Proposal: "single", "first", "last" contextual keywords in LINQ query "select" clauses

[orthoxerox]

Syntax

select-clause
    : 'select' expression
    | 'select' single-item expression
    | 'select' single-item expression 'or' 'default'
    ;

single-item
    : 'single'
    | 'first'
    | 'last'
    ;

Semantics

If single-item is present, then the following translation cases apply:

from x1 in e1
from x2 in e2
select single v

(e1).SelectMany(x1 => e2, (x1, x2) => v).Single()

from x1 in e1
join x2 in e2 on k1 equals k2
select single v

(e1).Join(e2, x1 => k1, x2 => k2, (x1, x2) => v).Single()

from x1 in e1
join x2 in e2 on k1 equals k2 into g
select single v

(e1).GroupJoin(e2, x1 => k1, x2 => k2, (x1, g) => v).Single()

from x in e
select single x;

(e).Single()

from x in e
where q
select single x;

(e).Single(x => q)

from x in e
select single v;

(e).Select(x => v).Single()

[sharwell]

💡 You probably meant to use 'or' 'default' instead of 'or default' in the grammar form.

[orthoxerox]

@sharwell Thanks, fixed

[phi1010]

The use of the variable name "v" does not seem to make any sense here, also, v is not declared in the transpiled code:

from x1 in e1
from x2 in e2
select single v

(e1).SelectMany(x1 => e2, (x1, x2) => v).Single()

I'd expect an assignment of the result, to let this make any sense:

var v = 
from x1 in e1
from x2 in e2
select single;
// results in:
var v = (e1).SelectMany(x1 => e2).Single()

[phi1010]

I'd also expect a definition of the return value -- sorry if I didn't see an approved suggestion yet, that describes implicit returning a tuple of all variables in scope, when the LINQ-query is terminated without a definition of the return value,

var v = 
    from x1 in e1
    from x2 in e2
    select single new {x1,x2};
// results in:
var v = (e1)
    .Select(x1 => new {x1}) // first "from" statement
    .SelectMany(scope => e2, (scope,x2) => new {scope.x1,x2} ) // second "from" statement
    .Single(scope => new {scope.x1,scope.x2}) // "select single" statement

[phi1010]

Also, instead of predefining those methods, could we allow the user to define a method delegate?

This could look like this (For readability, I also renamed the keyword single to only):

var v = 
    from x1 in e1
    from x2 in e2
    select only System.Linq.Enumerable.Single new {x1,x2};

var v = (e1)
    .Select(x1 => new {x1}) // first "from" statement
    .SelectMany(scope => e2, (scope,x2) => new {scope.x1,x2} ) // second "from" statement
    .Select(scope => new {x1 = scope.x1,x2 = scope.x2}) // select statement
    .Single() // "select only System.Linq.Enumerable.Single" modifier, written with extension method sugar

// simplify the FQN by prepending:
using static System.Linq.Enumerable;
// and make the code much shorter:

var v = 
    from x1 in e1
    from x2 in e2
    select only First new {x1,x2};

var v = (e1)
    .Select(x1 => new {x1}) // first "from" statement
    .SelectMany(scope => e2, (scope,x2) => new {scope.x1,x2} ) // second "from" statement
    .Select(scope => new {scope.x1,scope.x2}) // select statement
    .First() // "select only First" modifier

// We can use other definitions:
T SingleOrNull<T>(this IEnumerable<T> @this) { ... };

var v = 
    from x in e
    select only SingleOrNull x;

var v = (e)
    .Select(x => new {x}) // "from" statement
    .Select(scope.x => x) // "select" statement
    .SingleOrNull() // "only SingleOrNull" modifier

[orthoxerox]

v is not a variable, it's an expression. The code snippets aren't examples, they are transformation rules similar to the existing ones defined in the spec.

[phi1010]

Ah yes, I understand; that makes sense.

[amis92]

@orthoxerox I believe 2nd and 3rd example in OP lack the 'single' keyword; I've tried to make sense out of it as it is, and failed - so I assume it's a typo.

[orthoxerox]

@amis92 Thanks, fixed.