Constraint Handling Rules

A Tutorial for (Prolog) Programmers

Tom Schrijvers

Katholieke Universiteit Leuven, Belgium

[email protected]

ICLP 2008 – December 9-13, 2008

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 1 / 112

Overview 2/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 2 / 112

Overview 3/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 3 / 112

About This Tutorial

Expected background:
I know & like programming

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 4 / 112

About This Tutorial

Expected background:
I know & like programming

I know Prolog, . . .

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 4 / 112

About This Tutorial

Expected background:
I know & like programming

I know Prolog, . . .

I but have an itch that Prolog doesn’t scratch

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 4 / 112

About This Tutorial

Expected background:
I know & like programming

I know Prolog, . . .

I but have an itch that Prolog doesn’t scratch

I . . . or you will have at the end of this tutorial

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 4 / 112

About This Tutorial

Expected mindset:
I forget about Prolog for now

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 5 / 112

About This Tutorial

Expected mindset:
I forget about Prolog for now

I do not assume things to work as in Prolog

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 5 / 112

About This Tutorial

Expected mindset:
I forget about Prolog for now

I do not assume things to work as in Prolog

I really, they won’t be the same!!!

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 5 / 112

About This Tutorial

Expected mindset:
I forget about Prolog for now

I do not assume things to work as in Prolog

I really, they won’t be the same!!!

You’ve been warned!

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 5 / 112

About This Tutorial

Expected mindset:
I forget about Prolog for now

I do not assume things to work as in Prolog

I really, they won’t be the same!!!

You’ve been warned! Also:

I do not expect anything from the word “constraint”

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 5 / 112

About This Tutorial

Expected mindset:
I forget about Prolog for now

I do not assume things to work as in Prolog

I really, they won’t be the same!!!

You’ve been warned! Also:

I do not expect anything from the word “constraint”

I CHR constraints aren’t

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 5 / 112

About this tutorial

What you’ll learn:

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 6 / 112

About this tutorial

What you’ll learn:

1 how CHR works (operationally)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 6 / 112

About this tutorial

What you’ll learn:

1 how CHR works (operationally)
2 how to program in CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 6 / 112

About this tutorial

What you’ll learn:

1 how CHR works (operationally)
2 how to program in CHR
3 just maybe, what CHR is good for

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 6 / 112

Overview 7/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 7 / 112

Art 101: Mixing Paint

As novice, you have to start at the bottom of the ladder:

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 8 / 112

Art 101: Mixing Paint

As novice, you have to start at the bottom of the ladder:

I Learning how to obtain different colors

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 8 / 112

Art 101: Mixing Paint

As novice, you have to start at the bottom of the ladder:

I Learning how to obtain different colors

I by mixing paint

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 8 / 112

Art 101: Mixing Paint

As novice, you have to start at the bottom of the ladder:

I Learning how to obtain different colors

I by mixing paint

I Made easy by the Paint Mixing Rules (PMR) language

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 8 / 112

The Paint

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 9 / 112

Fewer Paint

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 10 / 112

The Rules

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 11 / 112

The Rules

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 11 / 112

The Rules

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 11 / 112

PMR as CHR

Under the hood, PMR is of course CHR:

PMR CHR

paints constraints

rules simplification rules (textual)

mixing bucket constraint store

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 12 / 112

Paint as Constraints

Declaring our paint colors, in textual form:

:- chr_constraint red. %

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 13 / 112

Paint as Constraints

Declaring our paint colors, in textual form:

:- chr_constraint red. %
:- chr_constraint blue. %

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 13 / 112

Paint as Constraints

Declaring our paint colors, in textual form:

:- chr_constraint red. %
:- chr_constraint blue. %
:- chr_constraint yellow. %
:- chr_constraint orange. %
...

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 13 / 112

Paint as Constraints?

Are paint colors really “constraints”?

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 14 / 112

Paint as Constraints?

Are paint colors really “constraints”?

No: not in the Constraint Programming sense

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 14 / 112

Paint as Constraints?

Are paint colors really “constraints”?

No: not in the Constraint Programming sense
Yes: in the CHR sense
I 1st-class entities in CHR

I rewritten by rules

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 14 / 112

Paint as Constraints?

Are paint colors really “constraints”?

No: not in the Constraint Programming sense
Yes: in the CHR sense
I 1st-class entities in CHR

I rewritten by rules

(Are Prolog predicates really “predicates”?)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 14 / 112

The Paint Rules in CHR

Simplification Rule: head <=> body .

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 15 / 112

The Paint Rules in CHR

Simplification Rule: head <=> body .

“head may be replaced with body ”

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 15 / 112

The Paint Rules in CHR

Simplification Rule: head <=> body .

“head may be replaced with body ”

%
red, blue <=> purple.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 15 / 112

The Paint Rules in CHR

Simplification Rule: head <=> body .

“head may be replaced with body ”

%
red, blue <=> purple.
%
red, yellow <=> orange.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 15 / 112

The Paint Rules in CHR

Simplification Rule: head <=> body .

“head may be replaced with body ”

%
red, blue <=> purple.
%
red, yellow <=> orange.
%
blue, yellow <=> green.
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 15 / 112
Time For. . .

. . . a cocktail break!
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 16 / 112
Cocktail Break

You can have as many heads as you like:

vodka, martini <=> vodka_martini.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 17 / 112

Cocktail Break

You can have as many heads as you like:

vodka, martini <=> vodka_martini.

tequila, cointreau, lime <=> margarita.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 17 / 112

Cocktail Break

You can have as many heads as you like:

vodka, martini <=> vodka_martini.

tequila, cointreau, lime <=> margarita.

gin, cherry_brandy, cointreau, grenadine,

pineapple, lemon, angostura_bitters
<=> singapore_sling.
but at least one!

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 17 / 112

Cocktail Break

Warning: too many heads may cause a headache!

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 18 / 112

Cocktail Break

Warning: too many heads may cause a headache!

vodka_martini, margarita, singapore_sling
<=> hangover, blackout.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 18 / 112

Cocktail Break

Warning: too many heads may cause a headache!

vodka_martini, margarita, singapore_sling
<=> hangover, blackout.
You can have many constraints in the body too.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 18 / 112

Running the Rules

Paint-Mixing Engine:
I Prolog (SWI-Prolog, . . . )

I Leuven CHR system

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 19 / 112

Running the Rules

Paint-Mixing Engine:
I Prolog (SWI-Prolog, . . . )

I Leuven CHR system

Everything in a file:
paint.pl
:- use_module(library(chr)).

:- chr_constraint red.
...

red, blue <=> purple.

...

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 19 / 112

Running the Rules

Paint-Mixing Engine:
I Prolog (SWI-Prolog, . . . )

I Leuven CHR system

Everything in a file:
paint.pl
:- use_module(library(chr)).

:- chr_constraint red.
...
?- [paint].
red, blue <=> purple.
...

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 19 / 112

Running the Rules

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red, blue.

I queries
I answers

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 20 / 112

Running the Rules

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red, blue.
purple

I queries ?- red, yellow.

I answers

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 20 / 112

Running the Rules

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red, blue.
purple

I queries ?- red, yellow.

orange
I answers
?- blue, yellow.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 20 / 112

Running the Rules

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red, blue.
purple

I queries ?- red, yellow.

orange
I answers
?- blue, yellow.
green

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 20 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , .
I query:
left-to-right

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 21 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , . I query:
left-to-right
I active
constraint:
blue

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 22 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

I query:
?- , . I
left-to-right
active
constraint:
blue
I put in
constraint
store

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 23 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

I active
?- , .
constraint:
blue
I fire first rule?
I no partner
red in
constraint
store

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 24 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

I active
?- , . constraint:
blue
I fire second
rule?
I no blue in
head

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 25 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

I active
constraint:
?- , . blue
I fire third
rule?
I no partner
yellow in
constraint
store

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 26 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , . I deactivate
constraint
blue

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 27 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , . I active
constraint:
yellow

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 28 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , . I active
constraint:
yellow
I put in
constraint
store

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 29 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , .
I active
constraint:
yellow
I fire first rule?
I no yellow in
head

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 30 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

I active
constraint:
?- , . yellow
I fire second
rule?
I no partner
red in
constraint
store

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 31 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , . I active
constraint:
yellow
I fire third
rule?

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 32 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

I active
?- , .
constraint:
yellow
I fire third rule!
I with partner
blue in
constraint
store

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 33 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , . I fire third rule!

I delete the
matched
head
constraints

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 34 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , . I fire third rule!

I add the body
to (front of)
query

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 35 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , . I active
constraint:
green

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 36 / 112

In Slow Motion

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- , .
to make a long
story short:
I ...
I final
constraint
store

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 37 / 112

What Happens?

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red, yellow.
orange

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 38 / 112

What Happens?

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red, yellow.
orange

?- yellow, red.
orange

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 38 / 112

What Happens?

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 39 / 112

What Happens?

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red.
red

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 39 / 112

What Happens?

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red, yellow, blue.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 40 / 112

What Happens?

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red, yellow, blue.

orange
blue

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 40 / 112

What Happens?

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red, blue, yellow.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 41 / 112

What Happens?

red, blue <=> purple.

red, yellow <=> orange.
blue, yellow <=> green.

?- red, blue, yellow.

purple
yellow

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 41 / 112

Summary

Simplification Rule:
I head <=> body .

I replace head with body

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 42 / 112

Summary

Simplification Rule:
I head <=> body .

I replace head with body

Queries:
I processed left-to-right

I incrementally

I active constraint looks for partners

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 42 / 112

Overview 43/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 43 / 112

Brown

Brown stays brown.

brown, orange <=> brown.
brown, green <=> brown.
brown, purple <=> brown.
...

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 44 / 112

Short Hand

Simpagation rule: headk \head r <=> body .

brown \ orange <=> true.
brown \ green <=> true.
brown \ purple <=> true.
...

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 45 / 112

Short Hand

Simpagation rule: headk \head r <=> body .

brown \ orange <=> true.
brown \ green <=> true.
brown \ purple <=> true.
...

I headk : kept head (1 or more)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 45 / 112

Short Hand

Simpagation rule: headk \head r <=> body .

brown \ orange <=> true.
brown \ green <=> true.
brown \ purple <=> true.
...

I headk : kept head (1 or more)

I headr : removed head (1 or more)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 45 / 112

Short Hand

Simpagation rule: headk \head r <=> body .

brown \ orange <=> true.
brown \ green <=> true.
brown \ purple <=> true.
...

I headk : kept head (1 or more)

I headr : removed head (1 or more)
I true : Prolog’s no-op

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 45 / 112

Lead to Gold

philosophers_stone \ lead
<=> gold.

?- philosophers_stone, lead.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 46 / 112

Lead to Gold

philosophers_stone \ lead
<=> gold.

?- philosophers_stone, lead.
philosophers_stone
gold

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 46 / 112

Lead to Gold

philosophers_stone \ lead
<=> gold.

?- philosophers_stone, lead.
philosophers_stone
gold

?- philosophers_stone, lead, lead.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 46 / 112

Lead to Gold

philosophers_stone \ lead
<=> gold.

?- philosophers_stone, lead.
philosophers_stone
gold

?- philosophers_stone, lead, lead.

philosophers_stone
gold
gold

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 46 / 112

All Combinations

philosophers_stone \ lead
<=> gold.

?- lead, lead.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 47 / 112

All Combinations

philosophers_stone \ lead
<=> gold.

?- lead, lead.
lead
lead

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 47 / 112

All Combinations

philosophers_stone \ lead
<=> gold.

?- lead, lead.
lead
lead

?- lead, lead, philosophers_stone.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 47 / 112

All Combinations

philosophers_stone \ lead
<=> gold.

?- lead, lead.
lead
lead

?- lead, lead, philosophers_stone.

philosophers_stone
gold
gold

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 47 / 112

All Combinations

philosophers_stone \ lead
<=> gold.

?- lead, lead.
lead
lead

?- lead, lead, philosophers_stone.

philosophers_stone
gold
gold
Rule fires with all combinations!
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 47 / 112
Summary

Simplification Rule:
I head <=> body .

I replace head with body

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 48 / 112

Summary

Simplification Rule:
I head <=> body .

I replace head with body

Simpagation Rule:
I headk \headr <=> body .

I replace headr with body

I in the presence of headk

I fires all possible combinations

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 48 / 112

Overview 49/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 49 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

I constraints can have arguments

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 50 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

I constraints can have arguments

I arity must be declared

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 50 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

I constraints can have arguments

I arity must be declared
I Prolog terms as arguments

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 50 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

?- piggy(5),piggy(1),piggy(4),piggy(2).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 51 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

?- 5€
,
1€
,
4€
,
2€
.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 51 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

?- ,
1€
,
4€
,
2€
.

5€

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 51 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

?- , ,
4€
,
2€
.

5€ 1€

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 51 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

?- , ,
4€
,
2€
.

6€

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 51 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

?- , , ,
2€
.

4€ 6€

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 51 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

?- , , ,
2€
.

10€

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 51 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

?- , , ,
6€
.

10€ 2€

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 51 / 112

Piggy Bank Merger

:- chr_constraint piggy/1.

piggy(I), piggy(J) <=> K is I + J, piggy(K).

?- , , ,
6€
.

12€

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 51 / 112

CHR(X )

CHR is an embedded language

I embedded in host language X

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 52 / 112

CHR(X )

CHR is an embedded language

I embedded in host language X

I here X = Prolog

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 52 / 112

CHR(X )

CHR is an embedded language

I embedded in host language X

I here X = Prolog

2-way communication:

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 52 / 112

CHR(X )

CHR is an embedded language

I embedded in host language X

I here X = Prolog

2-way communication:
I Prolog calls CHR constraints

e.g. all along from toplevel

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 52 / 112

CHR(X )

CHR is an embedded language

I embedded in host language X

I here X = Prolog

2-way communication:
I Prolog calls CHR constraints

e.g. all along from toplevel

I CHR calls Prolog

e.g. K is I + J

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 52 / 112

Not Just Numbers

:- chr_constraint value/1.

value(I), value(J) <=> append(I,J,K), value(K).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 53 / 112

Not Just Numbers

:- chr_constraint value/1.

value(I), value(J) <=> append(I,J,K), value(K).

?- value([a]), value([b]).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 53 / 112

Not Just Numbers

:- chr_constraint value/1.

value(I), value(J) <=> append(I,J,K), value(K).

?- value([a]), value([b]).
value([b,a]).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 53 / 112

Not Just Numbers

:- chr_constraint value/1.

value(I), value(J) <=> append(I,J,K), value(K).

?- value([a]), value([b]).
value([b,a]).

?- value([f(a),g(b)]), value([h(i,j)]).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 53 / 112

Not Just Numbers

:- chr_constraint value/1.

value(I), value(J) <=> append(I,J,K), value(K).

?- value([a]), value([b]).
value([b,a]).

?- value([f(a),g(b)]), value([h(i,j)]).
value([h(i,j),f(a),g(b)]).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 53 / 112

Not Just Numbers

:- chr_constraint value/1.

value(I), value(J) <=> append(I,J,K), value(K).

?- value([a]), value([b]).
value([b,a]).

?- value([f(a),g(b)]), value([h(i,j)]).
value([h(i,j),f(a),g(b)]).

Any Prolog terms!

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 53 / 112

Summary

CHR constraints
I zero or more arguments

I Prolog terms

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 54 / 112

Summary

CHR constraints
I zero or more arguments

I Prolog terms

Bodies
I CHR constraints and

I Prolog predicate calls

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 54 / 112

Overview 55/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 55 / 112

Count Down

Problem: write a program to enumerate values:

?- generate(5).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 56 / 112

Count Down

Problem: write a program to enumerate values:

?- generate(5).
value(5)
value(4)
value(3)
value(2)
value(1)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 56 / 112

Count Down

Solution:
:- chr_constraint generate/1.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 57 / 112

Count Down

Solution:
:- chr_constraint generate/1.
:- chr_constraint value/1.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 57 / 112

Count Down

Solution:
:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(N) <=>
value(N), M is N - 1, generate(M).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 57 / 112

Running the Program

:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(N) <=> value(N), M is N - 1, generate(M).

?- generate(5).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 58 / 112

Running the Program

:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(N) <=> value(N), M is N - 1, generate(M).

?- generate(5).
% waiting

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 58 / 112

Running the Program

:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(N) <=> value(N), M is N - 1, generate(M).

?- generate(5).
% waiting
% still waiting

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 58 / 112

Running the Program

:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(N) <=> value(N), M is N - 1, generate(M).

?- generate(5).
% waiting
% still waiting
ERROR: Out of local stack

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 58 / 112

Running the Program

:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(N) <=> value(N), M is N - 1, generate(M).

?- generate(5).
% waiting
% still waiting
ERROR: Out of local stack
Recursion without a base case!

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 58 / 112

Guarded Rules

:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(N) <=> N == 0 | true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 59 / 112

Guarded Rules

:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(N) <=> N == 0 | true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

Guarded rule:
I apply rule if guard succeeds

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 59 / 112

Guarded Rules

:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(N) <=> N == 0 | true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

Guarded rule:
I apply rule if guard succeeds

I guard is optional

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 59 / 112

Guarded Rules

:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(N) <=> N == 0 | true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

Guarded rule:
I apply rule if guard succeeds

I guard is optional

I guard may contain any Prolog, but may not bind any

variables from the head (pure check)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 59 / 112
Even Shorter

generate(N) <=> N > 0 | value(N),

M is N - 1, generate(M).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 60 / 112

Even Shorter

generate(N) <=> N > 0 | value(N),

M is N - 1, generate(M).

?- generate(5).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 60 / 112

Even Shorter

generate(N) <=> N > 0 | value(N),

M is N - 1, generate(M).

?- generate(5).
value(5)
value(4)
value(3)
value(2)
value(1)
generate(0)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 60 / 112

Summary

Simplification Rule:
I head <=> guard | body .

I replace head with body

I if guard succeeds

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 61 / 112

Summary

Simplification Rule:
I head <=> guard | body .

I replace head with body

I if guard succeeds

Simpagation Rule:
I headk \headr <=> guard | body .

I replace headr with body

I in the presence of headk

I if guard succeeds

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 61 / 112

Overview 62/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 62 / 112

CHR Style: Matching

:- chr_constraint generate/1.
:- chr_constraint value/1.

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

Matching:
I “inline” notation for guard

I same meaning as explicit guard

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 63 / 112

Meaning of Matching

Maching Guard

c(X) <=> true. c(X) <=> true | true.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 64 / 112

Meaning of Matching

Maching Guard

c(X) <=> true. c(X) <=> true | true.

c(a) <=> true. c(X) <=> X == a | true.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 64 / 112

Meaning of Matching

Maching Guard

c(X) <=> true. c(X) <=> true | true.

c(a) <=> true. c(X) <=> X == a | true.

c(f(A)) <=> true. c(X) <=> nonvar(X), X = f(A) | true.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 64 / 112

Meaning of Matching

Maching Guard

c(X) <=> true. c(X) <=> true | true.

c(a) <=> true. c(X) <=> X == a | true.

c(f(A)) <=> true. c(X) <=> nonvar(X), X = f(A) | true.

c(X,X) <=> true. c(X,Y) <=> X == Y | true.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 64 / 112

Meaning of Matching

Maching Guard

c(X) <=> true. c(X) <=> true | true.

c(a) <=> true. c(X) <=> X == a | true.

c(f(A)) <=> true. c(X) <=> nonvar(X), X = f(A) | true.

c(X,X) <=> true. c(X,Y) <=> X == Y | true.

c(X), d(X) <=> true. c(X), d(Y) <=> X == Y | true.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 64 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world).
Hello, World!

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world).
Hello, World!

?- p(Free).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world).
Hello, World!

?- p(Free).
Hello, World!
Free = world

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world).
Hello, World!

?- p(Free).
Hello, World!
Free = world

?- p(hello).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world).
Hello, World!

?- p(Free).
Hello, World!
Free = world

?- p(hello).
No

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world). ?- c(world).
Hello, World!

?- p(Free).
Hello, World!
Free = world

?- p(hello).
No

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world). ?- c(world).
Hello, World! Hello, World!

?- p(Free).
Hello, World!
Free = world

?- p(hello).
No

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world). ?- c(world).
Hello, World! Hello, World!

?- p(Free). ?- c(Free).
Hello, World!
Free = world

?- p(hello).
No

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world). ?- c(world).
Hello, World! Hello, World!

?- p(Free). ?- c(Free).
Hello, World! c(Free)
Free = world

?- p(hello).
No

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world). ?- c(world).
Hello, World! Hello, World!

?- p(Free). ?- c(Free).
Hello, World! c(Free)
Free = world

?- p(hello). ?- c(hello).
No

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Matching: CHR vs. Prolog

p(world) :- writeln(’Hello, World!’).

:- chr_constraint c/1.
c(world) <=> writeln(’Hello, World!’).
?- p(world). ?- c(world).
Hello, World! Hello, World!

?- p(Free). ?- c(Free).
Hello, World! c(Free)
Free = world

?- p(hello). ?- c(hello).
No c(hello)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 65 / 112

Summary

Matching
I short-hand for equality-based guards

I one-way unification
I only instantiates head

I does not instantiate constraints

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 66 / 112

Overview 67/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 67 / 112

Combining Rules

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

What’s the outcome?

?- generate(4).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 68 / 112

Combining Rules

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

What’s the outcome?

?- generate(4).
value(10)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 68 / 112

Combining Rules

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

What’s the outcome?

?- generate(4).
value(10)

?- generate(5).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 68 / 112

Combining Rules

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

What’s the outcome?

?- generate(4).
value(10)

?- generate(5).
value(15)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 68 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- 2 5

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- 5

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- 2 M is 2 - 1 M
5

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- M is 2 - 1 M 5

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- 1 5

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- 5

2 1

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- 1 M is 1 - 1 M
5

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- M is 1 - 1 M
5

2 1

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- M is 1 - 1 M
5

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- 0 5

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- 5

0 3

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

In Slow Motion

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

value(I), value(J) <=> K is I + J, value(K).

?- 5

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 69 / 112

Combining Rules

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

What’s the outcome?
?- generate(4).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 70 / 112

Combining Rules

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

What’s the outcome?
?- generate(4).
value(2)
value(3)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 70 / 112

Combining Rules

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

What’s the outcome?
?- generate(4).
value(2)
value(3)

?- generate(10).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 70 / 112

Combining Rules

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

What’s the outcome?
?- generate(4).
value(2)
value(3)

?- generate(10).
value(2)
value(3)
value(5)
value(7)
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 70 / 112
How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

1
8 9
0
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112
How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

How Does It Work?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 71 / 112

Alternatives?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4 Just added value(3).

What now?
5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 72 / 112

Alternatives?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4 Just added value(3).

What now?
5 6 7 I remove value(6) ?

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 72 / 112

Alternatives?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4 Just added value(3).

What now?
5 6 7 I remove value(6) ?

I remove value(9) ?

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 72 / 112

Alternatives?

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4 Just added value(3).

What now?
5 6 7 I remove value(6) ?

I remove value(9) ?

I unspecified in CHR!
8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 72 / 112

Alternate Ending

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 73 / 112

Alternate Ending

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 73 / 112

Alternate Ending

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 73 / 112

Alternate Ending

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 73 / 112

Alternate Ending

generate(1) <=> true.

generate(N) <=> N > 1 | value(N),
M is N - 1, generate(M).

value(I) \ value(J) <=> J mod I =:= 0 | true.

2 3 4

5 6 7

8 9 10

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 73 / 112

Summary

CHR Programs
I Programs consist of a sequence of rules

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 74 / 112

Summary

CHR Programs
I Programs consist of a sequence of rules
I Active constraint traverses the rules top-to-bottom to
find any that fire

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 74 / 112

Summary

CHR Programs
I Programs consist of a sequence of rules
I Active constraint traverses the rules top-to-bottom to
find any that fire
I Rest of query/body waits

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 74 / 112

Summary

CHR Programs
I Programs consist of a sequence of rules
I Active constraint traverses the rules top-to-bottom to
find any that fire
I Rest of query/body waits
I Unspecified which partner is found

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 74 / 112

Overview 75/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 75 / 112

Simpagation with Hk = ∅

Simpagation rule
headk \headr <=> guard | body .

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 76 / 112

Simpagation with Hk = ∅

Simpagation rule
headk \headr <=> guard | body .
What if head k = ∅?

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 76 / 112

Simpagation with Hk = ∅

Simpagation rule
headk \headr <=> guard | body .
What if head k = ∅?
Simplification rule
head r <=> guard | body .

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 76 / 112

Simpagation with Hr = ∅

Simpagation rule
headk \headr <=> guard | body .

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 77 / 112

Simpagation with Hr = ∅

Simpagation rule
headk \headr <=> guard | body .
What if head r = ∅?

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 77 / 112

Simpagation with Hr = ∅

Simpagation rule
headk \headr <=> guard | body .
What if head r = ∅?
Propagation rule
head k ==> guard | body .
I add body

I in the presence of head k

I if guard holds

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 77 / 112

Propagation

a, b ==> c.

?- a, b.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 78 / 112

Propagation

a, b ==> c.

?- a, b.
a
b
c

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 78 / 112

Propagation

a, b ==> c.

?- a, b.
a
b
c
?- a, b, b.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 78 / 112

Propagation

a, b ==> c.

?- a, b.
a
b
c
?- a, b, b.
a
b
b
c
c

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 78 / 112

Cellular Automaton

Writing Wolfram’s celluar automaton Rule 110 in CHR.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 79 / 112

Cellular Automaton
Writing Wolfram’s celluar automaton Rule 110 in CHR.

a
1 2 5
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 80 / 112
Cellular Automaton
Writing Wolfram’s celluar automaton Rule 110 in CHR.

a
1 2 5
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 80 / 112
Cellular Automaton
Writing Wolfram’s celluar automaton Rule 110 in CHR.

a
1 2 5
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 80 / 112
Cellular Automaton
Writing Wolfram’s celluar automaton Rule 110 in CHR.

a
1 2 5
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 80 / 112
Cellular Automaton
Writing Wolfram’s celluar automaton Rule 110 in CHR.

a
1 2 3 5
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 80 / 112
Cellular Automaton
Writing Wolfram’s celluar automaton Rule 110 in CHR.

a
1 2 3 4 5
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 80 / 112
Cellular Automaton
Writing Wolfram’s celluar automaton Rule 110 in CHR.

a
1 2 3 4 5
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 80 / 112
Cellular Automaton
Writing Wolfram’s celluar automaton Rule 110 in CHR.

a
1 2 3 4 5
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 80 / 112
Cellular Automaton
CHR cells
% black(Name,Below,Above,Generation).
:- chr_constraint black/4.

% white(Name,Below,Above,Generation).
:- chr_constraint white/4.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 81 / 112

Cellular Automaton
CHR cells
% black(Name,Below,Above,Generation).
:- chr_constraint black/4.

% white(Name,Below,Above,Generation).
:- chr_constraint white/4.
d
?- white(d,c,a,1), c
black(c,b,d,1),
white(b,a,c,1), b

white(a,d,b,1). a
1
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 81 / 112
Automaton Rule

white(C,_,_.G) ,
black(B,A,C,G) ,
white(A,_,_,G)
==> G < 10 | NG is G + 1, black(B,A,C,NG).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 82 / 112

Automaton Rule

white(C,_,_.G) ,
black(B,A,C,G) ,
white(A,_,_,G)
==> G < 10 | NG is G + 1, black(B,A,C,NG).
?- white(d,c,a,1),
black(c,b,d,1),
white(b,a,c,1),
white(a,d,b,1).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 82 / 112

Automaton Rule

white(C,_,_.G) ,
black(B,A,C,G) ,
white(A,_,_,G)
==> G < 10 | NG is G + 1, black(B,A,C,NG).
?- white(d,c,a,1),
black(c,b,d,1),
white(b,a,c,1),
white(a,d,b,1).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 82 / 112

Another Example of Propagation

generate(N) ==> value(2).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 83 / 112

Another Example of Propagation

generate(N) ==> value(2).

generate(N), value(I) ==>

I < N | J is I + 1, value(J).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 83 / 112

Another Example of Propagation

generate(N) ==> value(2).

generate(N), value(I) ==>

I < N | J is I + 1, value(J).

value(I) \ value(J) <=> J mod I =:= 0 | true.

?- generate(10).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 83 / 112

Another Example of Propagation

generate(N) ==> value(2).

generate(N), value(I) ==>

I < N | J is I + 1, value(J).

value(I) \ value(J) <=> J mod I =:= 0 | true.

?- generate(10).
value(2)
value(3)
value(5)
value(7)
generate(10)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 83 / 112

Summary

Simplification Rule: head <=> guard | body .

I replace head with body

Simpagation Rule: headk \headr <=> guard | body .

I replace headr with body
I in the presence of headk

Propagation Rule: head ==> guard | body .

I add body

I in the presence of head

I fires once for each combination (propagation history)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 84 / 112

Overview 85/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 85 / 112

Biased Coin?

coin <=> heads.

coin <=> tails.
What’s the outcome?
?- coin.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 86 / 112

Biased Coin?

coin <=> heads.

coin <=> tails.
What’s the outcome?
?- coin.
heads

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 86 / 112

Biased Coin?

coin <=> heads.

coin <=> tails.
What’s the outcome?
?- coin.
heads

Rules are tried in order!

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 86 / 112

Coin Flipping in Prolog

CHR Prolog

coin <=> heads. p_coin :- heads.

coin <=> tails. p_coin :- tails.

What’s the outcome?

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 87 / 112

Coin Flipping in Prolog

CHR Prolog

coin <=> heads. p_coin :- heads.

coin <=> tails. p_coin :- tails.

What’s the outcome?

?- coin.
heads

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 87 / 112

Coin Flipping in Prolog

CHR Prolog

coin <=> heads. p_coin :- heads.

coin <=> tails. p_coin :- tails.

What’s the outcome?

?- coin. ?- p_coin.
heads heads

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 87 / 112

Coin Flipping in Prolog

CHR Prolog

coin <=> heads. p_coin :- heads.

coin <=> tails. p_coin :- tails.

What’s the outcome?

?- coin. ?- p_coin.
heads heads ; y

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 87 / 112

Coin Flipping in Prolog

CHR Prolog

coin <=> heads. p_coin :- heads.

coin <=> tails. p_coin :- tails.

What’s the outcome?

?- coin. ?- p_coin.
heads heads ; y

tails

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 87 / 112

Coin Flipping in Prolog

CHR Prolog

coin <=> heads. p_coin :- heads.

coin <=> tails. p_coin :- tails.

What’s the outcome?

?- coin. ?- p_coin.
heads heads ; y

tails
I Prolog: backtracking
I CHR: committed choice for simplification rules

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 87 / 112

Propagation Coin Flipping

coin ==> heads.

coin ==> tails.
What’s the outcome?
?- coin.
heads
tails
coin

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 88 / 112

Propagation Coin Flipping

coin ==> heads.

coin ==> tails.
What’s the outcome?
?- coin.
heads
tails
coin

Rules are applied in sequence.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 88 / 112

Propagation Coin Flipping

coin ==> heads.

coin <=> tails.
coin ==> side.
What’s the outcome?
?- coin.
heads
tails

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 89 / 112

Propagation Coin Flipping

coin ==> heads.

coin <=> tails.
coin ==> side.
What’s the outcome?
?- coin.
heads
tails

Rules are applied in sequence, until the active constraint is removed.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 89 / 112

Exploiting Rule Order

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 90 / 112

Exploiting Rule Order

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

vs.
generate(0) <=> true.
generate(N) <=> value(N),
M is N - 1, generate(M).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 90 / 112

Exploiting Rule Order

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

vs.
generate(0) <=> true.
generate(N) <=> value(N),
M is N - 1, generate(M).

for queries ?- generate(N). with N ≥ 0

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 90 / 112

Exploiting Rule Order

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

vs.
generate(0) <=> true.
generate(N) <=> value(N),
M is N - 1, generate(M).

for queries ?- generate(N). with N ≥ 0

Better Style?

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 90 / 112

Exploiting Rule Order

generate(0) <=> true.

generate(N) <=> N > 0 | value(N),
M is N - 1, generate(M).

vs.
generate(0) <=> true.
generate(N) <=> value(N),
M is N - 1, generate(M).

for queries ?- generate(N). with N ≥ 0

Better Style?
Compare to:

generate(0) :- !.
generate(N) :- value(N), M is N - 1, generate(M).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 90 / 112

Summary

Rule Order
I rules are tried from top to bottom

I applied in sequence

I until the active constraint is removed

I committed choice: no alternatives explored for

simplification rules

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 91 / 112

Overview 92/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 92 / 112

Hello, World!

c(hello) <=> writeln(hello).

c(world) <=> writeln(world).

?- c(hello), c(world).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 93 / 112

Hello, World!

c(hello) <=> writeln(hello).

c(world) <=> writeln(world).

?- c(hello), c(world).
hello
world

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 93 / 112

Hello, World!

c(hello) <=> writeln(hello).

c(world) <=> writeln(world).

?- c(hello), c(world).
hello
world
?- c(X).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 93 / 112

Hello, World!

c(hello) <=> writeln(hello).

c(world) <=> writeln(world).

?- c(hello), c(world).
hello
world
?- c(X).
c(X)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 93 / 112

Hello, World!

c(hello) <=> writeln(hello).

c(world) <=> writeln(world).

?- c(hello), c(world).
hello
world
?- c(X).
c(X)
?- c(X), X = hello.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 93 / 112

Hello, World!

c(hello) <=> writeln(hello).

c(world) <=> writeln(world).

?- c(hello), c(world).
hello
world
?- c(X).
c(X)
?- c(X), X = hello.
hello

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 93 / 112

Hello, World!

c(hello) <=> writeln(hello).

c(world) <=> writeln(world).

?- c(hello), c(world).
hello
world
?- c(X).
c(X)
?- c(X), X = hello.
hello
?- c(X), c(world), X = hello.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 93 / 112

Hello, World!

c(hello) <=> writeln(hello).

c(world) <=> writeln(world).

?- c(hello), c(world).
hello Reactivation
world I constraints
?- c(X). suspend in
c(X) the constraint
?- c(X), X = hello. store
hello I unification
?- c(X), c(world), X = hello.
reactivates
world
suspended
hello
constraints
X = hello

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 93 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(a,a).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(a,a).
No

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(a,a).
No

?- neq(a,b).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(a,a).
No

?- neq(a,b).
Yes

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(a,a).
No

?- neq(a,b).
Yes

?- neq(A,B).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(A,B), A = B.
?- neq(a,a).
No

?- neq(a,b).
Yes

?- neq(A,B).
neq(A,B)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(A,B), A = B.
No
?- neq(a,a).
No

?- neq(a,b).
Yes

?- neq(A,B).
neq(A,B)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(A,B), A = B.
No
?- neq(a,a).
No
?- neq(A,B), A = a, B = a.
?- neq(a,b).
Yes

?- neq(A,B).
neq(A,B)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(A,B), A = B.
No
?- neq(a,a).
No
?- neq(A,B), A = a, B = a.
No
?- neq(a,b).
Yes

?- neq(A,B).
neq(A,B)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(A,B), A = B.
No
?- neq(a,a).
No
?- neq(A,B), A = a, B = a.
No
?- neq(a,b).
Yes
?- neq(A,B), A = a, B = b.
?- neq(A,B).
neq(A,B)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(A,B), A = B.
No
?- neq(a,a).
No
?- neq(A,B), A = a, B = a.
No
?- neq(a,b).
Yes
?- neq(A,B), A = a, B = b.
Yes
?- neq(A,B).
neq(A,B)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(A,B), A = B.
No
?- neq(a,a).
No
?- neq(A,B), A = a, B = a.
No
?- neq(a,b).
Yes
?- neq(A,B), A = a, B = b.
Yes
?- neq(A,B).
neq(A,B)
?- neq(A,B), A = f(C), B = f(D).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

A Constraint: Inequality

neq(X,X) <=> fail.

neq(X,Y) <=> X \= Y | true.

?- neq(A,B), A = B.
No
?- neq(a,a).
No
?- neq(A,B), A = a, B = a.
No
?- neq(a,b).
Yes
?- neq(A,B), A = a, B = b.
Yes
?- neq(A,B).
neq(A,B)
?- neq(A,B), A = f(C), B = f(D).
neq(f(C),f(D))

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 94 / 112

Using Reactivation for Finite Domains

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).
intersection(L1,L2,L3) :- ...

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 95 / 112

Using Reactivation for Finite Domains

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).
intersection(L1,L2,L3) :- ...

?- domain(X,[dog,fox,horse,snails,zebra]).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 95 / 112

Using Reactivation for Finite Domains

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).
intersection(L1,L2,L3) :- ...

?- domain(X,[dog,fox,horse,snails,zebra]).
domain(X,[dog,fox,horse,snail,zebra])

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 95 / 112

Using Reactivation for Finite Domains

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).
intersection(L1,L2,L3) :- ...

?- domain(X,[dog,fox,horse,snails,zebra]).
domain(X,[dog,fox,horse,snail,zebra])
?- domain(X,[d,f,h,s,z]), X = z.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 95 / 112

Using Reactivation for Finite Domains

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).
intersection(L1,L2,L3) :- ...

?- domain(X,[dog,fox,horse,snails,zebra]).
domain(X,[dog,fox,horse,snail,zebra])
?- domain(X,[d,f,h,s,z]), X = z.
X = z

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 95 / 112

Using Reactivation for Finite Domains

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).
intersection(L1,L2,L3) :- ...

?- domain(X,[dog,fox,horse,snails,zebra]).
domain(X,[dog,fox,horse,snail,zebra])
?- domain(X,[d,f,h,s,z]), X = z.
X = z
?- domain(X,[d,f,h,s,z]), domain(X,[c,s,z]).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 95 / 112

Using Reactivation for Finite Domains

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).
intersection(L1,L2,L3) :- ...

?- domain(X,[dog,fox,horse,snails,zebra]).
domain(X,[dog,fox,horse,snail,zebra])
?- domain(X,[d,f,h,s,z]), X = z.
X = z
?- domain(X,[d,f,h,s,z]), domain(X,[c,s,z]).
domain(X,[s,z])

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 95 / 112

Using Reactivation for Finite Domains

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).
intersection(L1,L2,L3) :- ...

?- domain(X,[dog,fox,horse,snails,zebra]).
domain(X,[dog,fox,horse,snail,zebra])
?- domain(X,[d,f,h,s,z]), X = z.
X = z
?- domain(X,[d,f,h,s,z]), domain(X,[c,s,z]).
domain(X,[s,z])
?- domain(X,[d,f,h,s,z]), domain(X,[c,z]).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 95 / 112

Using Reactivation for Finite Domains

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).
intersection(L1,L2,L3) :- ...

?- domain(X,[dog,fox,horse,snails,zebra]).
domain(X,[dog,fox,horse,snail,zebra])
?- domain(X,[d,f,h,s,z]), X = z.
X = z
?- domain(X,[d,f,h,s,z]), domain(X,[c,s,z]).
domain(X,[s,z])
?- domain(X,[d,f,h,s,z]), domain(X,[c,z]).
X = z

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 95 / 112

Summary

Reactivation
I rules may not fire because of lack of instantiation

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 96 / 112

Summary

Reactivation
I rules may not fire because of lack of instantiation
I upon instantiation, the rule may now fire

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 96 / 112

Summary

Reactivation
I rules may not fire because of lack of instantiation
I upon instantiation, the rule may now fire
I useful for implementing constraint solvers

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 96 / 112

Overview 97/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 97 / 112

Summary: CHR vs Prolog

Prolog CHR

heads 1 ≥1

rule selection unification matching & guard

different rules alternatives/backtracking try all in sequence

no rule failure delay

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 98 / 112

CHR and Backtracking

p :- a. a <=> c1. b <=> d1.

p :- b. a <=> c2. b <=> d2.

?- p.
c1

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 99 / 112

CHR and Backtracking

p :- a. a <=> c1. b <=> d1.

p :- b. a <=> c2. b <=> d2.

?- p.
c1 ; y
d1

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 99 / 112

CHR and Backtracking

p :- a. a <=> c1. b <=> d1.

p :- b. a <=> c2. b <=> d2.

I Prolog creates
choicepoints
?- p. I CHR does not
c1 ; y I Prolog
d1 backtracking
undoes CHR
changes

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 99 / 112

Backtracking for Labeling

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).

split(X) <=> ground(X) | true.

split(X), domain(X,[V|Vs]) <=> X = V ; domain(X,Vs).

?- domain(X,[d,z]), split(X).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 100 / 112

Backtracking for Labeling

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).

split(X) <=> ground(X) | true.

split(X), domain(X,[V|Vs]) <=> X = V ; domain(X,Vs).

?- domain(X,[d,z]), split(X).
X = d ; y
X = z

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 100 / 112

Backtracking for Labeling

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).

split(X) <=> ground(X) | true.

split(X), domain(X,[V|Vs]) <=> X = V ; domain(X,Vs).

?- domain(X,[d,z]), split(X).
X = d ; y
X = z

?- domain(X,[d,f,z]), split(X).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 100 / 112

Backtracking for Labeling

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).

split(X) <=> ground(X) | true.

split(X), domain(X,[V|Vs]) <=> X = V ; domain(X,Vs).

?- domain(X,[d,z]), split(X).
X = d ; y
X = z

?- domain(X,[d,f,z]), split(X).
X = d ; y
domain(X,[f,z])
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 100 / 112
Backtracking for Labeling

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).

indomain(X) <=> ground(X) | true.

% indomain(X), domain(X,[V1,...,Vn]) <=> X = V1;...;X = Vn.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 101 / 112

Backtracking for Labeling

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).

indomain(X) <=> ground(X) | true.

% indomain(X), domain(X,[V1,...,Vn]) <=> X = V1;...;X = Vn.
indomain(X), domain(X,Vs) <=> member(X,Vs).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 101 / 112

Backtracking for Labeling

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).

indomain(X) <=> ground(X) | true.

% indomain(X), domain(X,[V1,...,Vn]) <=> X = V1;...;X = Vn.
indomain(X), domain(X,Vs) <=> member(X,Vs).

?- domain(X,[d,f,z]), indomain(X).

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 101 / 112

Backtracking for Labeling

domain(X,L) <=> ground(X) | memberchk(X,L).

domain(X,[]) <=> fail.
domain(X,[V]) <=> X = V.
domain(X,L1), domain(X,L2) <=>
intersection(L1,L2,L), domain(X,L).

indomain(X) <=> ground(X) | true.

% indomain(X), domain(X,[V1,...,Vn]) <=> X = V1;...;X = Vn.
indomain(X), domain(X,Vs) <=> member(X,Vs).

?- domain(X,[d,f,z]), indomain(X).
X = d ; y
X = f ; y
X = z

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 101 / 112

Overview 102/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 102 / 112

Summary

You should now have an understanding of:

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 103 / 112

Summary

You should now have an understanding of:

I how CHR works.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 103 / 112

Summary

You should now have an understanding of:

I how CHR works.

I how CHR differs from Prolog.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 103 / 112

Summary

You should now have an understanding of:

I how CHR works.

I how CHR differs from Prolog.

I that CHR is not exclusively about constraints .

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 103 / 112

Overview 104/112

1 Introduction
2 The Art of CHR
3 Simpagation
4 For the Sake of Arguments
5 En Guarde
6 A Perfect Match
7 Bigger Programs
8 Propagation
9 Order in the Rules
10 Reactivation
11 CHR vs. Prolog
12 Tutorial Summary
13 Facts about CHR

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 104 / 112

History: Programming Highlights

1991 CHR is born, Thom Frühwirth

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 105 / 112

History: Programming Highlights

1991 CHR is born, Thom Frühwirth

1995 Christian Holzbaur implements CHR(SICStus)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 105 / 112

History: Programming Highlights

1991 CHR is born, Thom Frühwirth

1995 Christian Holzbaur implements CHR(SICStus)
2002 Leuven CHR is born

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 105 / 112

History: Programming Highlights

1991 CHR is born, Thom Frühwirth

1995 Christian Holzbaur implements CHR(SICStus)
2002 Leuven CHR is born
2002-2005 optimized compilation & program analysis (abstract
interpretation), PhDs of Gregory Duck and Tom Schrijvers

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 105 / 112

History: Programming Highlights

1991 CHR is born, Thom Frühwirth

1995 Christian Holzbaur implements CHR(SICStus)
2002 Leuven CHR is born
2002-2005 optimized compilation & program analysis (abstract
interpretation), PhDs of Gregory Duck and Tom Schrijvers
2004 refined semantics, Gregory Duck et al.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 105 / 112

History: Programming Highlights

1991 CHR is born, Thom Frühwirth

1995 Christian Holzbaur implements CHR(SICStus)
2002 Leuven CHR is born
2002-2005 optimized compilation & program analysis (abstract
interpretation), PhDs of Gregory Duck and Tom Schrijvers
2004 refined semantics, Gregory Duck et al.
2004 1st CHR workshop

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 105 / 112

History: Programming Highlights

1991 CHR is born, Thom Frühwirth

1995 Christian Holzbaur implements CHR(SICStus)
2002 Leuven CHR is born
2002-2005 optimized compilation & program analysis (abstract
interpretation), PhDs of Gregory Duck and Tom Schrijvers
2004 refined semantics, Gregory Duck et al.
2004 1st CHR workshop
2005-2008 computational complexity, PhD Jon Sneyers

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 105 / 112

History: Programming Highlights

1991 CHR is born, Thom Frühwirth

1995 Christian Holzbaur implements CHR(SICStus)
2002 Leuven CHR is born
2002-2005 optimized compilation & program analysis (abstract
interpretation), PhDs of Gregory Duck and Tom Schrijvers
2004 refined semantics, Gregory Duck et al.
2004 1st CHR workshop
2005-2008 computational complexity, PhD Jon Sneyers
2005- Leuven JCHR (Java), Peter Van Weert

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 105 / 112

History: Programming Highlights

1991 CHR is born, Thom Frühwirth

1995 Christian Holzbaur implements CHR(SICStus)
2002 Leuven CHR is born
2002-2005 optimized compilation & program analysis (abstract
interpretation), PhDs of Gregory Duck and Tom Schrijvers
2004 refined semantics, Gregory Duck et al.
2004 1st CHR workshop
2005-2008 computational complexity, PhD Jon Sneyers
2005- Leuven JCHR (Java), Peter Van Weert
2007 first concurrent system, Sulzmann & Lam

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 105 / 112

CHR Systems

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 106 / 112

CHR Programs

Thom Frühwirth, with his students:

I lots of example programs, constraint solvers and others
Thom Frühwirth & Tom Schrijvers,
I union-find
Jon Sneyers,
I Fibonacci heaps and Dijkstra’s shortest path
I Hopcroft’s DFA minimization
I Turing and RAM machine simulators
Henning Christiansen,
I meta-programming in CHR
...

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 107 / 112

CHR Theory & Applications

Declarative Semantics
I classical first-order logic (Frühwirth)
I linear logic (Betz)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 108 / 112

CHR Theory & Applications

Declarative Semantics
I classical first-order logic (Frühwirth)
I linear logic (Betz)
Rewriting Properties
I confluence (Frühwirth;Duck;Haemmerle)
I completion (Frühwirth&Abdennadher)
I termination (Frühwirth;Voets&Pilozzi)
I complexity (Frühwirth;De Koninck)

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 108 / 112

CHR Theory & Applications

Declarative Semantics
I classical first-order logic (Frühwirth)
I linear logic (Betz)
Rewriting Properties
I confluence (Frühwirth;Duck;Haemmerle)
I completion (Frühwirth&Abdennadher)
I termination (Frühwirth;Voets&Pilozzi)
I complexity (Frühwirth;De Koninck)
Applications
I type systems (Sulzmann & Stuckey)
I test case generation (Schrijvers)
I multi-agent systems (Alberti)
I ...
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 108 / 112
CHR Researchers around the World

Fig. 1. CHR research groups all over the world: 1. Ulm, Germany (Frühwirth, Meister, Betz,
Djelloul, Raiser, . . . );
2. Leuven, Belgium (Schrijvers, Demoen, Sneyers, De Koninck, Van Weert,
. . . ); 3. Melbourne, Australia (Duck, Stuckey, Garcı́a de la Banda, Wazny, Brand, . . . ); 4. Vi-
Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 109 / 112
CHR in Industry

multi-headed business rules

custom constraint solvers

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 110 / 112

CHR in Industry

multi-headed business rules

custom constraint solvers
I Scientific Software & Systems Ltd.
I stock brokering software
I Cornerstone Technology Inc
I injection mould design tool
I BSSE System and Software Engineering
I test generation
I MITRE Corporation
I optical network design

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 110 / 112

Further Reading

I CHR website (Google for)

I programs
I papers
I CHR Survey
I As Time Goes By: Constraint Handling Rules – A Survey of CHR
Research from 1998 to 2007, J. Sneyers, P. Van Weert, T. Schrijvers,
L. De Koninck.

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 111 / 112

Thank You!

My Questions: Get the Quiz!

Your Questions?

Tom Schrijvers (K.U.Leuven) Constraint Handling Rules 112 / 112