Principle: All APIs are library APIs #1280
Conversation
github-actions
bot
added
the
proposal rfc
| ### Built-in primitive types | ||
|
|
||
| We could follow the general outline of C++, where arithmetic and pointer types | ||
| are built-in. However, that would substantially erode the advantages outlined | ||
| above. We expect Carbon to have multiple kinds of pointers (for example, to | ||
| represent different kinds of ownership), and multiple kinds of arithmetic types | ||
| (for example, to handle overflow in different ways). They can't all be built-in, | ||
| so putting even the common-case types in the library helps ensure that Carbon | ||
| has enough expressive power for the uncommon-case library types. |
There was a problem hiding this comment.
Maybe worth calling out that C++ specifically has ended up with built-in types becoming inconsistent with user-defined types, and this has actually caused real problems with generic code (such as in templates) that is attempting to support both?
There was a problem hiding this comment.
LGTM (with the suggested typo fix)
Actually, I can just merge this with the typo fix, it already tracks who it is authored by. Of course, we can fix-forward if there is any issue uncovered here.
Sorry for the noise. I forgot I suggested an enhanced rationale for not following C++ here. Let's see if that makes sense or not before merging w/o it...
|
Oh, I also wanted to explicitly say I like describing this as "the APIs are library APIs" -- I think that does a great job of capturing the goal here. |
Co-authored-by: Richard Smith <[email protected]>
Actually, I can just merge this with the typo fix, it already tracks who it is authored by. Of course, we can fix-forward if there is any issue uncovered here. |
github-actions
bot
added
proposal accepted
Co-authored-by: Chandler Carruth <[email protected]> Co-authored-by: Richard Smith <[email protected]>
chandlerc
added
the
documentation
No description provided.