Re: [boost] Re: Re: New design proposal for boost::filesystem

From: Carlo Wood (carlo_at_[hidden])
Date: 2004-08-23 10:27:05

On Mon, Aug 23, 2004 at 09:47:52AM -0500, Patrick Bennett wrote:
> Correct (kind of), but I'd far prefer that std::string be used than for 
> some completely new type to be defined.

std::vector<char> comes to mind.

> For users of boost::filesystem, I can't personally think of a time when 
> a user would need to iterate the paths or files a character at a time.

Bzzzzt. Everything is always, somewhere, needed :p
I can't think of it either - but the number of times that I thought
something wasn't needed, or when I needed something and the original
author wrote that he thought it would never be needed, are countless.

> Because of UTF-8's nature, even if a user were to search for something 
> like '/', it would still work for find's, [], etc.  UTF-8 maps to 
> std::string extremely well. 
> I think there is also a fair amount of precendents already set for using 
> UTF-8 internally using std::string as the storage mechanism.

People associate a std::string with something printable and will be tempted
to use std::string::c_str() to pass the 'string' to some function.
I think that using a std::string of binary data is not a good idea when
it can contain '\0' in the middle.

> UTF-8 strings don't contain embedded nul's (std::string still works for 
> that though), ASCII characters remains ASCII characters, and you can 
> tell if you're in the middle of a multi-byte sequence.

Ah - I retract my objection ;).
If UTF-8 is garanteed not to have nul's embedded then there is no
reason not to use std::string for it imho.

-- 
Carlo Wood <carlo_at_[hidden]>