Union types in C# 15
devblogs.microsoft.com
devblogs.microsoft.com
Other languages have unions with named choices, where each name selects a type and the types are not necessarily all different. Rust, Haskell, Lean4, and even plain C unions are in this category (although plain C unions are not discriminated at all, so they’re not nearly as convenient).
I personally much prefer the latter design.
If you're interested, i can point you to specs i'm writing that address the area you care about :)
Just wanted to add just another opinion on a few things.
I think many people already mentioned it, but I also don't feel to good about non-boxed unions not being the default. I'd personally like the path of least resistance to lead to not boxing. Having to opt-in like the current preview shows it looks like a PITA that I'd quickly become tired of.
Also, ad-hoc union types could be really nice. At least in Python those are really nice, stuff like `def foo(x: str | int)` is just very nice. If I had to first give this union type a name I'd enjoy it way less.
But I'm aware that you are trying your best to find a good trade-off and I'm sure I don't know all the implications of the things I wish you'd do. But I just wanted to mention those to have another data point you can weigh into your decision process.
My belief is that we will have a `union struct` just like we have `record` and `record struct`. Where you can simply say you want value-type, non-boxing, behavior by adding the `struct` keyword. This feels very nice to me, and in line with how we've treated other similar types. You'll only have to state this on the decl point, so for each union type it's one and done.
The problem is that the only safe way for the compiler to generate non-boxed unions would require non-overlapping fields for most value types.
Specifically the CLR has a hard rule that it must know with certainty where all managed pointers are at all times, so that the GC can update them if it moves the referenced object. This means you can only overlap value types if the locations of all managed pointers line up perfectly. So sure, you can safely overlap "unmanaged" structs (those that recursively don't contain any managed pointers), but even for those, you need to know the size of the largest one.
The big problem with the compiler doing any attempt to overlap value types is that if the value types as defined at compile time may not match the definitions at runtime, especially for types defined in another assembly. A new library version can add more fields. This may mean one unmanaged struct has become too big to fit in the field, or that two types that were previously overlap compatible are not anymore.
Making the C# compiler jump though a bunch of hoops to try to determine if overlapping is safe and even then leaving room for an updated library at runtime to crash the whole things means that the compiler will probably never even try. I guess the primitive numeric types could be special cased, as their size is known and will never change.
Again, very rough. We go round and round on things. Loving to decompose, debate and determine how we want to tackle all these large and interesting areas :)
I admit that I haven’t actually used C# in 20 years or so.
An example that may show the difference: if you have a language with nullable types, then you basically have a language with union types like String|Null, where the Null type has a single value called `null` and String can not be null.
Now if you pass this around a function that itself may return `null`, then your type coalesces to String|Null still (you still get a nullable string, there is no doubly nullable). This is not true for Maybe/Option whatever you call types, where Some(None) (or Optional.of(Optional.empty())) is different from None only.
Rich Hickey once made a case that sort of became controversial in some FP circles, that the former can sometimes be preferred (e.g. at public API surfaces), as in for a parameter you take a non-nullable String but for returns you return a String|Null. In this case you can have an API-compatible change widening the input parameters' type, or restricting the return type - meanwhile with sum types you would have to do a refactor because Maybe String is not API compatible with String.
This is a really good point. I'd love to be able to have a sum type of two strings ("escaped" and "unescaped"); or any two kinds of the same type really, to model two kinds of the same type where one has already passed some sort of validation and the other one hasn't.
Edit to add: I figure what I want is for enums to be extended such that different branches are able to carry different properties.
Edit again (I should learn to think things through before posting. sorry): I suppose it can be faked using a union of different wrapper types, and in fact it might be the best way to do it so then methods can take just one of the types in arguments and maybe even provide different overloads.
I've seen exactly one Rust type which is actually a union, and it's a pretty good justification for the existence of this feature, but one isn't really enough. That type is MaybeUninit<T> which is a union of a T and the empty tuple. Very, very, clever and valuable, but I didn't run into any similarly good uses outside that.
That is essentially the motivation, primarily in the context of FFI where matching C's union behaviour using transmute is tricky and error-prone.
Also you can manually tag them and get s.th. more like other high level languages. It will just look ugly.
If you want disjoint types, something like Pascal's discriminated variants or Rust's enums is the way to go. It's embarrassing that C never had this.
Many bad design decision in C come from the fact that originally, separate compilation was really dumb, so the compiler would fit in small machines.
As for C, it is a sharp blade on its own.
I can't even think of any prominent C FFI problems where I'd reach for the union's C representation. Too many languages can't handle that so it seems less useful at an FFI edge.
One of the reasons .NET had Managed C++, replaced by C++/CLI (nowadays C++20 compliant, minus modules), is exactly that P/Invoke (and RCW/CCW) cannot represent everything.
Which they don't want to expose on .NET type system directly.
ColdString is easier to explain, the whole trick here is the "Maybe this isn't a pointer?" trick, ColdString might be a single raw pointer onto your heap with the rest of the data structure at the far end of the pointer, this case is expensive because nothing about the text lives inline, but... the other case is that your entire text was hidden in the pointer, on modern hardware that's 8 bytes of text, at no overhead, awesome.
CompactString is more like a drop-in replacement, it's much bigger, the same size as String, so 24 bytes on modern hardware, but that's all SSO, so text like "This will all fit nicely" fits inline, yet the out-of-line case has the usual affordances such as capacity and length in the data structure. This isn't doing the "Maybe this isn't a pointer?" trick but is instead relying on knowing that the last byte of a UTF-8 string can't have certain values by definition.
All ColdStrings which look like 8 bytes of UTF-8 text really are 8 bytes of UTF-8 text, just the type label on those 8 bytes isn't "[u8; 8]" an array of 8 bytes but instead "mut *u8" a raw pointer. "Validate" for example is 8 bytes of ASCII, thus UTF-8, and Rust is OK with us just saying we want a pointer on a 64-bit machine with those bytes. It's not a valid pointer, but it is a pointer and Rust is OK with that, we just need to be careful never to [unsafely] dereference the pointer because it's invalid
OK, so there are two cases left: First, what if there are fewer bytes of text? Zero even?
Since there are fewer than 8 bytes of text we can use the whole first byte to signal how many of the remainder are text, we use the UTF-8 over-long prefix indicator in which the top five bits of the byte are all set, bytes 0xF8 through 0xFF for this, there are eight of these bytes corresponding to our 8 lengths 0 through 7 inclusive. Because it's over-long this indicator isn't itself a valid UTF-8 prefix. Again we can pretend this is a pointer while knowing it's invalid.
Lastly, the seemingly trickiest problem, what if the string didn't fit inline? We use a heap allocation to store the text prefixed by a variable size integer length and we insist this allocation is aligned to 4 bytes. This means a valid pointer to our allocation has zeroes for the bottom two bits, then we rotate that pointer so those bottom two bits are at the top of the first byte position (depending on machine word layout) and we set the top bit. This is now always invalid UTF-8 because it has the continuation marker - the top bit is set but the next is not, which cannot happen in the first byte of any UTF-8 text, and so our code can detect this and reverse the transformation to get back a valid pointer using the strict provenance APIs if this marker is present.
This type is tomtomwombat's idea, credit to them:
Sure, but the comparable Rust feature is enum, not union.
I suppose an LLM could do a pretty good job at this.
- enum: essentially just a typed uint tag collection
- union: the plain data structure that can contain any of several types
- tagged union: combines enums and unions, so you can dispatch on its tag to get one of the union types
Read from this section and they appear in order:
It's trying to generalize - we might have exactly one T, fine, or a collection of T, and that's more T... except no, the collection might be zero of them, not at least one and so our type is really "OneOrMoreOrNone" and wow, that's just maybe some T.
https://hackage.haskell.org/package/oneormore or scala: https://typelevel.org/cats/datatypes/nel.html
it's for type purists, because sometimes you want the first element of the list but if you do that you will get T? which is stupid if you know that the list always holds an element, because now you need to have an unnecessary assertion to "fix" the type.
They could’ve done the Either type which would’ve been more correct or maybe EitherT (if the latter is even possible)
… wait, I've made a different mistake here while trying to explain the difference, haven't I? I was describing it as a sum type, but it's not really a sum type, it's really just set-theoretic union, right?
Which also means OneOrMore is unsound in a different way because it doesn't guarantee that T and IEnumerable<T> are disjoint; OneOrMore<object> initialized from [x] will always return [[x]] from AsEnumerable, won't it? If I'm interpreting the switch expression correctly and the first case predominates, since a list is-an object? I don't have a test setup handy; someone with actual C# experience, please tell me whether that's correct or whether the compiler signals an error here or something…
You are free to call it `public union Some<T>(T, IEnumerable<T>)`
If I understand correctly, it’s actually OneOrOneOrMoreOrNone. Because you have two different distinguishable representations of “one”.
The only reason to use this would be if you typically have exactly one, and you want to avoid the overhead of an enumeration in that typical case. In other words, AnyNumberButOftenJustOne<T>.
And you’re exactly right.
It’s not “one or more.”
It’s “one or not one.”
Need two or not two.
So IEnumerable<T> ? What's up with wrapping everything into fancy types just to arrive at the exact same place.
The problem with ad-hoc unions is that without discipline, it invariably ends in a mess that is very, very hard to wrap your head around and often requires digging through several layers to understand the source types.
In TS codebases with heavy usage of utility types like `Pick`, `Omit`, or ad-hoc return types, it is often exceedingly difficult to know how to correctly work with a shape once you get closer to the boundary of the application (e.g. API or database interface since shapes must "materialize" at these layers). Where does this property come from? How do I get this value? I end up having to trace through several layers to understand how the shape I'm holding came to be because there's no discrete type to jump to.
This tends to lead to another behavior which is lack of documentation because there's no discrete type to attach documentation to; there's a "behavioral slop trigger" that happens with ad-hoc types, in my experience. The more it gets used, the more it gets abused, the harder it is to understand the intent of the data structures because much of the intent is now ad-hoc and lacking in forethought because (by its nature) it removes the requirement of forethought.
"I am here. I need this additional field or this additional type. I'll just add it."
So even when I write TS and I have the option of using ad-hoc types and utility types, I almost always explicitly define the type. Same with types for props in React, Vue, etc; it is almost always better to just explicitly define the type, IME. You will thank yourself later; other devs will thank you.
Are you sure? This is a feature of OCaml but not F# IIUIR
Edit: https://github.com/fsharp/fslang-suggestions/issues/538
Yes and no. C# unions aren’t sealed types, that’s a separate feature. But they are strictly nominal - they must be formally declared:
union Foo(Bar, Baz);
Bar | Baz
We're very interesting in this space. And we're referring to it as, unsurprisingly, 'anonymous unions' (since the ones we're delivering in C#15 are 'nominal' ones).
An unfortunate aspect of lang design is that if you do something in one version, and not another, that people think you don't want the other (not saying you think that! but some do :)). That's definitely not the case. We just like to break things over many versions so we can get the time to see how people feel about things and where are limited resources can be spent best next. We have wanted to explore the entire space of unions for a long time. Nominal unions. Anonymous unions. Discriminated unions. It's all of interest to us :)
=> named sum type implicitly tagged by it's variant types
but not "sealed", as in no artificial constraints like that the variant types need to be defined in the "same place" or "as variant type", they can be arbitrary nameable types
> unions enable designs that traditional hierarchies can’t express, composing any combination of existing types into a single, compiler-verified contract.
To me that "compiler-verified" maps to "sealed", not "on the fly". Probably.
Their example is:
public union Pet(Cat, Dog, Bird);
Pet pet = new Cat("Whiskers");
- the union type is declared upfront, as is usually the case in c#. And the types that it contains are a fixed set in that declaration. Meaning "sealed" ?
As the language designer notes in the comments, these are named unions, as opposed to anonymous ones, but they are also working on the latter.
"Sealed" is probably not the correct word to use here, as it would be sealed in both case (it doesn't really make sense to "add" a type to the A | B union). The difference is that you have to add a definition and name it.
(Dog, Cat) pet = new Cat();
So without defining the union with an explicit name beforehand.
var someUser = new { Name = "SideburnsOfDoom", CommentValue = 3 };
A type can be given at compile-time in a declaration, or generated at compile-time by the compiler like this. But it is still "Compiler-verified" and not ad-hoc or at runtime.
the type (Dog, Cat) pet seems similar, it's known at compile-time and won't change. A type without a usable name is still a type.
Is this "ad-hoc"? It depends entirely on what you mean by that.
> Cat, Dog and Bird don't have to inherit from the union, you can declare a union of completely random types, as opposed to saying "Animal has three subtypes, no more, no less"
"Animal has three subtypes" is more like the c# "sealed" modifier on a class, meaning that subtyping is not allowed. Except in this case I guess for three existing subtypes.
Here is visual layout if anyone is interested - https://vectree.io/c/memory-layout-tagging-and-payload-overl...
F# offers actual field sharing for value-type (struct) unions by explicitly sharing field names across cases, which is as far as you can push it on the CLR without extra runtime support.
* https://devblogs.microsoft.com/dotnet/csharp-15-union-types/...
We have been pushing toward higher performance for years and this is a performance pitfall for unions would are often thought of as being lighter weight than inheritance hierarchies.
F# just stores a field-per-case, with the optimization that cases with the same type are unified which is still type safe.
I expect the same will old here. But given the former group is multiple orders of magnitude higher than the latter, we tend to design the language in that order accordingly.
Trust me, we're very intersted in the low-overhead space as well. But it will be for more advanced users that can accept the tradeoffs involved.
And, in the meantime, we're designing it in C#15 that you can always roll the perfect implementation for your use case, and still be thought of as a union from teh language.
"Never let perfect be the enemy of good"
It is relatively easy to find corporate libraries, or commercial products that still aren't using it, including Microsoft products still stuck in .NET Framework, or .NET standard 2.0.
If you want name shaming of commercial products with modern .NET, here is one, can provide more.
To be clear, this requires explicitly using the same field name as well.
This is essentially "sealed interface" from Java/Kotlin or "enum" in Rust
One thing I miss here (and admittedly I only skimmed through the post so if I missed this, please do correct me) is "ad hoc" unions.
It would be great to be able to do something like
public Union<TypeA, TypeB> GetThing()...
union Union<T1, T2>(T1, T2);
What a missed opportunity. I think really F# if you combine all of its features, and what it left out, was the way. Pulling them all into C# just makes C# seem like a big bag of stuff, with no direction.
F#'s features, and also what it did not included, gave it a style and 'terseness', that still can't really be done in C#.
I don't really get it. Was a functional approach really so 'difficult'? That it didn't continue to grow and takeover.
So far everything that was added to C# very much reduces the amount of dead boilerplate code other languages struggle with.
Really give it an honest try before you judge it based on the summation of headlines.
given that most of the thinks added seem more inspired by other languages then "moved over" from F# the "other languages struggle with" part makes not that much sense
like some languages which had been ahead of C# and made union type a "expected general purpose" feature of "some kind":
- Java: sealed interfaces (on high level the same this C# features, details differ)
- Rust: it's enum type (but better at reducing boilerplate due to not needing to define a separate type per variant, but being able to do so if you need to)
- TypeScript: untagged sum types + literal types => tagged sum types
- C++: std::variant (let's ignore raw union usage, that is more a landmine then a feature)
either way, grate to have it, it's really convenient to represent a `TYPE is either of TYPES` relationship. Which are conceptually very common and working around them without proper type system support is annoying (but very viable).
I also would say that while it is often associated with functional programing it has become generally expected even if you language isn't functional. Comparable to e.g. having some limited closure support.
I honestly have no idea where you would get this idea from. C# is a pretty opinionated language and it's worst faults all come from version 1.0 where it was mostly a clone of Java. They've been very carefully undoing that for years now.
It's a far more comfortable and strict language now than before.
`.ConfigureAwait(bool)` is another where it is relevant, but only in some contexts.
This is precisely because the language itself operates in many runtime scenarios.
But I'm not sure that's really a problem. Does the OP expect everyone to use an entirely different languages every single context? I have web applications and desktop applications that interact with Office that share common code.
Even `dynamic` is pretty nice as far as weird dynamic language features are concerned.
Interestingly enough `.ConfigureAwait(bool)` is entirely the opposite of `dynamic` -- it's not a language feature at all but instead a library call. I could argue that might instead be better as a keyword.
The reason I bring it up is that it is another one of those things where it matters in some cases depending on what you're doing.
Look at the depths that Toub had to go through to explain when to use it: https://devblogs.microsoft.com/dotnet/configureawait-faq/
David Fowl concludes in the comments:
> That’s correct, most of ASP.NET Core doesn’t use ConfigureAwait(false) and that was an explicit decision because it was deemed unnecessary. There are places where it is used though, like calls to bootstrap ASP.NET Core (using the host) so that scenarios you mention work. If you were to host ASP.NET Core in a WinForms or WPF application, you would end up calling StartAsync from the UI thread and that would do the right thing and use ConfigureAwait(false) internally. Request processing on the other hand is dispatching to the thread pool so unless some other component explicitly set a SynchronizationContext, requests are running on thread pool threads.
>
> Blazor on the other hand does have a SynchronizationContext when running inside of a Blazor component.
This is a good example of C# light-touch on language design. Async/await creates a state machine out of your methods but that's all it does. The language itself delegates entirely to platform/framework for the implementation. You can swap in your own implementation (just as it possible with this union feature)
> So I bring this up as a case of how supporting multiple platforms and runtime scenarios does indeed add some layer of complexity.
I agree that's true. A language that doesn't support multiple platforms and runtime scenarios can, indeed, be simpler. However that doesn't make the task simpler -- now you just have to use different languages entirely with potentially different semantics. If your task is just one platform and one runtime scenario, the mental cost here is still low. You don't actually need to know those other details.
Is it? F# code doesn't even need ConfigureAwait(false), one simply uses backgroundTask{} instead of task{} to ignore SynchronizationContext.Current, and this didn't require any language design changes at all (both are computation expressions), but it would for C# precisely because it delegates this choice to the framework.
This inspired the invokedynamic bytecode in the JVM, which has brought many benefits and much more use than the original .NET features, e.g. how lambdas get generated.
The only things that I wish for are: rusts borrow-checker and memory management. And the AOT story would be more natural.
Besides that, for me, it is the general purpose language.
Yes, C# is a jack of all trades and can be used at many things. Web, desktop mobile, microservices, CLI, embedded software, games. Probably is not fitted for writing operating systems kernels due to the GC but most areas can be tackled with C#.
Midori would like to have a word with you:
Just look at this feature: https://learn.microsoft.com/en-us/dotnet/csharp/whats-new/cs...
Was this needed? Was this necessary? It's reusing an existing keyword, fine. It's not hard to understand. But it adds a new syntax to a language that's already filled to the brim, just to save a few keystrokes?
Try teaching someone C# nowadays. Completely impossible. Really, I wish they would've given F# just a tenth of the love that C# got over the years. It has issues but it could've been so much more.
For you it may be fine to write:
List<string> strs = new List<string>();
And sure if you have been using C# for years you know all the things going on here.
But it shouldn’t be an argument that:
List<string> strs = [];
Is substentionally easier to grasp.
And that has been the theme of all changes.
The example you point out is the advanced case, someone only needs in a very specific case. It does not have a lot todo with learning the language.
The language design team is really making sure that the features work well throughout and I think that does deserve some credit.
I agree that = [] is perfectly fine syntax. But I would definitely argue that:
[with(capacity: values.Length * 2), ..
is non-intuitive and unnecessary. What other language is there that has this syntax? Alternatively, is this a natural way of writing this? I wouldn't say so.
My main language in my free time is Rust, a few years ago it was F#. So, I'm absolutely open to other syntax ideas. But I feel that there has to be a direction, things have to work together to make a language feel coherent.
Another example would be Clojure, which I started learning a few months ago (before we all got swept up in AI FOMO :D). Clojure as a language feels very coherent, very logical. I'm still a beginner, but every time I learn something about it, it just makes sense. It feels as if I could have guessed that it works this way. I don't get that feeling at all in many of the new features of C#.
> The example you point out is the advanced case, someone only needs in a very specific case. It does not have a lot todo with learning the language.
I disagree. When learning the language, you're going to have to read other people's code and understand it. It's the same basic principle, but, I'd argue, much worse in C++. Yes, in theory, you don't have to understand SFINAE and template metaprogramming and (now) concepts and all those things. You could just work in a subset of C++ that doesn't use those things. But in practice, you're always going to have issues if you don't.
> is non-intuitive and unnecessary.
intuitive is definitely in the eye of the beholder. When people saw:
`HashSet<string> people = [with(StringComparer.CaseInsensitiveComparer), .. group1, group2]`
they found it understandable. And this was also much nicer than what they'd have to write today (which would bring them out of the nice declarative collection-expression space).
Does that make it 'necessary'? Ultimately that's up to the individual. We felt like it was. Not being able to do simple things like this felt like a 'bitter pill'. Customization of collection construction is common (looking in codebases, it shows up about 7% of the time). So having to 'fall out' from the uniform collection-expr system into the much more verbose and clunky forms just for this common enough case felt 'necessary' to us.
>But I feel that there has to be a direction, things have to work together to make a language feel coherent.
I feel like this is conflicting feedback. Collection expressions made the language more coherent. Instead of 7 different ways of doing things (some of which were genuinely not efficient), we gave one uniform way of doing it. That makes things more coherent. Making it so you don't have to drop out of that for something as simple as configuring the collection makes things more coherent.
I tend to agree, but I also didn’t try it.
I also think the 7% number is wrong. 0.5% seems more realistic. It’s just that you are not able to see the majority of the code. Not everything is checked in or on GitHub.
C# doesn't need to have syntax sugar for every possible use case.
Some of the more recent features feel like the outcome of the team pressure to have new language features to announce in November every year.
I understand you feel this is ilke `!!`. We do not. We think being able to amke a dictionary, and pass in a custom comparer is deeply important. Analyzing code out there, we find that this happens in anywhere from 5-10% of all dicts. That is a ton of codebases and users impacted, and we've already heard from many of them about the friction this causes. Simply discarding that group greatly undercuts one of the core value props that collection expressions brings. A uniform and simple syntax that should suffice for nearly all collection needs.
You may feel differently. That's life in the design world :)
As such my opinion remains that in such context saving keystrokes isn't that high priority, when agents do the actual work.
Thanks for replying, though.
Collection expressions today are more the sort of thing that a code poet or golfer can do to prettify their code than something a newbie can count on using. It's tough to explain "you can only use this when the collection type is implied in that spot" to a newbie. The value of the base feature is still unproven for me. I'm not sure I agree, without some convincing, that collection expressions made the language more coherent rather than doing https://xkcd.com/927.
> Collection expressions made the language more coherent. Instead of 7 different ways of doing things (some of which were genuinely not efficient), we gave one uniform way of doing it.
I see your point on this. My dislike comes from a mixture of "I don't like how it looks" and "this language already has tons of features".
In terms of looks, I wish it could be more coherent with existing syntax.
List<int> = new {1, 2, 3} and List<int> = {1, 2, 3} are obviously taken up by anonymous types and blocks themselves. Would something like
List<int> = new(capacity: 10)[1, 2, 3]
have been possible? It feels like a combination of target-typed new and the initialization syntax. It involves the "new" keyword, which everybody already associates with constructor calls. It's short. Obviously, I don't know if this even works, maybe there's a parsing issue there (aren't those the most annoying issues in language design haha).
> they found it understandable
Kind of in my experience. Me and the people I've shown this to can easily remember it, but we all agree that it doesn't look like obvious syntax to them. Those two things are quite different to me. Contrast this to something like target-typed new, which immediately made sense to the same people. One might argue that that's fine enough and maybe it is, but I think, the less I have to remember about a language's syntax, the better. I'm going to have to remember many many other things anyway, better keep my memory free for the details of SynchronizationContext and async flow :)
I'm obviously aware that you get tons of bikeshedding comments like this all the time, so I'm sure you've gone through this. But to me, this invented syntax would have been fine. I just don't like the one that actually got in.
Now, the necessity on the other hand: May just be the company I'm working at, but my personal experience has never been that this is a big issue. Sure, it's nice to not have to fall back to explicit initialization a few more times. But personally, this doesn't pass my threshold of "painful enough to warrant additional syntax".
That's the core of my issue: Most, maybe all, of the new features in the language are fine to me in isolation. I may bikeshed about the explicit syntax (see: this thread). But my main issue is that the sum of complexity in the language and the issues beginners have when learning it are steadily increasing. I see this all the time at work.
As you said, this is definitely subjective. And in the end, language design is a very subjective process and maybe C# just won't be for me in the long run. But I wish it would, because at its core I like it, and .NET, a lot. Which is why I will continue to speak for my (subjective) viewpoint.
Well, this turned into a bit of an incoherent rant. I appreciate you exposing yourself to the HN acid pit ;)
Great question. And our design docs, and discussion with the community cover this. The reason that was eliminated as an option (we considered several dozen possible syntaxes) was that this syntax was actively confusing and misleading for people (for several reasons). These include (in no particular order):
1. the use of 'new' indicating that a new value was being allocated. That's not necessarily the case with collection expressions. The compiler is free to be smart here and not allocate if it doesn't need to. `[1, 2, 3]` for example, being constants, can in some cases just point at a data segment in the program.
2. the use of 'new' indicating that a constructor is being called ('new' has always meant that). That's not necessarily the case with collection expressions. Many collection forms (interfaces, immutables, spans, etc) do not go through constructors. This was actively confusing for people.
3. That syntax is already legal. It's an implicit objet creation that is being indexed into.
4. There was strong feedback from many in the community (and the design group, and lots that we talked to) that having things outside the boundary of the `[ ... ]` syntax was actively confusing. One could not easily tell what hte collection was and what wasn't part of it. The idea is that the `[ ... ]` is "the actual value". You know where it starts, where it ends, and what it represents.
--
Of course, at the end of the day, absolutely none of this may sway you. That's why we have a design process and we go through so many options. There were dozens considered here and we had many axes we were trying to optimize for. Overal, this struck a balance of working nicely, and having no major problems going for it (unlike other options).
> I'm obviously aware that you get tons of bikeshedding comments like this all the time, so I'm sure you've gone through this.
Yup :) Totally ok with us though.
> But personally, this doesn't pass my threshold of "painful enough to warrant additional syntax".
Sure. But that's why we look at the entire ecosystem. And we converse with people who have full codebases they haven't been able to move over because of the lack of this. And we look at the pain that this will cause esp when we get dictionary/key/value support. All of this motivated what was ultimately a tiny feature that cost very little to get in. It was medium bang for very low buck.
And that's worth explaining too. We are always working on some huge features. But they take up a ton of time and need tons of effort and runway. Small features like this are easy to slot in in gaps and help deal with papercuts and friction that are often annoying people.
https://docs.python.org/3/tutorial/datastructures.html#list-...
I'm also not sure that something not being intuitive or natural is necessarily a bad thing in of itself. You state it as if it's so, but you haven't demonstrated that this way of defining a list is worse. You also haven't made any attempt to understand any possible benefit, nor have you attempted any sort of analysis comparing the good and the bad aspects.
> I'm also not sure that something not being intuitive or natural is necessarily a bad thing in of itself. You state it as if it's so, but you haven't demonstrated that this way of defining a list is worse.
I would argue that a language having more features, without the feature being helpful, is a bad thing in itself. If the syntax isn't necessary or very convenient in many cases, it shouldn't exist. The syntax being natural (which, absolutely, is a very subjective thing) just makes it less of an issue, I'd say.
Every new syntax added to the language adds cognitive overhead to readers of code. But also, it adds possible interactions with other language features that may be added in the future. Now, the example I brought up doesn't really concern the second point, I'll concede that. But unions? That is a big concept to add to a language that already has decades of existing conventions and tons of other features. How will they interact with generics? Nullable reference types? And, just as importantly: How will they interact with any other features that might be added at some point that we don't even know about?
I'm not against adding syntax sugar. For example, I quite like primary constructors, which is another relatively new C# feature. I think it's a bit annoying that they were kind of added in a roundabout way, by first adding records and then adding primary constructors to classes, but this time they don't define properties but fields...but in the end, it's a nice convenience feature when using constructor injection. Which, whatever one may feel about this, is pretty common in C# code.
But the thing is: If every single feature that's nice for a few use cases gets added to a language, the language will explode. The best example for this is C++. C# is definitely not that bad, far from it, but my point is that I want it to stay that way :)
Dictionary<string, List<Tuple<string, string>>> foo = new Dictionary<string, List<Tuple<string, string>>>
Dictionary<string, List<Tuple<string, string>>> foo = DoSomeWork();
No-one wants to go back to that, and anyone who says C# is over-complicated I think is forgetting how rough it was in the earliest versions.
While introduction of auto-typing through "var" helped a lot with that, you'd still regularly have to fight if you wanted to properly initialise arrays with values, because the syntax was just not always obvious.
Collection literals are amazing, and now the ability to pass things into the constructor means they can be used when you need constructor parameters too, that's just a good thing as you say.
This is exactly how C++ landed where it is now. Every time it's "you only need to know that syntax if..." well it ends up everyone has to know that syntax because someone will use it and if you're a responsible programmer you'll end up reading a lot code written from other people.
But still there is a difference between learning and mastering.
I recently helped my partner learn for her CS class, and I feel very comfortable arguing that my previous statement holds up.
Mastering? No, in that case I agree with you.
I work on multiple applications with different versions of C# and/or Dotnet. I find it quite annoying to have to remember what syntax sugar is allowed in which versions.
If C# did not want verbose syntax, then Java was a poor choice to imitate.
Without the lawsuit, COM+ Runtime (aka .NET) would have used J++, as originally designed in the Ext-VOS paper.
Do you actually have a datapoint of someone failing to understand C# or are you just hyperbolically saying its a big language? The tooling, the ecosystem, the linting, the frameworks. Its a very easy language to get into...
So I'm happy to discuss the thinking here. It's not about saving keystrokes. It's about our decision that users shouldn't have 7 (yes 7) different ways of creating collections. They should just be able to target at least 99% of all cases where a collection is needed, with one simple and uniform syntax across all those cases.
When we created and introduced collection expressions, it was able to get close to that goal. But there were still cases left out, leaving people in the unenviable position of having to keep their code inconsistent.
This feature was tiny, and is really intended for those few percent of cases where you were stuck having to do things the much more complex way (see things like immutable builders as an example), just to do something simple, like adding an `IEqualityComparer<>`. This was also something that would become even more relevant as we add `k:v` support to our collections to be able to make dictionaries.
That isn't a reasonable take. Failing to teach a language by enumerating all its features is an indictment of the instructor and not the language.
You're right, it's not impossible and in general it's not among the hardest languages to teach. But I would argue, it is heading that way.
There are already so many ways to do things in C#. For example, try explaining the difference between fields and properties; sounds easy, but making it really stick is quite a challenge. And that's one of the simplest cases (and a feature I'm 100% in favor of).
And you will have to explain it at some point, because real codebases contain these features so at some point, it'll need to be taught. Learning a language doesn't stop when you can write a simple application, it continues up until at least you're comfortable with most of its features and their practical use. The quicker one can get people to that point, the easier the language is to teach, I'd argue.
One might also argue that learning never really stops, but that's beside the point :)
In general, my issue isn't any specific feature. C# has many features that are non-trivial to learn but still great: value types, generics, expression trees. Source generators are relatively new and I like them! I like most of the things they're doing in the standard library or the runtime. Spans everywhere is a nice improvement, most new APIs are sensible and nice to use and the runtime just keeps getting faster every release. Great. It's more the pure C# language side I have an issue with.
But every language has a budget of innovation and cognitive load that you can expect people to deal with, and C# is not using its budget very wisely in my opinion.
You should come engage with us on this then :)
We do all our design in the open on github. And a lot of us are available to chat and discuss all this stuff in Discord and the like :)
> C# is not using its budget very wisely in my opinion.
I can promise you. Every feature you think are great had similar detractors over the years. Every Single One :)
If they actually put effort in F#, it would have reached "unteachable" state already :)
I would've loved an F# that found a way to improve on the performance issues, especially when using computation expressions. That and, either, a deeper integration of .NETs native OOP subtyping, or some form of OCaml-like module system, would have been enough to make it an almost perfect language for my tastes.
Obviously, these are big, and maybe impossible, issues. But Microsoft as a whole never really dedicated enough resources to find out. I feel for the people still working on it, their work is definitely appreciated :)
So is it a F# issue or inherent to functional programming?
Looking at it, the MS docs contain something about this exact topic, so maybe it's better nowadays: https://learn.microsoft.com/en-us/dotnet/fsharp/language-ref...
Sadly, I haven't used F# for years at this point so I can't speak to the current state.
I actually think F# has received some "love" over the recent years contrary to some on this forum; that feature being an example. My view, maybe unpopular but in the age of AI maybe less so, is there is a diminishing returns to language features anyway w.r.t complexity and the use cases that new feature will actually apply for. F# in my mind and many other languages now for that matter is pretty much there or are almost there; the languages are converging. When I used F# I liked how it unified features and tried to keep things simple. Features didn't feel "tacked on" mostly with some later exceptions.
Last time I used F# a few libraries started adopting this for their CE's (e.g. IcedTasks library, etc).
Better than a new language for each task, like you have with Go (microservices) and Dart (GUI).
I'm using F# on a personal project and while it is a great language I think the syntax can be less readable than that of C#. C# code can contain a bit too much boilerplate keywords, but it has a clear structure. Lack of parenthesis in F# make it harder to grasp the structure of the code at a glance.
also called general purpose, general style langue
> that still can't really be done in C#
I would think about it more as them including features other more general purpose languages with a "general" style have adopted then "migrating F# features into C#, as you have mentioned there are major differences between how C# and F# do discriminated sum types.
I.e. it look more like it got inspired by it's competition like e.g. Java (via. sealed interface), Rust (via. enum), TypeScript (via structural typing & literal types) etc.
> Was a functional approach really so 'difficult'?
it was never difficult to use
but it was very different in most aspects
which makes it difficult to push, sell, adapt etc.
that the maybe most wide used functional language (Haskel) has a very bad reputation about being unnecessary complicated and obscure to use with a lot of CS-terminology/pseudo-elitism gate keeping doesn't exactly help. (Also to be clear I'm not saying it has this properties, but it has the reputation, or at least had that reputation for a long time)
Probably more this than any technical reason. More about culture and installed view points.
I don't want to get into the objects/function wars, but do think pretty much every technical problem can be solved better with functions. BUT, it would take an entire industries to re-tool. So think it was more about inertia.
Inertia won.
> I don't really get it
It does each reasonably well (with web APIs being where I think they truly shine).
> Was a functional approach really so 'difficult'
But I do agree. C# is heading to a weird place. At first glance C# looks like a very explicit language, but then you have all the hidden magical tricks: you can't even tell if a (x) => x will be a Func or Expression[0], or if a $"{x}"[1] will actually be evaluated, without looking at the callee's signature.
[0]: https://learn.microsoft.com/en-us/dotnet/csharp/advanced-top...
[1]: https://learn.microsoft.com/en-us/dotnet/csharp/advanced-top...
In the places where that is a thing, I've never needed to care. (Which is kind of the point)
Note that most of its development is still by the open source community and its tooling is an outsider for Visual Studio, where everything else is shared between Visual Basic and C#.
With the official deprecation of VB, and C++/CLI, even though the community keeps going with F#, CLR has changed meaning to C# Language Runtime, for all practical purposes.
Also UWP never officially supported F#, although you could get it running with some hacks.
Similarly with ongoing Native AOT, there are some F# features that break under AOT and might never be rewritten.
A lost opportunity indeed.
Not adding functional features to F# doesn't mean F# would have gained more usage. And if someone wants to use F#, no one is stopping him or her.
But, that would takes, schools changing, companies changing, everything. So it was really the installed base that won, not what was better.
We'd have to go back in time, and have some ML Language win over C++ .
For example, C# chose not to go down the route of type erasure for the sake of generics and because of that you don't get the same sort of runtime type issues that Java might have.
Agreed. Java is on the same trail.
I work much with C# these days and wish C# had as cohesive a syntax story. It often feels like "island of special syntax that makes you fall of a cliff".
For performance-sensitive scenarios where case types include value types, libraries can also implement the non-boxing access pattern by adding a HasValue property and TryGetValue methods. This lets the compiler implement pattern matching without boxing.Even outside of C#, this trend seems to show up everywhere — trying to reduce boilerplate while keeping things safe.
The only thing I wish now is for someone to build a functional Web framework for C#.
Active patterns, computation expressions, structural typing, statically resolved type parameters, explicit inlining, function composition, structural equality, custom operators and much richer generators.
Not sure I would want the last thing in C#, I think having boundaries at the function signature for that.
https://github.com/manifold-systems/manifold/tree/master/man...
In the article, the example with the switch works because it switches on the class of the instance.
The fact that they flatten a union of optional types into an optional union of the corresponding non-optional types is indeed a weird feature, which I do not like, because I think that a union must preserve the structural hierarchy of the united types, e.g. a union of unions must be different from a union of all types included in the component unions, and the same for a union of optional types, where an optional type is equivalent with a union between the void/null type and the non-optional type, but this C# behavior still does not make the C# unions anything else but discriminated unions, even if with a peculiar feature.
This is exactly the difference between simple union types and discriminated unions. This c# feature is what typescript has, not what Haskell/java/f#, etc.
"Discriminated" just means that at run time you can discriminate the values of a union type by their current type, so you can use them correctly in expressions that expect one of the component types.
I agree that the right implementation of discriminated types is that mentioned by you and which is that of many important programming languages, but even if I disapprove of this property that the C# unions have, which in my opinion may lead to unexpected behavior that can cause subtle bugs, the C# unions are still discriminated unions, where you can discriminate the current type at run-time, with a "switch".
In my opinion, one should avoid this weird behavior of C#, by always defining only unions of non-optional types. Where needed, one should then define an optional type having as base a union type. Then these unions will behave like the discriminated unions of other languages.
Whether you use or not this policy, there are types that the C# unions cannot express, but if you use this policy, at least the limitations become explicit.
You can see in the examples, how "switch" uses the implicit discriminant of the union to select the code branch that must be executed, depending on the current type of the value stored in an union.
The syntax of the "switch" seems acceptable, without too many superfluous elements. It could have been made slightly better, by not needing dummy variables that must be used for selecting structure members (or used in whichever else expression is put in the right hand side; the repetition of the dummy variable names could have been avoided if the name of the union variable could have been reused with a different meaning within a "switch", i.e. as a variable of the selected type).
I do not see what else do you want. Perhaps Rust has reused the traditional term "discriminated unions", which had been used for many decades before Rust, and which means the same thing as the more recent terms "tagged unions" or "sum types", with a non-standard meaning and you have that in mind.
> Union types — exhaustive matching over a closed set of types
> Closed hierarchies — exhaustive matching over a sealed class hierarchy
> Closed enums — exhaustive matching over a fixed set of enum values
I believe the last one would be sum types (disc. unions). This one allows overlapping types.
This is in contrast with what is called "union" in the C language, where you must know a priori the type of a value in order to use it correctly.
Moreover, if you can discriminate at run time which is the actual type, that means that you can discriminate between values that happen to have the same representation, but which come from distinct types, e.g. if you had a union between "signed integer" and "unsigned integer", you could discriminate between a signed "3" and an unsigned "3".
This property ensures that the number of possible values for a discriminated union type is the sum of the numbers of possible values for the component types, hence the alternative name "sum type", unlike for a non-discriminated union, where the number of possible values is smaller, because from the sum you must subtract the number of possible values of the intersection sets.
The C# unions described in the article are discriminated unions, except that their component types are not the types literally listed in their definition. If some of the component types are optional, than the true union components are the corresponding non-optional types, together with the null a.k.a. void type, which I find as a rather strange choice.
Or is it becoming a ball-of-mud/bad language compared to its contemporaries?
(Honest questions. I have never used .NET much. I'm curious)
Are there particular things about the ecosystem that you worry about (or have heard about)? Biggest complaint I would have is that it seems like many popular open source libraries in the .NET ecosystem decide to go closed source and commercial once they get popular enough.
(The amount of times I hear "the standard lib is great!" seems more to attempt to defend the plethora of commercial libraries, more than anything)
The community feels rather insular too? The 9-5 dayjob types with employers who don't understand or embrace open source? At my age I can respect that though
And is Postgresql a 2nd-class citizen? If so, your boss will tell you to use SQL Server surely?
I guess it's hard to get a grasp on the state/health of .NET as to me it seems 99.99999% of the code is in private repos companies, as it's not a popular choice for open source projects. Which itself seems like a proxy signal though
> And is Postgresql a 2nd-class citizen?
Microsoft maintains the Npgsql project[0] and I say that it is a very capable, feature rich adapter.
I have not used C# with SQL Server in almost a decade.
It is what it is but I wouldn't say its actually the fault of the language, especially now.
It serves many audiences so it can feel like the language is a jack of all trades and master of none (because it is) and because it is largely backwards compatible over its 20+ years of existence.
That said, I think people make a mountain out of a molehill with respect to keyword sprawl. Depending on what you're building, you really only need to focus on the slice of the language and platform you're working with. If you don't want to use certain language features...just don't use them?
I think it excels in a few areas: web APIs and EF Core being possibly the best ORM out there. For me, it is "just right". Excellent platform tooling, very stable platform, very good performance, hot reload (good, but not perfect), easy to pick up the language if you already know TypeScript[1]; there are many reasons it is a good language and platform.
[0] https://learn.microsoft.com/en-us/dotnet/csharp/advanced-top...
VB.NET's Object was created to support interfacing with COM interop easier. VB.NET's one key niche versus C# for many early years was COM interop through Object.
C#'s dynamic keyword was more directly added as a part of the DLR (Dynamic Language Runtime aka System.Dynamic) effort spurred by IronPython. It had the side benefit of making COM interop easier in C#, but the original purpose was better interop with IronPython, IronRuby, and any other DLR language. That's also why under the hood C#'s dynamic keyword supports a lot of DLR complexity/power. You can do a lot of really interesting things with `System.Dynamic.IDynamicMetaObjectProvider` [1]. The DLR's dependency on `System.Linq.Expressions` also points out to how much further in time the DLR was compared to VB.NET's Object which was "just" the VB7 rename of VB6 Variant originally (it did also pick up DLR support).
The DLR hasn't been invested into in a while, but it was really cool and a bit of an interesting "alternate universe" to still explore.
[0] https://learn.microsoft.com/en-us/dotnet/api/system.dynamic....
That's why I like it so much. And now, I can write mostly functional code.
>I think it excels in a few areas: web APIs and EF Core being possibly the best ORM out there
It's awesome for web stuff and microservices.
> It's awesome for web stuff and microservices.
Is it good at the wrong thing? Eg compare to strongly-typed query generators
> Is it good at the wrong thing?
Your original quote, verbatim:
> Eg compare to strongly-typed query generators
EF is precisely a code to structured query language (SQL) query generator and not a query to code generator.
It has a bad rep because Microsoft could Microsoft as they do.
.NET is a fantastic ecosystem. Has a decent build and dependency system (NuGet, dotnet run/build, declarative builds in XML). Massive standard library, with a consistent and wide focus on correctness, ergonomics, and performance across the board.
You can write everything in many languages, all on the same runtime: business logic in C#; hot paths interfacing with native libraries in C++/CLI; shell wrappers in PowerShell, document attachments with VB, data pipelines in F#.
I feel more people should use it, or at least try it, but sadly it is saddled with the perception that it is Windows-only, which hasn't been true for a decade (also, IMO, not necessarily a negative, because Windows is a decent OS, sue me).
In my experience it does not work very well outside of the sanctioned Linux distributions. Quirky heisenbugs and nonsensical crashes made it virtually unusable for me on Void. I doubt that's changed in the years that have since passed.
> not necessarily a negative, because Windows is a decent OS
Is a language runtime worth an operating system? I think that's a paradigm we left behind in the 1970s when the two were effectively inseperable (and interwoven with hardware!) I wouldn't expect someone to swap to using a Unix system because they really want a better Haskell experience.
I just don't see any actual interesting or meaningful reasons to care about .NET, I effectively feel the same way about it that I do about Go. Just not something that solves any problem I have, and doesn't have anything that interests me. Although effectively I did try it, so it's a moot point considering that's one of the outcomes you're wishing for.
It's open source. Did you follow the spirit of Linux to file a bug report of as much sense of the crashes as you could make? Most OSS only supports as many distros as people are willing to test and file accurate bug reports (and/or scratch the itch themselves and solve it). It seems a bit unfair to expect .NET to magically have a test matrix including every possible distro when almost nothing else does. (It's what keeps distro maintainers employed, testing other people's apps, too.)
It probably has gotten better since then, for what it is worth. .NET has gotten a lot of hardening on Linux and a lot of companies are relying on Linux servers for .NET apps now.
At the very least there are very tiny Alpine-based containers that run .NET considerably well and are very well tested, so Docker is always a strong option for .NET today no matter what Linux distro you want on the "bare metal" running Docker.
Linux distros don't differ too significantly from each other nowadays (systemd plus a different package manager most of the time), so I'm almost sure this is not the source of problems.
Nonetheless, I can only add that we have ridiculous slowdowns in some standard library network calls on Linux, and at that point it is just not true that it will "seamlessly run on Linux", unfortunately.
No, because the only reason I needed C#/.NET to work was to use an internal tool someone before me had written in C#/.NET. It was not really to explore C# or make it usable. I just threw out the old tool, wrote a new one in scheme so I could do my job, and moved on with my life. I don't particularly care about this spirit of Linux, and Microsoft's tooling being weirdly fragile isn't my problem. I assume they already know this is an architectural issue, hence why they specify supported distributions. On principle I believe solving the architectural issue is what they should be concerned about, rather than making new bandaids.
> Most OSS only supports as many distros as people are willing to test and file accurate bug reports
The problem is that most runtimes and standard libraries don't need to specify a notion of a "supported" distribution. At best, they just refer to platforms with pre-made packages while happily pointing other distributions to the git repo. Even complicated, highly abstract and weird ones don't make this kind of distinction. SWI-Prolog and its myriad of frameworks (which includes a full blown GNU Emacs clone) work out of the box anywhere. GHC and the RTS work flawlessly out of the box.
I understand (even if I don't feel the same way) why a comprehensive abstraction layer like .NET is evangelized. All the same I have to consider that it's a product of a multi-trillion dollar corporation, made to compete with the thing whose marketing tagline is "write once, run anywhere". That only makes the distro dependency stand in an even harsher relief, frankly.
You like .NET? Perfectly fine and valid, and I assume it actually works for you. Just indicating that "cross platform" is contingent on more than kernel and cpu architecture here, which is fairly unusual for this type of software. That's before we get into things like comparisons with ocaml, which I know is miserable on Windows and thus is often considered not really something you'd seriously consider using there. The .NET ecosystem essentially has the same problem outside of Windows where the grain and expectations of the tooling are counter-intuitive to the operating system and usual modus operandi of its users.
.NET has a huge kitchen sink standard library. Maybe the closer parallel is Python and Python has had periods where it only supported a few named distributions, too. That's not currently the case, but "how the sausage is made" is still a lot grosser than you might expect, with some Distribution Maintainers maintaining entire forks and a lot of the work not done by Python directly. Python is everywhere because it became one of the favorite shell scripting languages of Distribution Maintainers. (Which also exacerbated the Python 2 to 3 migration because entire distros got stuck on 2 while all their shell scripts got rewritten.) (But also if you want to compare Java's cross platform to .NET's I think we need a long digression into how many Java runtimes there are and the strange and subtle incompatibilities of different distro's affiliations to one or another. I also made the mistake of trying to use a Java application as a regular application in my youth of accidentally dealing with deep distro incompatibilities. That was also not fun.)
I get it, you don't have to like .NET. I just think you have an inflated view of what "cross platform" means when it comes to Linux. Linux isn't just one platform. Most things are rebuilt from source constantly because the binary interfaces especially libc's/glibc's under them are constantly shifting like quicksand. See also the messes that are Flatpak and Snap and how much work they've done to try to build around distro incompatibilities (by building increasingly more complex wish-it-were-VMs).
The build and dependency systems are an abysmal esoteric, poorly documented mix between csproj files, sln files, random scattered json files, etc.
The standard library in my experience sucks and has all sorts of issues, especially around Uris, DateTimes, etc.
And the ecosystem itself has such a low quality bar, ironically _especially_ with anything made by microsoft. For every nuget package that's well-designed, well-documented, and easy-to-use, there's five which have bugs and undocumented exceptions and poorly-designed APIs.
As much as I hate Microsoft, I admit they are doing great things for C#