This third edition is a revised and expanded version of Winston and Horn's best-selling introduction to the LISP programming language and to LISP-based applications, many of which are possible as a result of advances in Artificial Intelligence technology.
(NOTE: Each chapter ends with a Summary.
TABLE OF CONTENTS.
1. Understanding Symbol Manipulation.
Symbol Manipulation Is Like Working with Words and Sentences.
Lisp Helps Make Computers Intelligent.
Lisp Promotes Productivity and Facilitates Education.
Lisp Is the Right Symbol-Manipulation Language To Learn.
CommonLisp Is the Right Lisp To Learn.
Beware of False Myths.
2. Basic Lisp Primitives.
Lisp Means Symbol Manipulation.
Lisp Procedures and Data Are Symbolic Expressions.
Lists Are Like Bowls.
FIRST and REST Take Lists Apart.
Quoting Stops Evaluation.
Some Old Timers Use CARs and CDRs.
SETF Assigns Values to Symbols.
SETF Accepts Multiple Symbol-Value Pairs.
Certain Atoms Evaluate to Themselves.
CONS, APPEND and LIST Construct Lists.
CONS, APPEND, and LIST Do Not Alter Symbol Values.
NTHCDR, BUTLAST, and LAST Shorten Lists.
LENGTH and REVERSE Work on Top-Level Elements.
ASSOC Looks for Indexed Sublists.
Lisp Offers Integers, Ratios, and Floating-Point Numbers, among Others.
A Few Primitives for Numbers Round Out a Basic Repertoire.
3. Procedure Definition and Binding.
DEFUN is Lisp's Procedure-Definition Primitive.
Parameter Variable Values Are Isolated by Virtual Fences.
Special Variable Values Are Not Isolated by Virtual Fences.
Procedures Match Parameters to Arguments.
LET Forms Bind Parameters to Initial Values.
LET Forms Produce Nested Fences.
LET Forms Evaluate Initial-Value Forms in Parallel.
LET* Forms Evaluate Initial-Value Forms Sequentially.
Progressive Envelopment and Comment Translation Help Define New Procedures.
4. Predicates and Conditionals.
A Predicate Is a Procedure That Returns True or False.
EQUAL, EQ, EQL, and = Are Equality Predicates.
MEMBER Tests for List Membership.
Keyword Arguments Modify Behavior.
LISTP, ATOM, NUMBERP, and SYMBOLP Are Data-Type Predicates.
NIL Is Equivalent to the Empty List.
NULL and ENDP Are Empty-List Predicates.
There Are Many Number Predicates.
AND, OR, and NOT Enable Elaborate TestingPredicates Help IF, WHEN, and UNLESS Choose among Alternatives.
Predicates Help COND Choose among Alternatives.
CASE Is Still Another Conditional.
Conditionals Enable DEFUN To Do Much More.
Problem Reduction Helps Define New Procedures.
5. Procedure Abstraction and Recursion.
Procedure Abstraction Hides Details Behind Abstraction Boundaries.
Recursion Allows Procedures To Use ThemselvesRecursion Can Be Efficient.
Recursion Can Be Used To Analyze Nested Expressions.
Optional Parameters Eliminate the Need for Many Auxiliaries.
Advanced Programmers Use Rest, Key, and Aux ParametersOnly a Few Lisp Primitives Are Really Necessary.
6. Data Abstraction and Mapping.
Data Details Stifle Progress.
Data Abstraction Facilitates Progress.
You Should Use Readers, Constructors, and Writers Liberally.
It Is Useful To Transform and To Filter.
Recursive Procedures Can Transform and Filter.
Recursive Procedures Can Count and Find.
Cliches Embody Important Programming Knowledge.
MAPCAR Simplifies Transforming Operations.
REMOVE-IF and REMOVE-IF-NOT Simplify Filtering Operations.
COUNT-IF and FIND-IF Simplify Counting and Finding Operations.
FUNCALL and APPLY Also Take a Procedure Argument.
LAMBDA Defines Anonymous Procedures.
7. Iteration on Numbers and Lists.
DOTIMES Supports Iteration on Numbers.
DOLIST Supports Iteration on Lists.
DO Is More General than DOLIST and DOTIMES.
LOOP Never Stops, Almost.
PROG1 and PROGN Handle Sequences Explicitly.
8. File Editing, Compiling, and Loading.
Programs and Data Reside in Files.
File Specifications Tend to Have Baroque Forms.
ED Takes You to an Editor.
Emacs Is a Particularly Powerful Lisp Editor.
COMPILE-FILE Compiles Files.
LOAD Causes Lisp To Read from Files.
9. Printing and Reading.
PRINT and READ Facilitate Simple Printing and Reading.
FORMAT Enables Exotic Printing.
WITH-OPEN-FILE Enables Reading from Files.
Optional Arguments in READ Forms Specify End-of-File Treatment.
WITH-OPEN-FILE Enables Printing to Files.
READ Does Not Evaluate Expressions, but EVAL Evaluates Twice.
Special Primitives Manipulate Strings and Characters.
READ-LINE and READ-CHAR Read Strings and Characters.
10. Rules for Good Programming and Tools for Debugging.
Following Rules of Good Programming Practice Helps You To Avoid Bugs.
Big Programs Require Abstraction and Modularity.
Most Programmers use TRACE, STEP, and BREAK with Varying Frequency.
TRACE Causes Procedures To Print Their Arguments and Values.
STEP Causes Procedures To Proceed One Step at a Time.
BREAK Stops Evaluation so that You Can Evaluate Forms.
TIME, DESCRIBE, and DRIBBLE Are Helpful Too.
Debugging Is Implementation Specific.
11. Properties and Arrays.
Each Way of Storing Data Has Constructors, Readers, and Writers.
Properties Enable Storage in Symbolically Indexed Places.
GET and SETF are the Custodians of Properties.
Arrays Enable Storage in Numerically Indexed Places.
MAKE-ARRAY, AREF, and SETF are the Custodians of Arrays.
12. Macros and Backquote.
Macros Translate and Then Evaluate.
The Backquote Mechanism Simplifies Template Filling.
The Backquote Mechanism Simplifies Macro Writing.
Optional, Rest, and Key Parameters Enable More Powerful Macros.
Macros Deserve Their Own File.
Structure Types Facilitate Data Abstraction.
Structure Types Enable Storage in Procedurally Indexed Places.
Individual Structure Types Are New Data Types.
One Structure Type Can Include the Fields of Another.
Structure Types Are Important Components of Big Systems.
DESCRIBE Prints Descriptions.
DEFSTRUCTs Deserve Their Own File.
14. Classes and Generic Functions.
What to Do Depends on What You Do it to.
You Can Make Ordinary Procedures Data Driven, Albeit Awkwardly.
Methods Are Procedures Selected from Generic Functions by Argument Types.
Classes Resemble Structure Types but Resonate Better with Generic Functions.
Any Nonoptional Argument's Class Can Help Select a Method.
Classes Enable Method Inheritance.
The Most Specific Method Takes Precedence over the Others.
Parameter Order Helps Determine Method Precedence.
Simple Rules Approximate the Complicated Class Precedence Algorithm.
Methods Can Be Specialized to Individual Instances.
Method Selection Involves Three Steps.
Object-Oriented Programming Offers Advantages, Not Magic.
15. Lexical Variables, Generators, and Encapsulation.
LETs Produce Nested Fences.
Nested Fences Provide Variable Values.
Procedure Calls Usually Do Not Produce Nested Fences.
Nested Definitions do Produce Nested Fences.
Generators Produce Sequences.
Nameless Procedures Produce Nested Fences.
Nameless Procedures Can Be Assigned to Variables.
The Free Variables in Nameless Procedures Can Be Encapsulated.
Encapsulation Enables the Creation of Sophisticated Generators.
Generators Can Be Defined by other Procedures.
Nameless Procedures Become Lexical Closures.
16. Special Variables.
Bindings Could Be Kept on a Record of Calls.
Some Variables Are Declared To Be Special Forevermore.
Special-Variable Bindings Are Actually Kept on a Stack.
DEFVAR Can Assign as Well as Declare.
Some Variable Instances Can Be Special while Others Are Lexical.
Both Lexical and Special Variables Can Be Free Variables.
17. List Storage, Surgery, and Reclamation.
Lists Can Be Represented by Boxes and Pointers.
Boxes and Pointers Can Be Represented by Bytes.
CONS Builds New Lists by Depositing Pointers in Free Boxes.
APPEND Builds New Lists by Copying.
NCONC and DELETE Can Alter Box Contents Dangerously.
SETF Also Can Alter Box Contents Dangerously.
EQ Checks Pointers Only.
Garbage Collection Reclaims Abandoned Memory.
Lisp Allows You To Write Inefficient Procedures.
Simple Garbage Collectors Use the Mark and Sweep Approach.
Simulation Procedures Expose Garbage Collection Details.
MARK Places Marks on Useful Chunks.
SWEEP Collects Unmarked Chunks.
Marking Can Be Done without Recursion.
Our Nonrecursive Marking Procedure Leaves a Trail of Pointers.
Some Garbage Collectors Are Incremental.
18. Lisp in Lisp.
It Is Easy To Build a Simple Interpreter for a Lisplike Language.
MICRO-EVAL and MICRO-APPLY Work Together To Evaluate Forms.
Traces Show How MICRO-EVAL and MICRO-APPLY Work Together.
Closures Encapsulate Environments.
Special Variable Binding Can Be Arranged.
Lisp Does Call-by-Value Rather Than Call-by-Reference.
Lisp Can Be Defined in Lisp.
Fancy Control Structures Usually Start Out as Basic Lisp Interpreters.
19. Examples Involving Search.
Breadth-First and Depth-First Searches Are Basic Strategies.
Best-First Search and Hill-Climbing Require Sorting.
20. Examples Involving Simulation.
Projects Involve Events and Tasks.
Structures Can Represent Events and Tasks.
Simulation Procedures Can Propagate Event Times.
Event and Task Structures Require Special Printing Procedures.
An Event List Keeps Simulation in Step with the Simulated Project.
21. The Blocks World with Classes and Methods.
The Blocks-World Program Handles Put-On Commands.
Object-Oriented Programming Shifts Attention from Procedures to Objects.
Object-Oriented Programming Begins with Class Specification.
Slot Readers Are Generic Functions.
The Blocks-World Program's Methods Are Transparent.
Before and After Methods Simplify Bookkeeping.
Slot Writers Are Generic Functions.
Object-Oriented Programming Enables Automatic Procedure Assembly.
You Can Control How Instances Are Printed.
The Number-Crunching Methods Can Be Ignored.
The Blocks-World Program Illustrates Abstraction.
22. Answering Questions about Goals.
The Blocks-World Program Can Introspect into its Own Operation.
Remembering Generic Function Calls Creates a Goal Tree.
Macros Enable Method-Defining Procedures To Be Defined.
The Goal Tree Is Easy to Display.
The Goal Tree Answers Questions.
23. Constraint Propagation.
Constraints Propagate Numbers through Arithmetic Boxes.
Constraints Propagate Probability Bounds through Logic Boxes.
Classes Represent Assertions and Logical Constraints.
Generic Functions Enforce Constraints.
More Information Moves Probability Bounds Closer.
24. Symbolic Pattern Matching.
Matching Compares Patterns and Datums Element by Element.
MATCH Keeps Variable Bindings on an Association List.
Matching Is Easily Implemented by a Recursive Procedure.
Matching Is Better Implemented Using Procedure Abstraction.
Unification Is Generalized Matching.
25. Streams and Delayed Evaluation.
Streams Are Sequences of Data Objects.
We Can Represent Streams Using Lists.
We Can Delay Evaluation by Encapsulation.
We Can Represent Streams Using Delayed Evaluation.
26. Rule-Based Expert Systems and Forward Chaining.
Forward Chaining Means Working from Antecedents to Consequents.
We Use Streams To Represent Assertions and Rules.
Our First Pass Concentrates on MATCH and the Binding Stream.
Our Second Pass Concentrates on the Procedures that Surround MATCH.
Simple Rules Help Identify Animals.
Rules Facilitate Question Answering and Probability Computing.
Our Forward-Chaining Program Illustrates Abstraction.
27. Backward Chaining and PROLOG.
Our Backward Chainer Borrows Procedures from our Forward Chainer.
Backward Chaining Means Working from Consequents to Antecedents.
Our First Pass Concentrates on MATCH, UNIFY, and the Binding Stream.
Our Second Pass Concentrates on the Procedures that Surround MATCH and UNIFY.
Completing Our Backward-Chaining Program Involves a Few Auxiliary Procedures.
Simple Rules Help Identify Animals.
Our Backward Chainer Implements a Language like Prolog.
Our Backward-Chaining Program Illustrates Abstraction.
28. Interpreting Transition Trees.
Procedures Can Produce Multiple Values.
Natural-Language Interfaces Produce Database Commands.
Transition Trees Capture English Syntax.
A Transition Tree Interpreter Follows an Explicit Description.
Multiple-Valued Procedures Embody Transition Trees.
Our Interpreter Uses Explicit Transition-Tree Descriptions.
We Use a Macro To Simplify Tree Definition.
A Read, Analyze, and Report Loop Adds a Finishing Touch.
29. Compiling Transition Trees.
Transition Trees Can Be Compiled from Transparent Specifications.
Compilers Treat Programs as Data.
Compiled Programs Run Faster.
Compilers Are Usually Major Undertakings.
Lisp Itself Is Either Compiled or Interpreted.
30. Procedure-Writing Programs and Database Interfaces.
Grammars Can Be Sophisticated.
Answering Requests Is Done in Three Steps.
Most Database Commands Transform Relations into Relations.
English Questions Correspond to Database Commands.
Our Simulated Database Supports an Improved Grammar.
The Relational Database Can Be Faked.
The Database Illustrates Data Abstraction.
31. Finding Patterns in Images.
Generating All Possible Matches Helps Isolate the Correct Match.
Constraints Are Needed To Isolate the Correct Match.
The Search Tree Can Be Pruned Using Geometric Information.
Matches Have To Be Checked for Global Consistency.
Matching Is Harder if Mismatches Are Allowed.
Keeping Track of Mismatches Improves Efficiency.
The Cost of Filtering Has To Be Weighed against the Cost of Searching.
Multiple Matching Attempts Lead to Recognition.
Edges Provide More Constraint than Points.
32. Converting Notations, Manipulating Matrices, and Finding Roots.
It Is Easy to Translate Infix Notation to Prefix.
Sparse Matrices Can Be Represented as Lists of Lists.
Complex Numbers Constitute Another Numeric Data Type.
Roots of Quadratic Equations Are Easy To Calculate.
Roots of Cubic Equations Can Be Calculated.
Roots of Quartic Equations Are Harder To Calculate.
Appendix: The Computation of the Class Precedence List.
Make Initial Lists.
Make a List of Precedence Pairs.
Make a List of Precedence List Candidates.
Select a Candidate.
Shrink the List of Precedence Pairs. @AHEADS - Repeat.
Index of LISP Primitives Used in this Book.
Index of LISP Definitions.