Modula-2 Reloaded

A Modern Typesafe & Literate Programming Notation

Site Menu

Project

Specification

Implementation

Recommendations

Reference

Needs Updating

Work in Progress

Wastebasket

Wiki Manual

edit SideBar

Synthesis Of Non Scalar Type Conversion

WiP.SynthesisOfNonScalarTypeConversion History

Hide minor edits - Show changes to output

2010-04-07 16:10 by benjk -
Changed line 18 from:
should be synthesised by the compiler in the event that no direct conversion path exists from type @@T1@@ to type @@T2@@ as long as @@V@@ and @@W@@ have the same number of components.
to:
should be synthesised by the compiler in the event that no direct conversion path exists from type @@T1@@ to type @@T2@@, as long as @@V@@ and @@W@@ have the same number of components.
Restore
2010-04-07 16:09 by benjk -
Changed line 3 from:
The objective of the facility described here is to facilitate synthesis of type conversion between non-scalar types even if no direct conversion path exists between their component types.
to:
The objective of the facility described here is to facilitate synthesis of type conversion between non-scalar types even if no direct conversion path exists between them or their component types.
Restore
2010-04-07 16:08 by benjk -
Changed lines 56-57 from:
       V.toASCII( %ACCESSOR%(index, v), FALSE, significand, exponent );
        %MUTATOR%(index, w, W.fromASCII( TRUE, significand, exponent ));
to:
       V.toASCII( %ACCESSOR%(index, v), %decOrHexT1%, significand, exponent );
        %MUTATOR%(index, w, W.fromASCII( %decOrHexT2%, significand, exponent ));
Changed line 61 from:
where [@%MAX_LITERAL_LENGTH%@] is a compiler constant designating the maximum length allowed for literals, [@%ACCESSOR%@] is the accessor function and [@%MUTATOR%@] is the mutator procedure.
to:
where [@%MAX_LITERAL_LENGTH%@] is a compiler constant designating the maximum length allowed for literals, [@%decOrHexT1%@] and [@%decOrHexT2%@] are boolean values indicating the radix system of a type's internal representation, [@%ACCESSOR%@] is the accessor function and [@%MUTATOR%@] is the mutator procedure.
Restore
2010-04-07 16:04 by benjk -
Changed line 63 from:
NB: For the compiler to save stack space for the intermediate strings significand and exponent a means to obtain the exact required number of digits for a scalar type could be introduced. This is an optimisation issue and it has no impact on the semantics of the synthesis.
to:
NB: For the compiler to save stack space for the intermediate strings @@significand@@ and @@exponent@@ a means to obtain the exact required number of digits for a scalar type could be introduced. This is an optimisation issue and it has no impact on the semantics of the synthesis.
Restore
2010-04-07 16:04 by benjk -
Changed line 22 from:
The key to synthesising such a conversion is to decompose @@v@@ into its components, serialise the components using conversion primitive [@toASCII@] of type @@T1@@, then deserialise the components using conversion primitive [@fromASCII@] of type @@T1@@ and finally compose @@w@@ from the so converted components.
to:
The key to synthesising such a conversion is to decompose @@v@@ into its components, serialise the components using conversion primitive [@toASCII@] of type @@T1@@, then deserialise the components using conversion primitive [@fromASCII@] of type @@T2@@ and finally compose @@w@@ from the so converted components.
Restore
2010-04-07 16:03 by benjk -
Changed lines 50-52 from:
    index : CARDINAL;
     significand,
     exponent : ARRAY %MAX_LITERAL_LENGTH% OF CHAR;
to:
   index : CARDINAL;
    significand,
    exponent : ARRAY %MAX_LITERAL_LENGTH% OF CHAR;
Restore
2010-04-07 16:03 by benjk -
Changed lines 44-45 from:
[@
VAR v : V; w : W;
to:
[@VAR v : V; w : W;
Changed lines 49-54 from:
[@
VAR v : V; w : W;
        index : CARDINAL;
        significand,
        exponent : ARRAY %MAX_LITERAL_LENGTH% OF CHAR;
to:
[@VAR v : V; w : W;
    index : CARDINAL;
     significand,
    exponent : ARRAY %MAX_LITERAL_LENGTH% OF CHAR;
Changed lines 59-60 from:
END; (* IF *)
@]
to:
END; (* IF *)@]
Restore
2010-04-07 16:00 by benjk -
Changed line 22 from:
The key to synthesising such a conversion is to decompose @@v@@ into its components, serialise the components using @@v@@'s component type @@T1@@'s @@`toASCII@@ conversion primitive, then deserialise the components using @@w@@'s component type @@T2@@'s @@`fromASCII@@ conversion primitive and finally compose @@w@@ from the so converted components.
to:
The key to synthesising such a conversion is to decompose @@v@@ into its components, serialise the components using conversion primitive [@toASCII@] of type @@T1@@, then deserialise the components using conversion primitive [@fromASCII@] of type @@T1@@ and finally compose @@w@@ from the so converted components.
Restore
2010-04-07 15:57 by benjk -
Changed line 5 from:
Given variables @@v@@ of vector type @@V@@ with component type @@T1@@ and @@w@@ of vector type @@W@@ with component type @@T2@@ ...
to:
Given @@v@@ of vector type @@V@@ with component type @@T1@@ and @@w@@ of vector type @@W@@ with component type @@T2@@ ...
Restore
2010-04-07 15:56 by benjk -
Added lines 1-66:
!!!!Objective

The objective of the facility described here is to facilitate synthesis of type conversion between non-scalar types even if no direct conversion path exists between their component types.

Given variables @@v@@ of vector type @@V@@ with component type @@T1@@ and @@w@@ of vector type @@W@@ with component type @@T2@@ ...

[@TYPE
    V = OPAQUE RECORD ("V-Type") v1, v2, v3 : T1 END;
    W = OPAQUE RECORD ("V-Type") v1, v2, v3 : T2 END;

VAR
    v : V; w : W;@]

the conversion from @@v@@ to @@w@@ ...

[@w := v :: W;@]

should be synthesised by the compiler in the event that no direct conversion path exists from type @@T1@@ to type @@T2@@ as long as @@V@@ and @@W@@ have the same number of components.

!!!!Approach

The key to synthesising such a conversion is to decompose @@v@@ into its components, serialise the components using @@v@@'s component type @@T1@@'s @@`toASCII@@ conversion primitive, then deserialise the components using @@w@@'s component type @@T2@@'s @@`fromASCII@@ conversion primitive and finally compose @@w@@ from the so converted components.

To facilitate this, the non-scalar types must provide
* a function to return the number of components or highest index
* an accessor function to return the n-th component of a vector
* a mutator procedure to overwrite the n-th component of a vector

The component type can be deduced form the return type of the accessor function.

This could be done using the following bindings:

[@
PROCEDURE [HIGH] highestComponentIndex ( v : VECTOR ) : CARDINAL;
PROCEDURE [.] componentN ( n : CARDINAL; v : VECTOR ) : ComponentType;
PROCEDURE [!] setComponentN ( n : CARDINAL; VAR v : VECTOR; c : ComponentType );
@]

An example is given in [[WiP/TypeSampleVector]]


The generalised synthesis algorithm is:

[@
VAR v : V; w : W;
w := v :: W;@]

synthesises to =>

[@
VAR v : V; w : W;
        index : CARDINAL;
        significand,
        exponent : ARRAY %MAX_LITERAL_LENGTH% OF CHAR;

IF HIGH(v) = HIGH(w) THEN
    FOR index := 0 TO HIGH(v) DO
        V.toASCII( %ACCESSOR%(index, v), FALSE, significand, exponent );
        %MUTATOR%(index, w, W.fromASCII( TRUE, significand, exponent ));
    END; (* FOR *)
END; (* IF *)
@]

where [@%MAX_LITERAL_LENGTH%@] is a compiler constant designating the maximum length allowed for literals, [@%ACCESSOR%@] is the accessor function and [@%MUTATOR%@] is the mutator procedure.

NB: For the compiler to save stack space for the intermediate strings significand and exponent a means to obtain the exact required number of digits for a scalar type could be introduced. This is an optimisation issue and it has no impact on the semantics of the synthesis.
Restore