Relevant to this thread: https://github.com/apache/arrow/pull/10691 adds an enum replacement which includes compile-time to/from strings. This should reduce repetitive boilerplate in the bindings and elsewhere mapping to/from user provided string values.
On Mon, Jun 14, 2021 at 11:17 PM Micah Kornfield <emkornfi...@gmail.com> wrote: > > > > In light of this my preference would be: public enums must be scoped and > > named in CamelCase. The style guide would require kCamelCase but it does > so > > for consistency with public _unscoped_ enums. IMO we'd be better served > by > > requiring scoped enums, which obviates the 'k' prefix. > > > This seems to just add more rules to remember for not that much gain (and > an additional exception to the style guide). > > On Fri, Jun 4, 2021 at 8:48 PM Benjamin Kietzman <bengil...@gmail.com> > wrote: > > > My two cents: > > > > kSomeConstant includes the prefix 'k' so that regardless of what > expression > > it appears in, it is unambiguously a global constant. I'm restating the > > obvious here to clarify that the goal is making names obviously global. > > > > Note that a prefix is redundant for any qualified name (appears to the > > right of ::). For example, Type::Boolean unambiguously refers to a > global- > > static data member or a scoped enum value, since 'Type' is not a legal > > namespace. > > > > In light of this my preference would be: public enums must be scoped and > > named in CamelCase. The style guide would require kCamelCase but it does > so > > for consistency with public _unscoped_ enums. IMO we'd be better served > by > > requiring scoped enums, which obviates the 'k' prefix. > > > > Static data members may be referenced without qualification within the > > class' scope, so to ensure they are obviously global even there the 'k' > > prefix should still be applied. > > >
