Re: [boost] [filesystem] becoming worse than the standard?

On 8/15/22 17:10, Vinnie Falco wrote:

On Mon, Aug 15, 2022 at 3:37 AM Andrey Semashev via Boost wrote:

...

Could you describe why you need to construct/assign/append to paths from arbitrary containers and why you cannot use string types or iterators?

Yeah, and from your description it sounds like maybe it is just a user error on my part or a bug. The container in question is boost::urls::pct_encoded_view:

https://github.com/CPPAlliance/url/blob/cba301383791f67c31b3ccd52a401849f7cf...

This used to work:

pct_encoded_view v; boost::filesystem::path p;

p.append( v );

What should happen here is that a path segment will be appended to p formed by applying percent-decoding to the string referenced by 'v' (the percent-decoding happens in the iterator).

With std::filesystem::path that is what would happen, though pct_encoded_view::operator std::string(). With boost::filesystem::path, it will construct the path from a string that is obtained from [v.begin(), v.end()) range (which, I presume, will not be percent-encoded). Since the recent change, it will also give you a deprecated warning. The correct code is either use v.to_string() or explicitly use v.begin()/v.end(), depending on what you want. BTW, duplicate "as a" here: https://github.com/CPPAlliance/url/blob/cba301383791f67c31b3ccd52a401849f7cf...