News for hackers. Stuff that matters.

A comment on Two Weeks of Rust. Parent comment

Re:Two Weeks of Rust by pcwalton on Tuesday January 12, 2016 @12:35PM (#10888839)
> And guess what, they probably won't be converted to Rust either, but they will be better due to the new standards.

What we need is memory safety, and new standards don't provide that. Maybe in the future they will, but I have significant questions about how suitable the ISO Core C++ lifetime checker will be.

> Just ask the Python and Perl teams.

Python and Perl were not backwards compatible with anything when they were released.

Rust is not a new version of C++. It is a new language.

> And this is how C++ was able to build marketshare, by offering a transition path from C. What exactly is that path for C++ to Rust, rewriting everything?

No. It's a very comprehensive FFI, IMO among the best in any language. We even have (preliminary, quite hacky) C++ support.

We had to spend time developing this, because we actually use Rust for Servo and we had to develop incrementally, retaining large C++ components. It works great.

[ Parent | on HN ]
Re:Two Weeks of Rust by kbenson on Tuesday January 12, 2016 @12:52PM (#10888977)
>> Just ask the Python and Perl teams.

> Python and Perl were not backwards compatible with anything when they were released.

I think the implication is that the teams will tell you that the upgrades to Python 3 and Perl 6 were problematic because they didn't offer backwards compatibility. If so, it's a bit silly, because the point of those upgrades was to advance the languages beyond would could be done with backwards compatible iteration. Asking the teams would probably yield a bunch of people that were adamant that the decision was sound, and it needed to be done, but there were missteps in execution. Asking the communities would likely yield a more polarized set of opinions.

That said, IMO using either of these items as examples to bolster an argument is almost always a mistake, since their own discussions are so large and polarizing, they rarely ever make make a point more clear.

[ Parent | on HN ]
Re:Two Weeks of Rust by stcredzero on Tuesday January 12, 2016 @01:10PM (#10889128)
Python and Perl were not backwards compatible with anything when they were released.

Perl 5 (and older) was highly backwards compatible from a programmer standpoint. Much of Perl is an agglomeration of preexisting Unix command line tools and little languages into one language.

[ Parent | on HN ]
Re:Two Weeks of Rust by pcwalton on Tuesday January 12, 2016 @01:19PM (#10889209)
Well, it's closest semantically to awk, but awk is not syntactically compatible with Perl [1]. That's pretty similar to the situation with Rust and C++: semantically compatible but not syntactically compatible.

[1]: http://www.arl.wustl.edu/projects/fpx/references/perl/learn/...

[ Parent | on HN ]
Re:Two Weeks of Rust by stcredzero on Tuesday January 12, 2016 @01:33PM (#10889308)
"semantically compatible but not syntactically compatible" would have been a more impressive way of expressing it.

If Rust is highly semantically compatible with C++, then automated porting of C++ to Rust isn't such a far-out idea.

[ Parent | on HN ]
Re:Two Weeks of Rust by pcwalton on Tuesday January 12, 2016 @01:55PM (#10889502)
Definitely it'd be interesting. But I feel like such a thing is going to have the same issues I worry about with the ISO Core C++ lifetime checker: existing C++ code just isn't designed for lifetimes, so you will have to rework the logic a lot to get the static analysis to pass. When you have that much manual intervention, I wonder whether you're porting or really just rewriting—and if you're rewriting, syntactic compatibility doesn't matter so much.
[ Parent | on HN ]