> incidentally, how do we feel about adding a rustfmt check to the CI(s)?
yes we should be doing that. you can send another pr that adds the check to the Ci
subcommands
This commit changes the internal ID to a u64 which will allow for
greater optimizations down the road. In addition, it lays the ground
work for allowing users to use things like enum variants as argument
keys instead of strings.
The only downside is each key needs to be hashed (the implementation
used is an FNV hasher for performance). However, the performance gains
in faster iteration, comparison, etc. should easily outweigh the single
hash of each argument.
Another benefit of if this commit is the removal of several lifetime
parameters, as it stands Arg and App now only have a single lifetime
parameter, and ArgMatches and ArgGroup have no lifetime parameter.
The return value is only used in the error case, and only a single
String allocation, therefore the complexities introduced by using string
slices isn't worth it. clap primarily only needs to be as fast as
possible in the happy path, the error path can afford additional
allocations.
