RE: [boost] Filesystem Library: Formal Review

From: Keith Burton (kb_at_[hidden])
Date: 2002-09-20 09:10:45

-----Original Message-----
From: boost-bounces_at_[hidden]
[mailto:boost-bounces_at_[hidden]] On Behalf Of Vladimir Prus
Sent: 20 September 2002 12:54
To: boost_at_[hidden]
Subject: Re: [boost] Filesystem Library: Formal Review

>Keith Burton wrote:
>> If one uses
>> 
>> 	string relative_path("libs/locale");
>> 	
>> that is a relative path is remains a string until converted to a ( 
>> absolute ) path. The conversion to absolute can be relative to any 
>> absolute path.
>
>    path x(...) ;
>    string relative_path("libs/locale");
>    x << relative_path

>How will the third line work? Currently, relative_path will be
converted 
>to path, which will make it absolute in your scheme. Also, joining too 
>relative paths will become less convenient.

Works correctly if operator << ( const string & ) is defined
appropiately

>    string relative_path("libs/locale");
>    string relative_path2("ru_RU.KOI8-R");

>    relative_path += "/" + relative_path2 ;

The requirement to compose / decompose relative paths is a limitation of
my approach.  Is this a symptom of a real problem or a symptom of the
conventional view that relative and absolute paths are the same thing?
I obviously believe the latter and the problems caused by treating
relative and absolute paths as the same thing out way the need for a
change of mind set.

>> 
>> With regard to the UI , the absolute path may be unpleasant but it
may 
>> also be the saviour as it explains why the operation failed if the 
>> current directory was not as expected when the relative path was 
>> converted to absolute.

>But the programmer should be able to choose, I think.

Agreed but as the programmer is in control and need not use the text of
default message I do not see a problem.