The Last Hope for Scala's Infinity War [video]

I thought this was a potentially strong argument delivered weakly.

First, he poisons the well by talking about "integrity" and "intellectual honesty". This is a bad move at the beginning of a speech because it is far too easy to take it such that people with different conclusions are lacking in those qualities. Given his delivery, I suspect that he might actually believe that. It would be better if he just owned that this was his opinion and delivered it respectfully.

Second, the section on academic priorities was weak. He spent a large portion of the talk highlighting the mission statements of various languages, which is marketing language. But if Scala made no other changes other than its marketing language, it would not solve the problems he was highlighting. I thought too much time was spent on that part of the talk. There's another moment later in his rushed conclusion where he made stronger points that could have been fleshed out - how the academic influence leads to theoretical features that later get abandoned when the academic project is completed. (And I quite liked his revised marketing statement for Scala.)

I also didn't like that he brought up Cats vs scalaz as a flaw with the community, given that it became clear later in the talk that he had a clear bias on which was preferable. That seemed needlessly provocative.

I think the final conclusion of abandoning the OO path was interesting and fun to think about! I don't know if there's already an opinionated compiler plugin that discourages the more OO techniques while encouraging FP techniques, but I can see how that would be a lot of fun to play with, and educational in the same way that static analysis tools are. (I don't understand why "partial functions" was crossed out as something to abandon, though.)

Wise words for any project or product that starts in academia but stays there for too long (building for papers and PhDs rather than real customers). Brave to deliver such a cutting critique to this audience too.

As a potential newcomer to Scala, many of these points ring true. Hopefully the community takes a good hard look at itself to see what it can improve.

With some time off this past week I started hunting for things I should be learning. After doing some research on Scala's popularity and trending I came to may of the same conclusions John does.

Discouraging OO would be the worst thing Scala could do. Its one really unique capability is the fusion of OO and functional: it's a language that can go toe-to-toe with Haskell, yet also one that lets you switch from Java in a weekend. Take away the OO support and you'd have just another functional also-ran.

Sadly while the language itself has excellent OO support and Java interop, the community support just isn't there (by which I mainly mean: too many people like DeGoes will shout you down rather than trying to help you).

Partial functions are undefined outside of the part that is given, therefore the result of an input that falls in that zone is undefined. This is not good, how do you deal with it, exceptions? Termination? You must make the function total by extending it to a version that gives some result to manipulate whatever the input (or restrict the domain to make it again total.)

A minor note concerning the usual characterization of the Scala language itself as a fusion of OO and functional, it's definitely not an equal fusion. Not that you proposed that the fusion is equal, but I see it come up elsewhere.

The core language is very much OO. The functional aspects of Scala are either syntax sugar that desugars to OO stuff (e.g. automatic eta expansion, special syntax for defining FunctionN vs new classes, case syntax for partial functions, context bounds) or decisions made by the standard library (e.g. the prevalence of immutable data structures).

Understanding this has been the key for me to understanding where most of the warts of Scala come from. For example some of the weirdness of functions vs methods, weird performance issues, the existence of some of the encoding tricks that libraries like cats and scalaz use, and strange decisions made in the collection libraries all can be fundamentally traced back to this layering.

That being said you can do a lot of FP in Scala because of the combination of ecosystem and syntax sugar, but you have to always keep in mind that you're encoding FP concepts with OO tools (not as equals) otherwise Scala's limitations will come back to bite you.