Comment by MichaelGG | Hacker News Reader

Comment by MichaelGG | Hacker News Reader

Parent Full thread

I wouldn't call it strong continuing support. F#'s been a second class citizen both in resources given and marketing support. Community effort is great, but as long as MS pushes C# as the flagship language instead of giving it the equal-or-lesser footing it deserves, F# can't truly pull ahead.

It's good to see C# catching up with the past decades of language research. Maybe it's MS's DNA - they're still heavily pushing C++.

Tooling is the only real reason to ever use C# over F# - C# just doesn't do much (anything?) better. That, and legacy/enterprisey dev.

Heck, C# 7's tuple support is exactly what F# used to do, but then capitulated to MS's idea of making System.Tuple, a reference (heap allocated) type. Now in C# 7 since they finally got around to being a bit serious, they implement a new value-type tuple.

I guess we should be happy for any F# support we get. And indeed, tooling for functional languages is poor in general, so F# certainly leads...

You're seeing us strike a balance here. As I point out in the post, there are millions of C# developers, and tens of thousands of F# developers.

However, we think F# has awesome growth potential, and is great for .NET in general. So while we can't defend spending the same resources on it as we do on C#, we want to do what it takes to nurture it and keep it healthy and growing.

Being on the inside at Microsoft over the past years, it's been great to see more and more of the organization think of F# as part of the family.

In my opinion Microsoft should focus more on base tools for F#, such as Roslyn, integration with dotnet core, etc.

Integration with Visual Studio is nice, but if the language is to be adopted in hacker circles, without major Microsoft investment, it needs to provide very solid and flexible tools on top of which the community can build awesome things.

Example of small things Microsoft can help with: as far as I can see Nuclide doesn't work with dotnet core (only Mono). Throwing 1-2 devs that way would pay good dividends, in my opinion.

> Example of small things Microsoft can help with: as far as I can see Nuclide doesn't work with dotnet core (only Mono). Throwing 1-2 devs that way would pay good dividends, in my opinion.

I don't know anything about Nuclide, but if they want to work with dotnet core, they should look into working with omnisharp-roslyn[0]. VS Code[1] and Atom[2] both have extensions that work with it.

> In my opinion Microsoft should focus more on base tools for F#, such as Roslyn, integration with dotnet core, etc.

I think there's already work underway for F# support for dotnet core[6].

As far as F# support for editors go, have a look at ionide[4]. They only have extensions for VS Code and Atom at the moment.

> Integration with Visual Studio is nice, but if the language is to be adopted in hacker circles, without major Microsoft investment, it needs to provide very solid and flexible tools on top of which the community can build awesome things.

Have you seen omnisharp[5]? If so, what's missing from that?

[0] https://github.com/OmniSharp/omnisharp-roslyn

[1] https://github.com/OmniSharp/omnisharp-vscode

[2] https://github.com/OmniSharp/omnisharp-atom

[3] https://github.com/Microsoft/language-server-protocol

[4] http://ionide.io/

[5] http://www.omnisharp.net/

[6] https://github.com/dotnet/netcorecli-fsc

I meant Ionide. Ionide works with/on Mono.

About F# and .NET Core, you can read more info (usage/bugs/workaround) in the wiki https://github.com/dotnet/netcorecli-fsc/wiki/

The only ide who support f# and .net core is VSCode (with Ionide extension who add f# support)

see

- https://github.com/dotnet/netcorecli-fsc/wiki/.NET-Core-SDK-... for LTS of .net core 1.0 (project.json)

- https://github.com/dotnet/netcorecli-fsc/wiki/.NET-Core-SDK-... for msbuild based (latest bits)

just make folders work properly on F# solutions and I'll send beer/wine, whatever you want.

Striking a balance? It's the Innovator's Dilemma!

I'd like to see more of a focus on extension methods rather than the pipe operator in tutorials and such, and tupled args rather than curried args. Also make the OO aspects more visible. It's not so much a language thing as a presentation thing. I think these are the things that confuse mainstream developers, and it's silly to present them all up front when teaching the language. Intellisense and C# compatibility works way better if you go this route too.

If you want "OO aspects", you can use C#. F# should focus on being a functional language. Features like currying exist for a reason.

F# is multi paradigm though. It's just that, especially when looking locally at code, functional style is just way better.

C# is also multi-paradigm, or certainly well on its way, as it gets more and more functional features. F# needs to be functional-first.

In four years full-time F# use, I still find standard LOB apps work better with an OO SOA structure at the high level. I've tried various methods of designing FP-style architectures for those apps but nothing has ever ended up nice and clean as a standard interface-based SOA (in F#) and a standard OO DI utility. At the low-level, F# is awesome for immutable records, ADTs, combinators, etc., in a way that C# would be hard to replicate just because the syntax would be so foreign.

So I think there's an opportunity for the F# market to explode as C# devs start to realize the utility of such things at the bottom level (and also for making interesting combinator-based libraries like suave). But in order to get there, I think the stepping stone has to be a change of emphasis in the F# world, to show off F# as a "better C#", starting with standard SOA-style OO-style frameworks, emphasizing a similar style (tupled explicitly-typed fn params, ext methods rather than pipe, C# naming conventions), and then just bumping the lower-level implementations of things and the domain model to ML style.

Going head-first functional-first I think leaves F# useful to just a handful of developers, where everyone else just wants their objects and DI frameworks back.

What exactly do you mean by "OO SOA structure at the high level"? Do you mean essentially a function from a document to a result? If so, then what exactly is object-oriented about it? I'm no expert on industry lingo, but "Service Oriented Architecture" sounds to me much more functional than object-oriented.

That you feel the need for DI frameworks is a great shame and more F#'s failing than functional programming per se. If F# had inherited the module system from OCaml, then there would be much less need to use objects. In OCaml, DI comes for free with module functors, no framework necessary. As it stands, I agree that you are somewhat forced to use objects as a module system, as F# has no other. That is what MS should fix.

I mean `class Database implements IPersist`, `class Redis implements ICache`, `class MyLogger implements ILogger`, etc.

So classes/interfaces to each service your app uses. I just use very basic autofac, declaring which implementations to use in code. (Really nothing I couldn't wire up by hand; autofac is just a tad less cumbersome).

I've heard about OCaml's module system and am intrigued by it but never used it.

But that's a catch-22 isn't it? There's few users in large part from not just tooling but messaging. MS has pitched F# as a specialised language for scientists and financial work.

But I do appreciate the work MS does - F# is a leader in FP tooling. And C# is getting easier to deal with :).

While I'm already one of the few that use F# full time, I think this is a rather self-fulfilling statement. Looking back in time, Microsoft has continually cubbyholed the language, almost to the point of counter-marketing its general purpose feature set.

Simply giving it a fair share per-developer after so many years of mismanaged commercial handling seems to fall short. You say yourself, F# has awesome growth potential and I'm sure you'd agree that the growth potential has been there for years now. Is this time around going to be different?

I hope so.

Think how many developers you could add by adding Higher Kinded Types to F#. You could poach a ton from Scala/Haskell land. Alas, many of us can't leave until that happens. :)

It really needs to be added at the dotnet level. I'd love to have it in C# as well. LINQ expressions should work on arrays, ienumerables, lists, iobservables, subjects, nullables, etc, basically anything monadplus, and return back what you passed in. It's silly to have to write the same expression for each of those cases, or to convert everything to ienumerable/iobservable and back by hand. I'm surprised they didn't take the opportunity to do that in dotnetcore. It's mainstream enough now that I'm curious why they didn't; whether the perceived need was low or whether there are other problems it would have introduced.

My reply here [1] may be of interest :)

[1] https://news.ycombinator.com/item?id=13547181

Agree with other comments, back then MS invested a lot in C# while there was then way more VB users, now it should invest a lot on F# while there are way more C# users.

C# isn't going to dwindle because F# expands, if F# expands C# wins because F# has ability to appeal to developers on non .net platforms, where C# isn't seen as appealing, and F# has ability to bring outstanding projects and idioms to .net eco-system.

Take the returns on investment made with C# all those years and invest a fair share in F#.

The main reasons F# is not picking up have been stated, and I face this situation at my work where my use of F# is put on hold because Microsoft is not investing in it much and my colleagues have the feeling C# is "good enough" which is very short sighted perspective.

You have a business to keep running, I respect that. But if you truely want to bring innovation to the regular developer, bump up your F# efforts.

It needs a significant effort to educate developers to develop in a truly functional way. You are on track to implement many FP constructs and patterns into C#, so I'm sure the F# efforts would not be wasted at all.

Preface: I really love what you guys are doing but...

I really don't think you guys appreciate how many Scala/Clojure/Haskell devs would love to work in F# but can't because of the tooling.

Also equating the type of work being done with C# to the type of work being done F# by just comparing developer numbers is incredibly naive.

I think they do, and this is why they giving it any resources at all while admitting they only see it used by tens of thousands, compared to millions using c#

At the end of the day, MS is a big corp, decisions needs to be justified and supported by both quantitative and qualitative measures

Also on the bright side, in programming, more developers doesn't really mean it will evolve faster, actually with a smaller team f# may evolve faster, it this wasn't true no language other than c# or java would have stood a chance

I agree with the other comments as well, that treating F# as a second-class citizen is a self-fulfilling prophecy.

As someone who was telling myself, "One day, I should learn F#", this thread makes me pause.

Throw all your weight behind it, make it a first-class citizen, and Microsoft may be the first to make functional programming mainstream. That would be exciting.

When I was doing .Net full-time I really wanted to get into F#. But...

- I wanted to use it for websites. Sorry, not really supported in MVC.

- I thought it'd be great for doing .Net stuff in SQL-Server and SSIS, like you can with VB and C#. But it's not supported there either.

They always positioned it as "use F# for the extra hard stuff and C# for everything else," but in the end C# isn't terrible for hard stuff, and if that's what you use all day then for you, C# is probably better.

Not even just about web projects. Even standard GUI apps have no ready templates at least in VS 2015. Just Console apps. Even if guides online show how you can easily hack together both GUI apps as well as e.g ASP .NET Web API's!

I'd like to see some love there. These simple things introduces needless uncertainty about the language.

There are a few F# WPF templates. Here is one: https://marketplace.visualstudio.com/items?itemName=DanielMo...

About web:

- use Suave (https://suave.io/), best from f#, works on .netcore too

- use Aspnet Core Mvc (just `dotnet new -l fsharp -t web` in dotnetcore), or just aspnet core

I feel like this was addressed a bit in the linked blog post: the support is getting better and it's now "in the family" as opposed to a primarily community and research-supported "toy".

I'm looking for the nearing releases where F# starts to share much more of the same Roslyn platform that C# and VB have been using for a while now. That should be some very interesting tooling to have at F#'s disposal.

C# is always going to be the flagship for the same reason that C/C++/Java/et al are still some of the most common programming languages in the world. That doesn't mean that we can't expect good things from F#, though.

F# has shipped with VS for the better part of a decade now though, yet tooling doesn't reflect that.

Hopefully this will improve. It's just frustrating seeing C# slowly adopt stuff that was known to be good decades ago...

I am interested in F# for enterprisey web development. Is there a good resource/guide for using F# for ASP.NET MVC?

Reply on news.ycombinator.com