[gdal-dev] Using a "standard" argument parser for command line utilities?

[Andrew C Aitchison]

On Fri, 8 Mar 2024, Even Rouault via gdal-dev wrote:

> Hi,
>
> Our command line C++ utilities use ad-hoc manual parsing, which means that:
>
> -  the usage message must be manually composed,
>
> -  you must take care to check that there are enough remaining arguments for 
> the ones that take value to avoid out-of-bounds accesses (tests like argc + 1 
> < argn)
>
> - detection for duplicated arguments when only a single occurrence is allowed 
> must be manually done, nd thus is often not done, confusing users, cf 
> https://github.com/OSGeo/gdal/issues/9415
>
> - etc.
>
> I've come across https://github.com/p-ranav/argparse which fit all my 
> requirements at first sight: compatible with our C++ requirements (C++17), 
> MIT license, easily usable (single header), well documented, and enough 
> feature-full. From a quick testing, it seems to work well. It looks also as 
> it has taken some inspiration from the Python argparse module.
>
> I'd be tempted to give that a try to retrofit our existing utilities 
> (probably starting with the ones with the less options :-)). Opinions? I 
> guess there must be a plethora of similar projects, due to the absence of a 
> std::argparse module... At least I see it is in the list of (9) alternatives 
> mentioned at 
> https://en.cppreference.com/w/cpp/links/libs?source=post_page---------------------------#Configuration:Command_Line

Some options have related environment variables, eg --debug /CPL_DEBUG.
Options are also passed through to several drivers.
I assume that the implications for these are not significant ?

-- 
Andrew C. Aitchison                      Kendal, UK
                    andrew at aitchison.me.uk