On Tuesday, 24 February 2026 at 19:01:35 UTC, monkyyy wrote:
On Tuesday, 24 February 2026 at 18:49:58 UTC, user1234 wrote:
On Tuesday, 24 February 2026 at 16:56:57 UTC, monkyyy wrote:
[...]
The tagged union approach is actually faster when you deal
with a situation such as
- you only care about final types
- your final types have several bases
- you would cast a lot
- you don't care about nicely looking declarations (e.g SDL
event, XEvent, etc 🤢)
Consider the `old` function, a tagged union would add a layer
of indirection that does nothing. This wont be faster, it may
only tie if the compiler untangles the mess that was made.
A 3d engine will have lots of things, but then have a draw loop
that like "I dont care whats your model id and position"
I see.. I thought to my reply yesterday and saw how unperfect it
was. What I mean is really like when you have to deal with a lot
of let'say base "Object"s then a tag can be faster than a dynamic
cast. Problem is we cant really merge struct members...so even
with the tag there's might be indirections, compared to using a
tag for class casts. I hope this is clearer now.
