Re: Boost.URL: Something Wicked This Way Comes!

From: Gavin Lambert (boost_at_[hidden])
Date: 2021-10-14 01:52:47

On 14/10/2021 14:28, Vinnie Falco wrote:
> Thinking out loud here, it seems that the segments container models the path as
> 
>      vector< string >
> 
> When it actually needs to model the container as
> 
>     struct
>     {
>         bool leading_slash;
>         vector< string >
>      };
> 
> Should I just add this function
> 
>      void segments::set_absolute( bool path_should_be_absolute );
> 
> and keep
> 
>      bool segments::is_absolute() const noexcept;
> 
> ?

I guess the answer to that depends on how you want 
clear-and-multiple-push_back to behave.  Does that preserve or alter 
absoluteness of the path?

Most of my experience with manipulating URIs (other than just using 
hard-coded strings) is via the .NET Uri class, and FWIW with those:

1. Instances are immutable; the only way to get a new Uri is to parse a 
new one from scratch or resolve a relative one from a base.  (Both the 
base and the 'relative' can actually be either absolute or relative.)

2. Its 'Segments' property does include trailing slashes in the 
components -- so { "/", "foo/", "bar/", "index.html" } and is also 
read-only.  (And also doesn't support relative URIs, though I don't know 
why not as they wouldn't be ambiguous.)

I'm not saying these are necessarily good choices for your 
implementation as well; just as a point of comparison.