Nim: Hashes: Adding Macros to Make an User Defined object Hashable

[loloicci]

I want macro used as pragma who makes proc hash(x: UserDefinedObject): Hash.

example:
(hashable is the name of the macro)

```example1.nim
import hashes

type
MyObject {.hashable.} = object
foo: int
bar: string

is converted to
```converted1.nim
import hashes

type
  MyObject = object
    foo: int
    bar: string

proc hash(x: MyObject): Hash =
  ## Computes a Hash from `x`.
  var h: Hash = 0
  h = h !& hash(x.foo)
  h = h !& hash(x.bar)
  result = !$h

another example:

```example2.nim
import hashes

type
MyKind = enum
A, B
MyObject {.hashable.} = object
case kind: MyKind
of A:
foo: int
of B:
bar: string

is converted to
```converted2.nim
import hashes

type
  MyKind = enum
    A, B
  MyObject = object
    case kind: MyKind
    of A:
      foo: int
    of B:
      bar: string

proc hash(x: MyObject): Hash =
  ## Computes a Hash from `x`.
  var h: Hash = 0
  h = h !& hash(x.kind)
  case x.kind
  of A:
    h = h !& hash(x.foo)
  of B:
    h = h !& hash(x.bar)
  result = !$h

[data-man]

And...?

[loloicci]

I hope a new feature like the examples' hashable macro will be implemented.
Sorry for my poor English ;(

[yglukhov]

Somewhat duplicate of #6696

[mratsim]

Why is it won't fix @data-man ? I could see that in sugar.nim

[data-man]

@mratsim
Sorry for my poor English!

[dom96]

I don't think this is possible. Your hashable macro generates code that is outside the scope of the type block. You can already do something like this, which I think makes more sense:

import hashes

hashable:
  type
    MyObject = object
      foo: int
      bar: string

[Bulat-Ziganshin]

@dom96 and on the next day you will need to generate hashable+comparable :)

Until the language itself solved the problem, it may be better to provide meta-macro:

derive hashable, comparable, serializable("BE"):
  type
    MyObject = object
      foo: int
      bar: string

where each submacro only generates extra procedures and derive collects their results + type itself.

[Araq]

@Bulat-Ziganshin sugar.derive seems entirely possible with today's Nim. No new feature required.

[Araq]

Useful feature, but not scheduled for 2020 and the existing workarounds are perfectly acceptable.