Case Study: The Design And Development Of C C Language: Motivation

Case Study: The Design and
Development of C
CS386L - Programming Languages
Dr. Greg Lavender
Department of Computer Sciences
The University of Texas at Austin
CS386L: Lecture-02 - Case Study: The Design & Development of C
C Language: motivation

designed as a systems implementation language

for the development of an operating system, compiler, &
utilities

created by a small group of software engineers at Bell
Labs in early 1970s

Ken Thompson, Dennis Richie, Brian Kernighan, et al.

born out of frustration with “big” OS projects and
large complex languages

Multics, PL/I, Algol 68

originally designed for small memory machines (PDP-7,
PDP-11)

design influenced by small machine architectural limitations

word size, small main memory size, small register set, etc.

later ported to mainframes & made “portable”
2/2/04
2
1

CS386L: Lecture-02 - Case Study: The Design & Development of C
C Language: influences

indirect: Fortran, Algol, PL/I, Algol68

direct: BCPL, B, NB (new B)

Ken Thompson invented B

B is C without types

B is derived from BCPL

BCPL invented at MIT by Martin Richards

B “… is BCPL squeezed into 8K bytes of memory…”

“… and filtered through [Ken] Thompson’s brain.”

BCPL, B and C all differ in some aspects of syntax &
semantics

B & C do not allow nested procedure scopes, BCPL does

all support separate compilation and allow “include” files

B is “simpler” than BCPL because of small memory machine

very compact syntax required so compiler could work in one pass
2/2/04
3
CS386L: Lecture-02 - Case Study: The Design & Development of C
C Language: syntax

BCPL used ‘:=‘ for assignment (like Algol)

B dropped the ‘:’ and just used the ‘=‘

how many times have you made the following mistake in
C/C++/Java !!

if (a = b) then … else …

B uses /*…*/ for comments, BCPL uses //

/*…*/ comes from PL/I

note that C++ re-introduced //

Bjarne Stroustrup used BCPL during his PhD research

B declarations begin with ‘auto’ or ‘static’. C
introduced type specifiers, but auto is optional and
assumed in C for local variables

static int x = 0;

int f() { auto int y; … }
2/2/04
4
2

CS386L: Lecture-02 - Case Study: The Design & Development of C
C Language: syntax

generalized assignment operations

BCPL/B/early C: x =+ y;

simplified lexical analysis

later C: x += y; // and the others: -=,*=,/=

pre/post increment operators: ++, --

appeared in B as a more compact syntax for parsing

++x: x = x + 1;

y--: y = y - 1;

Question: what is the meaning of ++x + x++ in C/C++?

TMG - TransMoGrifier

an early top-down recursive descent compiler-
compiler by McClure

lead to the development of lex/yacc in C
2/2/04
5
CS386L: Lecture-02 - Case Study: The Design & Development of C
C Language: semantics

B kept much of BCPL semantics

equivalence of pointers and arrays, pointer
arithmetic

BCPL: let V = vec 10

V!I to index to ith element

B/C: auto V[10]

*(V+i) or V[i] to index to ith element of an array

note: because of equivalence of pointers and arrays, no
auto array bounds checking as in Pascal & Java

has led to tens of thousands of programming errors

strings

BCPL: first byte contains the length

B: last byte contains a special “end of string” character
2/2/04
6
3

CS386L: Lecture-02 - Case Study: The Design & Development of C
C Language: pragmatics

B deficiencies motivated an improved compiler

B generated threaded-code, not native-code

basically, a byte-code interpreter using a stack

B used packed strings causing inefficient string manipulation

overhead in dealing with pointers

“New B” (NB) invented by Richie to be more efficient

introduced int, char & their array types, and pointer types

compiled to native code

C fully equated typed arrays with typed pointers

int a[10]; … x = a[i]; means x = (a+i*sizeof(int)) where a is
pointer to the first memory cell of the array
2/2/04
7
CS386L: Lecture-02 - Case Study: The Design & Development of C
C Language: types

C primarily improved on B by adding type
declarations & type rules

syntax influenced by Algol 68

int i, *pi, **ppi;

int f(), *f(), (*pf)();

int ai[10], *api[10], (*pai)[10];

also structs and unions
2/2/04
8
4

CS386L: Lecture-02 - Case Study: The Design & Development of C
C Language: standardization

C language definition complete in 1973

Unix kernel rewritten in C

new features added from 1973-1980

unsigned, long, union, enums

porting of C compiler to other machines

K&R C book published 1978

de facto language standard for 10 years

Explosion of C compilers in early 1980s

for mainframes, minicomputers, workstations & PCs

ANSI C standardization 1989

key additions:

function prototypes as in C++

const and volatile keywords introduced

but omitted the builtin ‘bool’ primitive as in C++ and Java

bool finally added to the 1999 update to the C Standard

see <stdbool.h>
2/2/04
9
CS386L: Lecture-02 - Case Study: The Design & Development of C
C Language: evolution

variants & descendants

Concurrent C, Objective C, C*, C++

Java, C#

used as a “portable assembly” language

early C++ and Modula-3 and Eiffel all “compile” to C

leverages back-end (code gen) of existing C compilers

but makes debugging more difficult

C compilers generate very efficient code
today

for all kinds of devices

at last count, GNU gcc generated native code for 70+
different instruction set architectures

most systems & networking software is written in C
2/2/04
10
5

CS386L: Lecture-02 - Case Study: The Design & Development of C
C Language: Challenge Problem!!

Write the shortest C program that prints an exact
syntactic replica of itself

you may not just print the original source file

this is an example of self-reference in a language

Douglas Hofstadter, in the book Godel, Escher, Bach, calls this
kind of program a “quine” after W.V.O Quine (1908-2000), a
famous philosopher/logician (see www.wvquine.org)

first correct and shortest solution wins a small prize!

submit your best solution to lavender@cs.utexas.edu

unix% gcc quine.c -o quine

unix% quine > quine.copy

unix% diff quine.c quine.copy

smallest C quine is 64 characters.

You are on your honor to solve the problem with no help from any
source other than your own mind.

challenge problems are not officially homework, but they are
very relevant to the course. For more fun, try writing a quine in
other languages (e.g., Scheme, ML, Java, Perl, …) and compare
them.
2/2/04
11
6