in the javadoc, that the both last are equivalent, if no percent char is available. This reduces fear of complexity by app developers.
Marc, if I remeber right, you talked about a next API version. When should it come? Is it right, that this should not change?