←back to thread

In Defense of C++

(dayvster.com)
185 points todsacerdoti | 1 comments | 16 Sep 25 19:40 UTC | HN request time: 0.379s | source
Show context
beej71 ◴[16 Sep 25 22:36 UTC] No.45269149[source]▶
It doesn't mention the horrific template error messages. I'd heard that this was an area targeted for improvement a while ago... Is it better these days?
replies(1): >>45270360 #
jandrewrogers ◴[17 Sep 25 01:12 UTC] No.45270360[source]▶
Qualitatively better. C++20 'concepts' obviated the need for the arcane metaprogramming tricks responsible for generating the vast majority of that template vomit.

Now you mostly get an error to the effect of "constraint foo not satisfied by type bar" at the point of use that tells you specifically what needs to change about the type or value to satisfy the compiler.

replies(1): >>45281622 #
1. einpoklum ◴[17 Sep 25 21:30 UTC] No.45281622[source]▶
> C++20 'concepts' obviated the need for the arcane metaprogramming tricks responsible for generating the vast majority of that template vomit.

1. Somewhat exaggerated claim. It reduced that need; and for when you can assumpe everyting is C++20 or later.

2. Even to the extent the need for TMP was obviated in principle - it will take decades for TMP to go away in popular libraries and in people's application code. At that time, maybe, we would stoopp seeing these endless compliation artifacts.