I get how important UX is, and I still struggle to fathom what in structopt even generates 20x more machine code and 30x slower compilation versus a simpler argument parser which can do 95% of the job. Someone at some time thought that every line of code had to be there for some reason, but it just doesn’t seem like a reasonable default choice to me for most apps, because it’s a huge penalty just to handle some edge cases that won’t come up most of the time.
ripgrep, your example of something demands a complex arguments parsing system, seems to have had problems with unnecessary changes in clap (on which structopt is based) slowing down compilation for no reason[0][1] and currently uses none of its configurable features except suggestions[2].
ripgrep, your example of something demands a complex arguments parsing system, seems to have had problems with unnecessary changes in clap (on which structopt is based) slowing down compilation for no reason[0][1] and currently uses none of its configurable features except suggestions[2].
[0] https://github.com/BurntSushi/ripgrep/commit/c8e755f11f31b6d...
[1] https://github.com/clap-rs/clap/commit/739e7048a50d14023ef9e...
[2] https://github.com/BurntSushi/ripgrep/blob/def993bad1a9275cd...