You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Matthijs pointed out in #149 that it's a bit difficult to read the code for generating stuff when you deal with all of the variants in internal_fn in a variety of different places depending on which function it is. We should move to all of these being labeled CompilerInternal, and possibly delete some of them from the variant and just use the StanLib function name where appropriate.
The motivating example was that negative_infinity is actually a StanLib funcction and we'd construct it in the compiler using the StanLib function kind and the FnNegInf variant. But perhaps the thing to actually do is just use the actual string negative_infinity and not create an internal_fn variant for it? OR to call it CompilerInternal and pretend that it doesn't have a StanLib equivalent. I'm not sure. What do you all think? I'm leaning towards one of those two options now - having these special variants sometimes labeled StanLib and sometimes CompilerInternal is not helpful.
The text was updated successfully, but these errors were encountered:
seantalts
changed the title
Make all Fn* stuff actually labeled as CompilerInternal
Generate all Fn* internal_fn variant FunApps as CompilerInternalJun 6, 2019
Matthijs pointed out in #149 that it's a bit difficult to read the code for generating stuff when you deal with all of the variants in
internal_fn
in a variety of different places depending on which function it is. We should move to all of these being labeledCompilerInternal
, and possibly delete some of them from the variant and just use the StanLib function name where appropriate.The motivating example was that
negative_infinity
is actually a StanLib funcction and we'd construct it in the compiler using theStanLib
function kind and theFnNegInf
variant. But perhaps the thing to actually do is just use the actual stringnegative_infinity
and not create aninternal_fn
variant for it? OR to call itCompilerInternal
and pretend that it doesn't have a StanLib equivalent. I'm not sure. What do you all think? I'm leaning towards one of those two options now - having these special variants sometimes labeled StanLib and sometimes CompilerInternal is not helpful.The text was updated successfully, but these errors were encountered: