Mirror of forum.nim-lang.org

237 :: unions

• Mat2 (orginal) [2013-09-14T15:59:42+02:00] view original

  Hello, I need combined access to memory areas both at word as byte basis (unions). At current my solution is to bit-mask selected bytes of a given word which have the disadvantage to be endian dependent. I know it is possible to define a variant object (which should be something similar to a variant record) in Nimrod but not sure if translation into an union structure is granted, so what would be an elegant way to implement unions in Nimrod?

---

• MFlamer (orginal) [2013-09-14T18:19:51+02:00] view original

  I have read that the internal representation of an object variant is indeed a union. You can also look into the {.pure.} pragma which erases the type identification at compile time so you don't have to carry that baggage around. Araq's intention is to have nimrod compile to fast and efficient C, so if he says variants are unions I wouldn't worry about it too much. But I also am a nimrod n00b so take my advice with a grain of salt.

---

• Araq (orginal) [2013-09-15T16:13:53+02:00] view original

  Variant objects are mapped to a union but that doesn't mean that you can use them as a union for implicit type casting: All accesses are checked at runtime. You need to use 'cast' to get different views on the data.

---