Proposal: 'T(literal)' should construct T using the appropriate literal protocol if possible

That makes sense. Thanks for sharing your reasoning.

···

On Thu, Jun 2, 2016 at 1:38 PM, John McCall <rjmccall@apple.com> wrote:

On Jun 2, 2016, at 11:55 AM, Austin Zheng <austinzheng@gmail.com> wrote:
I think we should actually go further.

If this proposal is accepted, disallow "as" as a way of specifying the
type to construct from a literal.

I see no reason to restrict "as" like this. We're not going to stop using
type context as a way of determining the type of a literal, and "as" is a
general feature for providing type context.

If this proposal isn't accepted, disallow using literal values as the
argument to one-unlabeled-argument constructors.

I can't imagine accepting this, either; it would be a major regression in
the usefulness of unlabeled initializers. At best, this would be
appropriate only when the type conforms to an appropriate literal protocol.

In general, Swift gives unlabeled initializers an idiomatic meaning: they
coerce the argument to the target type in a nominally value-preserving
way. One way of viewing this proposal is that it recognizes that there is
an obvious way to do that when the target type supports being directly
formed from the given kind of literal. But if the target type doesn't
support that, coercing a value of some other appropriate literal type is
still a completely reasonable behavior.

John.

Austin

On Thu, Jun 2, 2016 at 11:46 AM, Austin Zheng <austinzheng@gmail.com> > wrote:

+1.

The primary advantage is that it aligns the language semantics with how
most programmers expect this common C-language-family idiom to behave and
removes a potential source of silently wrong code.

The primary disadvantage is that it introduces special-case behavior to
certain types of initializers (although, to be fair, this special-case
behavior is easily recognizable: unlabeled one-argument initializer with a
literal as the argument).

I think the advantage outweighs the disadvantage.

This problem should be addressed one way or another. I prefer this
solution, but if it is rejected for whatever reason we should at least
explicitly outlaw A(literal) syntax in favor of "literal as A".

Austin

On Thu, Jun 2, 2016 at 9:08 AM, John McCall via swift-evolution < >> swift-evolution@swift.org> wrote:

The official way to build a literal of a specific type is to write the
literal in an explicitly-typed context, like so:
    let x: UInt16 = 7
or
    let x = 7 as UInt16

Nonetheless, programmers often try the following:
    UInt16(7)

Unfortunately, this does *not* attempt to construct the value using the
appropriate literal protocol; it instead performs overload resolution using
the standard rules, i.e. considering only single-argument unlabelled
initializers of a type which conforms to IntegerLiteralConvertible. Often
this leads to static ambiguities or, worse, causes the literal to be built
using a default type (such as Int); this may have semantically very
different results which are only caught at runtime.

In my opinion, using this initializer-call syntax to build an
explicitly-typed literal is an obvious and natural choice with several
advantages over the "as" syntax. However, even if you disagree, it's clear
that programmers are going to continue to independently try to use it, so
it's really unfortunate for it to be subtly wrong.

Therefore, I propose that we adopt the following typing rule:

  Given a function call expression of the form A(B) (that is, an
*expr-call* with a single, unlabelled argument) where B is an
*expr-literal* or *expr-collection*, if A has type T.Type for some type
T and there is a declared conformance of T to an appropriate literal
protocol for B, then the expression is always resolves as a literal
construction of type T (as if the expression were written "B as A") rather
than as a general initializer call.

Formally, this would be a special form of the argument conversion
constraint, since the type of the expression A may not be immediately known.

Note that, as specified, it is possible to suppress this typing rule by
wrapping the literal in parentheses. This might seem distasteful; it would
be easy enough to allow the form of B to include extra parentheses. It's
potentially useful to have a way to suppress this rule and get a normal
construction, but there are several other ways of getting that effect, such
as explicitly typing the literal argument (e.g. writing "A(Int(B))").

A conditional conformance counts as a declared conformance even if the
generic arguments are known to not satisfy the conditional conformance.
This permits the applicability of the rule to be decided without having to
first decide the type arguments, which greatly simplifies the type-checking
problem (and may be necessary for soundness; I didn't explore this in
depth, but it certainly feels like a very nasty sort of dependence). We
could potentially weaken this for cases where A is a direct type reference
with bound parameters, e.g. Foo<Int>() or the same with a typealias, but
I think there's some benefit from having a simpler specification, both for
the implementation and for the explicability of the model.

John.

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution

So, you think that this syntax is enticing to new developers who naturally think that the feature works the way that I'm proposing it should work, and you think that the right solution is to make the syntax illegal so that you can more conveniently tell them it doesn't work that way? :)

I think the difference between a cast (which merely reinterprets a value as a compatible type) and a fullwidth conversion (which creates a similar instance of an incompatible type) is very important to understanding how to write Swift, and we shouldn't muddy the waters by creating a magic syntax.

My point is that a literal doesn't really have a pre-existing type. Your explanation relies on a very strange model in which we assign it an arbitrary type just so we can introduce unwanted semantic conversions from that type.

You can still tell them that it's a struct and you're calling an initializer on it; it's just that the initializer chosen is the special literal initializer because the argument is a literal.

If you're planning to change `IntegerLiteralConvertible` and friends to require a fullwidth conversion initializer like `init(_ value: IntegerLiteralType)`, then this is simply an overload resolution rule. In that case, I think your proposal is fine.

This does not work. Literal construction is not necessarily as simple as converting from an existing type, and it's important to use a semantically distinguished initializer.

But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.

It sounds like you're not in favor of the proposal, then.

John.

···

On Jun 2, 2016, at 4:22 PM, Brent Royal-Gordon <brent@architechies.com> wrote:

The IntegerLiteral type idea might be worth exploring. It does seem to
provide some additional consistency. For example, wasn't it clarified on
this list just recently that literals don't have a type and adopt one based
on context? It'd be nice to point out that 42 is an IntegerLiteral when
explaining that it's not an Int.

···

On Thu, Jun 2, 2016 at 18:22 Brent Royal-Gordon via swift-evolution < swift-evolution@swift.org> wrote:

> So, you think that this syntax is enticing to new developers who
naturally think that the feature works the way that I'm proposing it should
work, and you think that the right solution is to make the syntax illegal
so that you can more conveniently tell them it doesn't work that way? :)

I think the difference between a cast (which merely reinterprets a value
as a compatible type) and a fullwidth conversion (which creates a similar
instance of an incompatible type) is very important to understanding how to
write Swift, and we shouldn't muddy the waters by creating a magic syntax.

> You can still tell them that it's a struct and you're calling an
initializer on it; it's just that the initializer chosen is the special
literal initializer because the argument is a literal.

If you're planning to change `IntegerLiteralConvertible` and friends to
require a fullwidth conversion initializer like `init(_ value:
IntegerLiteralType)`, then this is simply an overload resolution rule. In
that case, I think your proposal is fine.

But if you're going to call `init(integerLiteral:)` like it's `init(_:)`,
I don't think that's a good idea. Parameter labels are supposed to be
significant; we don't want to lose that.

--
Brent Royal-Gordon
Architechies

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution

> Like I said, the standard library and compiler conspire to make sure that easy cases like this are caught at compile time, but that would not help non-standard types that conform to IntegerLiteralConvertible.
>
> Also, even for standard types, the syntax only works statically if the literal fits in the range of Int, which may not be a superset of the desired type. For example, UInt64(0x10000000000) would not work on a 32-bit platform. It is diagnosed statically, however.

I believe I understand the problem you described, but I really can't figure out how this problem can produce a unexpected behavior and run-time errors, as was stated in your initial message. So, this is why I was asking for any code that can prove this. The example with UInt64(0x10000000000) on 32bit systems will raise error at _compilation_ time. Could someone provide any code to illustrate the possible problems at run-time?

It's sufficient to do the same thing as the standard library, just without the @transparent tricks that enable static detection of the failure:

  struct WidgetCount : IntegerLiteralConvertible {
    var value: UInt64
    init(integerLiteral value: UInt64) {
      self.value = value
    }
    init(_ value: Int) {
      self.value = UInt64(value)
    }
  }

  let count = WidgetCount(-1)

John.

···

On Jun 2, 2016, at 10:28 PM, Vladimir.S <svabox@gmail.com> wrote:

For others, who not really understand the issue(probably I'm not the one, I hope ;-) ) : If we'd have Int128 type, we can't create an instance of it in this form:
let x = Int128(92233720368547758070)
(92233720368547758070 == Int.max * 10)
as '92233720368547758070' literal will be treated always as of Int type.

In more general description, the difference between UIntN(xxx) and yyy as UIntN is that xxx will be treated as Int(so it can't be greater than Int.max for example) then this Int will be sent to UIntN(:Int) initializer and then we have UIntN as a result of call to initializer; yyy will be treated as UIntN literal just by its definition, no calling to any initializer, yyy can be of any value allowed for UIntN.

But again, could someone help with examples when this can produce problems at run-time?

On 03.06.2016 1:12, John McCall wrote:

On Jun 2, 2016, at 2:36 PM, Vladimir.S <svabox@gmail.com> wrote:
Well, I understand that it seems like there is some problem with UIntN() initialization, but can't find any simple code that will demonstrate this..

All below works as expected:

var x1: Int32 = 0
var x2 = Int32(0)

print(x1.dynamicType, x2.dynamicType) // Int32 Int32

// integer overflows when converted from 'Int' to 'UInt16'
//var x = UInt16(100_000)
//var x = UInt16(-10)

// negative integer cannot be converted to unsigned type 'UInt64'
// var x = UInt64(-1)

So, what code will produce some unexpected behavior / error at runtime?

Like I said, the standard library and compiler conspire to make sure that easy cases like this are caught at compile time, but that would not help non-standard types that conform to IntegerLiteralConvertible.

Also, even for standard types, the syntax only works statically if the literal fits in the range of Int, which may not be a superset of the desired type. For example, UInt64(0x10000000000) would not work on a 32-bit platform. It is diagnosed statically, however.

John.

On 03.06.2016 0:25, John McCall wrote:

On Jun 2, 2016, at 1:56 PM, Vladimir.S <svabox@gmail.com> wrote:

Often
this leads to static ambiguities or, worse, causes the literal to be built
using a default type (such as Int); this may have semantically very
different results which are only caught at runtime.

Seems like I'm very slow today.. Could you present a couple of examples where such initialization(like UInt16(7)) can produce some unexpected behavior / error at runtime?

UIntN has unlabeled initializers taking all of the standard integer types, including itself. The literal type will therefore get defaulted to Int. The legal range of values for Int may not be a superset of the legal range of values for UIntN. If the literal is in the legal range for an Int but not for the target type, this might trap at runtime. Now, for a built-in integer type like UInt16, we will recognize that the coercion always traps and emit an error at compile-time, but this generally won't apply to other types.

John.

On 02.06.2016 19:08, John McCall via swift-evolution wrote:

The official way to build a literal of a specific type is to write the
literal in an explicitly-typed context, like so:
let x: UInt16 = 7
or
let x = 7 as UInt16

Nonetheless, programmers often try the following:
UInt16(7)

Unfortunately, this does /not/ attempt to construct the value using the
appropriate literal protocol; it instead performs overload resolution using
the standard rules, i.e. considering only single-argument unlabelled
initializers of a type which conforms to IntegerLiteralConvertible. Often
this leads to static ambiguities or, worse, causes the literal to be built
using a default type (such as Int); this may have semantically very
different results which are only caught at runtime.

In my opinion, using this initializer-call syntax to build an
explicitly-typed literal is an obvious and natural choice with several
advantages over the "as" syntax. However, even if you disagree, it's clear
that programmers are going to continue to independently try to use it, so
it's really unfortunate for it to be subtly wrong.

Therefore, I propose that we adopt the following typing rule:

Given a function call expression of the form A(B) (that is, an
/expr-call/ with a single, unlabelled argument) where B is
an /expr-literal/ or /expr-collection/, if A has type T.Type for some type
T and there is a declared conformance of T to an appropriate literal
protocol for B, then the expression is always resolves as a literal
construction of type T (as if the expression were written "B as A") rather
than as a general initializer call.

Formally, this would be a special form of the argument conversion
constraint, since the type of the expression A may not be immediately known.

Note that, as specified, it is possible to suppress this typing rule by
wrapping the literal in parentheses. This might seem distasteful; it would
be easy enough to allow the form of B to include extra parentheses. It's
potentially useful to have a way to suppress this rule and get a normal
construction, but there are several other ways of getting that effect, such
as explicitly typing the literal argument (e.g. writing "A(Int(B))").

A conditional conformance counts as a declared conformance even if the
generic arguments are known to not satisfy the conditional conformance.
This permits the applicability of the rule to be decided without having to
first decide the type arguments, which greatly simplifies the type-checking
problem (and may be necessary for soundness; I didn't explore this in
depth, but it certainly feels like a very nasty sort of dependence). We
could potentially weaken this for cases where A is a direct type reference
with bound parameters, e.g. Foo<Int>() or the same with a typealias, but
I think there's some benefit from having a simpler specification, both for
the implementation and for the explicability of the model.

John.

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution

.

The official way to build a literal of a specific type is to write the
literal in an explicitly-typed context, like so:

  let x: UInt16 = 7
or
  let x = 7 as UInt16

Nonetheless, programmers often try the following:
  UInt16(7)

Unfortunately, this does not attempt to construct the value using the
appropriate literal protocol; it instead performs overload resolution
using the standard rules, i.e. considering only single-argument
unlabelled initializers of a type which conforms to
IntegerLiteralConvertible. Often this leads to static ambiguities or,
worse, causes the literal to be built using a default type (such as
Int); this may have semantically very different results which are only
caught at runtime.

In my opinion, using this initializer-call syntax to build an
explicitly-typed literal is an obvious and natural choice with several
advantages over the "as" syntax. However, even if you disagree, it's
clear that programmers are going to continue to independently try to
use it, so it's really unfortunate for it to be subtly wrong.

Therefore, I propose that we adopt the following typing rule:

Given a function call expression of the form A(B) (that is, an
expr-call with a single, unlabelled argument) where B is an
expr-literal or expr-collection, if A has type T.Type for some type T
and there is a declared conformance of T to an appropriate literal
protocol for B, then the expression is always resolves as a literal
construction of type T (as if the expression were written "B as A")
rather than as a general initializer call.

Formally, this would be a special form of the argument conversion
constraint, since the type of the expression A may not be immediately
known.

I realize this is somewhat tangential, but... IMO this may not be entirely
about literals.

We have a standard that full-width type conversions are written as a
label-free initializer
<Swift.org - API Design Guidelines.
I believe that is partly responsible for setting up the expectation that
Int(42) works as one would expect. It gets ultra-weird when you can
convert from type A to type B using B(someA) but you can't write
B(someB). We should automatically generate a label-free “copy
initializer” for value types, to complete implementation of the expected
mental model.

That may also be a good idea, but it won't magically be preferred for
literal construction if the type has any other constructors of
literal-convertible type.

I know. I'm saying, fixing this for literals without giving value types
copy initializers leaves us with only a partial realization of a larger
mental model to which I believe people are programming.

That's fair. Would you like me to incorporate that into my proposal, then? I see the relation, but it's a pretty significant jump in scope.

John.

···

On Jun 7, 2016, at 4:25 PM, Dave Abrahams <dabrahams@apple.com> wrote:
on Tue Jun 07 2016, John McCall <rjmccall-AT-apple.com> wrote:

On Jun 5, 2016, at 5:18 PM, Dave Abrahams via swift-evolution <swift-evolution@swift.org> wrote:
on Thu Jun 02 2016, John McCall <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:

Note that, as specified, it is possible to suppress this typing rule
by wrapping the literal in parentheses. This might seem distasteful;
it would be easy enough to allow the form of B to include extra
parentheses. It's potentially useful to have a way to suppress this
rule and get a normal construction, but there are several other ways
of getting that effect, such as explicitly typing the literal argument
(e.g. writing "A(Int(B))").

A conditional conformance counts as a declared conformance even if the
generic arguments are known to not satisfy the conditional
conformance. This permits the applicability of the rule to be decided
without having to first decide the type arguments, which greatly
simplifies the type-checking problem (and may be necessary for
soundness; I didn't explore this in depth, but it certainly feels like
a very nasty sort of dependence). We could potentially weaken this
for cases where A is a direct type reference with bound parameters,
e.g. Foo<Int>() or the same with a typealias, but I think there's
some benefit from having a simpler specification, both for the
implementation and for the explicability of the model.

John.
_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution

--
-Dave

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org <mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution

--
Dave

The IntegerLiteral type idea might be worth exploring. It does seem to provide some additional consistency. For example, wasn't it clarified on this list just recently that literals don't have a type and adopt one based on context? It'd be nice to point out that 42 is an IntegerLiteral when explaining that it's not an Int.

I think that's a very promising way of thinking about literals. Writing a literal creates a notional value whose type is the informal, infinite-precise type of all integer/FP/collection/etc. literals, which (1) can be implicitly converted to any type that implements the appropriate protocol and (2) in fact *must* be converted to some such type (possibly the default type for that literal) in order for the code to be executable. You can then think about this proposal as saying that an explicit conversion from that informal type to a literal-convertible type follows the same path as an implicit conversion would have.

John.

···

On Jun 2, 2016, at 4:33 PM, Xiaodi Wu <xiaodi.wu@gmail.com> wrote:

On Thu, Jun 2, 2016 at 18:22 Brent Royal-Gordon via swift-evolution <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:
> So, you think that this syntax is enticing to new developers who naturally think that the feature works the way that I'm proposing it should work, and you think that the right solution is to make the syntax illegal so that you can more conveniently tell them it doesn't work that way? :)

I think the difference between a cast (which merely reinterprets a value as a compatible type) and a fullwidth conversion (which creates a similar instance of an incompatible type) is very important to understanding how to write Swift, and we shouldn't muddy the waters by creating a magic syntax.

> You can still tell them that it's a struct and you're calling an initializer on it; it's just that the initializer chosen is the special literal initializer because the argument is a literal.

If you're planning to change `IntegerLiteralConvertible` and friends to require a fullwidth conversion initializer like `init(_ value: IntegerLiteralType)`, then this is simply an overload resolution rule. In that case, I think your proposal is fine.

But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.

--
Brent Royal-Gordon
Architechies

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org <mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution

The IntegerLiteral type idea might be worth exploring. It does seem to provide some additional consistency. For example, wasn't it clarified on this list just recently that literals don't have a type and adopt one based on context? It'd be nice to point out that 42 is an IntegerLiteral when explaining that it's not an Int.

I think that's a very promising way of thinking about literals. Writing a literal creates a notional value whose type is the informal, infinite-precise type of all integer/FP/collection/etc. literals, which (1) can be implicitly converted to any type that implements the appropriate protocol and (2) in fact *must* be converted to some such type (possibly the default type for that literal) in order for the code to be executable. You can then think about this proposal as saying that an explicit conversion from that informal type to a literal-convertible type follows the same path as an implicit conversion would have.

It sounds like the reason you don't want to remove the label is because it distinguishes the literal initializer and you don't want it called with a variable accidentally right?

What if we do something like this:

init(_ value: IntegerLiteralType, literal: IntegerLiteral = #literal)

The idea here is that the IntegerLiteral type and friends are special types that have no members and cannot be used except by the literal convertible protocols and their conformances. There is nothing you can do with the value at all.

The trick is that by defaulting it to #literal we are indicating the compiler synthesizes the logical value, as with #file, #line, etc. The compiler would refuse to synthesize this value when a literal is not used. This means it is never selected when the user provides a variable to a converting initializer. If an independent initializer accepting a value of the same type also exists that would be selected. However, when a literal *is* used and the type conforms to the relevant literal convertible protocol the compiler always synthesized the value making it always the most specific overload.

Of course no runtime value would actually exist. This is just a logical value marking the fact that a literal was used to call the initializer.

This approach solves the same problem while retaining semantic consistency with the language (no label elision or short circuit of overload resolution). The magic is arguably a lot more restrained - types for which values can only be supplied by the compiler. We could artificially restrict usage of these types if we wanted to, but we wouldn't have to. Nothing could be accomplished by using the types anywhere else so nobody do so and it wouldn't be actively harmful to allow them to be used anywhere other types can be used. Only the ability to create values of the type needs to be restricted and we can already write types like that by marking the initializers private.

Any thoughts on this approach?

···

Sent from my iPad

On Jun 2, 2016, at 6:51 PM, John McCall via swift-evolution <swift-evolution@swift.org> wrote:
On Jun 2, 2016, at 4:33 PM, Xiaodi Wu <xiaodi.wu@gmail.com> wrote:

John.

On Thu, Jun 2, 2016 at 18:22 Brent Royal-Gordon via swift-evolution <swift-evolution@swift.org> wrote:
> So, you think that this syntax is enticing to new developers who naturally think that the feature works the way that I'm proposing it should work, and you think that the right solution is to make the syntax illegal so that you can more conveniently tell them it doesn't work that way? :)

I think the difference between a cast (which merely reinterprets a value as a compatible type) and a fullwidth conversion (which creates a similar instance of an incompatible type) is very important to understanding how to write Swift, and we shouldn't muddy the waters by creating a magic syntax.

> You can still tell them that it's a struct and you're calling an initializer on it; it's just that the initializer chosen is the special literal initializer because the argument is a literal.

If you're planning to change `IntegerLiteralConvertible` and friends to require a fullwidth conversion initializer like `init(_ value: IntegerLiteralType)`, then this is simply an overload resolution rule. In that case, I think your proposal is fine.

But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.

--
Brent Royal-Gordon
Architechies

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution

The IntegerLiteral type idea might be worth exploring. It does seem to provide some additional consistency. For example, wasn't it clarified on this list just recently that literals don't have a type and adopt one based on context? It'd be nice to point out that 42 is an IntegerLiteral when explaining that it's not an Int.

I think that's a very promising way of thinking about literals. Writing a literal creates a notional value whose type is the informal, infinite-precise type of all integer/FP/collection/etc. literals, which (1) can be implicitly converted to any type that implements the appropriate protocol and (2) in fact *must* be converted to some such type (possibly the default type for that literal) in order for the code to be executable. You can then think about this proposal as saying that an explicit conversion from that informal type to a literal-convertible type follows the same path as an implicit conversion would have.

It sounds like the reason you don't want to remove the label is because it distinguishes the literal initializer and you don't want it called with a variable accidentally right?

What if we do something like this:

init(_ value: IntegerLiteralType, literal: IntegerLiteral = #literal)

The idea here is that the IntegerLiteral type and friends are special types that have no members and cannot be used except by the literal convertible protocols and their conformances. There is nothing you can do with the value at all.

The trick is that by defaulting it to #literal we are indicating the compiler synthesizes the logical value, as with #file, #line, etc. The compiler would refuse to synthesize this value when a literal is not used. This means it is never selected when the user provides a variable to a converting initializer. If an independent initializer accepting a value of the same type also exists that would be selected. However, when a literal *is* used and the type conforms to the relevant literal convertible protocol the compiler always synthesized the value making it always the most specific overload.

Of course no runtime value would actually exist. This is just a logical value marking the fact that a literal was used to call the initializer.

This approach solves the same problem while retaining semantic consistency with the language (no label elision or short circuit of overload resolution). The magic is arguably a lot more restrained - types for which values can only be supplied by the compiler. We could artificially restrict usage of these types if we wanted to, but we wouldn't have to. Nothing could be accomplished by using the types anywhere else so nobody do so and it wouldn't be actively harmful to allow them to be used anywhere other types can be used. Only the ability to create values of the type needs to be restricted and we can already write types like that by marking the initializers private.

Any thoughts on this approach?

This is still a special-case type-checking rule, which means that it's still basically my proposal except, instead of just rewriting the call to use a labeled literal initializer, it rewrites the call to use a magic initializer which cannot be used by users because it requires arguments that users cannot create. I just don't understand the motivation here. It's not a simpler language model to use, explain, or implement; it purports to be conceptually simpler and less magical while actually inventing two magic new language concepts (IntegerLiteral and #literal) on top of the same special cases.

Literals are always going to involve some magic behind the scenes. The non-trivial literals, for example, all actually traffic through some sort of builtin protocol (e.g. BuiltinIntegerLiteralConvertible) which cannot be implemented by users and whose construction necessarily involves a private API contract between the standard library and compiler that we reserve the right to change at will without bringing in swift-evolution. All of the supposedly simple explanations eventually reach, "okay, but if Int isn't a special type, how does the literal turn into an Int in the first place", and that is just unanswerable, because it turns out that standard library types *are* actually special in the sense that they get to conform to BuiltinIntegerLiteralConvertible and nobody else does.

John.

···

On Jun 2, 2016, at 6:48 PM, Matthew Johnson <matthew@anandabits.com> wrote:
On Jun 2, 2016, at 6:51 PM, John McCall via swift-evolution <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:

On Jun 2, 2016, at 4:33 PM, Xiaodi Wu <xiaodi.wu@gmail.com <mailto:xiaodi.wu@gmail.com>> wrote:

John.

On Thu, Jun 2, 2016 at 18:22 Brent Royal-Gordon via swift-evolution <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:
> So, you think that this syntax is enticing to new developers who naturally think that the feature works the way that I'm proposing it should work, and you think that the right solution is to make the syntax illegal so that you can more conveniently tell them it doesn't work that way? :)

I think the difference between a cast (which merely reinterprets a value as a compatible type) and a fullwidth conversion (which creates a similar instance of an incompatible type) is very important to understanding how to write Swift, and we shouldn't muddy the waters by creating a magic syntax.

> You can still tell them that it's a struct and you're calling an initializer on it; it's just that the initializer chosen is the special literal initializer because the argument is a literal.

If you're planning to change `IntegerLiteralConvertible` and friends to require a fullwidth conversion initializer like `init(_ value: IntegerLiteralType)`, then this is simply an overload resolution rule. In that case, I think your proposal is fine.

But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.

--
Brent Royal-Gordon
Architechies

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org <mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org <mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution

John McCall wrote:

I think that's a very promising way of thinking about literals. Writing
a literal creates a notional value whose type is the informal, infinite-
precise type of all integer/FP/collection/etc. literals, which (1) can be
implicitly converted to any type that implements the appropriate protocol
and (2) in fact *must* be converted to some such type (possibly the
default type for that literal) in order for the code to be executable.

Could you allow IntegerLiteralConvertible.IntegerLiteralType associatedtype
to override the default Swift.IntegerLiteralType typealias iff there's more
than one unlabelled init(_:) to choose from? Then you can call the "correct"
init(_:) instead of calling init(integerLiteral:) as a "special case".

[stdlib/public/core/CompilerProtocols.swift]

    /// Conforming types can be initialized with integer literals.
    public protocol IntegerLiteralConvertible {
        associatedtype IntegerLiteralType : _BuiltinIntegerLiteralConvertible
        /// Create an instance initialized to `value`.
        init(integerLiteral value: IntegerLiteralType)
    }

[stdlib/public/core/Policy.swift]

    /// The default type for an otherwise-unconstrained integer literal.
    public typealias IntegerLiteralType = Int

-- Ben

Like I said, the standard library and compiler conspire to make sure that easy cases like this are caught at compile time, but that would not help non-standard types that conform to IntegerLiteralConvertible.

Also, even for standard types, the syntax only works statically if the literal fits in the range of Int, which may not be a superset of the desired type. For example, UInt64(0x10000000000) would not work on a 32-bit platform. It is diagnosed statically, however.

I believe I understand the problem you described, but I really can't figure out how this problem can produce a unexpected behavior and run-time errors, as was stated in your initial message. So, this is why I was asking for any code that can prove this. The example with UInt64(0x10000000000) on 32bit systems will raise error at _compilation_ time. Could someone provide any code to illustrate the possible problems at run-time?

It's sufficient to do the same thing as the standard library, just without the @transparent tricks that enable static detection of the failure:

Thank you, but I was awaiting the example where using of UIntN/IntN() with integer literal produces *unexpected* error at run-time or unexpected behavior (as was stated in your initial message). I don't understand how the example you sent is connected with this. Could some one provide a sample code to understand in which situations we can have described problems when using UIntN/IntN with integer literals?

About your example : you have init(:Int), you sent -1 to it, UInt64(value) will generate *expected* run time error in this case.

···

On 03.06.2016 21:09, John McCall wrote:

On Jun 2, 2016, at 10:28 PM, Vladimir.S <svabox@gmail.com> wrote:

  struct WidgetCount : IntegerLiteralConvertible {
    var value: UInt64
    init(integerLiteral value: UInt64) {
      self.value = value
    }
    init(_ value: Int) {
      self.value = UInt64(value)
    }
  }

  let count = WidgetCount(-1)

John.

For others, who not really understand the issue(probably I'm not the one, I hope ;-) ) : If we'd have Int128 type, we can't create an instance of it in this form:
let x = Int128(92233720368547758070)
(92233720368547758070 == Int.max * 10)
as '92233720368547758070' literal will be treated always as of Int type.

In more general description, the difference between UIntN(xxx) and yyy as UIntN is that xxx will be treated as Int(so it can't be greater than Int.max for example) then this Int will be sent to UIntN(:Int) initializer and then we have UIntN as a result of call to initializer; yyy will be treated as UIntN literal just by its definition, no calling to any initializer, yyy can be of any value allowed for UIntN.

But again, could someone help with examples when this can produce problems at run-time?

On 03.06.2016 1:12, John McCall wrote:

On Jun 2, 2016, at 2:36 PM, Vladimir.S <svabox@gmail.com> wrote:
Well, I understand that it seems like there is some problem with UIntN() initialization, but can't find any simple code that will demonstrate this..

All below works as expected:

var x1: Int32 = 0
var x2 = Int32(0)

print(x1.dynamicType, x2.dynamicType) // Int32 Int32

// integer overflows when converted from 'Int' to 'UInt16'
//var x = UInt16(100_000)
//var x = UInt16(-10)

// negative integer cannot be converted to unsigned type 'UInt64'
// var x = UInt64(-1)

So, what code will produce some unexpected behavior / error at runtime?

Like I said, the standard library and compiler conspire to make sure that easy cases like this are caught at compile time, but that would not help non-standard types that conform to IntegerLiteralConvertible.

Also, even for standard types, the syntax only works statically if the literal fits in the range of Int, which may not be a superset of the desired type. For example, UInt64(0x10000000000) would not work on a 32-bit platform. It is diagnosed statically, however.

John.

On 03.06.2016 0:25, John McCall wrote:

On Jun 2, 2016, at 1:56 PM, Vladimir.S <svabox@gmail.com> wrote:

Often
this leads to static ambiguities or, worse, causes the literal to be built
using a default type (such as Int); this may have semantically very
different results which are only caught at runtime.

Seems like I'm very slow today.. Could you present a couple of examples where such initialization(like UInt16(7)) can produce some unexpected behavior / error at runtime?

UIntN has unlabeled initializers taking all of the standard integer types, including itself. The literal type will therefore get defaulted to Int. The legal range of values for Int may not be a superset of the legal range of values for UIntN. If the literal is in the legal range for an Int but not for the target type, this might trap at runtime. Now, for a built-in integer type like UInt16, we will recognize that the coercion always traps and emit an error at compile-time, but this generally won't apply to other types.

John.

On 02.06.2016 19:08, John McCall via swift-evolution wrote:

The official way to build a literal of a specific type is to write the
literal in an explicitly-typed context, like so:
let x: UInt16 = 7
or
let x = 7 as UInt16

Nonetheless, programmers often try the following:
UInt16(7)

Unfortunately, this does /not/ attempt to construct the value using the
appropriate literal protocol; it instead performs overload resolution using
the standard rules, i.e. considering only single-argument unlabelled
initializers of a type which conforms to IntegerLiteralConvertible. Often
this leads to static ambiguities or, worse, causes the literal to be built
using a default type (such as Int); this may have semantically very
different results which are only caught at runtime.

In my opinion, using this initializer-call syntax to build an
explicitly-typed literal is an obvious and natural choice with several
advantages over the "as" syntax. However, even if you disagree, it's clear
that programmers are going to continue to independently try to use it, so
it's really unfortunate for it to be subtly wrong.

Therefore, I propose that we adopt the following typing rule:

Given a function call expression of the form A(B) (that is, an
/expr-call/ with a single, unlabelled argument) where B is
an /expr-literal/ or /expr-collection/, if A has type T.Type for some type
T and there is a declared conformance of T to an appropriate literal
protocol for B, then the expression is always resolves as a literal
construction of type T (as if the expression were written "B as A") rather
than as a general initializer call.

Formally, this would be a special form of the argument conversion
constraint, since the type of the expression A may not be immediately known.

Note that, as specified, it is possible to suppress this typing rule by
wrapping the literal in parentheses. This might seem distasteful; it would
be easy enough to allow the form of B to include extra parentheses. It's
potentially useful to have a way to suppress this rule and get a normal
construction, but there are several other ways of getting that effect, such
as explicitly typing the literal argument (e.g. writing "A(Int(B))").

A conditional conformance counts as a declared conformance even if the
generic arguments are known to not satisfy the conditional conformance.
This permits the applicability of the rule to be decided without having to
first decide the type arguments, which greatly simplifies the type-checking
problem (and may be necessary for soundness; I didn't explore this in
depth, but it certainly feels like a very nasty sort of dependence). We
could potentially weaken this for cases where A is a direct type reference
with bound parameters, e.g. Foo<Int>() or the same with a typealias, but
I think there's some benefit from having a simpler specification, both for
the implementation and for the explicability of the model.

John.

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution

.

.

I'm not sure it should be in the same proposal; I just want it to be a
part of the discussion. We should be able to make incremental progress,
but let's make sure we understand how our changes fit into the big picture.

···

on Wed Jun 08 2016, John McCall <rjmccall-AT-apple.com> wrote:

On Jun 7, 2016, at 4:25 PM, Dave Abrahams <dabrahams@apple.com> wrote:
on Tue Jun 07 2016, John McCall <rjmccall-AT-apple.com> wrote:

On Jun 5, 2016, at 5:18 PM, Dave Abrahams via swift-evolution <swift-evolution@swift.org> wrote:
on Thu Jun 02 2016, John McCall <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:

The official way to build a literal of a specific type is to write the
literal in an explicitly-typed context, like so:

  let x: UInt16 = 7
or
  let x = 7 as UInt16

Nonetheless, programmers often try the following:
  UInt16(7)

Unfortunately, this does not attempt to construct the value using the
appropriate literal protocol; it instead performs overload resolution
using the standard rules, i.e. considering only single-argument
unlabelled initializers of a type which conforms to
IntegerLiteralConvertible. Often this leads to static ambiguities or,
worse, causes the literal to be built using a default type (such as
Int); this may have semantically very different results which are only
caught at runtime.

In my opinion, using this initializer-call syntax to build an
explicitly-typed literal is an obvious and natural choice with several
advantages over the "as" syntax. However, even if you disagree, it's
clear that programmers are going to continue to independently try to
use it, so it's really unfortunate for it to be subtly wrong.

Therefore, I propose that we adopt the following typing rule:

Given a function call expression of the form A(B) (that is, an
expr-call with a single, unlabelled argument) where B is an
expr-literal or expr-collection, if A has type T.Type for some type T
and there is a declared conformance of T to an appropriate literal
protocol for B, then the expression is always resolves as a literal
construction of type T (as if the expression were written "B as A")
rather than as a general initializer call.

Formally, this would be a special form of the argument conversion
constraint, since the type of the expression A may not be immediately
known.

I realize this is somewhat tangential, but... IMO this may not be entirely
about literals.

We have a standard that full-width type conversions are written as a
label-free initializer
<Swift.org - API Design Guidelines
<Swift.org - API Design Guidelines.
I believe that is partly responsible for setting up the expectation that
Int(42) works as one would expect. It gets ultra-weird when you can
convert from type A to type B using B(someA) but you can't write
B(someB). We should automatically generate a label-free “copy
initializer” for value types, to complete implementation of the expected
mental model.

That may also be a good idea, but it won't magically be preferred for
literal construction if the type has any other constructors of
literal-convertible type.

I know. I'm saying, fixing this for literals without giving value types
copy initializers leaves us with only a partial realization of a larger
mental model to which I believe people are programming.

That's fair. Would you like me to incorporate that into my proposal,
then? I see the relation, but it's a pretty significant jump in
scope.

--
Dave

The IntegerLiteral type idea might be worth exploring. It does seem to provide some additional consistency. For example, wasn't it clarified on this list just recently that literals don't have a type and adopt one based on context? It'd be nice to point out that 42 is an IntegerLiteral when explaining that it's not an Int.

I think that's a very promising way of thinking about literals. Writing a literal creates a notional value whose type is the informal, infinite-precise type of all integer/FP/collection/etc. literals, which (1) can be implicitly converted to any type that implements the appropriate protocol and (2) in fact *must* be converted to some such type (possibly the default type for that literal) in order for the code to be executable. You can then think about this proposal as saying that an explicit conversion from that informal type to a literal-convertible type follows the same path as an implicit conversion would have.

It sounds like the reason you don't want to remove the label is because it distinguishes the literal initializer and you don't want it called with a variable accidentally right?

What if we do something like this:

init(_ value: IntegerLiteralType, literal: IntegerLiteral = #literal)

The idea here is that the IntegerLiteral type and friends are special types that have no members and cannot be used except by the literal convertible protocols and their conformances. There is nothing you can do with the value at all.

The trick is that by defaulting it to #literal we are indicating the compiler synthesizes the logical value, as with #file, #line, etc. The compiler would refuse to synthesize this value when a literal is not used. This means it is never selected when the user provides a variable to a converting initializer. If an independent initializer accepting a value of the same type also exists that would be selected. However, when a literal *is* used and the type conforms to the relevant literal convertible protocol the compiler always synthesized the value making it always the most specific overload.

Of course no runtime value would actually exist. This is just a logical value marking the fact that a literal was used to call the initializer.

This approach solves the same problem while retaining semantic consistency with the language (no label elision or short circuit of overload resolution). The magic is arguably a lot more restrained - types for which values can only be supplied by the compiler. We could artificially restrict usage of these types if we wanted to, but we wouldn't have to. Nothing could be accomplished by using the types anywhere else so nobody do so and it wouldn't be actively harmful to allow them to be used anywhere other types can be used. Only the ability to create values of the type needs to be restricted and we can already write types like that by marking the initializers private.

Any thoughts on this approach?

This is still a special-case type-checking rule, which means that it's still basically my proposal except, instead of just rewriting the call to use a labeled literal initializer, it rewrites the call to use a magic initializer which cannot be used by users because it requires arguments that users cannot create. I just don't understand the motivation here. It's not a simpler language model to use, explain, or implement; it purports to be conceptually simpler and less magical while actually inventing two magic new language concepts (IntegerLiteral and #literal) on top of the same special cases.

I understand that this is still a special case and also that the implementation is pretty much identical. But I think the way it is presented to users is important.

The motivation is to come up with a design that lets us implement your proposal without magically eliding a parameter label. There is no other case in Swift where you can directly call any function (initializer or otherwise) without including labels for all parameters that have external labels. Creating a special case to allow that here is inconsistent. Why wouldn’t look for a way to do this that treats the signature consistently with the rest of the language and is more in line with how we handle other compiler magic (i.e. #file, #line, etc)?

I am not the only one who feels that way. Quoting Brent:

"But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.”

I agree with this. That is the motivation for my suggestion. I think it’s at least worth discussing as an alternative to magically allowing an external parameter label to be omitted. Brent, what do you think of my suggestion?

Literals are always going to involve some magic behind the scenes. The non-trivial literals, for example, all actually traffic through some sort of builtin protocol (e.g. BuiltinIntegerLiteralConvertible) which cannot be implemented by users and whose construction necessarily involves a private API contract between the standard library and compiler that we reserve the right to change at will without bringing in swift-evolution. All of the supposedly simple explanations eventually reach, "okay, but if Int isn't a special type, how does the literal turn into an Int in the first place", and that is just unanswerable, because it turns out that standard library types *are* actually special in the sense that they get to conform to BuiltinIntegerLiteralConvertible and nobody else does.

Understood. But Swift does a good job of making the magic feel consistent with the rest of the language. That’s a good thing. Allowing omission of a parameter label is a break from that IMO.

I believe this is an important problem to solve but I also think we should try to find a way to solve it that doesn’t violate rules that apply everywhere else in the language (specifically regarding parameter labels).

-Matthew

···

On Jun 3, 2016, at 1:36 PM, John McCall <rjmccall@apple.com> wrote:

On Jun 2, 2016, at 6:48 PM, Matthew Johnson <matthew@anandabits.com <mailto:matthew@anandabits.com>> wrote:
On Jun 2, 2016, at 6:51 PM, John McCall via swift-evolution <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:

On Jun 2, 2016, at 4:33 PM, Xiaodi Wu <xiaodi.wu@gmail.com <mailto:xiaodi.wu@gmail.com>> wrote:

John.

John.

On Thu, Jun 2, 2016 at 18:22 Brent Royal-Gordon via swift-evolution <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:
> So, you think that this syntax is enticing to new developers who naturally think that the feature works the way that I'm proposing it should work, and you think that the right solution is to make the syntax illegal so that you can more conveniently tell them it doesn't work that way? :)

I think the difference between a cast (which merely reinterprets a value as a compatible type) and a fullwidth conversion (which creates a similar instance of an incompatible type) is very important to understanding how to write Swift, and we shouldn't muddy the waters by creating a magic syntax.

> You can still tell them that it's a struct and you're calling an initializer on it; it's just that the initializer chosen is the special literal initializer because the argument is a literal.

If you're planning to change `IntegerLiteralConvertible` and friends to require a fullwidth conversion initializer like `init(_ value: IntegerLiteralType)`, then this is simply an overload resolution rule. In that case, I think your proposal is fine.

But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.

--
Brent Royal-Gordon
Architechies

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org <mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org <mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution

This is essentially already how it works. The literal protocols are not invoked via
overload resolution; Swift always invokes the initializer that satisfies the protocol
requirement.

That is, you can provide ten different init(integerLiteral: T) initializers, but that will
just prevent the compiler from inferring the associated type, so you'll have to
declare it. Once you do, that associated type will determine the initializer that
satisfies the requirement, and that'll always be the initializer chosen.

John.

···

On Jun 3, 2016, at 5:31 PM, Ben Rimmington via swift-evolution <swift-evolution@swift.org> wrote:
John McCall wrote:

I think that's a very promising way of thinking about literals. Writing
a literal creates a notional value whose type is the informal, infinite-
precise type of all integer/FP/collection/etc. literals, which (1) can be
implicitly converted to any type that implements the appropriate protocol
and (2) in fact *must* be converted to some such type (possibly the
default type for that literal) in order for the code to be executable.

Could you allow IntegerLiteralConvertible.IntegerLiteralType associatedtype
to override the default Swift.IntegerLiteralType typealias iff there's more
than one unlabelled init(_:) to choose from? Then you can call the "correct"
init(_:) instead of calling init(integerLiteral:) as a "special case".

I think your original proposal is good, but just in case:

[Currently]

    // Literal `7` is converted to `Swift.IntegerLiteralType` typealias.
    UInt16(7) calls UInt16.init(_ value: Int)

[Suggested]

    // Literal `7` is converted to `UInt16.IntegerLiteralType` associatedtype.
    UInt16(7) calls UInt16.init(_ value: UInt16)

Have "user-defined literals" already been rejected for Swift?
<C++11 - Wikipedia;
<C++14 - Wikipedia;

-- Ben

···

On 4 Jun 2016, at 02:20, John McCall <rjmccall@apple.com> wrote:

On Jun 3, 2016, at 5:31 PM, Ben Rimmington via swift-evolution <swift-evolution@swift.org> wrote:
John McCall wrote:

I think that's a very promising way of thinking about literals. Writing
a literal creates a notional value whose type is the informal, infinite-
precise type of all integer/FP/collection/etc. literals, which (1) can be
implicitly converted to any type that implements the appropriate protocol
and (2) in fact *must* be converted to some such type (possibly the
default type for that literal) in order for the code to be executable.

Could you allow IntegerLiteralConvertible.IntegerLiteralType associatedtype
to override the default Swift.IntegerLiteralType typealias iff there's more
than one unlabelled init(_:) to choose from? Then you can call the "correct"
init(_:) instead of calling init(integerLiteral:) as a "special case".

This is essentially already how it works. The literal protocols are not invoked via
overload resolution; Swift always invokes the initializer that satisfies the protocol
requirement.

That is, you can provide ten different init(integerLiteral: T) initializers, but that will
just prevent the compiler from inferring the associated type, so you'll have to
declare it. Once you do, that associated type will determine the initializer that
satisfies the requirement, and that'll always be the initializer chosen.

John.

Sure thing. I agree that that is an interesting way of conceptualizing how
this proposal works, at least for value types where the initializer is synthesized:
you could think of it as saying that there's a rule to prefer the "copy initializer"
when the argument is a literal and the type is literally constructible.

The flip side is that that argument wouldn't apply to class types or (presumably)
generic type parameters, but we'd still want e.g. T(1) to work if T: IntegerLiteralConvertible.

John.

···

On Jun 8, 2016, at 11:15 AM, Dave Abrahams <dabrahams@apple.com> wrote:
on Wed Jun 08 2016, John McCall <rjmccall-AT-apple.com> wrote:

On Jun 7, 2016, at 4:25 PM, Dave Abrahams <dabrahams@apple.com> wrote:
on Tue Jun 07 2016, John McCall <rjmccall-AT-apple.com> wrote:

On Jun 5, 2016, at 5:18 PM, Dave Abrahams via swift-evolution <swift-evolution@swift.org> wrote:
on Thu Jun 02 2016, John McCall <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:

The official way to build a literal of a specific type is to write the
literal in an explicitly-typed context, like so:

let x: UInt16 = 7
or
let x = 7 as UInt16

Nonetheless, programmers often try the following:
UInt16(7)

Unfortunately, this does not attempt to construct the value using the
appropriate literal protocol; it instead performs overload resolution
using the standard rules, i.e. considering only single-argument
unlabelled initializers of a type which conforms to
IntegerLiteralConvertible. Often this leads to static ambiguities or,
worse, causes the literal to be built using a default type (such as
Int); this may have semantically very different results which are only
caught at runtime.

In my opinion, using this initializer-call syntax to build an
explicitly-typed literal is an obvious and natural choice with several
advantages over the "as" syntax. However, even if you disagree, it's
clear that programmers are going to continue to independently try to
use it, so it's really unfortunate for it to be subtly wrong.

Therefore, I propose that we adopt the following typing rule:

Given a function call expression of the form A(B) (that is, an
expr-call with a single, unlabelled argument) where B is an
expr-literal or expr-collection, if A has type T.Type for some type T
and there is a declared conformance of T to an appropriate literal
protocol for B, then the expression is always resolves as a literal
construction of type T (as if the expression were written "B as A")
rather than as a general initializer call.

Formally, this would be a special form of the argument conversion
constraint, since the type of the expression A may not be immediately
known.

I realize this is somewhat tangential, but... IMO this may not be entirely
about literals.

We have a standard that full-width type conversions are written as a
label-free initializer
<Swift.org - API Design Guidelines
<Swift.org - API Design Guidelines.
I believe that is partly responsible for setting up the expectation that
Int(42) works as one would expect. It gets ultra-weird when you can
convert from type A to type B using B(someA) but you can't write
B(someB). We should automatically generate a label-free “copy
initializer” for value types, to complete implementation of the expected
mental model.

That may also be a good idea, but it won't magically be preferred for
literal construction if the type has any other constructors of
literal-convertible type.

I know. I'm saying, fixing this for literals without giving value types
copy initializers leaves us with only a partial realization of a larger
mental model to which I believe people are programming.

That's fair. Would you like me to incorporate that into my proposal,
then? I see the relation, but it's a pretty significant jump in
scope.

I'm not sure it should be in the same proposal; I just want it to be a
part of the discussion. We should be able to make incremental progress,
but let's make sure we understand how our changes fit into the big picture.

The IntegerLiteral type idea might be worth exploring. It does seem to provide some additional consistency. For example, wasn't it clarified on this list just recently that literals don't have a type and adopt one based on context? It'd be nice to point out that 42 is an IntegerLiteral when explaining that it's not an Int.

I think that's a very promising way of thinking about literals. Writing a literal creates a notional value whose type is the informal, infinite-precise type of all integer/FP/collection/etc. literals, which (1) can be implicitly converted to any type that implements the appropriate protocol and (2) in fact *must* be converted to some such type (possibly the default type for that literal) in order for the code to be executable. You can then think about this proposal as saying that an explicit conversion from that informal type to a literal-convertible type follows the same path as an implicit conversion would have.

It sounds like the reason you don't want to remove the label is because it distinguishes the literal initializer and you don't want it called with a variable accidentally right?

What if we do something like this:

init(_ value: IntegerLiteralType, literal: IntegerLiteral = #literal)

The idea here is that the IntegerLiteral type and friends are special types that have no members and cannot be used except by the literal convertible protocols and their conformances. There is nothing you can do with the value at all.

The trick is that by defaulting it to #literal we are indicating the compiler synthesizes the logical value, as with #file, #line, etc. The compiler would refuse to synthesize this value when a literal is not used. This means it is never selected when the user provides a variable to a converting initializer. If an independent initializer accepting a value of the same type also exists that would be selected. However, when a literal *is* used and the type conforms to the relevant literal convertible protocol the compiler always synthesized the value making it always the most specific overload.

Of course no runtime value would actually exist. This is just a logical value marking the fact that a literal was used to call the initializer.

This approach solves the same problem while retaining semantic consistency with the language (no label elision or short circuit of overload resolution). The magic is arguably a lot more restrained - types for which values can only be supplied by the compiler. We could artificially restrict usage of these types if we wanted to, but we wouldn't have to. Nothing could be accomplished by using the types anywhere else so nobody do so and it wouldn't be actively harmful to allow them to be used anywhere other types can be used. Only the ability to create values of the type needs to be restricted and we can already write types like that by marking the initializers private.

Any thoughts on this approach?

This is still a special-case type-checking rule, which means that it's still basically my proposal except, instead of just rewriting the call to use a labeled literal initializer, it rewrites the call to use a magic initializer which cannot be used by users because it requires arguments that users cannot create. I just don't understand the motivation here. It's not a simpler language model to use, explain, or implement; it purports to be conceptually simpler and less magical while actually inventing two magic new language concepts (IntegerLiteral and #literal) on top of the same special cases.

I understand that this is still a special case and also that the implementation is pretty much identical. But I think the way it is presented to users is important.

The motivation is to come up with a design that lets us implement your proposal without magically eliding a parameter label. There is no other case in Swift where you can directly call any function (initializer or otherwise) without including labels for all parameters that have external labels.

Don't think of it as omitting a parameter label. Constructing a value from a literal is a complex and well-defined process that happens to end by calling an initializer with a labeled argument. The proposal says that we follow that entire process as normal, not just that we silently rewrite the call to add a label, which is not how it works.

Creating a special case to allow that here is inconsistent. Why wouldn’t look for a way to do this that treats the signature consistently with the rest of the language and is more in line with how we handle other compiler magic (i.e. #file, #line, etc)?

It's not more consistent with the rest of the language. You're talking about introducing a default argument that only defaults when the compiler recognizes a special syntactic form in the call site. I don't understand why inventing novel and massively invasive language mechanisms like that is somehow clearer or more consistent than explaining that a particular syntactic form is recognized as performing the well-defined literal-construction process.

There is simply no way that this is going to feel 100% consistent with the ordinary language while still prohibiting the literal initializer from being called as an unlabelled initializer in the ordinary language, which is explicitly something we do not want to allow. It will be a special case. It will not be consistent. Special cases are not consistent. You cannot thread this needle.

John.

···

On Jun 3, 2016, at 12:41 PM, Matthew Johnson <matthew@anandabits.com> wrote:

On Jun 3, 2016, at 1:36 PM, John McCall <rjmccall@apple.com <mailto:rjmccall@apple.com>> wrote:

On Jun 2, 2016, at 6:48 PM, Matthew Johnson <matthew@anandabits.com <mailto:matthew@anandabits.com>> wrote:
On Jun 2, 2016, at 6:51 PM, John McCall via swift-evolution <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:

On Jun 2, 2016, at 4:33 PM, Xiaodi Wu <xiaodi.wu@gmail.com <mailto:xiaodi.wu@gmail.com>> wrote:

I am not the only one who feels that way. Quoting Brent:

"But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.”

I agree with this. That is the motivation for my suggestion. I think it’s at least worth discussing as an alternative to magically allowing an external parameter label to be omitted. Brent, what do you think of my suggestion?

Literals are always going to involve some magic behind the scenes. The non-trivial literals, for example, all actually traffic through some sort of builtin protocol (e.g. BuiltinIntegerLiteralConvertible) which cannot be implemented by users and whose construction necessarily involves a private API contract between the standard library and compiler that we reserve the right to change at will without bringing in swift-evolution. All of the supposedly simple explanations eventually reach, "okay, but if Int isn't a special type, how does the literal turn into an Int in the first place", and that is just unanswerable, because it turns out that standard library types *are* actually special in the sense that they get to conform to BuiltinIntegerLiteralConvertible and nobody else does.

Understood. But Swift does a good job of making the magic feel consistent with the rest of the language. That’s a good thing. Allowing omission of a parameter label is a break from that IMO.

I believe this is an important problem to solve but I also think we should try to find a way to solve it that doesn’t violate rules that apply everywhere else in the language (specifically regarding parameter labels).

-Matthew

John.

John.

On Thu, Jun 2, 2016 at 18:22 Brent Royal-Gordon via swift-evolution <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:
> So, you think that this syntax is enticing to new developers who naturally think that the feature works the way that I'm proposing it should work, and you think that the right solution is to make the syntax illegal so that you can more conveniently tell them it doesn't work that way? :)

I think the difference between a cast (which merely reinterprets a value as a compatible type) and a fullwidth conversion (which creates a similar instance of an incompatible type) is very important to understanding how to write Swift, and we shouldn't muddy the waters by creating a magic syntax.

> You can still tell them that it's a struct and you're calling an initializer on it; it's just that the initializer chosen is the special literal initializer because the argument is a literal.

If you're planning to change `IntegerLiteralConvertible` and friends to require a fullwidth conversion initializer like `init(_ value: IntegerLiteralType)`, then this is simply an overload resolution rule. In that case, I think your proposal is fine.

But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.

--
Brent Royal-Gordon
Architechies

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org <mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org <mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution

I am not the only one who feels that way. Quoting Brent:

"But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.”

I agree with this. That is the motivation for my suggestion. I think it’s at least worth discussing as an alternative to magically allowing an external parameter label to be omitted. Brent, what do you think of my suggestion?

I think it could be simpler:

  public struct Literal<LiteralType> {
    public let value: LiteralType
    internal init(_value value: LiteralType)
  }
  
  public protocol IntegerLiteralConvertible {
    associatedtype IntegerLiteralType
    init(_ literal: Literal<IntegerLiteralType>)
  }

Only the standard library can create a Literal, which it would do by constructing the IntegerLiteralType with its magic builtin literal stuff and wrapping it. You still have your magic parameter aspect, but without any actual magic, just access control.

···

--
Brent Royal-Gordon
Architechies

The IntegerLiteral type idea might be worth exploring. It does seem to
provide some additional consistency. For example, wasn't it clarified on
this list just recently that literals don't have a type and adopt one based
on context? It'd be nice to point out that 42 is an IntegerLiteral when
explaining that it's not an Int.

I think that's a very promising way of thinking about literals. Writing a
literal creates a notional value whose type is the informal,
infinite-precise type of all integer/FP/collection/etc. literals, which (1)
can be implicitly converted to any type that implements the appropriate
protocol and (2) in fact *must* be converted to some such type (possibly
the default type for that literal) in order for the code to be executable.
You can then think about this proposal as saying that an explicit
conversion from that informal type to a literal-convertible type follows
the same path as an implicit conversion would have.

It sounds like the reason you don't want to remove the label is because it
distinguishes the literal initializer and you don't want it called with a
variable accidentally right?

What if we do something like this:

init(_ value: IntegerLiteralType, literal: IntegerLiteral = #literal)

The idea here is that the IntegerLiteral type and friends are special
types that have no members and cannot be used except by the literal
convertible protocols and their conformances. There is nothing you can do
with the value at all.

The trick is that by defaulting it to #literal we are indicating the
compiler synthesizes the logical value, as with #file, #line, etc. The
compiler would refuse to synthesize this value when a literal is not used.
This means it is never selected when the user provides a variable to a
converting initializer. If an independent initializer accepting a value of
the same type also exists that would be selected. However, when a literal
*is* used and the type conforms to the relevant literal convertible
protocol the compiler always synthesized the value making it always the
most specific overload.

Of course no runtime value would actually exist. This is just a logical
value marking the fact that a literal was used to call the initializer.

This approach solves the same problem while retaining semantic consistency
with the language (no label elision or short circuit of overload
resolution). The magic is arguably a lot more restrained - types for which
values can only be supplied by the compiler. We could artificially
restrict usage of these types if we wanted to, but we wouldn't have to.
Nothing could be accomplished by using the types anywhere else so nobody do
so and it wouldn't be actively harmful to allow them to be used anywhere
other types can be used. Only the ability to create values of the type
needs to be restricted and we can already write types like that by marking
the initializers private.

Any thoughts on this approach?

This is still a special-case type-checking rule, which means that it's
still basically my proposal except, instead of just rewriting the call to
use a labeled literal initializer, it rewrites the call to use a magic
initializer which cannot be used by users because it requires arguments
that users cannot create. I just don't understand the motivation here.
It's not a simpler language model to use, explain, or implement; it
purports to be conceptually simpler and less magical while actually
inventing two magic new language concepts (IntegerLiteral and #literal) on
top of the same special cases.

I understand that this is still a special case and also that the
implementation is pretty much identical. But I think the way it is
presented to users is important.

The motivation is to come up with a design that lets us implement your
proposal without magically eliding a parameter label. There is no other
case in Swift where you can directly call any function (initializer or
otherwise) without including labels for all parameters that have external
labels. Creating a special case to allow that here is inconsistent. Why
wouldn’t look for a way to do this that treats the signature consistently
with the rest of the language and is more in line with how we handle other
compiler magic (i.e. #file, #line, etc)?

I am not the only one who feels that way. Quoting Brent:

"But if you're going to call `init(integerLiteral:)` like it's `init(_:)`,
I don't think that's a good idea. Parameter labels are supposed to be
significant; we don't want to lose that.”

I agree with this. That is the motivation for my suggestion. I think
it’s at least worth discussing as an alternative to magically allowing an
external parameter label to be omitted. Brent, what do you think of my
suggestion?

Literals are always going to involve some magic behind the scenes. The
non-trivial literals, for example, all actually traffic through some sort
of builtin protocol (e.g. BuiltinIntegerLiteralConvertible) which cannot be
implemented by users and whose construction necessarily involves a private
API contract between the standard library and compiler that we reserve the
right to change at will without bringing in swift-evolution. All of the
supposedly simple explanations eventually reach, "okay, but if Int isn't a
special type, how does the literal turn into an Int in the first place",
and that is just unanswerable, because it turns out that standard library
types *are* actually special in the sense that they get to conform to
BuiltinIntegerLiteralConvertible and nobody else does.

Understood. But Swift does a good job of making the magic feel consistent
with the rest of the language. That’s a good thing. Allowing omission of
a parameter label is a break from that IMO.

I believe this is an important problem to solve but I also think we should
try to find a way to solve it that doesn’t violate rules that apply
everywhere else in the language (specifically regarding parameter labels).

Would it help if we had an attribute or contextual keyword to describe this
magic? By this I mean, what if we had `init(_: @literal Int)` or something
similar? The magic would be isolated to the fact that you'd have to supply
an integer literal to such an argument, which through behind-the-scenes
stuff becomes an Int.

···

On Fri, Jun 3, 2016 at 2:41 PM, Matthew Johnson <matthew@anandabits.com> wrote:

On Jun 3, 2016, at 1:36 PM, John McCall <rjmccall@apple.com> wrote:
On Jun 2, 2016, at 6:48 PM, Matthew Johnson <matthew@anandabits.com> > wrote:
On Jun 2, 2016, at 6:51 PM, John McCall via swift-evolution < > swift-evolution@swift.org> wrote:
On Jun 2, 2016, at 4:33 PM, Xiaodi Wu <xiaodi.wu@gmail.com> wrote:

-Matthew

John.

John.

On Thu, Jun 2, 2016 at 18:22 Brent Royal-Gordon via swift-evolution < > swift-evolution@swift.org> wrote:

> So, you think that this syntax is enticing to new developers who
naturally think that the feature works the way that I'm proposing it should
work, and you think that the right solution is to make the syntax illegal
so that you can more conveniently tell them it doesn't work that way? :)

I think the difference between a cast (which merely reinterprets a value
as a compatible type) and a fullwidth conversion (which creates a similar
instance of an incompatible type) is very important to understanding how to
write Swift, and we shouldn't muddy the waters by creating a magic syntax.

> You can still tell them that it's a struct and you're calling an
initializer on it; it's just that the initializer chosen is the special
literal initializer because the argument is a literal.

If you're planning to change `IntegerLiteralConvertible` and friends to
require a fullwidth conversion initializer like `init(_ value:
IntegerLiteralType)`, then this is simply an overload resolution rule. In
that case, I think your proposal is fine.

But if you're going to call `init(integerLiteral:)` like it's `init(_:)`,
I don't think that's a good idea. Parameter labels are supposed to be
significant; we don't want to lose that.

--
Brent Royal-Gordon
Architechies

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org
https://lists.swift.org/mailman/listinfo/swift-evolution

The IntegerLiteral type idea might be worth exploring. It does seem to provide some additional consistency. For example, wasn't it clarified on this list just recently that literals don't have a type and adopt one based on context? It'd be nice to point out that 42 is an IntegerLiteral when explaining that it's not an Int.

I think that's a very promising way of thinking about literals. Writing a literal creates a notional value whose type is the informal, infinite-precise type of all integer/FP/collection/etc. literals, which (1) can be implicitly converted to any type that implements the appropriate protocol and (2) in fact *must* be converted to some such type (possibly the default type for that literal) in order for the code to be executable. You can then think about this proposal as saying that an explicit conversion from that informal type to a literal-convertible type follows the same path as an implicit conversion would have.

It sounds like the reason you don't want to remove the label is because it distinguishes the literal initializer and you don't want it called with a variable accidentally right?

What if we do something like this:

init(_ value: IntegerLiteralType, literal: IntegerLiteral = #literal)

The idea here is that the IntegerLiteral type and friends are special types that have no members and cannot be used except by the literal convertible protocols and their conformances. There is nothing you can do with the value at all.

The trick is that by defaulting it to #literal we are indicating the compiler synthesizes the logical value, as with #file, #line, etc. The compiler would refuse to synthesize this value when a literal is not used. This means it is never selected when the user provides a variable to a converting initializer. If an independent initializer accepting a value of the same type also exists that would be selected. However, when a literal *is* used and the type conforms to the relevant literal convertible protocol the compiler always synthesized the value making it always the most specific overload.

Of course no runtime value would actually exist. This is just a logical value marking the fact that a literal was used to call the initializer.

This approach solves the same problem while retaining semantic consistency with the language (no label elision or short circuit of overload resolution). The magic is arguably a lot more restrained - types for which values can only be supplied by the compiler. We could artificially restrict usage of these types if we wanted to, but we wouldn't have to. Nothing could be accomplished by using the types anywhere else so nobody do so and it wouldn't be actively harmful to allow them to be used anywhere other types can be used. Only the ability to create values of the type needs to be restricted and we can already write types like that by marking the initializers private.

Any thoughts on this approach?

This is still a special-case type-checking rule, which means that it's still basically my proposal except, instead of just rewriting the call to use a labeled literal initializer, it rewrites the call to use a magic initializer which cannot be used by users because it requires arguments that users cannot create. I just don't understand the motivation here. It's not a simpler language model to use, explain, or implement; it purports to be conceptually simpler and less magical while actually inventing two magic new language concepts (IntegerLiteral and #literal) on top of the same special cases.

I understand that this is still a special case and also that the implementation is pretty much identical. But I think the way it is presented to users is important.

The motivation is to come up with a design that lets us implement your proposal without magically eliding a parameter label. There is no other case in Swift where you can directly call any function (initializer or otherwise) without including labels for all parameters that have external labels.

Don't think of it as omitting a parameter label. Constructing a value from a literal is a complex and well-defined process that happens to end by calling an initializer with a labeled argument. The proposal says that we follow that entire process as normal, not just that we silently rewrite the call to add a label, which is not how it works.

This is an implementation detail. It is important to the implementation but is inconsistent on the surface, which is what the programmer sees. The fact is that the programmer-visible surface of the language currently treats it as a parameter label. There are *no* exceptions to parameter labels and at least some of us would prefer to keep it that way if we can find an acceptable alternative.

Creating a special case to allow that here is inconsistent. Why wouldn’t look for a way to do this that treats the signature consistently with the rest of the language and is more in line with how we handle other compiler magic (i.e. #file, #line, etc)?

It's not more consistent with the rest of the language. You're talking about introducing a default argument that only defaults when the compiler recognizes a special syntactic form in the call site. I don't understand why inventing novel and massively invasive language mechanisms like that is somehow clearer or more consistent than explaining that a particular syntactic form is recognized as performing the well-defined literal-construction process.

I was never fully comfortable but wanted to prompt discussion about some alternative way to mark the initializer that doesn’t contradict the rules for parameter labels. I can see how this compiler generated default would not behave like other compiler generated defaults now and agree it also has problems.

Fortunately it looks like I succeeded in prompting some other alternatives which are better options IMO.

There is simply no way that this is going to feel 100% consistent with the ordinary language while still prohibiting the literal initializer from being called as an unlabelled initializer in the ordinary language, which is explicitly something we do not want to allow. It will be a special case. It will not be consistent. Special cases are not consistent. You cannot thread this needle.

The way I see it, the question is how to find a way to handle the special case that feels the least inconsistent. Using an external parameter label in a declaration and allowing it to be omitted at the call site does not feel like the right answer to me. I think we can find a better way to present this syntactically.

···

On Jun 3, 2016, at 3:17 PM, John McCall <rjmccall@apple.com> wrote:

On Jun 3, 2016, at 12:41 PM, Matthew Johnson <matthew@anandabits.com <mailto:matthew@anandabits.com>> wrote:

On Jun 3, 2016, at 1:36 PM, John McCall <rjmccall@apple.com <mailto:rjmccall@apple.com>> wrote:

On Jun 2, 2016, at 6:48 PM, Matthew Johnson <matthew@anandabits.com <mailto:matthew@anandabits.com>> wrote:
On Jun 2, 2016, at 6:51 PM, John McCall via swift-evolution <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:

On Jun 2, 2016, at 4:33 PM, Xiaodi Wu <xiaodi.wu@gmail.com <mailto:xiaodi.wu@gmail.com>> wrote:

John.

I am not the only one who feels that way. Quoting Brent:

"But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.”

I agree with this. That is the motivation for my suggestion. I think it’s at least worth discussing as an alternative to magically allowing an external parameter label to be omitted. Brent, what do you think of my suggestion?

Literals are always going to involve some magic behind the scenes. The non-trivial literals, for example, all actually traffic through some sort of builtin protocol (e.g. BuiltinIntegerLiteralConvertible) which cannot be implemented by users and whose construction necessarily involves a private API contract between the standard library and compiler that we reserve the right to change at will without bringing in swift-evolution. All of the supposedly simple explanations eventually reach, "okay, but if Int isn't a special type, how does the literal turn into an Int in the first place", and that is just unanswerable, because it turns out that standard library types *are* actually special in the sense that they get to conform to BuiltinIntegerLiteralConvertible and nobody else does.

Understood. But Swift does a good job of making the magic feel consistent with the rest of the language. That’s a good thing. Allowing omission of a parameter label is a break from that IMO.

I believe this is an important problem to solve but I also think we should try to find a way to solve it that doesn’t violate rules that apply everywhere else in the language (specifically regarding parameter labels).

-Matthew

John.

John.

On Thu, Jun 2, 2016 at 18:22 Brent Royal-Gordon via swift-evolution <swift-evolution@swift.org <mailto:swift-evolution@swift.org>> wrote:
> So, you think that this syntax is enticing to new developers who naturally think that the feature works the way that I'm proposing it should work, and you think that the right solution is to make the syntax illegal so that you can more conveniently tell them it doesn't work that way? :)

I think the difference between a cast (which merely reinterprets a value as a compatible type) and a fullwidth conversion (which creates a similar instance of an incompatible type) is very important to understanding how to write Swift, and we shouldn't muddy the waters by creating a magic syntax.

> You can still tell them that it's a struct and you're calling an initializer on it; it's just that the initializer chosen is the special literal initializer because the argument is a literal.

If you're planning to change `IntegerLiteralConvertible` and friends to require a fullwidth conversion initializer like `init(_ value: IntegerLiteralType)`, then this is simply an overload resolution rule. In that case, I think your proposal is fine.

But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.

--
Brent Royal-Gordon
Architechies

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org <mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution

_______________________________________________
swift-evolution mailing list
swift-evolution@swift.org <mailto:swift-evolution@swift.org>
https://lists.swift.org/mailman/listinfo/swift-evolution

I am not the only one who feels that way. Quoting Brent:

"But if you're going to call `init(integerLiteral:)` like it's `init(_:)`, I don't think that's a good idea. Parameter labels are supposed to be significant; we don't want to lose that.”

I agree with this. That is the motivation for my suggestion. I think it’s at least worth discussing as an alternative to magically allowing an external parameter label to be omitted. Brent, what do you think of my suggestion?

I think it could be simpler:

  public struct Literal<LiteralType> {
    public let value: LiteralType
    internal init(_value value: LiteralType)
  }
  
  public protocol IntegerLiteralConvertible {
    associatedtype IntegerLiteralType
    init(_ literal: Literal<IntegerLiteralType>)
  }

Only the standard library can create a Literal, which it would do by constructing the IntegerLiteralType with its magic builtin literal stuff and wrapping it. You still have your magic parameter aspect, but without any actual magic, just access control.

I like this solution as well as Xiaodi’s `@literal` idea better than the idea I came up with. I think I like `@literal` best because it feels more appropriate to restrict an attribute to a specific context (signatures for literal convertible initializers) than anything else. We don’t want to allow this in any other signatures and no types are restricted in that way. The solution I came up with has that problem as well as the `#literal` “value” that is also restricted unlike other compiler generated values.

I think our decision should be based upon which syntactic construct feels least inconsistent with other similar syntactic constructs and therefore feels the least arbitrary. Restricting the usage of a generic `Literal` type to literal convertible initializers feels a little bit less arbitrary than allowing the call site to omit a label, but it feels a little bit more arbitrary than introducing an attribute that has a very specialized context of applicability.

···

On Jun 3, 2016, at 4:05 PM, Brent Royal-Gordon <brent@architechies.com> wrote:

--
Brent Royal-Gordon
Architechies