Personal tools

User:Zzo38/Proposal for more-notation

From HaskellWiki

< User:Zzo38
Revision as of 20:29, 3 September 2011 by Zzo38 (Talk | contribs)

Jump to: navigation, search

This document is proposal for more-notation in Haskell.

Contents

1 Syntax

Anywhere that a more-notation is allowed (see Uses), you can use the syntax:

  • more [(name_of_enumeration)] more_name {parameters}
  • name_of_enumeration: A name of any type which is an enumeration. It can be another one using more-notation, but it must be able to fully determine it before this more-notation can be fully determined.
  • more_name: A capitalized name. It uses the same namespace as constructors, so you are not allowed to have a constructor of the same name (although it is OK to have types of the that name).
  • parameters: Optional lowercased names. These names are in local scope, and may already exist in this scope, although they do not have to.

A more-declaration is a module-level declaration of this syntax:

  • [ numeric_literal | (enumeration_constructor) ] more_name {parameters} = contents { | contents } ;
  • numeric_literal: It should be a natural number. Omitted is the same as zero. This number is called the "order" of the declaration.
  • enumeration_constructor: A constructor of the enumeration that was specified in the more-notation. If an enumeration is specified, the enumeration constructor is required here. This number is called the "order" of the declaration.
  • parameters: Lowercased names which will be scoped locally to this declaration. The number of parameters must match the number of parameters specified in the more-notation. An unused parameter can be replaced by an underscore.
  • contents: The contents depends on where the more-notation is.

2 Semantics

In the place where the more-notation is, all contents of more-declarations of that name will be reordered and corrected before being placed in place of the more-notation.

Automatic reordering is overridden by the enumeration constructors or numeric literals in front of more-declarations.

3 Uses

3.1 Do-blocks

A statement in a do-block may be replaced by a more-notation. Semantics of more-declarations becomes as follows:

  • Multiple statements in a single more-declaration stay in that order and consecutive.

Example:

x = (2 *);
main = do {
  x <- return 21;
  more Doing x;
  print x;
};
Doing a = a <- return $ x a;

becomes

x = (2 *);
main = do {
  x <- return 21;
  x <- return $ Main.x x;
  print x; -- results in 42 (the answer to life, the universe, and everything)
};

3.2 Case alternatives

An alternative for a case block may be replaced by a more-notation. Semantics of more-declarations becomes as follows:

  • Cases within an order are reordered so that the most specific ones come first.

Example:

X = (_, False, True) -> 1;
X = (True, True, True) -> 2;
X = _ -> 3;
X = (False, True, True) -> 4;
y x = case x of {
  more X;
};

becomes:

y x = case x of {
  (True, True, True) -> 2;
  (False, True, True) -> 4;
  (_, False, True) -> 1;
  _ -> 3;
};

3.3 Case guards

Cases can have guards and case alternatives in more-declarations can also have guards that use more-notation. There is no guarantee to the order except:

  • Guards listed in a single more-declaration keep that order.
  • Duplicate guards are removed.
  • Duplicate guard conditions that to not have matching result expressions are errors.
  • You can list guard conditions without the result values; these cause the order to be forced if such guard conditions exist, but does not create them if they do not exist.

Example:

y x = case x of {
  more Cases;
};
Cases = (a, b) | more X a b;
X a b = a > 5 -> b;
X a b = a < 5 -> a;
X a b = a == 5 -> 0;

3.4 Datatype declarations

A constructor definition in a data declaration is allowed to be replaced by a more-notation. Semantics of more-declarations becomes as follows:

  • Duplicate constructors are removed.
  • Duplicate constructors that do not match are errors.
  • Multiple constructors in a single more-declaration are guaranteed to keep the order given.

Example:

data T = Zero | more T deriving (Eq);
T = Two | Three | Four;
T = One | Two;

becomes

data T = Zero | One | Two | Three | Four deriving (Eq);

3.5 Fields in record syntax

A field in either a data type declaration or in a value construction or update syntax can be replaced by more-notation. Semantics of more-declarations becomes as follows:

  • Duplicate fields are removed.
  • Duplicate fields that do not match are errors.
  • A more-declaration can specify both the type and the value. Types are only used in data type declarations and values only in the other places.

Example:

data Abc = Abc { name_of :: String, more AbcField };
makeAbc :: Abc;
makeAbc = Abc { name_of = "", more AbcField };
AbcField = age_of :: Int = 0;
AbcField = existing :: Bool = True;

becomes

data Abc = Abc { name_of :: String, age_of :: Int, existing :: Bool };
makeAbc :: Abc;
makeAbc = Abc { name_of = "", age_of = 0, existing = True }; 

3.6 Lists

A list item can be replaced by a more-notation. Semantics of more-declarations becomes as follows:

  • Multiple items in a single more-declaration stay in that order and consecutive.

Example:

k :: [String];
k = [more K];
1 K = "One";
5 K = "Five";
3 K = "Three";
K = "Zero";

becomes:

k :: [String];
k = ["Zero", "One", "Three", "Five"];

4 Combining with Template Haskell

4.1 Uses in quotations

4.2 Manipulation in splices

4.3 Types and constructors

* Dec
** MoreCaseD Name MoreOrder [Name] [Match]
** MoreDataD Name MoreOrder [Name] [Con]
** MoreDoD Name MoreOrder [Name] [Stmt]
** MoreListD Name MoreOrder [Name] [Exp]
** MoreRecD Name MoreOrder [Name] [MoreField]
* Info
** MoreI Int MoreMode Name [Dec]
* MoreField = (Name, Maybe StrictType, Maybe Exp)
* MoreMode
** MoreCase
** MoreData
** MoreDo
** MoreList
** MoreRec Bool Bool
* MoreOrder
** NoOrder
** NamedOrder Name
** NumericOrder Int