Research Directions in Object Oriented Languages* Bharat Jayaraman Professor and Chair Department of Computer Science & Engineering University at Buffalo,

Research Directions in Object Oriented Languages*

Bharat Jayaraman

Professor and ChairDepartment of Computer Science & Engineering

University at Buffalo, NY 14260

http://www.cse.buffalo.edu/LRG

* Joint work with recent PhD graduates Pallavi Tambay (2003) and Paul Gestwicki (2005)

and several MS students at Buffalo

Programming Languages

Imperative Declarative

Procedural Data Abstraction

C …Fortran ObjectsModules

Ada … Java …

Relational Functional

Constraints Logic ML …

CLP … Prolog …

… …

Two Research Projects

I. Modeling Complex Systems

II. Interactive Visualization

I. Modeling Complex Systems

• Motivating Domains Engineering Organizational Biological …

• Common Characteristics Assembly of components Laws of behavior Natural visual representations

Building Elements

(Kalay et al 1998)

Examples of Laws/Constraints

• “The load on a wall ≤ its load bearing capacity.”

• “The sum of the forces at a joint should be zero.”

• Various other constraints, such as:- ‘mate’ constraints- ‘flush’ constraints- ‘angle’ constraints

etc.

Our Principles of Modeling

• compositional specification of structure

objects• declarative specification of behavior

constraints• visual specification of assembly

diagrams

I. Constrained Objects

• An object is a container of data that is accessed via a well-defined interface of procedures. (Imperative)

• A constrained object contains data that is governed by laws, or constraints. (Declarative)

• Examples: resistor in a circuit a joint in a truss a cell in a spreadsheet

Constraints• A constraint is a declarative relation among

variables, e.g.,F = 32 + 1.8*C

• Constraints are additive, and collectively enforce the conditions among variables

D ≥ 1 D ≤ 31 D ≤ 29 ← M = 2 /\ leapyear(Y)

• Many algorithms for constraint satisfaction

1. Circuits as Constrained Objects

V1 = I1 * R1

V2 = I2 * R2

V3 = I3 * R3

V1 =V2 = V3 = V

I1 + I2 + I3 = I

1/R1 + 1/R2 + 1/R3 =1/R

V = I * R

Overview of Cob Programs

class class_id [ extends class_id ] { attributes attributes

constraints constraints

constructor constructor_clauses

}

Class inheritance

equations, inequations,conditional,quantified,aggregate

Cob Class Definitions

class component { attributes Real V, I, R; constraints V = I * R; constructor component(V1, I1, R1) { V = V1; I = I1; R = R1; }}

Ohm’s Law

class parallel extends component {attributes component [ ] C; constraints forall X in C: (X.V = V); (sum X in C: X.I) = I; (sum X in C: 1/X.R) = 1/R; constructor parallel(P) { C = P; }}

2. Truss as Constrained Object

Load Load

Beam Joint

class joint { attributes bar [ ] Bars; load [ ] Loads; constraints (sum X in Bars : X.B.F * sin(X.A)) + (sum L in Loads : L.F * sin(L.A)) = 0;

(sum Y in Bars : Y.B.F * cos(Y.A)) + (sum M in Loads : M.F * cos(M.A)) = 0;

constructor joint(B1, L1) { Bars = B1; Loads = L1; }}

Law of Equilibrium of Forces

class bar {attributes beam B; real A; constraints 0 A; A 360; constructor bar (B1, A1)

{ B = B1; A = A1; }}

class load { attributes real F, A; constraints 0 A; A 360; constructor load (F1, A1)

{ F = F1; A = A1; }}

class beam { attributes real E, Sy, L, W, H, F_bn, F_bk, F_t, Sigma, I, F; constraints Pi = 3.141; I = F_bk * L* L / (Pi * Pi * E) ; I = W * H * H * H / 12; F_t = Sy * W * H; Sigma = H * I * F_bn / (8 * I); F = F_t :- F > 0; F = F_bk :- F < 0; constructor beam(E1, Sy1, L1, W1, H1,F1) { E = E1; L = L1; H=H1; W=W1; F = F1; }}

ConditionalConstraint

Overall Modeling Scenario

modeling ::= define_classes ;

build_instance ;

solve_instance ;

[ [ modify ; re_solve ] +;

query * ] +

Modeling Tools • Constrained Object Language (Cob)

Text-based language

• Constraint-Based UML (CUML) Domain-independent graphical modeling language

• Domain-Specific Visual Programming Diagrammatic notations customized for specific application

domains, e.g., circuits, trusses, etc.

(with R. Jotwani)

CUML: Constraint-based UML

CUML Cob

(with A. Dev and N. Menon)

`

Domain-Specific Visual Language

cse@buffalo13

resistor value

extraconstraint

changeresistorbehavior

R1 isunknown

outputconstraint

Compiled CUML Code

Compiled Circuit Diagram Code

Cob Compiler

CLP® Code

Optimizer

Optimized Code

ConstraintEngine

Answers

DiagramManager

CUML Class Diagram

Compiler

Textual Cob Code

By Chinchani, Pramanik, Dutta

Circuit Diagram

Compiler

Textual Code

Optimized Code (by partial evaluation of generated code)

V1 = V2 = I1 * 10 = I2 * 20 V3 = V4 = I3 * 20 = I4 * 20 S.I = I1 + I2 = I3 + I41/P1.R = 1/10 + 1/ 20V3 = S.I * P2 .R1/P2.R = 1/20 + 1/2030 = S.I + S.R30 = V1 + V3S.R = P1.R + P2.R

Building Elements (Kalay 1998)

Overall Class Diagram

class building { attributes level[] Ls; real Area; constraints Area = sum X in Ls: X.Area; constructor building(L) { Ls = L; }}

Class Building

Area of the building is the sum of areas of each level

Class Slabclass slab { attributes level Pl; beam[ ] Peripheralbeams; real Z; constraints forall B in Peripheralbeams: B.Z = Z; constructor slab(L,B) { Pl = L; Peripheralbeams = B; }}

The height of all beams surrounding the slab must be the same

Two-level House (Kalay et al)

Cob Representation for a Space space Spaces00; slab[] Slabs0; slab[] Slabs1; surface[] Surfaces00; beam[] Beams01; slab Slabs00; surface Surfaces003, Surfaces002; surface Surfaces001, Surfaces000; horizontalSurface Ceiling00; horizontalSurface Floor00; beam B014, B013,B012, B011 ; edge Edge003, Edge002; edge Edge001, Edge000; wall Wall007, Wall002; wall Wall001, Wall000; column C8,C4, C6, C2; vertex V8, V7, V6, V5; vertex V26, V30, V36, V31; level Level0;

All the elements needed to depict Spaces00 (green)

Wall007 = new wall( V7, V8, 3.0, B014 );Wall002 = new wall( V6, V7, 3.0, B013 );Wall001 = new wall( V5, V6, 3.0, B012 );Wall000 = new wall( V8, V5, 3.0, B011 );Slabs10 = new slab( Level1, Beams10 );Slabs00 = new slab( Level0, Beams00 );B012 = new beam( V5, V6, B014, _ );B011 = new beam( V8, V5, B013, _ );B014 = new beam( V7, V8, B012, _ );B013 = new beam( V6, V7, B011, _ );C4 = new column( V6, V50 );C6 = new column( V8, V51 );C2 = new column( V5, V46 );C8 = new column( V7, V55 );Edge003 = new edge( Surfaces003, V7, Edge011 );Edge002 = new edge( Surfaces002, V6, _ );Edge001 = new edge( Surfaces001, V5, _ );Edge000 = new edge( Surfaces000, V8, _ );V36 = new vertex( V55, C8, 5.0, 5.0, 3.0 );V8 = new vertex( V31, _, 5.0, 0.0, 0.0 );V31 = new vertex( V51, C6, 5.0, 0.0, 3.0 );V7 = new vertex( V36, _, 5.0, 5.0, 0.0 );V30 = new vertex( V50, C4, 0.0, 5.0, 3.0 );V6 = new vertex( V30, _, 0.0, 5.0, 0.0 );V5 = new vertex( V26, _, 0.0, 0.0, 0.0 );V26 = new vertex( V46, C2, 0.0, 0.0, 3.0 );dump([Level0 .Vol]);

Cob Representation (cont)

Observations• Constrained objects represent nicely the structure and constraints

arising in intelligent building representations.

• The model described in this presentation has been implemented and works in a predictable way.

• The current approach, even when applied to a simple two-level house with five spaces, is still tedious and error prone.

• There is a need for a higher-level specification from which Cob code can be generated Component-based Design

• Inconsistency analysis and debugging are very important

Conclusions and Further Work

• Constrained Objects ≥ Constraints + Objects• Applicable in many domains• Facilitates true visual programming

• Model Analysis and Fault Detection• Dynamic and Evolving Systems• Intelligent Visualizations

II. Interactive Visualization of OO Programs

• Conventional Execution Model: data objects and procedure activations are in

separate spaces.• Object-Oriented Execution Model: procedure activations occur inside data objects,

i.e., objects are execution environments• Several more criteria for a good visualization

system for OOPs …

Binary Search Tree in Java

class Tree {public Tree(int n) { value = n; left = null; right = null; }protected int value;protected Tree left;protected Tree right;}

public void insert(int n) { if (value == n) return; if (value < n) if (right == null) right = new Tree(n); else right.insert(n); else if (left == null) left = new Tree(n); else left.insert(n); }

Tree

Tree Tree

Tree Tree Tree Tree

1

2 3

4 5 6 7

Tree

Tree Tree

Tree Tree Tree Tree

1

2 3

4 5 6 7

insert1

insert3

2insert

Treei

value

left

right

Tree

insert

int

Tree

Tree

const

proc

insertj

n

rpdl

int

insert in Tree k n

Tree i

Visualization Criteria• Object as environments• Multiple View of Objects• Aesthetic Drawing• Forward and Reverse Execution• History of Execution• Support Run-time Queries• Use Existing Compiler and Run-time System

JIVE: Java Interactive Visualization Environment

JIVE supports Source Code Highlighting, Break Points, andForward as well asReverse Steppingof Programs

JIVE Interaction Controls

Forward and Fast Forward Buttons

Reverse and Fast Reverse Buttons

Minimize, Maximize,Compact View

Call Path View

Displayingonly the‘Call Path’

Object Diagram without Inherited Objects

Minimized Object Diagram

Final Object Diagram

Final Minimized Diagram

TimeSequenceDiagram- to capturethe historyof methodcalls betweenobjects

Tree Program with GUI front-end

Tree Programwith GraphicalUser Interface

SelectiveExpansionofObjectDiagram

Querying the Runtime State

Runtime

Query

“Which variable has the value 50?”

Answer toQuery

Visual Display of Query Result

Object BST:2has the answer

Exact pointof assignment

Visual Display of Query Result

“Variable V in Object BST:2 has the value 50!”

Show all changes to variable ‘left’

Concluding Remarks• Diagrams clarify Java semantics

handles advanced features• Visualization environment

pedagogic tool visual debugger

• JPDA is key to the implementation• Some research issues

Aesthetic drawing Efficient reverse stepping

Current and Future Work

• Reverse execution and efficient state-saving• Drawing algorithms• Visual Queries• Relationship between programs and diagrams• Applications in Security