Szathmáry László honlapja @ DEIK | Nim2 / named tuple with custom type

Previously, we saw this example:

func getMovieInfo(): (string, int, float) =
return ("Total Recall", 1990, 7.5)

let (title, year, score) = getMovieInfo() # !!! on the left side, use parentheses !!!

echo title # Total Recall
echo year # 1990
echo score # 7.5

Let's put a type on the tuple:

type
Movie = tuple[title: string, year: int, score: float]

func getMovieInfo(): Movie =
return (title: "Total Recall", year: 1990, score: 7.5)
# return ("Total Recall", 1990, 7.5) # it would also work

let m = getMovieInfo()

echo m.title # Total Recall
echo m.year # 1990
echo m.score # 7.5

You can also use the following syntax. It's equivalent to the previous:

type
Movie = tuple
title: string
year: int
score: float

However, we might ask ourself: why do we use a tuple here? Why don't we use a normal object?

Same thing with a "class" and object

type
Movie = object
title: string
year: int
score: float

func getMovieInfo(): Movie =
return Movie(title: "Total Recall", year: 1990, score: 7.5)

let m = getMovieInfo()

echo m.title # Total Recall
echo m.year # 1990
echo m.score # 7.5

The main difference is in the construction of the tuple/object:

# tuple:
return ("Total Recall", 1990, 7.5)
# or:
return (title: "Total Recall", year: 1990, score: 7.5)

# object:
return Movie(title: "Total Recall", year: 1990, score: 7.5)

Unfortunately, Nim does not support positional object construction by default. You must use the named field syntax.

Let's add a constructor procedure

What if we don't want to indicate the field names every time when we create a Movie object?

Well, there is an idiomatic Nim pattern for object construction — defining a newTypeName proc that acts as a constructor.

Here it is:

type
Movie = object
title: string
year: int
score: float

# new thing:
proc newMovie(title: string, year: int, score: float): Movie =
return Movie(title: title, year: year, score: score)

func getMovieInfo(): Movie =
#return Movie(title: "Total Recall", year: 1990, score: 7.5)
return newMovie("Total Recall", 1990, 7.5)

let m = getMovieInfo()

echo m.title # Total Recall
echo m.year # 1990
echo m.score # 7.5

Now, the object creation is very similar to the tuple creation:

# tuple:
return ("Total Recall", 1990, 7.5)

# object:
return newMovie("Total Recall", 1990, 7.5)

I have the impression that in most cases, using an object is cleaner and more extensible. However, tuples still have their place when you need destructuring.

Destructuring: tuples can be unpacked with let (title, year, score) = …, while objects cannot.