From: Gennadiy Rozental (gennadiy.rozental_at_[hidden])
Date: 2006-04-24 11:10:24
"Rene Rivera" <grafik.list_at_[hidden]> wrote in message
>> Why would I want to do manual editing if there is a program that does
> Because some times there aren't such programs (I said usually above).
This scenario does not apply. We are talking about files that are both human
and application editable.
> And more often you can't use those programs, for many reasons, like not
> having access to the UI system the run on.
So are you going to change application options and don't even run it to test
that it works?
>> Why would I choose to subject myself to the possibility of making error,
>> the routine work of satisfying particular file format?
> It has nothing to do with choice, it's the current situation in many
> cases. And if you are just starting on a *new* program would you write
> the configuration GUI before you write your program?
I would use human-editable only config at first. And switched to
application-editable only once it's ready.
>> Not saying that it
>> complicate my application development since now it's need to handle both
>> proper format and anything I will manage to type in file.
> But not if such a library as the Property Tree library which would do
> that for you.
I doubt it will be flexible enough. And also any such solution would be more
complecated (eventually - error prone) in comparison with one that doesnt
need to handle human input.
>> If I edit config
>> file - that mean I don't have a tool to handle this for me. I am not
>> this is completely impossible. But shouldn't be that wide spread need.
> OK, I can only surmise that you don't visit the Unix world often.
These days I develop exclusively on *nix systems. I don't have any GUI
though. So all my config files are human editable.
>>> So I must ask; How can you think this domain is limited?
>> In a big picture of program runtime parameter handling - it's just a
>> friction IMO.
> Perhaps, but vector is a *very* small fraction of the data structure
> doamin. Yet we still have such a utility.
I think vector should be insulted by your assesment ;) I may disagree also.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk