Grammar

Japanese version

Contents

Symbol and Comment

In LiLFeS the following is treated as token Comments can be used in the following ways:

Type, Features & Variables

As mentioned in chapter 3, all constants in LiLFeS are treated as TYPE. Also predicates becomes subtype of pred. Alphanueric, whitespace, 2 byte characters etc... can be used as strings, only quote needs to be delimited with ' to use quote ', delimit it with '
Types that can be used in Program
    head_subject_schema
    Tokyo Univ
    'ABC'                  % type ABC
    'Good morning!'        % type Good morning! 
    'don''t'               % type don't
Feature Name can be any string which is alphanumeric (valid characters are alphanumeric, underscope, 2byte char). Like Type, strings within quotes can also be used. it is necessary to add \ suffix. White space or newline character is not allowed between the name and \.
Features that can be used in Program                
    F1\
    feature\
    
    'Let me see...'\
variable can be any string starting with   $ (valid characters are uppercase alphanumeric, underscore, 2 byte char)
Variables that can be used in Program               
    X
    $VAR
    $1
    $VARIABLE

Type Definition

All predicates and types have to be defined as supertype. It is not possible to use the same name more than two times.
Type <- [ supertype1 , supertype2,cc ] + [ Feature1\ Type1(Priority1), Feature2\ Type2( Priority2),...].
Type Definition
(1) p <- pred.
(2) Square <- [Rectangle , Rhombus].
(3) employ <- [act] + [EMPLOYER\, EMPLOYEE\].
(4) book <- [bot] + [TITLE\string(-5),AUTHOR\name(10),PUBLISHER\co(20)].
Left hand side of <-  is the new type. The right hand side can be multiple supertypes separated by +. If the supertype is only one then [] can be omitted (e.g., 1). It is possible to separate multiple types by comma (e.g., 2).  If feature are to be added, its priority can be omitted. if Type is omitted it will be treated as bot. the order of features will be alphabetical if priority is omitted. e.g., 3 it will be EMPLOYEE, EMPLOYER. In e.g., 4 it will be TITLE, AUTHOR, PUBLISHER based on the priority specified.

Feature

Feature can be defined as shown:

FEATURE AVM
employ & EMPLOYER\"John" & EMPLOYEE\"Mary" 
|~employ         ~|
| EMPLOYEE:"Mary" |
|_EMPLOYER:"John"_|

Racing Horse & Name\ Dance in the Dark &
         Lineage\ 'Hail to Reason' &
         Father\'(Stud Horse & Name\Sunday Science &
                      Lineage\ 'Hail to Reason')

|~Racing Horse                      ~|
| Name: Dance in the Dark            |
| Lineage:Hail to Reason             |
|        |~Stud Horse             ~| |
| Father:| Name :Sunday Science    | |
|        |_Lineage: Hail to Reason_| |
|_Mother:bot                        _|

Unification & Predicate

Unification is 
(Type or variable) = (Type or variable)
Unification Result
X = man X can be used instead of Man
Equilateral = Rectangle Not possible
$Horse & Lineage\$1 = Father\Lineage\$1 $Horse will be horse with same Father and Lineage

Predicate is same syntax as in Prolog

Predicate Name( Param1, Param2, ...)

Query

Same syntax as Prolog
(1) Predicate .
(2) Predicate :- structure1 , structure2,.... .
(3) :- structure .

query

?- structure1 ,  structure2 ,....
?-  is not a prompt. Unlike Prolog, LiLFeS is not interpreted.

difference in ?- & :- 
> p <- [pred].  q <- [pred].               
> a <- [bot].  b <- [bot].
> p(a,b).
> p(X,X) :- q(X).
> p(b).

> ?- p(X,Y), q(Y).
X: a
Y: b
Enter ';' for more choices, otherwise press ENTER --> ;
X: [$1] b
Y: [$1] ...
Enter ';' for more choices, otherwise press ENTER --> ;
no
 (in case of query, variable's values are shown, inputting ; will backtrack)
> :- p(X,Y), X=Y, q(Y), print(Y).
b
 (direct execution, nothing is displayed. only print output is shown
  and thus one can print only selected variables but cannot backtrack)

Modules

An example Module in LiLFeS is shown below:

LiLFeS Program
    :- module("list:reverse").

    :- module_interface.
    % read required modules
    :- ensure_loaded("append").
    % make the scope global
    reverse <- [pred].

    :- module_implementation.
    % make the scope local
    reverse_sub <- [pred].
    % Predicate Definition
    reverse_sub([], $X, $X).
    reverse_sub([$A|$B], $C, $D) :-
        reverse_sub($B, [$A|$C], $D).
    reverse($X, $Y) :-
        reverse($X, [], $Y).

module, module_interface, module_implementation etc. are predefined keywords and are also referred to as directives, which can be used to write programs.

First the module name has to be specified in the file. The module name can be anything, but the "/" in file path is to be converted to ":". For example, if LILFES_PATH is ~/lilfes/ Cand  module is ~/lilfes/list/reverse.lil  then the module name will be list:reverse.

Next is the module_interface which defines the global scope of the module. for example  reverse predicate is global. This can be accessed from other modules without module path i.e., list:reverse.

Finally is the module_implementation which defines the local scope of the modules. In order to view reverse_sub, the full path has to be specified such as list:reverse:reverse_sub.

In order to use predicates from other modules, use  ensure_loaded directive to load the module. In the above example, list:append module is loaded and all predicates defined in that module can be used in this program. ensure_loaded can be used both in  module_interface or module_implementation.

Special Types

List

List can be defined as: (list subtype)

[Element1 , Element2 , ...]
[Head | Tail List]

Numeric

integer, float subtype 

String

String is any character enclosed within " " is of string  subtype.

Other

Don't care

variables starting with underscore (_) are treated as Don't care and will always succeed in unification. Hence predicates which are not referred to can be preceded by _ to ignore them. The unification part is exactly same as trying to unify with bot.

Inference

, is AND  ; is OR
p(a) = q(x), q(y) ; r(z).
will become:
p(a) = q(x), q(y).
p(a) = r(z).

if 

Condition1 , Condition2 , ... -> Predicate on success; Predicate on failure
P :- Q -> R ; S.
will be treated as
P :- Q , ! , R.
P :- S.
if
> p <- [pred].
> check <- [pred].
> a <- [bot].
> b <- [bot].
> p(a).
> check(X) :- p(X) -> strprintln("succeed") ; strprintln("failed").
> ?- check(a).
succeed
; (Current occupation: 9 heap, 3 stack, 10 trail cell(s), 39 instructions)
yes
> ?- check(b).
failed
; (Current occupation: 9 heap, 3 stack, 8 trail cell(s), 62 instructions)
yes

Negative

Like Prolog, using \+, the failure of a predicate can be negated to success
Negative
> p <- [pred].
> a <- [bot].
> b <- [bot].
> p(a).
> ?- \+ p(a).
; (Current occupation: 15 heap, 23 stack, 16 trail cell(s), 101 instructions)
no
> ?- \+ p(b).
; (Current occupation: 11 heap, 3 stack, 10 trail cell(s), 122 instructions)
yes

Unification not possible cases

It is possible to allow unification not possible cases by using  \= which is opposite of  = , and can be used when unification fails. Note, the left hand side value and \= should have a white space. If there is no white space, it will be treated as feature.

Unification not possible
> mammal <- bot.
> birl <- bot.
> cat <- mammal.
> ?- mammal \= cat.
; (Current occupation: 10 heap, 20 stack, 16 trail cell(s), 12 instructions)
no
> ?- mammal \= bird.
; (Current occupation: 10 heap, 3 stack, 12 trail cell(s), 27 instructions)
yes
Prev: Types and Features Next: Error Messages
Table of Contents LiLFeS Documents LiLFeS Home Page Tsujii Lab.