Difference between revisions of "Default and explicit variance"

m (Better terminology)
m (Better terminology)
Line 12: Line 12:
  
 
The first is a weaker precondition and the second is a stronger postconidtion.
 
The first is a weaker precondition and the second is a stronger postconidtion.
Only the second statment is a new semantic in Eiffel.
+
Only the second rule is a new possibility in Eiffel.
  
 
* Contravariant redefinition of request result is allowed for non-conforming inheritance
 
* Contravariant redefinition of request result is allowed for non-conforming inheritance

Revision as of 00:14, 10 May 2013

Research: This page describes research about Eiffel, not the actual language specification.

Introduction

This solution enables covariance and contravariance redefinition. The default behavior is detected automatically. When there is a cat-call risk the programmer must explicitly declare the variance.

Default variance

Feature redefinition

  • Covariant redefinition of request result is allowed
  • Contravariant redefinition of feature arguments is allowed

The first is a weaker precondition and the second is a stronger postconidtion. Only the second rule is a new possibility in Eiffel.

  • Contravariant redefinition of request result is allowed for non-conforming inheritance
  • Covariant redefinition of feature arguments is allowed for non-conforming inheritance

Generic conformance

Generic conformance is close from redefinition.

  • Generic used only on feature argument is contravariant.
  • Generic used only on request result is covariant.
  • Generic used both on feature argument and request result is novariant.

And then generic not used is covariant and contravariant.

The first checkpoint "Generic lists" is checked. As explain for Usage-site variance the generic 'OPEN_ARGS' from ROUTINE, PROCEDURE, FUNCTION, PREDICATE must be contravariant and the generics of TUPLE must be covariants.

However 'OPEN_ARGS' is novariant. Indeed the generic is used on request type and feature arguments.

The class must probably be redesigned. Indeed if contravariant is allowed then there may be a new contravariant cat-call on request result.

In these classes there are only two requests using the generic as type:

operands: detachable OPEN_ARGS
 
empty_operands: OPEN_ARGS
	do create Result ensure ... end

The second request is not used internally and may be problematic for creation: How create a tuple of not self-initialized and attached types? The first is used in these features : 'target', 'is_equal', 'set_operands', 'copy', 'apply'

'is_equal', 'set_operands' and 'copy' rely on data model. 'apply' and 'target' are more sensitive. Indeed the problem is when there is an opened target.

'apply' could be restricted for no opened arguments.

apply
	require
		no_opened_operands: open_count = 0
	do
		call (Void)
	end

'target' could be modfied:

target: ANY
	require
		is_target_closed
	do
		-- ...
	end

Or maybe it is better to separate opened target and closed target in two abstractions.

The redisgn of agent classes is an opportunity to solve agent problems (see Minor-ECMA-problems, Agents in SCOOP).

Another solution will be mentioned later.

Note: The generic 'RESULT_TYPE' of FUNCTION class is used only on request result. Then it is a covariant generic. It is an expected point.

Sub-conclusion

The default semantic use no new keyword.

The lack of expressivity for generic variance can be reduced with a prefixed mark for a genric (see below).

The addition in TYPE class for reflexivity and dynamic object test is little. Indeed two simple booleans "is_contravariant" and "is_covariant" for each generic is required.

generic_parameter_contravariant (i: INTEGER): BOOLEAN
	-- Is `i'-th generic parameter contravariant?
 
generic_parameter_covariant (i: INTEGER): BOOLEAN
	-- Is `i'-th generic parameter covariant?

Variant typing

What is the advantages of covariance compared to a novariant typing?

  • explicit and adaptative interface
  • One routine

Feature redefinition

These rules concern the conforming inheritance.

  • Covariant redefinition of feature argument requires a 'variant' typing on redfined feature or first feature definition.
  • Contravariant redefinition of request result requires a 'variant' typing on first feature definition.

The contravariant redefinition has no interests for non-generic and little interests for generics. It can be avoided with a right abstraction.

Variant typing for Covariant redefinition:

  • A variant type argument has the most restricted type: The type of the previous definition
  • A variant type argument requires a simple object test to use the argument with the expected type.
  • A variant type argumpent can be assigned to a formal and variant type argument if the most restricted type of the argument is conform to the msot restricted type of the formal argument.

example 1: covariant redefinition of feature argument with 'variant' typing on redfined feature

Note that the type is not repeated in the object test. In the interface of the current type the 'variant' is removed. Indeed, it is not possible to call 'eat' on a COW instance with a parameter of type FOOD.

class
	ANIMAL
 
feature -- Access
	last: FOOD
 
faeture -- Eating
	eat (f: like last)
		require
			True
		do
			last := f
		ensure
			True
		end
end
class
	COW
 
inherit
	ANIMAL
		redefine all end
 
feature -- Access
	last: GRASS
 
faeture -- Eating
	eat (f: variant like last) -- or eat (f: variant FOOD)
		require else
			True
			-- 'f' type = expected type. Here: GRASS
		do
			-- 'f' type = type of the previous definition. Here: FOOD
			if attached f as g then
				-- 'g' type = expected type. Here: GRASS
				last := g
			end
		ensure then
			True
			-- 'f' type = type of the previous definition. Here: FOOD
		end
 
end

example 2: covariant redefinition of feature argument with 'variant' typing on first feature definition.

class
	ANIMAL
 
feature -- Access
	last: FOOD
 
faeture -- Eating
	eat (f: variant like last)
		do
			if attached f as safe then
				last := safe
			end
		end
end
class
	COW
 
inherit
	ANIMAL
		redefine last end
 
feature -- Access
	last: GRASS
 
end

example 3: contravariant redefinition

Z inherits of Y

Y inherits of X

class
	A
 
feature -- Access
	something: variant Z
 
feature -- Other
	do_something
		do
			-- 'something' type = ANY
			if attached something as expected then
				-- 'expected' type = feature result type. Here: Z
			end
		end
 
end
class
	B
 
inherit
	A
		redefine something end
 
feature -- Access
	something: variant Y
 
end
class
	C
 
inherit
	B
		redefine something, do_something end
 
feature -- Access
	something: X
 
feature -- Other
	do_something
		do
			-- 'something' type = X
		end
end

Generic conformance

The variant typing can be used to change the default variance of generic type.

  • Generic used only on feature argument or request result with variant typing is contravariant.
  • Generic used only on request result or feature argument with variant typing is covariant.

example: agents Then for agents it is possible to decalre requests 'operands' and 'empty_operands' as variant.

operands: variant detachable OPEN_ARGS
 
empty_operands: variant OPEN_ARGS
	do create Result ensure ... end

'apply' and 'target' should be write again. For example:

apply
	do
		-- 'operands' type = constraint inheritance. Here: detachable TUPLE
		if attached operands as args then
			-- 'operands' type = request result type. Here: OPEN_ARGS
			call (args)
		else
			call (Void)
		end
	end

Sub-conclusion

This proposition reuses an existing keyword and is very close from the Detachable types proposition, but without the detachable constraint and extended on the generics. The contravariant redefinition of request result can be remote for simplicity.

The lack of expressivity for generic variance can be reduced with a prefixed mark for a genric (see just below).

Ensure the generic variance

Sometimes the programmer wishes ensure a certain behavior. For example for agents: it is expected 'OPEN_ARGS' be a contravariant generic.

A generic prefixed with the 'variant' keyword cannot be novariant. It can be covariant or contravariant or both. This mark is optional (Backward compatibility and simplicity).

example: agents

class
	ROUTINE [BASE_TYPE, variant OPEN_ARGS -> detachable TUPLE create default_create end]
-- ...
end

The compiler must ensure 'OPEN_ARGS' is not novariant.

Discussions

Greater flexibility for generics

A lot of generics could be novariant, encouraging to propose a solution to have a new flexibility, but safe.

Wildcard generics

The request result type is the constrained inheritance type. And the type of the feature argument is (attached) NONE. The new semantic of 'Void' is considered: Void is not a NONE instance.

local
	a: ARRAY [?]; b: ARRAY [STRING]
	o: ANY
do
	-- ...
	a := b
	o := a.item (1)
	a.put ("try") -- invalid call. The type expected is NONE.
end

However the wildcard generics will be used on request result or feature argument. A more power and elegent solution could be the parametrized routines.

Parametrized routines

do_something [G] (a: ARRAY [G]): G
	require
		a.count >= 1
	do
		Result := a.item (1)
	end

Right abstraction

The wildcard generics and the parametrized routines introduce new constructs for Eiffel. Is there another solution avoiding this?

With a right abstraction it is possible to have flexible classes. It is comparable to the imuutable cocncept.

example: V_CONTAINER class of Eiffel Base 2 To obtain the genric covariance behavior it is necessary to have: - V_ITERATOR must have a covariant generic. - An immutable abstraction of tuple IMMUTABLE_TUPLE should be wrote.

In V_CONTAINER class only two features should use the variant typing:

occurrences (v: variant G): INTEGER
		do
			-- G -> ANY then 'v' type = ANY
			across Current as it loop
				if it.item = v then
					Result := Result + 1
				end
			end
		end
 
has (v: variant G)
	local
		it: like new_cursor
	do
		it := new_cursor
		it.search_forth (v) -- {V_ITERATOR [G]}.search_forth (v: variant G)
		Result := not it.after
	ensure
		occurrences (v) = 1	
	end

In V_ITERATOR only two features should use the variant typing without object test!

With these changes the next code is valid:

local
	a: V_CONTAINER [ANY]; b: V_CONTAINER [STRING]
do
	a := b
 
end

Sub-conclusion

With a right abstraction and a smart use of avraint typing for generics, it is possible to obtain a greater flexibility keeping a fully aivailable interface.

Export status restrictions

Since the ECMA Eiffel Standard forbids the export restriction with conforming inheritance, it is not a problem. However, the semantic can be changed to enable this restriction on conforming inheritance.

Restrict exportation should not cause a cat-call. The mechanism could be used just to change the class interface.

deferred class
	ANIMAL
feature
	is_vegetarian: BOOLEAN
		deferred end
end
class
	COW
inherit
	ANIMAL
		export {NONE}
			is_vegetarian
		end
feature {NONE}
	is_vegetarian: BOOLEAN = True
end
local
	an_animal: ANIMAL; a_cow: COW
	b: BOOLEAN
do
	create a_cow
	b := a_cow.is_vegetarian -- invalid call
 
	an_animal := a_cow
	b := an_animal .is_vegetarian -- valid call
end

The class interface is easier.

Conclusion

The proposition uses no new keyword and solve the cat-call problem. It passes all Catcall checkpoints.

The default generic conformance give a natural safe and flexible static typing. The variant typing enables to create adaptive interfaces keeping a safe static typing.

With a fine abstraction and the use of variant typing for generic, the genric conformance flexibility is little restricted. The contravariant for request result can then be discarded for simplicity.

There is no interface restriction (Interval types or Usage-site variance). Class interfaces are fully available.