CLIPS

1
CLIPS

• Bob McKay
• School of Computer Science and Engineering
• College of Engineering
• Seoul National University
• Partly based on
• CLIPS documentation
• Lecture notes by Dr X Yao

2
Outline

• CLIPS as a rule-based system
• CLIPS functions
• CLIPS Pattern Matching

3
References

• Giarratano and Riley Expert Systems Principles
  and Programming, 4th Edition, Course Technology,
  ISBN 0-534-38447-1

4
What is CLIPS?

• The CLIPS system is a forward-chaining rule-based
  expert system shell
• That is, a reasoning system for positive definite
  clauses
• But with a lot of extensions
• It uses the forward-chaining RETE algorithm
• It is written in C
• Syntax is very similar to LISP
• It runs under UNIX, Windows and Macintosh

5
Running CLIPS

• On unix, just type clips
• The CLIPSgt prompt will appear
• This is the top level of CLIPS
• On windows and macintosh, just click the CLIPS
  icon
• To exit, enter (exit)
• Note that all CLIPS commands and functions are
  enclosed in a pair of brackets
• Comments start with a

6
FACTS

• A fact is one or more fields enclosed within
  matching left and right parentheses
• (isa John student)
• (isa Tom teacher)

7
Asserting Facts

• CLIPS keeps track of the facts it believes to be
  true through the facts list
• It is part of CLIPS working memory
• You can create facts and add them to the fact
  list with assert commands
• CLIPSgt (assert (isa John student))
• CLIPSgt (assert (isa Tom teacher))
• CLIPSgt (facts) this is a comment
• f-0 (isa John student)
• f-1 (isa Tom teacher)
• The (facts) command displays all the facts
  currently in the fact list.

8
Retracting Facts

• You can delete a fact from the fact list with the
  retract command
• You can delete all the facts with the clear
  command
• CLIPSgt (clear)
• CLIPSgt (assert (isa Mary student) (isa Joe
  student) (isa Kay student))
• CLIPSgt (facts)
• f-0 (isa Mary student)
• f-1 (isa Joe student)
• f-2 (isa Kay student)
• CLIPSgt (retract 0 2)
• CLIPSgt (facts)
• f-1 (isa Joe student)
• CLIPSgt (clear)
• CLIPSgt (facts)
• CLIPSgt

9
Initial Facts

• Instead of entering facts from CLIPS directly, we
  can also define a set of initial facts to be
  loaded into CLIPS
• These initial facts can be written and edited in
  a file
• The deffacts function is used for defining
  initial facts
• (deffacts facts-name comments fact-1 fact-2 ...
  fact-n)
• Note that the deffacts function can also be
  entered under CLIPS
• It will only be added to CLIPS's fact list when
  CLIPS is reset
• The reset command always adds a special fact
  (initial-fact) with identifier f-0.
• To remove a deffacts function from memory, use
• (undeffacts facts-name)
• We can even remove the system-defined
  initial-fact
• (undeffacts initial-fact)

10
Rules

• Rules are defined using the defrule command
• (defrule rule-name comment
• pattern-1 ... pattern-n
• gt
• action-1 ... action-m)
• Note that the entire rule must be surrounded by
  parentheses
• Each pattern and action must be too
• CLIPS matches the patterns of rules against facts
  in the fact list
• If all the patterns of a rule match, the rule is
  put on the agenda
• The agenda is a priority list of activated rules
  awaiting execution
• Once a rule is selected by CLIPS for execution,
  it is fired and the actions in the RHS
  (right-hand-side) of the rule are carried out
• The rule is removed from the agenda
• The program stops when no activations are on the
  agenda

11
Conflict Resolution

• By default, the rules on the agenda are in stack
  order
• The rules are read bottom to top, which means
  that the earliest rules in the KB will be at the
  top of the stack
• But you can control the rule priorities another
  way
• CLIPS uses salience to order rules on the
  agenda.
• The rule with largest salience value has the
  highest priority.
• Salience is represented by a value between
• -10,000 (lowest priority)
• 10,000 (highest priority)
• You set the salience in the rule definition
• (defrule rule-name comment (declare (salience
  25))
• pattern-1 ... pattern-n
• gt
• action-1 ... action-m)
• The default salience value is 0

12
The Watch Command

• With the watch command, you can watch a number of
  things
• facts, activations, rules
• CLIPSgt (reset)
• CLIPSgt (watch facts)
• CLIPSgt (watch activations)
• CLIPSgt (watch rules)
• CLIPSgt (assert (just testing))
• gt f-1 (just testing)
• CLIPSgt (defrule rule-1 "have a try" (declare
  (salience 20))
• (just testing) this is the
  pattern
• gt the arrow
• (printout t "it works!" crlf))
  a print function

13
The Watch Command

• Activation 20 rule-1 f-1
• CLIPSgt (agenda)
• 20 rule-1 f-1
• For a total of 1 activation
• CLIPSgt (run)
• FIRE 1 rule-1 f-1
• it works!
• CLIPSgt (save "example1.clp") only saves the
  defrule or deffacts
• CLIPSgt (unwatch all)
• CLIPSgt (agenda)
• CLIPSgt (run)
• CLIPSgt

14
Other facts commands

• (save-facts)
• saves the current fact list to a file
• (load-facts)
• reads a file of facts into the fact list
• Different from the (load) command which reads a
  file into CLIPS' memory
• (rules)
• tells us how many rules are in CLIPS' memory
• what their names are
• (pprule rule-name)
• displays the definition of the rule called
  rule-name

15
Variables

• Variables names
• "?" followed by a symbol which is the variable
  name
• A variable can be bound to any value
• CLIPSgt (reset)
• CLIPSgt (assert (isa Joe student) (isa Bo
  student))
• CLIPSgt (defrule rule-2
• ?x lt- (isa ?whoever student)
• gt
• (retract ?x)
• (assert (isa ?whoever teacher)))

16
Variables

• CLIPSgt (run)
• CLIPSgt (facts)
• f-0 (initial-fact)
• f-1 (isa Joe teacher)
• f-2 )isa Bo teacher)
• For a total of 3 facts
• The left arrow "lt-" binds a variable to an entire
  fact

17
Wildcards

• A question mark "?" without anything else is a
  single field wildcard
• It matches anything in a single field
• It is used to match something which we are not
  interested in
• The multi-field wildcard is "?
• It matches zero or more fields
• It can also have a name like ?here

18
Constraints

• Negation constraint
• represented by the prefix .
• x matches anything other than x
• Or constraint
• represented by the infix "_".
• x_y matches either x or y.
• And constraint.
• represented by the infix "".
• Normally used only with other constraints
• otherwise it's not much practical use.
• ?xy_z matches either y or z and binds x to it
• (careful ?xred_yellow)
• matches
• (careful red)
• and
• (careful yellow)

19
Mathematical Functions

• CLIPS uses prefix form for arithmetic
• CLIPSgt (defrule addition
• (number ?x ?y)
• gt
• (assert (sum ( ?x ?y))))
  is important here
• CLIPSgt (assert (number 2 3))
• CLIPSgt (run)
• CLIPSgt (facts)
• f-0 (initial-fact)
• f-1 (number 2 3)
• f-2 (sum 5)
• For a total of 3 facts

20
Bind

• You can also use bind to set a variable to a
  value
• CLIPSgt (defrule addition-again
• (numbers ?x ?y)
• gt
• (bind ?sum ( ?x ?y))
• (printout t "The sum is " ?sum
  crlf))
• CLIPSgt (run)
• The sum is 5
• "" or bind in the above tells CLIPS to
  perform an evaluation

21
Type Conversion

• CLIPSgt ( 1 2)
• 3
• CLIPSgt ( 1 2.0)
• 3.0
• CLIPSgt (integer ( 1 2.9))
• 3
• CLIPSgt (float ( 1 2))
• 3.0

22
The test function

• The test function is used to compare values
• (test (function arg-1 arg-2 ... arg-n) )
• the function can be
• eq (equal)
• neq (not equal)
• (numeric equal)
• ! (numeric not equal)
• gt
• gt
• lt
• lt
• ! (not)
• (and)
• __ (or)

23
Test

• (defrule same-height
• (height ?x ?hx)
• (height ?y ?hy)
• (test ( ?hx ?hy))
• gt
• (printout t ?x " and " ?y " are of the same
  height." crlf))

24
Pattern Connectives

• Not the same as field constraints!
• There are three explicit pattern connectives
  and, or, not.
• (defrule tall-person
• (or (and (isa ?x female)
• (height ?x ?hx)
• (test (gt ?hx 170)))
• (and (isa ?y male)
• (height ?y ?hy)
• (test (gt ?hy 180))) )
• gt
• (printout t "He/She is tall." crlf))

25
User-Defined Functions

• You can define your own functions
• (deffunction function-name comment
• (?arg1 ?arg2 ... ?argn)
• ltaction-1gt
• ltaction-2gt
• ...
• ltaction-mgt )
• (deffunction summation (?x ?y ?z)
• (bind ?t ( ?x ?y ?z))
• (printout t "The summation is " ?t crlf))
• To remove a user-defined function, use
  undeffunction

26
I/O File Open and Close

• Files must be opend before they can be used
• (open "file-name" logical-name mode)
• where mode is one of
• "w" (write)
• "r" (read)
• "r" (read and write)
• "a" (append)
• Closing a file
• (close logical-name)

27
I/O Output

• We have already used printout for output to the
  terminal
• printout can output to a file if we give the
  logical name of the file
• (open "myfile.dat" writeit "w")
• (fprintout writeit "Here is the message." crlf)
• writeit is the logical name of the file
  myfile.dat
• Open function creates a writable ("w") file in
  the current directory

28
I/O Input

• (read logical-name) only read one field
• (readline logical-name) read one line as a
  string
• If you leave out logical-name (or put t), it
  means the terminal
• (defrule echo
• no patterns! It matches initial-fact.
• gt
• (fprintout t "Please enter your name
  below " crlf)
• (bind ?a (readline))
• (fprintout t "Aha, your name is " ?a))
• (reset)
• (run)

29
Pattern Matching by Generate Test

• Pattern matching is used extensively in CLIPS
• Especially on the left hand side (LHS) of rules
• usually makes use of
• Variables
• field constraints
• pattern connectives
• Generate-and-test often used for pattern matching

30
Generate Test Example

• Suppose we have a fact list with all the student
  records
• (student John 80)
• (student Joe Smith 70)
• (student Tom Bruce Smith 48)
• We want to print out names of passing students
• their marks are at least 50
• One possible solution is the following rule
• (defrule pass-student "passes"
• (student ?name ?marks) generate
• (test (gt ?marks 50)) test
• gt
• (printout t ?name " passed." crlf))

31
Example Remarks

• generate a structure, then test one or more of
  its arguments
• Note use of a multi field variable (wild card)
• It can be very inefficient.
• Beware that
• a rule will not fired twice for the same set of
  facts
• but it will fire again if you retract a fact and
  assert it again
• CLIPS regards the fact as a new one
• although it's literally the same as the retracted
  one
• Infinite Loops
• Generate and Test can easily generate infinite
  loops

32
Generate Test Infinite Loops

• Given an initial fact list as
• (student John 80)
• (student Joe Smith 70)
• (student Tom Bruce Smith 48)
• (counter 0)
• (sum 0)
• We want to calculate the average mark

33
Infinite Loops (cont)

• Will the this work?
• (defrule summation "get total first"
• (student ?name ?marks)
• ?sum lt- (sum ?total)
• ?counter lt- (counter ?StuNum)
• gt
• (retract ?sum) why ?sum
• (retract ?counter)
• (assert (sum ( ?total ?marks)))
• (assert(counter ( ?StuNum 1))))
• defrule average "get average mark"
• (sum ?total)
• (counter ?StuNum)
• gt
• (assert (average (/ ?total ?StuNum))))

34
Infinite Loops (cont)

• What's the result of running this rule?
• summation will loop infinitely on the first
  student
• Because the sum and counter facts change on
  each iteration
• So the rule can fire again with the same student
• Solutions
• Simplest
• delete each student after the rule fires
• crude, since we may need the students for other
  parts of the system
• More complicated, but cleaner
• Use an auxiliary fact

35
Infinite Loops Auxiliary Facts

• Replace the summation rule with the following
  two
• (defrule auxiliary-sum (student ?name ?marks)
• gt
• (assert (mark ?marks))
• (defrule summation "get the total marks first"
• ?m lt- (mark ?marks)
• ?sum lt- (sum ?total)
• ?counter lt- (counter ?StuNum)
• gt
• (retract ?m ?counter ?sum)
• (assert (sum ( ?total ?marks)))
• (assert (counter ( ?StuNum 1))))

36
Global Variables

• (bind ?x ( 3.14 3.14)) 9.8596
• (bind ?x (- ?x 1)) What would be the result?
• Global variables defined by defglobal
• (defglobal
• global-var-name expression
• global-var-name expression
• ... ... )
• global-var-name must be of the form ?symbol.
• (defglobal ?x 3.14 ?y 1) two global
  variables
• (bind ?x ( ?x ?x)) 9.8596
• (bind ?x (- ?x ?y)) 8.8596
• (bind ?x "abc")

37
Use of Global Variables

• Global variables may be used in almost any place
  that a local variable may occur
• There are two exceptions
• Global variables may not be used on the LHS of a
  rule to bind a value
• the following rule is illegal
• (defrule example (fact ?x) gt (assert (done)))
• Global variables may not be used as a parameter
  for a deffunction
• (like most other computing languages)

38
Non-Ordered Facts

• In ordered facts, fields are distinguished by
  position
• (what we have seen so far)
• In non-ordered facts, fields are distinguished by
  name
• The deftemplate function is used to define them
• (deftemplate deftemplate-name optional-comment
• single-field-definition ...
• multifield-definition
• single-field-definition ...)

39
Deftemplate example

• (deftemplate student-record
• (field name (type SYMBOL) (default unknown))
• (field year (type INTEGER) (default 2))
• (field major (type STRING) (default "Comp
  Sci"))
• (multifield hobby (type SYMBOL) (default
  football swim)))
• (assert (student-record))
• (assert (student-record (year 3) (hobby
  basketball)))

40
Manipulating Deftemplates

• ppdeftemplate
• Displays the text of a given deftemplate
• (ppdeftemplate ltdeftemplate-namegt)
• list-deftemplates
• Displays the list of all deftemplates
• (list-deftemplates)
• undeftemplate
• Deletes a deftemplate
• (undeftemplate ltdeftemplate-namegt)

41
CLIPS Functions

• Predicate Functions
• Mathematical Functions
• Multifield Functions
• String Functions
• Procedural Functions
• Generating Symbols

42
Predicate Functions

• A function which returns a FALSE or non-FALSE
  value
• The colon followed by a predicate function is
  called a predicate constraint
• Predefined Predicate Functions
• They all end with letter p

43
Built-in Predicate Functions

• (evenp arg) even number
• (floatp arg) floating-point number
• (integerp arg) integer
• (numberp arg) float or integer
• (oddp arg) odd number
• (stringp arg) string
• (symbolp arg) symbol

44
Predicate Function Examples

• (student ?name ?mark(numberp ?mark))
• (evenp 3.2) FALSE
• (oddp 3.2) TRUE
• (symbolp x) TRUE
• (symbolp 45) FALSE
• (stringp "hello") TRUE

45
Built-In Mathematical Functions

• round
• round to closest
• integer
• truncate to integer
• random
• return random integer
• div
• integer division
• max
• maximum value
• min
• minimum value
• abs
• absolute value
• float
• convert to a float

46
Mathematical Functions (cont)

• (pi)
• return 3.1415926...
• Note that the brackets are necessary
• (otherwise interpreted as a variable)
• sqrt
• square root
• power, x to power y
• exp
• exponential, e to power x
• log
• log e
• log10
• log10
• mod
• modulus

47
Mathematical Function Examples

• (mod 5 2)
• returns 1
• (max 1 3.4 (pi) (log 2.81) ( 2 10))
• (mod (random) 2)
• (float (integer (pi)))

48
Multifield Functions

• Multifield function names end with
• create
• append any number of fields to create multifield
  value
• nth
• return the value of the nth field
• member
• test whether a value is a member of a multifield
  value
• delete
• delete the nth member of the multifield value
• subsetp
• subset testing

49
Multifield Function Examples

• (nth 3 (create a b c d e))
• returns c
• (member d (create a b c d e))
• returns 4
• (subsetp (create a d) (create a b c d e))
• returns TRUE
• (delete 2 (create a b c d e))
• returns (a c d e)

50
String Functions

• str-cat
• string concatenation
• sym-cat
• symbol concatenation
• sub-string
• returns a sub-string in the range
• str-index
• returns the position of the first matching
  character
• eval
• evaluate the string
• build
• similar to eval but used with constructs
• upcase
• change all letters to upper case
• lowcase
• change all letters to lower case
• str-compare
• compare two strings according to dictionary order
  
• str-length

51
String Function Examples

• (str-cat "abc" OK 2 3.3 ( 3 4))
• "abcOK23.37"
• (sym-cat "abc" OK 2 3.3 ( 3 4))
• abcOK23.37
• (sub-string 2 4 "abcdefg")
• "bcd"
• (str-index "bcd" "abcdefg")
• 2
• (str-index "bcd" "abc")
• FALSE
• (eval "(create x y z)")
• (x y z)
• (build "(defrule r1 (x y z) gt (assert (done)))")
  
• TRUE

52
String Function Examples

• (upcase "abcOK23.37")
• "ABCOK23.37"
• (lowcase "abcOK23.37")
• "abcok23.37"
• (str-compare "abc" "abd")
• -1
• (str-length "abcOK23.37")
• 10

53
Procedural Functions

• if-then-else
• while
• progn
• return
• break

54
The if-then-else Function

• very similar to the if-then-else statement in a
  procedural language
• Like Java, C, .
• (if expression
• then action1 action2 ...
• else action-1 action-2 ...)
• where an action may be any function
• The else part is optional
• This function returns the value of the last
  expression or action evaluated

55
if-then-else Example

• (defrule print-marks
• ?record lt- (student ?name ?marks)
• gt
• (if (lt ?marks 50)
• then
• (printout t ?name " fails" crlf)
• else
• (printout t ?name " passes" crlf)))
• Note that the above rule can be replaced by two
  rules without using if-then-else.

56
The while Function

• provides us with a loop structure
• (while expression do do is optional
• action1 action2 ...)
• For example
• (deffunction print-sqrt (?n)
• (while (gt ?n 0)
• (printout t "Sqrt of " ?n " is " (sqrt ?n) crlf)
• (bind ?n (- ?n 1))))
• (print-sqrt 3)
• we should get 4 results

57
The progn Function

• evaluates all of its arguments and returns the
  value of the last argument
• (progn ( 2 3) ( 1 (pi)))
• returns 4.1415926535...

58
The return Function

• immediately terminates the currently executing..
• Deffunction
• Defrule
• RHS
• Etc
• can have an optinal argument, the return value
• (deffunction sign (?x)
• (if (gt ?x 0) then (return 0) else (return 1)))

59
The break Function

• can be used to terminate
• while loop
• progn execution
• etc
• (deffunction print-sqrt-2 (?n)
• (while TRUE do
• (printout t "Sqrt of " ?n " is " (sqrt ?n) crlf)
• (if (lt ?n 0) then (break))
• (bind ?n (- ?n 1))))

60
Symbol-Generating Functions

• gensym
• used to generate a symbol of the form genn
• where n is a positive integer
• (create (gensym) (gensym) (gensym))
• (gen1 gen2 gen3)
• gensym
• generates a unique symbol not currently in the
  CLIPS environment
• setgen
• sets the starting integer after gen

61
Symbol-Generating Example

• (setgen 3)
• returns 3 as its function value
• (create (gensym) (gensym) (gensym))
• returns (gen3 gen4 gen5)
• (gensym)
• returns gen6