[FS-1030] - Anonymous records #4499

dsyme · 2018-03-12T19:32:50Z

Replaces #2671

This adds anonymous records to the F# compiler per RFC FS-1030.

Please discuss on the RFC or below - if the discussion gets long we will move to the RFC discussion thread.

The implementation is now complete (apart from copy-and-update) and I believe it to be stable and usable.

Implementation:

Testing:

Unit testing for language service features
LINQ queries
Add tests for negative case error messages
Add more testing where anon types are mentioned in types but not expressions
Add more testing for cross-assembly cases
Testing for Symbols.* FCS API

For trial:

Test suitability of {| X = 1; Y = 2 |} types for working with known common libraries accepting C# anonymous types (e.g. ASP.NET routing? ML.NET?)

Here is the view of an anonymous value in the VS2017 debugger:

If anyone wants to give a hand with the testing that would be great. I think it's really important that someone else besides myself do serious testing on this feature. I'd love someone to try this out and give early feedback on usability and "fit and polish".

dsyme · 2018-03-12T22:19:17Z

Updated to master, and now F# Interactive now works properly:

> let a = {|  X = 1 |};;
val a : {|X : int|} = {X = 1;}

> a = {|  X = 1 |};;
val it : bool = true

> #q;;

majocha · 2018-03-13T16:48:49Z

What is the story with type inference?
At the moment stuff like this

let y = {| V = "hi!" |}
let f x = x.V

doesn't work (asks for type annotation) but lambdas are inferred fine. Is this by design?

smoothdeveloper · 2018-03-13T18:28:25Z

@majocha I think in the example you give, the type inference will only work with record types.

I don't see why it would infer from anonymous type that happens to be defined in same scope.

Lambdas are infered because they actually carry the type of the expression at their usage spot.

the following type annotation should work:

let y = {| V = "hi!" |}
let f (x: {|V:string|}) = x.V

dsyme · 2018-03-13T18:37:03Z

@majocha Yes, type annotations are necessary in cases where type information has not flowed to the point of use. This is a major limitation of the feature

You can leave the types incomplete like this

let f (x: {|V: _|} ) = x.V

but there is no row polymorphism. The addition of row-polymorphism for inlined-code could be considered separately

…o anon-1

dsyme · 2018-03-14T17:59:22Z

OK, this feature is now ready. It would be great if someone could give this a full review, and if people could read the RFC very, very carefully.

Savelenko · 2018-03-14T19:50:26Z

The RFC is a bit unclear about adding fields, seemingly putting this out of scope/orthogonal in one section and giving examples in another. So is it or is it not possible to add fields with with?
Why no pattern matching, is there some technical limitation? Is this somehow connected to “no row polymorphism”, as nominal record pattern matching allows leaving out fields? Especially with nested pattern matching this seems like an “ad-hoc” limitation given that other pattern matching forms are so nicely combinable. I think this special case may be quite annoying to many F# programmers and limits expressivity. For example, not being able to do this:

| Some {| X = x |} -> f x
| None -> y

dsyme · 2018-03-15T00:07:28Z

@Savelenko Thanks for that feedback - I will clarify the RFC and think about pattern matching. You are right that matching a subset of fields is already valid for records and makes sense for anonymous records too. It would not be hard to add.

dsyme · 2018-03-15T00:17:13Z

@Savelenko See fsharp/fslang-design#265, thanks

forki · 2018-03-15T07:52:27Z

I personally would prefer if would just print the set difference here.

isaacabraham · 2018-03-15T07:52:30Z

The RFC mentions abbreviations a couple of times with regards to anonymous records - but I couldn't find any examples of how one would actually define such an abbreviation?

Related to this - how will this function with regards to type inference e.g. imagine calling a function foo:

let foo bar =
    bar.X + bar.Y

where bar is an anonymous record created elsewhere (not in scope). What are our options for type hint to tell the compiler that this is an anonymous record? Will the compiler just assume that this is an anonymous type (which would be a change to current behaviour which would be a compiler error)? If not, how would we indicate this to the compiler - like this?

let foo (bar : {| X : int; Y : int |}) =
    bar.X + bar.Y

forki · 2018-03-15T07:54:48Z

we are missing suggestions here.

forki · 2018-03-15T07:58:45Z

Is it expected that we don't have the pipes in ToString() ?

forki · 2018-03-15T08:07:21Z

maybe we should print the field name in this case?

realvictorprm · 2018-03-15T08:30:54Z

Would be so much fun to change this but I've not enough time :( I agree with your feedback so far @fork! Also pattern matching would be superb. Anyways I'm still not sure how often I'm going to use anonymous records... Am 15.03.2018 9:07 vorm. schrieb "Steffen Forkmann" < notifications@github.com>:

…

[image: image] <https://user-images.githubusercontent.com/57396/37450935-38481280-2830-11e8-8b24-083f326ea556.png> maybe we should print the field name in this case? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#4499 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AYPM8IG-6fdHd3YqZJV77PpeGMWJj28zks5teiE-gaJpZM4SnP8Z> .

realvictorprm · 2018-03-15T08:52:04Z

The CI failes with:

17:16:59 1>    Compiling assembly FSharp.Core, Version=4.4.3.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a (CLR v4.0.30319) ...
17:16:59 2>    Compiling assembly System.ValueTuple, Version=4.0.2.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51 (CLR v4.0.30319) ...
17:16:59 Failed to load dependency System.Threading.Tasks.Dataflow of assembly Microsoft.Build, Version=15.0.0.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a because of the following error : The system cannot find the file specified. (Exception from HRESULT: 0x80070002)
17:16:59 2>    Compiling assembly System.ValueTuple, Version=4.0.2.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51 (CLR v4.0.30319) ...
17:17:03 2>    Compiling assembly FSharp.Core, Version=4.4.3.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a (CLR v4.0.30319) ...
17:17:03 1>    Compiling assembly FSharp.Compiler.Private, Version=10.1.1.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a (CLR v4.0.30319) ...
17:17:03 3>    Compiling assembly FSharp.Core, Version=4.4.3.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a (CLR v4.0.30319) ...
17:17:03 2>    Compiling assembly D:\j\w\release_ci_pa---866fd2c3\release\net40\bin\FSharp.Build.dll (CLR v4.0.30319) ...
17:17:03 3>    Compiling assembly FSharp.Compiler.Server.Shared, Version=15.6.0.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a (CLR v4.0.30319) ...
17:17:33 1>    Compiling assembly D:\j\w\release_ci_pa---866fd2c3\release\net40\bin\fsc.exe (CLR v4.0.30319) ...
17:17:33 2>    Compiling assembly FSharp.Compiler.Private, Version=10.1.1.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a (CLR v4.0.30319) ...
17:17:33 2>    Compiling assembly D:\j\w\release_ci_pa---866fd2c3\release\net40\bin\fsi.exe (CLR v4.0.30319) ...
17:17:33 Failed to load dependency System.Threading.Tasks.Dataflow of assembly Microsoft.Build, Version=15.0.0.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a because of the following error : The system cannot find the file specified. (Exception from HRESULT: 0x80070002)
17:17:33 1>    Compiling assembly FSharp.Core, Version=4.4.3.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a (CLR v4.0.30319) ...
17:17:34 2>    Compiling assembly System.ValueTuple, Version=4.0.2.0, Culture=neutral, PublicKeyToken=cc7b13ffcd2ddd51 (CLR v4.0.30319) ...
17:17:39 1>    Compiling assembly D:\j\w\release_ci_pa---866fd2c3\release\net40\bin\FSharp.Build.dll (CLR v4.0.30319) ...
17:17:39 2>    Compiling assembly FSharp.Core, Version=4.4.3.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a (CLR v4.0.30319) ...
17:17:39 3>    Compiling assembly FSharp.Compiler.Server.Shared, Version=15.6.0.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a (CLR v4.0.30319) ...
17:17:39 2>    Compiling assembly FSharp.Compiler.Private, Version=10.1.1.0, Culture=neutral, PublicKeyToken=b03f5f7f11d50a3a (CLR v4.0.30319) ...
17:18:15 1>    Compiling assembly D:\j\w\release_ci_pa---866fd2c3\release\net40\bin\fsiAnyCpu.exe (CLR v4.0.30319) ...

gibranrosa · 2018-03-15T12:40:31Z

@isaacabraham some posts before @dsyme has already clarified this:
"@majocha Yes, type annotations are necessary in cases where type information has not flowed to the point of use. This is a major limitation of the feature

You can leave the types incomplete like this

let f (x: {|V: _|} ) = x.V"

isaacabraham · 2018-03-15T12:53:46Z

@gibranrosa cheers, I thought as much, but wasn't 100% sure what "abbreviations" meant though :-)

dsyme · 2018-03-15T20:06:40Z

@dotnet-bot Test Windows_NT Release_ci_part2 Build please

cartermp · 2018-11-06T17:10:45Z

@TIHan @KevinRansom can we get a review here? I really want to get this into 16.0 preview 2.

forki · 2018-11-06T18:23:46Z

wow!!! it's merged. It's in. Can't believe it. @ncave @alfonsogarciacaro did you see that?

forki · 2018-11-06T18:24:25Z

ok one is not enough.

cartermp · 2018-11-06T18:28:07Z

This should also be in nightlies for VS shortly. We'd really appreciate further validation work, as we do have some runway with upcoming VS 2019 previews to fix things prior to releasing a new language version.

baronfel · 2018-11-06T18:29:04Z

would be great to get a prerelease of FSharp.Compiler.Tools for us non-VS-usable folks to test as well. What's the process there?

forki · 2018-11-06T18:32:15Z

Yes it needs to go to fcs and then we test it in fable. Transpile to dynamic language will be best test that you can get. Am Di., 6. Nov. 2018, 19:29 hat Chet Husk <notifications@github.com> geschrieben:

…

would be great to get a prerelease of FSharp.Compiler.Tools for us non-VS-usable folks to test as well. What's the process there? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#4499 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AADgNAWTxhwiOdvk5iXzRTW0rj6qo4U8ks5usdTwgaJpZM4SnP8Z> .

vasily-kirichenko · 2018-11-06T18:36:45Z

To test it, we need FSharp.Compiler.Tools built from master.

boeremak · 2018-11-06T19:39:31Z

Wow, this is great, I was having an issue where I could really use this feature and now I come to find out it has been merged an hour ago. Way to go guys. :)

cartermp · 2018-11-10T19:15:18Z

@vasily-kirichenko Feel free to build FCT out of master.

@forki you can build FCS out of master today - AFAIK this is what @Krzysztof-Cieslak does for some Ionide releases.

We'll pull in other PRs related for an F# 4.6 and rev the versions of all assets so we can start shipping an F# 4.6 preview through the .NET CLI as well.

dsyme · 2018-11-13T17:03:54Z

As we've discussed before in a few places it is fairly difficult to roll out updates to all of the required tooling for all editor stacks instantly, even for master branch. FCT (fsharp/fsharp) and FCS (fsharp/FSharp.Compiler.Service) are built and released from master but I only tend to do that once the corresponding VS release corresponding to that master is near final.

But I understand why it would be great to roll these out immediately.

cmeeren · 2019-01-22T10:54:38Z

Has this been released in relative silence? I'm using VS 15.9.5 (no VF# nightlies) and am able to use anonymous records, but I can't find anything about this fantastic feature in any VS release notes.

cartermp · 2019-01-22T16:12:45Z

@cmeeren If you're using it then you have either a nightly build or a self-built VSIX installed. This was not released yet.

cmeeren · 2019-01-22T16:41:18Z

Hm, strange. Does the information below indicate a version other than the one shipped with VS?

Microsoft Visual Studio Professional 2017 
Version 15.9.5
VisualStudio.15.Release/15.9.5+28307.280
Microsoft .NET Framework
Version 4.7.03056

Installed Version: Professional

...

Visual F# Tools 10.2 for F# 4.5   15.8.0.0.  Commit Hash: ace88e6d3ffaf5e4ada0153a56e78ad0a75e82c8.
Microsoft Visual F# Tools 10.2 for F# 4.5

...

cartermp · 2019-01-22T16:53:42Z

Check the extensions page (Tools > Extensions and Updates) to see the version there.

cmeeren · 2019-01-22T17:13:23Z

Would you look at that, I did in fact have a nightly installed... 😅😳 I was certain I had uninstalled it long ago. Sorry for the trouble!

alfonsogarciacaro · 2019-02-01T10:20:40Z

I'm a bit sad about this. Without getting any hint from the F# team and after spending a lot of time to investigate this myself, I've finally realized this has shipped without including anonymous records in the AST exposed by FSharpImplementationFileContents (src/fsharp/symbols/Exprs.fs hasn't be updated except for a couple of lines) so anonymous records remain invisible to Fable. There's only support for symbols in editors 😞

cartermp · 2019-02-01T10:55:02Z

@dsyme any idea what it would take to add the relevant case to this so that fsExpr is updated for Fable?

@alfonsogarciacaro Since it's still in preview things aren't quite baked yet. Though even if this were shipped I believe we could update FCS after the fact.

alfonsogarciacaro · 2019-02-01T23:00:28Z

@cartermp All the signs are the fsharp/FSharp.Compiler.Service repo and the corresponding nuget package are being abandoned. When someone wants to use FCS to test latest features they're encouraged to build it from this repository. However, there's no incentive to maintain the FCS API (beyond the symbols for editor support) when the developing in this repository, as the current situation has shown.

The feeling is that anyone with a project relying on the Expression tree API from FCS should be prepared to become a maintainer of this repository to make sure new features get implemented in the Expression tree too. Something that my already stretched-too-thin time budget for OSS doesn't allow me to do.

cartermp · 2019-02-01T23:29:36Z

I acknowledge the current issue is annoying, but it's important to stress that the feature is still in preview and has only ever been advertised as such, so there's time to make sure we emit something Fable can consume before anything is considered completed.

As for the FCS package/repo, we can put together a proposal to publish from this repository and make it a part of our release checklist. We can discuss the issue of maintaining the Expression Tree API as a part of that, if that's something that would be desirable.

Krzysztof-Cieslak · 2019-02-02T07:41:31Z

As for the FCS package/repo, we can put together a proposal to publish from this repository and make it a part of our release checklist.

I’d rather not have release cycle of FCS coupled with VF#. Don has given NuGet release access to @baronfel so maybe we will see new releases more often. And TBF, we probably just need to write some automated sync process (for example using GitHub actions when they are out of closed preview).

And as of omitting anon records in expression tree... it seems that was just small mistake, no need to make this as example of some deeper problem (https://twitter.com/dsyme/status/1091311874514321408?s=21)

baronfel · 2019-02-02T21:09:52Z

I'd echo what @Krzysztof-Cieslak has said. I'm on top of (and motivated to!) merge FCS with a more regular cadence and I look forward to automating/documenting that flow even moreso than has already been done in that repo. I'm working on making the periodic resyncs of FCS and fsharp/fsharp more painless for the community, but the first step from my perspective to to establish a routine of doing them quickly and painlessly, before automating it all away.

simon-reynolds mentioned this pull request Mar 13, 2018

Migrations support efcore/EFCore.FSharp#2

Closed

anonymous record types implementation

b952b72

dsyme force-pushed the anon-1 branch from 38c5b58 to b952b72 Compare March 13, 2018 18:36

dsyme added 5 commits March 13, 2018 19:07

stable tests

4335cff

fix build

6000c34

Merge branch 'master' of http://github.com/Microsoft/visualfsharp int…

375b422

…o anon-1

copy and update

1ffbee7

more tests

4f424c2

dsyme mentioned this pull request Mar 14, 2018

[WIP] Nested Record Field Copy and Update Expression #4511

Closed

16 tasks

KevinRansom added the Candidate for F# 4.5 label Mar 14, 2018

merge master

b73595a

update ilxgen

9b9f6bf

dsyme reopened this Oct 31, 2018

cartermp requested review from KevinRansom and TIHan November 6, 2018 17:10

KevinRansom approved these changes Nov 6, 2018

View reviewed changes

KevinRansom merged commit e709f52 into dotnet:master Nov 6, 2018

cartermp added this to the 16.0 milestone Nov 6, 2018

cartermp mentioned this pull request Nov 6, 2018

Support anonymous record types [ RFC FS-1030 ] fsharp/fslang-suggestions#207

Closed

This comment has been minimized.

Sign in to view

alfonsogarciacaro mentioned this pull request Feb 1, 2019

Added ParseAndCheckFileInProject fable-compiler/Fable#1714

Merged

[FS-1030] - Anonymous records #4499

[FS-1030] - Anonymous records #4499

Conversation

dsyme commented Mar 12, 2018 • edited by cartermp Loading

dsyme commented Mar 12, 2018

majocha commented Mar 13, 2018

smoothdeveloper commented Mar 13, 2018

dsyme commented Mar 13, 2018 • edited Loading

dsyme commented Mar 14, 2018

Savelenko commented Mar 14, 2018

dsyme commented Mar 15, 2018

dsyme commented Mar 15, 2018

forki commented Mar 15, 2018

isaacabraham commented Mar 15, 2018

forki commented Mar 15, 2018

forki commented Mar 15, 2018

forki commented Mar 15, 2018

realvictorprm commented Mar 15, 2018 via email

realvictorprm commented Mar 15, 2018

gibranrosa commented Mar 15, 2018 • edited Loading

isaacabraham commented Mar 15, 2018

dsyme commented Mar 15, 2018

cartermp commented Nov 6, 2018

forki commented Nov 6, 2018

forki commented Nov 6, 2018

cartermp commented Nov 6, 2018

baronfel commented Nov 6, 2018

forki commented Nov 6, 2018 via email

vasily-kirichenko commented Nov 6, 2018

boeremak commented Nov 6, 2018

cartermp commented Nov 10, 2018

This comment has been minimized.

dsyme commented Nov 13, 2018

cmeeren commented Jan 22, 2019

cartermp commented Jan 22, 2019

cmeeren commented Jan 22, 2019 • edited Loading

cartermp commented Jan 22, 2019

cmeeren commented Jan 22, 2019

alfonsogarciacaro commented Feb 1, 2019

cartermp commented Feb 1, 2019

alfonsogarciacaro commented Feb 1, 2019

cartermp commented Feb 1, 2019

Krzysztof-Cieslak commented Feb 2, 2019

baronfel commented Feb 2, 2019

dsyme commented Mar 12, 2018 •

edited by cartermp

Loading

dsyme commented Mar 13, 2018 •

edited

Loading

gibranrosa commented Mar 15, 2018 •

edited

Loading

cmeeren commented Jan 22, 2019 •

edited

Loading