Rebased. There may be review work pending around hygiene? Actually, it looks like I addressed some comments from a year ago. But there's a bootstrap issue with try relying on library code.

Would it be possible to submit the part of this that fixes parsing of try { 1; 1 } + 1 catch from the part that changes how catch expressions are translated? I can't remember the context for the latter part, and I'm a bit hesitant about having catch sometimes translating into something heavyweight (an applyOrElse call with a lambda as an argument).

7-year itch, scratched.

7 years later.

:head_scratch: is this addressing scala/bug#5887?

@dwijnand this is the catch part. It doesn't typecheck correctly yet, mea culpa. Dotty requires a Function instead of PartialFunction, but this is an improvement over the status quo.

My previous joke about Try is relevant. That is, prefer Try to try for this sort of safety.

I'll add a lint if that will assuage our collective conscience, but I think the catch function is quite niche.

The use case for linting would be, when I write the handler, they never made it easy for me to specify the type:

val handler: PartialFunction[Throwable, String] = { case _ => "hello, world" }

val danger_? = (t: Throwable) => "goodbye, cruel world"

try e catch ???

Using handler doesn't warn. Or rather, defining it does not warn.

Gah, lost my comment edit when I clicked on the validate failure! which was spurious, of course.

To recap, now it issues the usual warning on catch with total function.

My comment also complained about stretching this PR into a long tail.

To answer Seth's question, the usual tractable syntax is the block of case clauses. Anyone doing something else syntactically will know what they're doing.

I wanted to make the case that in a Scalable language, I should be able to say try f() catch println in REPL. But I don't want to get into a philosophical debate about it either.

This change goes to regularity.

@lrytz care to weigh in about the design issues here?

Closing because I don't care any more.

Allowing Function instead of requiring PartialFunction only encourages catching Throwable, which is pretty much never a good idea (because of e.g. OutOfMemoryError). What desirable code pattern does this enable?

This is a good point.

Opened lampepfl/dotty#11019, also that Scala 3 doesn't have the "catches all Throwables" warning at all.

The good point was addressed in the follow-up.

👍, I noticed the warning. So it would be a new, handy feature for yolo-hacking, but discouraged from the beginning for larger projects.

(To be clear, I'm not opposed, as long as it warns. But can't get excited about it either.)

I'm slightly torn; it's generally good to align Scala 2 and 3, and I also think having the warning is right. But then we introduce a new feature that gives a warning as soon and as long as you use it. Does that make sense?

Are we certain that the change in dotty was fully intentioned?

We can adjust later to the status quo established by the dotty ticket.

It's not like everyone will start writing try e catch f everywhere. Probably no one notices the difference.

But it's much more regular than the previous state of affairs.

I had an example of a feature that was introduced and immediately deprecated, but I don't remember it off the cuff.

For the record, now I believe the compiler should emit no warnings at all, because someone will always complain either that the warning is too strict or too lax. Leave it to 3rd party tooling to express preferences and unwarranted fears about style.

I re-opened this PR because it's an obvious good.

there hasn't been any response on lampepfl/dotty#11019

Lukas and I are both on the fence, but Dale likes it, and Som is still behind this even after having had six years to think about it (😬), so let's merge

I'm going to delete this branch and go back to sleep.

I'm grateful there was no need to come up with an option name like for splain.

I'm sad Seth is not excited.

I see there were a few previous summations. My footnote would be to echo my previous comment: really, people on the fence want to say, Don't use try, use Try instead. Why don't we say that?

Because of the pun, "on the fens", I have learned from wikipedia that a fen is a type of mire, but less acidic than a bog. Please keep this in mind when saying casually, "I'm bogged down," "in a mire," etc, let alone, "on the fens." Let's use language precisely.

a fen is a type of mire, but less acidic than a bog

as for me, just yesterday I learned what “muskeg” is

This PR was definitely muskegged.

No relation to "musk egg," "Muskogee," or "beer keg." The keg hauled by moose on rescue missions would be moose keg.