Object expression

[humhei]

Any ideas about this feature?
See Tests.ObjectExpression for details

[humhei]

This will support follow syntax

///provider.fs
type IColor =
    abstract member Color : string 

type IBarcode =
    abstract member Barcode : string 

type ISize =
    abstract member Size : int

type IItem = 
    abstract member Id : int
    abstract member Name : string
    abstract member Art : string
    abstract member Number : int
[<CLIMutable>]    
type EOrder=
  { Id: int
    Items : IItem list
    OrderNumRange: string }   


///customer.fs 
type Item1 =
    inherit IItem
    inherit IBarcode

type Item2 = 
    inherit IItem
    inherit ISize
    inherit IColor
let item1 = 
    { new Item1 with 
        member this.Id = 0
        member this.Art = "art"
        member this.Name = "name"
        member this.Number = 1000
        member this.Barcode = "7254301" }
let item2 = 
    { new Item2 with 
        member this.Id = 0
        member this.Art = field
        member this.Name = "name"
        member this.Number = 1000
        member this.Color = "red" 
        member this.Size = 39}
let tp = item1.GetType()
FSharpJsonConverter.registerInheritedConverterType<IItem>(tp)            
let eorder = { Id = 1; Items = [item1];  OrderNumRange = "" }

let queryedEOrder =
    db 
    |> LiteRepository.insertItem eorder
    |> LiteRepository.query<EOrder> 
    |> LiteQueryable.first

match queryedEOrder.Items with 
| [item] -> 
    let t = item :? IBarcode
    match item with 
    | :? IBarcode as item1 -> 
        printfn "%A" item1.Barcode
        pass()
    | _ -> fail()    
| _ -> fail()    

the items in queryedEOrder is type of Item1 but not IItem

[Zaid-Ajaj]

Hello @humhei this is very interesting and useful! I need to look closely at the code, at first sight there are a couple of things,

• Can you rename "$itName" to something more descriptive like "$interfaceName" (if that is what it refers to)
• inheritedConverterTypes doesn't have to be concurrent because registrations are done once at startup
• Can you add a couple of examples to the Readme as well?

Thanks a lot!

[humhei]

• Can you rename "$itName" to something more descriptive like "$interfaceName" (if that is what it refers to)

• Can you add a couple of examples to the Readme as well?

I will do it

• inheritedConverterTypes doesn't have to be concurrent because registrations are done once at startup

Are concurrentDictionary has any disadvantage?
I prefer not to fix it
Because i love the syntax of AddOrUpdate and GetOrAdd

[Zaid-Ajaj]

Are concurrentDictionary has any disadvantage?

In practice no, it is ok to keep it 😄

let me know when you finish, and I will merge/publish etc.

[humhei]

I wil fix it tomorrow.
Now i have to go to sleep

[Zaid-Ajaj]

Sure buddy, good night!

[humhei]

@Zaid-Ajaj there are some limitations of fsharp(I think) that restraint this feature
I will explain them in Readme(Just wait for a while)
Please read it and tell me if we open a issue in fsharp suggestion?

[Zaid-Ajaj]

Limitation 1

Yes this is a limitation of F# that is does not have structural equality by default but is this a limitation for this feature?

Please read it and tell me if we open a issue in fsharp suggestion?

No need for a fsharp language feature just for this, it doesn't make sense in the first place to do structural equality on arbitrary interfaces

Limitation 2

That is not a limitation, the code is correct: the list is generated only once, whether you generate it inside the object expression or outside, it doesn't matter

Looking the decompiled C# from compiled IL isn't helpful as this is the result of many compiler optimizations

Can you please remove the details on decompiled C# from the readme? if this is an actual problem (I don't think that is the case) then open an issue

[humhei]

https://github.com/humhei/LiteDB.FSharp/tree/object-expression#the-best-way-i-found-to-solve-this-limitations-now

Why this way cannot be treated by fsharp as default

The codes is verbose but corrently serialized and deserialized

[Zaid-Ajaj]

@humhei It is the same code, the list is generated once, no limitation of F# or anything

[humhei]

@Zaid-Ajaj
The anonymous type generated by fsharp is quite different
Which we use to serialization and deserialization

[humhei]

@Zaid-Ajaj Let's merge it?
This is a feature may only used for myself
But also have no conflits to current repo
So i move the docs to inner docs folder?
Do you have other things propose to change?

[humhei]

Limitation 1
Yes this is a limitation of F# that is does not have structural equality by default but is this a limitation for this feature?
Please read it and tell me if we open a issue in fsharp suggestion?
No need for a fsharp language feature just for this, it doesn't make sense in the first place to do structural equality on arbitrary interfaces

Open a issue in fsharp suggestion may be the last strategy i would take
It seems there already exists too much suggestions about interface and OO feature
Before it i would compile fsharp source code myself try to fix it,
If it's a actual problem, i will open a issue

[Zaid-Ajaj]

Yeah it is ok to add the feature here because it does not conflict with the main api, but in the future I would rather not add it unless many users ask for it (the same with LiteRepository module). So please, next time, let us open a discussion first on feature to decide whether or not to add it before you start working on it

[humhei]

@Zaid-Ajaj finally find the way worked for me
Not using object expression(It's really a hard way to treat with anonymous type)
But using normal type inherientance
It worked well without any limitations

See Tests.typeinherited.fs for detail

[Zaid-Ajaj]

@humhei Nice! Why don't you send a PR then 😉