support F# anonymous types

[enricosada]

Support F# anonymous types in query like

let res = db.QuerySingle<{| Tid: int; Name: string; X:string; Y: int |}>("select Tid=20, Name='fdsf', X='hfds', Y=15")  

• The search for constructor fallback to search by names sorted, if not found by position.
• The instantiation of the constructor must respect order or constructor argument

To initialize the constructor argument correctly, i first tried to reoder the il fields doing the read from datareader to match constructor arguments position, not datareader name position, but i got:

System.InvalidOperationException : Invalid attempt to read from column ordinal '0'.  With CommandBehavior.SequentialAccess, you may only read from column ordinal '2' or greater.

so cannot be done like this, because read need to be sequential by column index.

To maintain the read order, before initializing the constructor, if needed, it will:

• pop from the stack who contains the constructor args to new local variables (the order is known, because is always by column names index)
• push back in the stack the values from the local variables, in the correct order based on constructor arguments position

fix https://github.com/StackExchange/Dapper/issues/1226

[enricosada]

done, @mgravell @NickCraver ready to review

[enricosada]

@mgravell anything i can help with for this PR?

[enricosada]

@NickCraver @mgravell any specific issue with this PR i need to address? happy to help to have it merged

sorry but anonymous types in f# are really handy with dapper, but are really bugged without this PR

[mgravell]

I totally understand. It isn't personal - I simply haven't had chance to look yet - I'm juggling time between multiple projects. The existence of tests in the PR is great. I'll try and look for the next drop. Sorry for delays.

[enricosada]

Np at all @mgravell , i know it may happen, it was just to check if there was something todo to help

meanwhile i'll rebase it

thanks a lot!

[NickCraver]

@enricosada did you see Marc's comment above? We'd love to get this in, but it could be a fair bit more efficient if intent matches.

[enricosada]

@NickCraver no, i missed that sry, let me check. thx for ping

[enricosada]

@mgravell i added some comments, because i not understood your feedback @NickCraver happy to fix, but dunno how to improve this atm, it fix the bug atm and i didnt see a big perf hit for F#

[mgravell]

Hi @enricosada - I got nudged by a mutual friend to remind me to look at this again; sorry for delay. Can I check - the confusion we're talking about - is this the zip vs dual sort vs arrays thing?

My understanding - and please do correct me if I'm wrong here - is that we have two arrays; one of the names, one of the types, matching by index; so if we have 3 columns, we might have

names: "def", "ghi", "abc" types: Int32, String, Single

as I read the code, the intent here is to sort by the names, preserving the respective positions in both arrays, so we end up with:

names: "abc", "def", "ghi" types: Single, Int32, String

is that correct? If so, that's exactly what the dual array version of Array.Sort does:

        string[] names = { "def", "ghi", "abc" };
        Type[] types = { typeof(int), typeof(string), typeof(float) };

        Array.Sort(names, types);

        Console.WriteLine($"names: {string.Join(", ", names)}");
        Console.WriteLine($"types: {string.Join(", ", types.Select(x => x.Name))}");

        /* output:
            names: abc, def, ghi
            types: Single, Int32, String
         */

Have I misunderstood?

[enricosada]

Have I misunderstood?

hi @mgravell

Maybe i didnt understood what the dual array version of Array.Sort does. I tried locally when you mentioned it but didnt understood.

I think that's exactly what i want, sort two arrays based on values of first array and maintaining the sync.

Let me rebase and use Array.Sort

[mgravell]

sort two arrays based on values of first array and maintaining the sync

to confirm, yes: that is an accurate description of what Array.Sort does with the overload that takes two arrays

[enricosada]

@mgravell rebased and applied your feedback, it's now using Array.Sort instead of zip+sort I tested locally with other cases and seems fine too. So if appveyor is green, should be ok (the PR contains a test anyway)

Thanks a lot btw @mgravell , TIL about Array.Sort overload :D

[TravisLeithVeloqx]

Can I ask about the status of this? I would like to use this feature in my project and can clone this branch for now, but only if it is expected to make it into master.

[atlemann]

I just got hit by this as well. Would be awesome if it made it in a new release!

[enricosada]

@mgravell i think this is ready for review and merge.

happy to do any change required

[TravisLeithVeloqx]

I keep coming up against this issue. Is there a release schedule in mind?

[mgravell]

will re-review; sorry for dropping between cracks

[mgravell]

Code seems good; checking the tests locally for regressions - seems fine; you're happy that it is working from F#?

[TravisLeithVeloqx]

@mgravell @enricosada Until now I have not actually attempted to use this. Today I pulled this branch and built it. I added the Dapper.dll to my the project I am currently working on and attempted to use an anonymous record. I get the same parameterless default constructor error.

The offending code is

let parameters = Map ["ids", box ids; "dates", box dates]
use conn = new MySqlConnection(connectionString)

conn
|> dapperMapParametrizedQuery<{|id:int; date:DateTime; variance:float|}> sql parameters
|> Seq.groupBy (fun x -> x.date)
|> Seq.map (fun (dt, sq) -> dt, sq |> Seq.map (fun x -> {id = x.id; variance = x.variance})) |> Ok

So either I have built it wrong, or am using it wrong.

[enricosada]

Let me check your repro

From mobile

-- Enrico Sada @enricosada

---

Da: TravisLeithVeloqx [email protected] Inviato: Thursday, April 16, 2020 8:59:59 AM A: StackExchange/Dapper [email protected] Cc: Enrico Sada [email protected]; Mention [email protected] Oggetto: Re: [StackExchange/Dapper] support F# anonymous types (#1233)

@mgravellhttps://github.com/mgravell @enricosadahttps://github.com/enricosada Until now I have not actually attempted to use this. Today I pulled this branch and built it. I added the Dapper.dll to my the project I am currently working on and attempted to use an anonymous record. I get the same parameterless default constructor error.

The offending code is

let parameters = Map ["ids", box ids; "dates", box dates] use conn = new MySqlConnection(connectionString)

conn |> dapperMapParametrizedQuery<{|id:int; date:DateTime; variance:float|}> sql parameters |> Seq.groupBy (fun x -> x.date) |> Seq.map (fun (dt, sq) -> dt, sq |> Seq.map (fun x -> {id = x.id; variance = x.variance})) |> Ok

So either I have built it wrong, or am using it wrong.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/StackExchange/Dapper/pull/1233#issuecomment-614453188, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AABD6K5GVU7DEJWOPXLI77DRM2UG7ANCNFSM4HECYHLQ.

[travis-leith]

@enricosada is this working?

[NickCraver]

I updated this branch to latest main - still prepared to merge but we don't use F# and would really appreciate some community feedback on whether this works as intended.

[enricosada]

@NickCraver thanks a lot.

I’ll check tomorrow morning

[brase]

@enricosada How can I test it? Should it work with unordered ctor parameters?

[realvictorprm]

pew, has been a while. Maybe I should take a look into that.

[realvictorprm]

@enricosada How can I test it? Should it work with unordered ctor parameters?

So general suggestion from my side, I would have gone down the route and added another test folder with an F# project that depends on Dapper and uses it with anon records to make sure that the feature is functional. Anything against going that far? Alternatively an F# library project could be added, that contains an anon record with a type alias, such that it can be consumed by C# and used for directly piping into the C# functions to test.

[realvictorprm]

@NickCraver are you still open for having this change? If yes, could we reiterate on what you want to see from your side to be changed to get it merged, considering how long it has been since work has been done on this PR & topic?