Relational algebra: Difference between revisions
EndreyMark (talk | contribs) (Adding reference to pointfree (in shorly describing Oliveira's mentioned paper)) |
EndreyMark (talk | contribs) (→Just a thought: : an early, immature thought of mine to represent relational algebra expressions) |
||
Line 1: | Line 1: | ||
__TOC__ | __TOC__ | ||
== Pointfree == | |||
José Nuno Oliveira: [http://www.di.uminho.pt/~jno/ps/_.pdf First Steps in Pointfree Functional Dependency Theory]. A concise and deep approach, it is [[pointfree]]. See also [http://www.di.uminho.pt/~jno/html/ the author's homepage] and also [http://www.di.uminho.pt/~jno/html/jnopub.html his many other papers] -- many materials related to in this topic can be found. | José Nuno Oliveira: [http://www.di.uminho.pt/~jno/ps/_.pdf First Steps in Pointfree Functional Dependency Theory]. A concise and deep approach, it is [[pointfree]]. See also [http://www.di.uminho.pt/~jno/html/ the author's homepage] and also [http://www.di.uminho.pt/~jno/html/jnopub.html his many other papers] -- many materials related to in this topic can be found. | ||
== Just a thought == | |||
An early, immature thought of mine to represent relational algebra expressions: | |||
<haskell> | |||
data Query :: * -> * -> * where | |||
Identity :: Scheme a => Query a a | |||
Restrict :: (Scheme a, Scheme b) => Expr b Bool -> Query a b -> Query a b | |||
Project :: (Scheme a, Scheme b, Scheme b', Sub b' b) => b' -> Query a b -> Query a b' | |||
Rename :: (Scheme a, Scheme b, Scheme b', Iso b b') => Query a b -> Query a b' | |||
Product :: (Scheme a, Scheme b1, Scheme b2, Scheme b, Sum b1 b2 b) => | |||
Query a b1 -> Query a b2 -> Query a b | |||
Union :: (Scheme a, Scheme b, Id b) => Query a b -> Query a b -> Query a b | |||
Difference :: (Scheme a, Scheme b, Id b) => Query a b -> Query a b -> Query a b | |||
</haskell> | |||
... using the concepts / ideas of | |||
* [[generalised algebraic datatype]] | |||
* a sort of differential approach (I think I took it from [[Zipper]]). | |||
The case of <hask>Restrict</hask> uses <hask>Expr</hask>. I think, the concept of <hask>Expr</hask> is an ''inside'' approach (making the relational algebra -- regarded as an embedded language -- richer, more autonome from the host language, but also more restricted): | |||
<haskell> | |||
data Expr :: * -> * -> * where | |||
Constant :: (Scheme sch, Literal a) => a -> Expr sch a | |||
Attribute :: (Scheme sch, Match attr a, Context attr sch) => attr -> Expr sch a | |||
Not :: Scheme sch => Expr sch Bool -> Expr sch Bool | |||
And :: Scheme sch => Expr sch Bool -> Expr sch Bool -> Expr sch Bool | |||
Or :: Scheme sch => Expr sch Bool -> Expr sch Bool -> Expr sch Bool | |||
Equal :: (Scheme sch, Eq a) => Expr sch a -> Expr sch a -> Expr sch Bool | |||
Less :: (Scheme sch, Ord a) => Expr sch a -> Expr sch a -> Expr sch Bool | |||
</haskell> | |||
Maybe an ''outside'' approach (exploiting the host language more, thus enjoying more generality) would be also appropriate: | |||
<haskell> | |||
data Query :: * -> * -> * where | |||
... | |||
Restrict :: (Scheme a, Scheme b, Record br, On b br) => (br -> Bool) -> Query a b -> Query a b | |||
... | |||
Rename :: (Scheme a, Scheme b, Scheme b', Iso b b') => (b -> b') -> Query a b -> Query a b' | |||
</haskell> | |||
[[Category:Theoretical foundations]] | [[Category:Theoretical foundations]] |
Revision as of 10:19, 17 June 2006
Pointfree
José Nuno Oliveira: First Steps in Pointfree Functional Dependency Theory. A concise and deep approach, it is pointfree. See also the author's homepage and also his many other papers -- many materials related to in this topic can be found.
Just a thought
An early, immature thought of mine to represent relational algebra expressions:
data Query :: * -> * -> * where
Identity :: Scheme a => Query a a
Restrict :: (Scheme a, Scheme b) => Expr b Bool -> Query a b -> Query a b
Project :: (Scheme a, Scheme b, Scheme b', Sub b' b) => b' -> Query a b -> Query a b'
Rename :: (Scheme a, Scheme b, Scheme b', Iso b b') => Query a b -> Query a b'
Product :: (Scheme a, Scheme b1, Scheme b2, Scheme b, Sum b1 b2 b) =>
Query a b1 -> Query a b2 -> Query a b
Union :: (Scheme a, Scheme b, Id b) => Query a b -> Query a b -> Query a b
Difference :: (Scheme a, Scheme b, Id b) => Query a b -> Query a b -> Query a b
... using the concepts / ideas of
- generalised algebraic datatype
- a sort of differential approach (I think I took it from Zipper).
The case of Restrict
uses Expr
. I think, the concept of Expr
is an inside approach (making the relational algebra -- regarded as an embedded language -- richer, more autonome from the host language, but also more restricted):
data Expr :: * -> * -> * where
Constant :: (Scheme sch, Literal a) => a -> Expr sch a
Attribute :: (Scheme sch, Match attr a, Context attr sch) => attr -> Expr sch a
Not :: Scheme sch => Expr sch Bool -> Expr sch Bool
And :: Scheme sch => Expr sch Bool -> Expr sch Bool -> Expr sch Bool
Or :: Scheme sch => Expr sch Bool -> Expr sch Bool -> Expr sch Bool
Equal :: (Scheme sch, Eq a) => Expr sch a -> Expr sch a -> Expr sch Bool
Less :: (Scheme sch, Ord a) => Expr sch a -> Expr sch a -> Expr sch Bool
Maybe an outside approach (exploiting the host language more, thus enjoying more generality) would be also appropriate:
data Query :: * -> * -> * where
...
Restrict :: (Scheme a, Scheme b, Record br, On b br) => (br -> Bool) -> Query a b -> Query a b
...
Rename :: (Scheme a, Scheme b, Scheme b', Iso b b') => (b -> b') -> Query a b -> Query a b'