Unformatted version of **ai12s/ai12-0339-1.txt version 1.1**

Other versions for file**ai12s/ai12-0339-1.txt**

Other versions for file

!standard A.18.2(8/5) 19-06-16 AI12-0339-1/01

!class Amendment 19-06-16

!status work item 19-06-16

!status received 19-06-12

!priority Low

!difficulty Easy

!subject Empty function for Container aggregates

!class Amendment 19-06-16

!status work item 19-06-16

!status received 19-06-12

!priority Low

!difficulty Easy

!subject Empty function for Container aggregates

!summary

Define appropriate Empty functions for all the containers. For containers
types that have an associated Capacity, the corresponding Empty function accepts a
Capacity parameter to specify this value. The Capacity parameter will have a
system defined default value. For other container types that do not have the
notion of capacity, the Empty function will be parameterless.

!problem

AI12-0212-1 defines the aggregate aspect, which adds the capability of specifying
container aggregates. One part of the aggregate aspect specification, Empty,
identifies the name that denotes a constant of the container type, or a function with a
result of the container type that is used to produce an empty container object.
Currently, the containers name the preexisting constant object defined in Ada 2012,
for Empty. However, these constant objects, for container types that
have the notion of capacity, have a capacity of 0 elements, but container
aggregates often are used to create container aggregates that have 1 or more
elements. It would be more efficient, for cases where multiple elements are
needed, if the constructed container object was initially sized to have a
capacity that could accommodate all the elements
associated with the aggregate. It can be very inefficient for a container such as
a vector container to start with 0 capacity, and increase the capacity as each
element of the aggregate is inserted into the container. Should a mechanism
be provided to allow container objects to be initially sized to a given
capacity, allowing for more efficient construction of container objects? (Yes.)

!proposal

We propose to add an Empty function to all the containers that have the Aggregate
aspect, and modify the Aggregate aspect such that the Empty field denotes the new
function instead of the original constant object. For those containers that
have the notion of capacity, the Empty function will accept a Capacity parameter to
indicate the initial capacity of the container object of the result of the Empty
function. The Capacity parameter will have a system defined default value.
Otherwise, for containers that do not have the notion of capacity, the Empty
function will be a parameter-less function, that returns a container object with
zero elements.

!wording

Modify A.18.2 (8/5)

Insert before A.18.2 (13/5)

function Empty

(Capacity : Count_Type := Implementation Defined) return Vector

with Pre => (if Capacity > Maximum_Length then raise Constraint_Error),

Post =>
Empty'Result.Capacity = Capacity and then
not Tampering_With_Elements_Prohibited (Empty'Result) and then
not Tampering_With_Cursors_Prohibited (Empty'Result) and then
Empty'Result.Length = 0;

Modify A.18.2 (78.2/5)

Insert before A.18.2 (99/5)

Insert before A.18.3 (9.1/5)

Modify A.18.5 (3/5)

Insert before A.18.5 (6.1/5)

function Empty

(Capacity : Count_Type := Implementation Defined) return Map

with Post =>

Empty'Result.Capacity = Capacity and then
not Tampering_With_Elements_Prohibited (Empty'Result) and then
not Tampering_With_Cursors_Prohibited (Empty'Result) and then
Empty'Result.Length = 0;

Modify A.18.5 (37.3/5)

Insert before A.18.5 (47/5)

Insert before A.18.6 (7.1/5)

Modify A.18.8 (3/5)

Insert before A.18.8 (6.1/5)

function Empty

(Capacity : Count_Type := Implementation Defined) return Map

with Post =>

Empty'Result.Capacity = Capacity and then
not Tampering_With_Elements_Prohibited (Empty'Result) and then
not Tampering_With_Cursors_Prohibited (Empty'Result) and then
Empty'Result.Length = 0;

Modify A.18.8 (59.2/5)
type Set (Base : not null access Hashed_Sets.Set) is

tagged limited private
with Constant_Indexing => Constant_Reference,

Default_Iterator => Iterate,
Iterator_Element => Element_Type,
Aggregate => (Empty => Empty[_Set],

Add_Unnamed => Include),

Stable_Properties => (Length),
Global => (in Set.Base.all, synchronized out Set.Base.all),
Default_Initial_Condition => Length (Set) = 0;

pragma Preelaborable_Initialization(Set);

Insert before A.18.8 (69/5)

Insert before A.18.9 (7.1/5)

Insert before A.18.10 (11.1/5)

Insert before A.18.18 (7.1/5)

Insert before A.18.20 (7/5)

The function Empty is replaced by:

function Empty

(Capacity : Count_Type := Implementation Defined) return List

with Post =>

Empty'Result.Capacity = Capacity and then
not Tampering_With_Elements_Prohibited (Empty'Result) and then
not Tampering_With_Cursors_Prohibited (Empty'Result) and then
Empty'Result.Length = 0;

Returns a List with a length of 0 and a capacity of Capacity.

Insert before A.18.22 (7/5)

The function Empty is replaced by:

function Empty

(Capacity : Count_Type := Implementation Defined) return Map

with Post =>

Returns a Map with a length of 0 and a capacity of Capacity.

Insert before A.18.24 (7/5)

The function Empty is replaced by:

function Empty

(Capacity : Count_Type := Implementation Defined) return Set

with Post =>

Returns a Set with a length of 0 and a capacity of Capacity.

Insert before A.18.25 (8/3)

The function Empty is replaced by:

function Empty

(Capacity : Count_Type := Implementation Defined) return Tree

with Post =>

Empty'Result.Capacity = Capacity and then
not Tampering_With_Elements_Prohibited (Empty'Result) and then
not Tampering_With_Cursors_Prohibited (Empty'Result) and then
Empty'Result.Is_Empty;

Returns an Tree with a capacity of Capacity and for which Is_Empty returns True.

!discussion

We considered whether the Empty function should be added to the various
synchronized queue containers. Those containers do not define stable views,
do not have the Aggregate aspect, nor do they define an Empty_Queue object,
so it seemed best to not attempt to define an empty function for these
containers.

!ASIS

!ACATS test

ACATS B and C-Tests are needed to check that the new capabilities are supported.

!appendix

From: Tucker Taft Sent: Wednesday, June 12, 2019 12:23 AM A comment from an GNAT user mentioned that we have not provided in Ada 202X "Empty" functions that take a capacity, to support the capability available to container aggregates. Perhaps we should do so to make the container aggregate construction process more efficient! Seems particularly important for vectors, but could be useful for any case where the compiler can predict the size of the aggregate being created. **************************************************************** From: Randy Brukardt Sent: Wednesday, June 12, 2019 7:59 AM Seems like it would be necessary for bounded containers, esp. if someone wanted to use :=. **************************************************************** From: Tucker Taft Sent: Wednesday, June 12, 2019 9:47 AM It seems important for dynamic containers as well, from an efficiency point of view. Otherwise the container might be expanded for each element, or over-expanded well beyond the size needed. **************************************************************** From: Brad Moore Sent: Sunday, June 16, 2019 1:29 AM Here is a new AI that we talked about yesterday (AI12-0338?) that defines an Empty function for most of the Ada containers. [This is version /01 of the AI - Editor.] ****************************************************************

Questions? Ask the ACAA Technical Agent