.\" Define Superscript Macro .de SUP \v'-.4m'\s-3\\$1\s+3\v'.4m' .. .\" Define Subscript Macro .de SUB \*<\s-3\\$1\s+3\*> .. .\" Define Horizontal Line Macro .nr VS 15p .VS .de HL .sp 0.5v \D'l \\n[.ll]u 0' .sp 0.5v .. .TL Users\(aq Reference to B .AU Ken Thompson .AI Bell Telephone Laboratories .AB .LP B is a computer language intended for recursive, primarily non-numeric applications typified by system programming. B has a small, unrestrictive syntax that is easy to compile. Because of the unusual freedom of expression and a rich set of operators, B programs are often quite compact. .br .br This manual contains a concise definition of the language, sample programs, and instructions for using the PDP-11 version of B. .AE .LP .br .br .HL .PP [ \f[I] On January 7, 1972 Ken Thompson published the original manual for the B language on the PDP-11. In July of 1997, Dennis Ritchie created an HMTL rendition of that document after scanning, OCR, and editing it, and published it on his Bell Labs website. Nokia, as of this writing, still maintains that website \f[CR]https://www.nokia.com/bell-labs/about/dennis-m-ritchie/kbman.html\fP .PP \f[I]This modern \f[CR]groff(1)\fP rendition was created in April of 2026 by Clem Cole, by using Unix \f[CR]pandoc\fP(1) tool to convert from Dennis's HTML to the \f[CR]-ms\fP macro packaged as supplied by the newest tool suite. There was some minor editing, including adding some of Ken's section numbers, which are not in Dennis's rendition. Note to the careful reader, Ken never created a section 11.0, as he skips from section 10.0 to 12.0. Since Dennis did not repair that error in the original document, I did not either. All of the versions have been placed under \f[CR]Documents\fP directory of the Unix Historical Society (TUHS), in PDF form, as well HTML files as appropriate. and source to this document. The scan of the original that Dennis started with is called \f[CR]kbman.pdf\fP. An image of the original document is available in PDF: just under 1MB. .PP \f[I]Also in that directory is CSTR #8, which is a report by Steve Johnson and Brian Kernighan describing the B implementation on Honeywell equipment. Dennis created two PDFs from this document, one is called: \f[CR]btut.pdf\fP \(em .UL "A Tutorial Introduction to the Language B" , and the other \f[CR]bref.pdf\fP \(em .UL "User's Reference to B on MH-TSS" .PP \f[I]The language being described was really the same, but it\(aqs interesting to look at the differences in description and environment. To describe the BNF syntax, Ken used a variant that depends on super- and subscripts to say \(dqhow many\(dq, which in the original were stacked above each other when they occurred simultaneously. This doesn\(aqt look too awful in the HTML rendering, but was probably better in the original. .PP \f[I]The other thing that is observable is the degree to which the GCOS version had to struggle to work in that environment. Its library description is full of odd concepts like \(dqthe AFT\(dq the *SRC file, and other peculiarities. ] .LP .HL .SH 4 1.0 Introduction .LP .br .br B is a computer language directly descendant from BCPL [1,2]. B 1s running at Murray Hill on the DEC PDP-11 computer under the UNIX-11 time sharing system [3,4]. B is good for recursive, non-numeric, machine independent applications, such as system and language work. .br .br B, compared to BCPL, is syntactically rich in expressions and syntactically poor in statements. A look at the examples in section 9 of this document will give a flavor of the language. .br .br B was designed and implemented by D. M. Ritchie and the author. .SH 4 2.0 Syntax .LP .br .br The syntactic notation in this manual is basically BNF with the following exceptions: .IP "\f[B]1.\f[R]" 3 The metacharacters \f[CR]<\f[R] and \f[CR]>\f[R] are removed. Literals are underlined [in typewriter font here -- DMR] to differentiate them from syntactic variables. .RS 3 .RE .IP "\f[B]2.\f[R]" 3 The metacharacter \f[CR]|\f[R] is removed. Each syntactic alternative is placed on a separate line. .RS 3 .RE .IP "\f[B]3.\f[R]" 3 The metacharacters \f[CR]{\f[R] and \f[CR]}\f[R] denote syntactic grouping. .RS 3 .RE .IP "\f[B]4.\f[R]" 3 A syntactic group followed by numerical sub- and superscripts denote repetition of the group as follows: .RS 3 .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] {..}\* m,m+1,... .sp 0.5v {..}\*\s-3\h'-\w'm'u'\s+3\*{n\*} m,m+1,...,n \f[] .fi .RS 3 .RE .RE .SH 4 2.1 Canonical Syntax .LP .br .br The syntax given in this section defines all the legal constructions in B without specifying the association rules. These are given later along with the semantic description of each construction. .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] program ::= {definition}\*<0\*> definition ::= .sp 0.5v name {[\h'-\w'['u'\l'\w'['u\(ul' {constant}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} ]\h'-\w']'u'\l'\w']'u\(ul'}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} {ival {,\h'-\w','u'\l'\w'['u\(ul' ival}\*<0\*>}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} ;\h'-\w';'u'\l'\w';'u\(ul' .sp 0.5v name (\h'-\w'('u'\l'\w'('u\(ul' {name {,\h'-\w','u'\l'\w','u\(ul' name}\*<0\*>}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} )\h'-\w')'u'\l'\w')'u\(ul' statement ival ::= .sp 0.5v constant .sp 0.5v name .bp statement ::= .sp 0.5v auto\h'-\w'auto'u'\l'\w'auto'u\(ul' name {constant}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} {,\h'-\w','u'\l'\w','u\(ul' name {constant}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*}}\*<0\*> ;\h'-\w';'u'\l'\w';'u\(ul' statement .sp 0.5v extrn\h'-\w'extrn'u'\l'\w'extrn'u\(ul' name {,\h'-\w','u'\l'\w','u\(ul' name}\*<0\*> ;\h'-\w';'u'\l'\w';'u\(ul' statement .sp 0.5v name :\h'-\w':'u'\l'\w':'u\(ul' statement .sp 0.5v case\h'-\w'case'u'\l'\w'case'u\(ul' constant :\h'-\w':'u'\l'\w':'u\(ul' statement .sp 0.5v {\h'-\w'{'u'\l'\w'{'u\(ul' {statement}\*<0\*> }\h'-\w'}'u'\l'\w'}'u\(ul' .sp 0.5v if\h'-\w'if'u'\l'\w'if'u\(ul' (\h'-\w'('u'\l'\w'('u\(ul' rvalue )\h'-\w')'u'\l'\w')'u\(ul' statement {else\h'-\w'else'u'\l'\w'else'u\(ul' statement}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} .sp 0.5v while\h'-\w'while'u'\l'\w'while'u\(ul' (\h'-\w'('u'\l'\w'('u\(ul' rvalue )\h'-\w')'u'\l'\w')'u\(ul' statement .sp 0.5v switch\h'-\w'switch'u'\l'\w'switch'u\(ul' rvalue statement .sp 0.5v goto\h'-\w'goto'u'\l'\w'goto'u\(ul' rvalue ;\h'-\w';'u'\l'\w';'u\(ul' .sp 0.5v return\h'-\w'return'u'\l'\w'return'u\(ul' {(\h'-\w'('u'\l'\w'('u\(ul' rvalue )\h'-\w')'u'\l'\w')'u\(ul'}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} ;\h'-\w';'u'\l'\w';'u\(ul' .sp 0.5v {rvalue}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} ;\h'-\w';'u'\l'\w';'u\(ul' rvalue ::= .sp 0.5v (\h'-\w'('u'\l'\w'('u\(ul' rvalue )\h'-\w')'u'\l'\w')'u\(ul' .sp 0.5v lvalue .sp 0.5v constant .sp 0.5v lvalue assign rvalue .sp 0.5v inc-dec lvalue .sp 0.5v lvalue inc-dec .sp 0.5v unary rvalue .sp 0.5v &\h'-\w'&'u'\l'\w'&'u\(ul' lvalue .sp 0.5v rvalue binary rvalue .sp 0.5v rvalue ?\h'-\w'?'u'\l'\w'?'u\(ul' rvalue :\h'-\w':'u'\l'\w':'u\(ul' rvalue .sp 0.5v rvalue (\h'-\w'('u'\l'\w'('u\(ul'{rvalue {,\h'-\w','u'\l'\w','u\(ul' rvalue}\*<0\*> }\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} )\h'-\w')'u'\l'\w')'u\(ul' assign ::= .sp 0.5v =\h'-\w'='u'\l'\w'='u\(ul' {binary}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} .bp inc-dec ::= .sp 0.5v ++\h'-\w'++'u'\l'\w'++'u\(ul' .sp 0.5v --\h'-\w'--'u'\l'\w'--'u\(ul' unary ::= .sp 0.5v -\h'-\w'-'u'\l'\w'-'u\(ul' .sp 0.5v !\h'-\w'!'u'\l'\w'!'u\(ul' binary ::= .sp 0.5v |\h'-\w'|'u'\l'\w'|'u\(ul' .sp 0.5v &\h'-\w'&'u'\l'\w'&'u\(ul' .sp 0.5v ==\h'-\w'=='u'\l'\w'=='u\(ul' .sp 0.5v !=\h'-\w'!='u'\l'\w'!='u\(ul' .sp 0.5v <\h'-\w'<'u'\l'\w'<'u\(ul' .sp 0.5v <=\h'-\w'<='u'\l'\w'<='u\(ul' .sp 0.5v >\h'-\w'>'u'\l'\w'>'u\(ul' .sp 0.5v >=\h'-\w'>='u'\l'\w'>='u\(ul' .sp 0.5v <<\h'-\w'<<'u'\l'\w'<<'u\(ul' .sp 0.5v >>\h'-\w'>>'u'\l'\w'>>'u\(ul' .sp 0.5v -\h'-\w'-'u'\l'\w'-'u\(ul' .sp 0.5v +\h'-\w'+'u'\l'\w'+'u\(ul' .sp 0.5v %\h'-\w'%'u'\l'\w'%'u\(ul' .sp 0.5v *\h'-\w'*'u'\l'\w'*'u\(ul' .sp 0.5v /\h'-\w'/'u'\l'\w'/'u\(ul' lvalue ::= .sp 0.5v name .sp 0.5v *\h'-\w'*'u'\l'\w'*'u\(ul' rvalue .sp 0.5v rvalue [\h'-\w'['u'\l'\w'['u\(ul' rvalue ]\h'-\w']'u'\l'\w']'u\(ul' .bp constant ::= .sp 0.5v {digit}\*<1\*> .sp 0.5v \(aq\h'-\w'\(aq'u'\l'\w'\(aq'u\(ul' {char}\*<1\*>\s-3\h'-\w'1'u'\s+3\*{2\*} \(aq\h'-\w'\(aq'u'\l'\w'\(aq'u\(ul' .sp 0.5v \(dq\h'-\w'\(dq'u'\l'\w'\(dq'u\(ul' {char}\*<0\*> \(dq\h'-\w'\(dq'u'\l'\w'\(dq'u\(ul' name ::= .sp 0.5v alpha {alpha-digit}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{7\*} alpha-digit ::= .sp 0.5v alpha .sp 0.5v digit .sp 0.5v \f[] .fi .RS 3 .RE .SH 4 2.2 Comments and Character Sets .LP .br .br Comments are delimited as in PL/I by /* and */. .br .br In general, B requires tokens to be separated by blanks, comments or newlines, however the compiler infers separators surrounding any of the characters \f[CR](){}[],;?:\f[R] or surrounding any maximal sequence of the characters \f[CR]+-*/<>&|!\f[R]. .br .br The character set used in B is ANSCII. .br .br The syntactic variable \(aqalpha\(aq is not defined. It represents the characters \f[CR]A\f[R] to \f[CR]Z\f[R], \f[CR]a\f[R] to \f[CR]z\f[R], \f[CR]_\f[R], and backspace. .br .br The syntactic variable \(aqdigit\(aq is not defined. It represents the characters \f[CR]0\f[R], \f[CR]1\f[R], \f[CR]2\f[R], \&... \f[CR]9\f[R]. .br .br The syntactic variable \(aqchar\(aq is not defined. It is essentially any character in the set plus the escape character \f[CR]*\f[R] followed by another character to represent characters not easily represented in the set. The following escape sequences are currently defined: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] *0 null *e end-of-file *( { *) } *t tab ** * *\(aq \(aq *\(dq \(dq *n new line \f[] .fi .RS 3 .RE .LP All keywords in the language are only recognized in lower case. Keywords are reserved. .SH 4 3.0 Rvalues and Lvalues .LP .br .br An rvalue is a binary bit pattern of a fixed length. On the PDP-11 it is 16 bits. Objects are rvalues of different kinds such as integers, labels, vectors and functions. The actual kind of object represented is called the type of the rvalue. .br .br A B expression can be evaluated to yield an rvalue, but its type is undefined until the rvalue is used in some context. It is then assumed to represent an object of the required type. For example, in the following function call .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] (b?f:g[i])(1,x>1) \f[] .fi .RS 3 .RE .LP The expression \f[CR](b?f:g[i])\f[] is evaluated to yield an rvalue which is interpreted to be of type function. Whether \f[CR]f\f[] and \f[CR]g[i]\f[] are in fact functions is not checked. Similarly, \f[CR]b\f[] is assumed to be of type truth value, \f[CR]x\f[] to be type integer \fIetc\fR. .br .br There is no check to insure that here are no type mismatches. Similarly, there are no wanted, or unwanted, type conversions. .br .br An lvalue is a bit pattern representing a storage location containing an rvalue. An lvalue is a type in B. The unary operator * can be used to interpret an rvalue as an lvalue. Thus .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] *X \f[] .fi .RS 3 .RE .LP evaluates the expression \f[CR]x\fP to yield an rvalue, which is then interpreted as an lvalue. If it is then used in an rvalue context, the application of \f[CR]*\f[R] yields the rvalue therein stored. The operator \f[CR]*\f[R] can be thought of as indirection. .br .br The unary operator \f[CR]&\f[R] can be used to interpret an lvalue as an rvalue. Thus .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] &x \f[] .fi .RS 3 .RE .LP evaluates the expression \f[CR]x\fP as an lvalue. The application of \f[CR]&\f[R] then yields the lvalue as an rvalue. The operator \f[CR]&\f[R] can therefore be thought of as the address function. .br .br The names lvalue and rvalue come from the assignment statement which requires an lvalue on the left and an rvalue on the right. .SH 4 4.0 Expression Evaluation .LP .br .br Binding of expressions (lvalues and rvalues) is in the same order as the sub-sections of this section except as noted. Thus expressions referred to as operands of \f[CR]+\f[R] (section 4.4) are expressions defined in sections 4.1 to 4.3. The binding of operators at the same level (left to right, right to left) is specified in each sub-section. .SH 4 4.1 Primary Expressions .IP "\f[B]1.\f[R]" 3 A name is an lvalue of one of three storage classes (automatic, external and internal). .RS 3 .RE .IP "\f[B]2.\f[R]" 3 A decimal constant is an rvalue. It consists of a digit between \f[CR]1\fP and \f[CR]9\fP followed by any number of digits between \f[CR]O\fP and \f[CR]9\fP. The value of the constant should not exceed the maximum value that can be stored in an object. .RS 3 .RE .IP "\f[B]4.\f[R]" 3 An octal constant is the same as a decimal constant except that it begins with a zero. It is then interpreted in base 8. Note that \f[CR]09\fP (base 8) is legal and equal to \f[CR]011\fP. A character constant is represented by \f[CR]\(aq\f[R] followed by one or two characters (possibly escaped) followed by another \f[CR]\(aq\f[R]. It has an rvalue equal to the value of the characters packed and right adjusted. .RS 3 .RE .IP "\f[B]5.\f[R]" 3 A string is any number of characters between \f[CR]\(dq\f[R] characters. The characters are packed into adjacent objects (lvalues sequential) and terminated with the character \(aq\f[CR]*e\fP\(aq. The rvalue of the string is the lvalue of the object containing the first character. See section 8.0 for library functions used to manipulate strings in a machine independent fashion. .RS 3 .RE .IP "\f[B]6.\f[R]" 3 Any expression in () parentheses is a primary expression. Parentheses are used to alter order of binding. .RS 3 .RE .IP "\f[B]7.\f[R]" 3 A vector is a primary expression followed by any expression in \f[CR][]\fP brackets. The two expressions are evaluated to rvalues, added and the result is used as an lvalue. The primary expression can be thought of as a pointer to the base of a vector, while the bracketed expression can be thought of as the offset in the vector. Since \f[CR]E1[E2\fP] is identical to \f[CR]*(E1+E2)\fP, and addition is commutative, the base of the vector and the offset in the vector can swap positions. .RS 3 .RE .IP "\f[B]8.\f[R]" 3 A function is a primary expression followed by any number of expressions in \f[CR]()\f[R] parentheses separated by commas. The expressions In parentheses are evaluated (in an unspecified order) to rvalues and assigned to the function\(aqs parameters. The primary expression is evaluated to an rvalue (assumed to be type function). The function is then called. Each call is recursive at little cost in time or space. .RS 3 .RE .IP "\f[B]\f[R]" 3 Primary expressions are bound left to right. .RS 3 .RE .SH 4 4.2 Unary Operators .LP .br .br .IP "\f[B]1.\f[R]" 3 The rvalue (or indirection) prefix unary operator \f[CR]*\f[R] is described in section 3.0. Its operand is evaluated to rvalue, and then used as an lvalue. In this manner, address arithmetic may be performed. .RS 3 .RE .IP "\f[B]2.\f[R]" 3 The lvalue (or address) prefix unary operator \f[CR]&\fP is also described in section 3.0. Note that \f[CR]&*x\fP is identically \f[CR]x\fP, but \f[CR]*&x\fP is only \f[CR]x\fP if \f[CR]x\fP is an lvalue. .RS 3 .RE .IP "\f[B]3.\f[R]" 3 The operand of the negate prefix unary operator \f[CR]-\fP is interpreted as an integer rvalue. The result is an rvalue with opposite sign. .RS 3 .RE .IP "\f[B]4.\f[R]" 3 The \f[CR]NOT\fP prefix unary operator \f[CR]!\fP takes an integer rvalue operand. The result is zero if the operand is non-zero. The result is one if the operand is zero. .RS 3 .RE .IP "\f[B]5.\f[R]" 3 The increment \f[CR]++\f[R] and decrement \f[CR]--\f[R] unary operators may be used either in prefix or postfix form. Either form requires an lvalue operand. The rvalue stored in the lvalue is either incremented or decremented by one. The result is the rvalue either before or after the operation depending on postfix or prefix notation respectively. Thus if \f[CR]x\fP currently contains the rvalue \f[CR]5\fP, then \f[CR]++x\fP and \f[CR]x++\fP both change \f[CR]x\fP to \f[CR]6\fP. The value of \f[CR]++x\fP is \f[CR]6\fP while \f[CR]x++\fP is \f[CR]5\fP. .RS 3 .RE .IP "\f[B]\f[R]" 3 Similarly, \f[CR]--x\fP and \f[CR]x--\fP store \f[CR]4\fP in \f[CR]x\fP. The former has rvalue result \f[CR]4\fP, the latter \f[CR]5\fP. .RS 3 .RE .LP .br .br Unary operators are bound right to left. Thus \f[CR]-!x+\fP+ is bound \f[CR]-(!(x++))\fP. .SH 4 4.3 Multiplicative Operators .LP .br .br The multiplicative binary operators \f[CR]*\f[R], \f[CR]/\f[R], and \f[CR]%\f[R], expect rvalue integer operands. The result is also an integer. .IP "\f[B]1.\f[R]" 3 The operator * denotes multiplication. .RS 3 .RE .IP "\f[B]2.\f[R]" 3 The operator \f[CR]/\f[R] denotes division. The result is correct if the first operand is divisible by the second. If both operands are positive, the result is truncated toward zero. Otherwise the rounding is undefined, but never greater than one. .RS 3 .RE .IP "\f[B]3.\f[R]" 3 The operator \f[CR]%\f[R] denotes modulo. If both operands are positive, the result is correct. It is undefined otherwise. .RS 3 .RE .LP .br .br The multiplicative operators bind left to right. .SH 4 4.4 Additive Operators .LP .br .br The binary operators \f[CR]+\f[R] and \f[CR]-\f[R] are add and subtract. The additive operators bind left to right. .SH 4 4.5 Shift Operators .LP .br .br The binary operators \f[CR]<<\f[R] and \f[CR]>>\f[R] are left and right shift respectively. The left rvalue operand is taken as a bit pattern. The right operand is taken as an integer shift count. The result is the bit pattern shifted by the shift count. Vacated bits are filled with zeros. The result is undefined if the shift count negative or larger than an object length. The shift operators bind left to right. .SH 4 4.6 Relational Operators .LP .br .br The relational operators \f[CR]<\f[R] (less than), \f[CR]<=\f[R] (less than or equal to), \f[CR]>\f[R] (greater than), and \f[CR]>=\f[R] ( greater than or equal to) take integer rvalue operands. The result is one if the operands are in the given relation to one another. The result is zero otherwise. .SH 4 4.7 Equality Operators .LP .br .br The equality operators \f[CR]==\f[R] (equal to) and \f[CR]!=\f[R] (not equal to) perform similarly to the relation operators. .SH 4 4.8 \f[CR]AND\fP operator .LP .br .br The \f[CR]AND\fP operator \f[CR]&\f[R] takes operands as bit patterns. The result is the bit pattern that is the bit-wise \f[CR]AND\fP of the operands. The operator binds and evaluates left to right. .SH 4 4.9 The \f[CR]OR\fP operator .LP .br .br The \f[CR]OR\fP operator \f[CR]|\fP performs exactly as \f[CR]AND\fP, but the result is the bit-wise inclusive \f[CR]OR\fP of the operands. The \f[CR]OR\fP operator also binds and evaluates left to right. .SH 4 4.10 Conditional Expression .LP .br .br Three rvalue expressions separated by \f[CR]?\f[R] and \f[CR]:\f[R] form a conditional expression. The first expression (to the left of the \f[CR]?\fP) is evaluated. If the result is non-zero, the second expression is evaluated and the third ignored. If the value is zero, the second expression is ignored and the third is evaluated. The result is either the evaluation of the second or third expression. .br .br Binding is right to left. Thus \f[CR]a?b:c?d:e \f[] is \f[CR]a?b:(c?d:e)\f[]. .SH 4 4.11 Assignment Operators .LP .br .br There are 16 assignment operators in B. All have the form .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] lvalue op rvalue \f[] .fi .RS 3 .RE .LP The assignment operator \f[CR]=\fP merely evaluates the rvalue and stores the result in the lvalue. The assignment operators \f[CR]=|\f[R], \f[CR]=&\f[R], \f[CR]===\f[R], \f[CR]=|=\f[R], \f[CR]=<\f[R], \f[CR]=<=\f[R], \f[CR]=>\f[R], \f[CR]=>=\f[R], \f[CR]=<<\f[R], \f[CR]=>>\f[R], \f[CR]=+\f[R], \f[CR]=-\f[R], \f[CR]=%\f[R], \f[CR]=*\f[R], and \f[CR]=/\f[R] perform a binary operation (see sections 4.3 to 4.9) between the rvalue stored in the assignment\(aqs lvalue and the assignment\(aqs rvalue. The result is then stored in the lvalue. The expression \f[CR]x=*10\fP is identical to \f[CR]x=x*10\fP. Note that this is not \f[CR]x= *10\fP. The result of an assignment is the rvalue. Assignments bind right to left, thus \f[CR]x=y=0\fP assigns zero to \f[CR]y\fP, then \f[CR]x\fP, and returns the rvalue zero. .SH 4 5.0 Statements .LP .br .br Statements define program execution. Each statement is executed by the computer in sequence. There are, of course, statements to conditionally or unconditionally alter normal sequencing. .SH 4 5.1 Compound Statement .LP .br .br A sequence of statements in \f[CR]{}\f[R] braces is syntactically a single statement. This mechanism is provided so that where a single statement is expected, any number of statements can be placed. .SH 4 5.2 Conditional Statement .LP .br .br A conditional statement has two forms. The first: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] if\h'-\w'if'u'\l'\w'if'u\(ul'(rvalue) statement\*<1\*> \f[] .fi .RS 3 .RE .LP evaluates the rvalue and executes statement\*<1\*>, if the rvalue is non-zero. If the rvalue is zero, statement\*<1\*> is skipped. The second form: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] if\h'-\w'if'u'\l'\w'if'u\(ul'(rvalue) statement\*<1\*> else\h'-\w'else'u'\l'\w'else'u\(ul' statement\*<2\*> \f[] .fi .RS 3 .RE .LP is defined as follows in terms of the first form: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] if\h'-\w'if'u'\l'\w'if'u\(ul'(x=(rvalue)) statement\*<1\*> if\h'-\w'if'u'\l'\w'if'u\(ul'(!x) statement\*<2\*> \f[] .fi .RS 3 .RE .LP Thus, only one of the two statements is executed, depending on the value of rvalue. In the above example, \f[CR]x\fP is not a real variable, but just a demonstration aid. .SH 4 5.3 While Statement .LP .br .br The while statement has the form: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] while\h'-\w'while'u'\l'\w'while'u\(ul'(rvalue) statement \f[] .fi .RS 3 .RE .LP The execution is described in terms of the conditional and goto statements as follows: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] x: if\h'-\w'if'u'\l'\w'if'u\(ul'(rvalue) { statement goto\h'-\w'goto'u'\l'\w'goto'u\(ul' x; ] \f[] .fi .RS 3 .RE .LP Thus the statement is executed repeatedly while the rvalue is non-zero. Again, \f[CR]x\fP is a demonstration aid. .SH 4 5.4 Switch Statement .LP .br .br The switch statement is the most complicated statement in B. The switch has the form: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] switch\h'-\w'switch'u'\l'\w'switch'u\(ul' rvalue statement\*<1\*> \f[] .fi .RS 3 .RE .LP Virtually always, statement\*<1\*> above is a compound statement. Each statement in statement\*<1\*> may be preceded by a case as follows: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] case\h'-\w'case'u'\l'\w'case'u\(ul' constant:\h'-\w':'u'\l'\w':'u\(ul' \f[] .fi .RS 3 .RE .LP During execution, the rvalue is evaluated and compared to each case constant in undefined order. If a case constant is equal to the evaluated rvalue, control is passed to the statement following the case. If the rvalue matches none of the cases, statement\*<1\*> is skipped. .SH 4 5.5 Goto Statement .LP .br .br The goto statement is as follows: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] goto\h'-\w'goto'u'\l'\w'goto'u\(ul' rvalue ;\h'-\w';'u'\l'\w';'u\(ul' \f[] .fi .RS 3 .RE .LP The rvalue is expected to be of type label. Control is then passed to the corresponding label. Goto\(aqs cannot be executed to labels outside the currently executing function level. .SH 4 5.6 Return Statement .LP .br .br The return statement is used in a function to return control to the caller of a function. The first form simply returns control. .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] return\h'-\w'return'u'\l'\w'return'u\(ul' ;\h'-\w';'u'\l'\w';'u\(ul' \f[] .fi .RS 3 .RE .LP The second form returns an rvalue for the execution of the function. .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] return\h'-\w'return'u'\l'\w'return'u\(ul' (\h'-\w'('u'\l'\w'('u\(ul' rvalue )\h'-\w')'u'\l'\w')'u\(ul' ;\h'-\w';'u'\l'\w';'u\(ul' \f[] .fi .RS 3 .RE .LP The caller of the function need not use the returned rvalue. .SH 4 5.7 Rvalue Statement .LP .br .br Any rvalue followed by a semicolon is a statement. The two most common rvalue statements are assignment and function call. .SH 4 5.8 Null Statement .LP .br .br A semicolon is a null statement causing no execution. It is used mainly to carry a label after the last executable statement in a compound statement. It sometimes has use to supply a null body to a while statement. .SH 4 6.0 Declarations .LP .br .br Declarations in B specify storage class of variables. Such declarations also, in some circumstances, specify initialization. .br .br There are three storage classes in B. Automatic storage is allocated for each function invocation. External storage is allocated before execution and is available to any and all functions. Internal storage is local to a function and is available only to that function, but is available to all invocations of that function. .SH 4 6.1 External Declaration .LP .br .br The external declaration has the form: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] extrn\h'-\w'extrn'u'\l'\w'extrn'u\(ul' name\*<1\*> ,\h'-\w','u'\l'\w','u\(ul' name\*<2\*> ... ;\h'-\w';'u'\l'\w';'u\(ul' \f[] .fi .RS 3 .RE .LP The external declaration specifies that each of the named variables is of the external storage class. The declaration must occur before the first use of each of the variables. Each of the variables must also be externally defined. .SH 4 6.2 Automatic Declaration .LP .br .br The automatic declaration also constitutes a definition: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] auto\h'-\w'auto'u'\l'\w'auto'u\(ul' name\*<1\*> {constant}\*<0\*>}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} ,\h'-\w','u'\l'\w','u\(ul' name\*<2\*> {constant]\*<0\*>}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} ... ;\h'-\w';'u'\l'\w';'u\(ul' \f[] .fi .RS 3 .RE .LP In absence of the constant, the automatic declaration defines the variable to be of class automatic. At the same time, storage is allocated for the variable. When an automatic declaration is followed by a constant, the automatic variable is also initialized to the base of an automatic vector of the size of the constant. The actual subscripts used to reference the vector range from zero to the value of the constant less one. .SH 4 6.3 Internal Declaration .LP .br .br The first reference to a variable not declared as external or automatic constitutes an internal declaration, All internal variables not defined as labels are flagged as undefined within a function. Labels are defined and initialized as follows: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] name :\h'-\w':'u'\l'\w':'u\(ul' \f[] .fi .RS 3 .RE .SH 4 7.0 External Definitions .LP .br .br A complete B program consists of a series of external definitions. Execution is started by the hidden sequence .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] main(); exit(); \f[] .fi .RS 3 .RE .LP Thus, it is expected that one of the external definitions is a function definition of main. (Exit is a predefined library function. See section 8.0) .SH 4 7.1 Simple Definition .LP .br .br The simple external definition allocates an external object and optionally initializes it: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] name {ival ,\h'-\w','u'\l'\w','u\(ul' ival ...}\*<0\*> ;\h'-\w''u'\l'\w''u\(ul' \f[] .fi .RS 3 .RE .LP If the external object is defined with no initialization, it is initialized with zero. A single initialization with a constant initializes the external with the value of the constant. Initialization with a name initializes the external to the address of that name. Many such initializations may be accessed as a vector based at &name. .SH 4 7.2 Vector Definitions .LP .br .br An external vector definition has the following form: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] name [\h'-\w'['u'\l'\w'['u\(ul' {constant}\*<0\*>}\*<0\*>\s-3\h'-\w'0'u'\s+3\*{1\*} ]\h'-\w']'u'\l'\w']'u\(ul' {ival ,\h'-\w','u'\l'\w','u\(ul' ival ...}\*<0\*> ;\h'-\w';'u'\l'\w';'u\(ul' \f[] .fi .RS 3 .RE .LP The name is initialized with the lvalue of the base of an external vector. If the vector size is missing, zero is assumed. In either case, the vector is initialized with the list of initial values. Each initial value is either a constant or a name. A constant initial value initializes the vector element to the value of the constant. The name initializes the element to the address of the name. The actual size is the maximum of the given size and the number of initial values. .SH 4 7.3 Function Definitions .LP .br .br Function definitions have the following form: .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] name (\h'-\w'('u'\l'\w'('u\(ul' arguments )\h'-\w')'u'\l'\w')'u\(ul' statement \f[] .fi .RS 3 .RE .LP The name is initialized to the rvalue of the function. The arguments consist of a list of names separated by commas. Each name defines an automatic lvalue that is assigned the rvalue of the corresponding function call actual parameters. The statement (often compound) defines the execution of the function when invoked. .SH 4 8.0 Library Functions .LP .br .br There is a library of B functions maintained in the file /f[CR]/etc/libb.a\fP. The following is a list of those functions currently in the library. See section II of [4] for complete descriptions of the functions marked with an *. .IP "\f[B]c = char(string, i);\f[R]" 3 .br The \f[CR]i\fP-th character of the \f[CR]string\fP is returned. .RS 3 .RE .IP "\f[B]error = chdir(string) ;\f[R]" 3 .br The path name represented by the \f[CR]string\fP becomes the current directory. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]error = chmod(string, mode);\f[R]" 3 .br The file specified by the \f[CR]string\fP has its \f[CR]mode\fP changed to the mode argument. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]error = chown(string, owner);\f[R]" 3 .br The file specified by the \f[CR]string\fP has its owner changed to the \f[CR]owner\fP argument. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]error = close(file) ;\f[R]" 3 .br The open file specified by the file argument is closed. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]file = creat(string, mode);\f[R]" 3 .br The file specified by the \f[CR]string\fP is either truncated or created in the \f[CR]mode\fP specified depending on its prior existence. In both cases, the file is opened for writing and a file descriptor is returned. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]ctime(time, date);\f[R]" 3 .br The system time (60-ths of a second) represented in the two-word vector \f[CR]time\fP is converted to a 16-character \f[CR]date\fP in the 8-word vector date. The converted date has the following format: \(dq\f[CR]Mmm dd hh:mm:ss\fP\(dq. .RS 3 .RE .IP "\f[B]execl(string, arg0, arg1, ..., 0);\f[R]" 3 .br The current process is replaced by the execution of the file specified by \f[CR]string\fP. The \f[CR]arg\fP-i strings are passed as arguments. A return indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]execv(string, argv, count);\f[R]" 3 .br The current process is replaced by the execution of the file specified by \f[CR]string\fP. The vector of \f[CR]argv\fP strings of length \f[CR]count\fP are passed as arguments. A return indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]exit( ) ;\f[R]" 3 .br The current process is terminated. (*) .RS 3 .RE .IP "\f[B]error = fork( ) ;\f[R]" 3 .br The current process splits into two. The child process is returned a zero. The parent process is returned the process ID of the child. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]error = fstat(file, status);\f[R]" 3 .br The \f[CR]i-node\fP of the open file designated by file is put in the 20-word vector status. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]char = getchar( ) ;\f[R]" 3 The next character form the standard input file is returned. The character \(aq*e\(aq is returned for an \f[CR]end-of-file\fP. .RS 3 .RE .IP "\f[B]id = getuid();\f[R]" 3 .br The user-ID of the current process is returned. (*) .RS 3 .RE .IP "\f[B]error = gtty(file, ttystat);\f[R]" 3 .br The teletype modes of the open file designated by file is returned in the 3-word vector ttstat. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]lchar(string, i, char);\f[R]" 3 .br The character char is stored in the \f[CR]i\fP-th character of the \f[CR]string\fP. .RS 3 .RE .IP "\f[B]error = link(string1, string2);\f[R]" 3 .br The pathname specified by \f[CR]string2\fP is created such that it is a link to the existing file specified by \f[CR]string1\fP. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]error = mkdir(string, mode);\f[R]" 3 .br The directory specified by the \f[CR]string\fP is made to exist with the specified access mode. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]file = open(string, mode);\f[R]" 3 .br The file specified by the \f[CR]string\fP is opened for reading if mode is zero, for writing if mode is not zero. The open file designator is returned. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]printf(format, argl, ...);\f[R]" 3 .br See section 9.3 below. .RS 3 .RE .IP "\f[B]printn(number, base);\f[R]" 3 .br See section 9.1 below. .RS 3 .RE .IP "\f[B]putchar(char) ;\f[R]" 3 .br The character char is written on the standard output file. .RS 3 .RE .IP "\f[B]nread = read(file, buffer, count);\f[R]" 3 .br The \f[CR]count\fP number bytes are read into the vector \f[CR]buffer\fP from the open file designated by \f[CR]file\fP. The actual number of bytes read are returned. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]error = seek(file, offset, pointer);\f[R]" 3 .br The I/O pointer on the open file designated by \f[CR]file\fP is set to the value of the designated \f[CR]pointer\fP plus the \f[CR]offset\fP. A \f[CR]pointer\fP of zero designates the beginning of the file. A \f[CR]pointer\fP of one designates the current I/O pointer. A \f[CR]pointer\fP of two designates the end of the file. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]error = setuid(id);\f[R]" 3 .br The user-ID of the current process is set to \f[CR]id\fP. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]error = stat(string, status);\f[R]" 3 .br The \f[CR]i-node\fP of the file specified by the \f[CR]string\fP is put in the 20-word vector \f[CR]status\fP. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]error = stty(file, ttystat);\f[R]" 3 .br The teletype modes of the open file designated by \f[CR]file\fP is set from the 3-word vector \f[CR]ttystat\fP. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]time(timev);\f[R]" 3 .br The current system time is returned in the 2-word vector \f[CR]timev\fP. (*) .RS 3 .RE .IP "\f[B]error = unlink(string);\f[R]" 3 .br The link specified by the \f[CR]string\fP is removed. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]error = wait( );\f[R]" 3 .br The current process is suspended until one of its child processes terminates. At that time, the child\(aqs process-ID is returned. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .IP "\f[B]nwrite = write(file, buffer, count);\f[R]" 3 .br The \f[CR]count\fP number of bytes are written out of the vector \f[CR]buffer\fP on the open file designated by \f[CR]file\fP. The actual number of bytes written are returned. A negative number returned indicates an \f[CR]error\fP. (*) .RS 3 .RE .LP .br .br Besides the functions available from the library, there is a predefined external vector named argv included with every program. The size of \f[CR]argv\fP is \f[CR]argv[0]+1\fP. The elements \f[CR]argv[1]...argv[argv[0]]\fP are the parameter strings as passed by the system in the execution of the current process. (See shell in II of [4]) .SH 4 9.0 Examples .LP .br .br The examples appear exactly as given to B. .SH 4 9.1 Print a Non-Negative Number .LP .br .br .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] /* The following function will print a non-negative number, n, to the base b, where 2<=b<=10, This routine uses the fact that in the ANSCII character set, the digits O to 9 have sequential code values. */ printn(n,b) { extrn putchar; auto a; if(a=n/b) /* assignment, not test for equality */ printn(a, b); /* recursive */ putchar(n%b + \(aq0\(aq); } \f[] .fi .RS 3 .RE .SH 4 9.2 Calculate the Constant \fIe\fP-2 .LP .br .br .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] /* The following program will calculate the constant e-2 to about 4000 decimal digits, and print it 50 characters to the line in groups of 5 characters. The method is simple output conversion of the expansion 1/2! + 1/3! + ... = .111.... where the bases of the digits are 2, 3, 4, . . . */ main() { extrn putchar, n, v; auto i, c, col, a; i = col = 0; while(i= b111 => b120 =/ \f[] .fi .RS 3 .RE .RE .IP "\f[B]4.\f[R]" 3 Initialization of external variables with addresses of other externals is not possible due to a loader deficiency. .RS 3 .RE .IP "\f[B]5.\f[R]" 3 Since externals are implemented as globals with names preceded by \(aq\f[CR].\fP\(aq, the external names \(aq\f[CR]byte\fP\(aq, \(aq\f[CR]endif\fP\(aq, \(aq\f[CR]even\fP\(aq and \(aq\f[CR]globl\fP\(aq conflict with assembler pseudooperations and should be avoided. .RS 3 .RE .SH 4 14.0 Diagnostics .LP .br .br Diagnostics consist of two letters, an optional name, and a source line number. Due to the free format of the source, the number might be high. The following is a list of the diagnostics. .IP "\f[B]\f[R]" 3 .IP .nf \f[CR] error name meaninq $) -- {} imbalance () -- () imbalance */ -- /* */ imbalance [] -- [] imbalance >c -- case table overflow (fatal) >e -- expression stack overflow (fatal) >i -- label table overflow (fatal) >s -- symbol table overflow (fatal) ex -- expression syntax lv -- rvalue where lvalue expected rd name name redeclaration sx keyword statement syntax un name undefined name xx -- external syntax \f[] .fi .RS 3 .RE .LP [ signature line ] .SH 4 References .LP .br .br .IP "\f[B]1.\f[R]" 3 Richards, M. .UL "The BCPL Reference Manual." Multics repository M0099. .RS 3 .RE .IP "\f[B]2.\f[R]" 3 Canaday, R.H. and Ritchie, D.M. .UL "Bell Laboratories BCPL." MM 69-1371/1373-7/12. .RS 3 .RE .IP "\f[B]3.\f[R]" 3 Ritchie, D.M. .UL "The UNIX Time Sharing System." MM 71-1273-4. .RS 3 .RE .IP "\f[B]4.\f[R]" 3 Thompson, K. and Ritchie, D.M. .UL "UNIX Programmer\(aqs Manual. Available by arrangement." .RS 3 .RE .LP .br .br [\f[I]Two of these references\(emthe one to Richards\(aqs early BCPL manual, and to the first edition of the UNIX Programmer\(aqs manual, are available under my main home page. The ultimate content of \(dqThe UNIX Time Sharing System evolved into the C.ACM paper, there (in further revised form) into the BSTJ paper available there as well.\f[R]] .br .br Copyright \(co 1996 Lucent Technologies Inc. All rights reserved.