shithub: purgatorio

ref: b5bc6572b0a82c52ab04bcf25e1ed76700deb246
dir: /man/1/sh-expr/

View raw version
.TH SH-EXPR 1
.SH NAME
expr, ntest, mpexpr \- shell module for simple arithmetic.
.SH SYNOPSIS
.B load expr
OR
.B load mpexpr

.B ${expr
[
-r
.I radix
]
[
.I arg...
]
.B }
.br
.B ntest
.I num
.br
.SH DESCRIPTION
.I Expr
and
.I mpexpr
are loadable modules for
.IR sh (1)
that provide support for integer arithmetic.
.I Expr
uses 64-bit signed integers;
.I mpexpr
uses arbitrary-precision signed integers.
They each provide the same interface:
a command
.IR ntest ,
which performs a simple boolean test
on its integer argument, and the
substitution operator
.IR expr ,
which takes an expression in Reverse Polish
notation, and yields its result.
.PP
.I Ntest
returns true if its argument
.I num
is non-zero,
and false otherwise.
.PP
.I Expr
evaluates each
.I arg
in turn; if it is an integer it gets pushed onto
the stack; otherwise it should name
one of the operators below, whereupon
the appropriate number of operands are
popped off the stack, evaluated as arguments
to the operator, and the result pushed back onto
the stack. Arguments are passed to the operator
first-pushed first, so, for instance,
.B ${expr 2 1 -}
yields 1, not -1.
Alternative names are given for some operators;
this is to avoid the necessity of quoting operators
that contain
.IR sh (1)
metacharacters. Integers are given in the same form acceptable
to Limbo. The relational operators yield either
1 (true) or 0 (false). If the
.B -r
option is given,
.I radix
specifies an output base for printed numbers.
It may be from 2 to 36;
.I mpexpr
also allows 64 to specify base64 notation.
Numbers are printed in a form suitable for re-interpretation
by
.IR expr .
.PP
When all its arguments have been evaluated,
.B expr
yields all the values remaining on its stack, first pushed
first. Note that bitwise operators treat their operands as if they
were stored in two's complement form. The operators supported by expr are as follows (the number
of operands required in is given parentheses).
.TP 15
.BR + \ (2)
Addition
.TP
.BR - \ (2)
Subtraction
.TP
.BR x ,\  * \ (2)
Multiplication
.TP
.BR / \ (2)
Division. Division by zero raises a
.B divide by zero
exception.
.TP
.BR % \ (2)
Modulus. A zero modulus will cause a
.B divide by zero
exception to be raised.
.TP
.BR and \ (2)
Bitwise-and.
.TP
.BR or \ (2)
Bitwise-or.
.TP
.BR xor \ (2)
Bitwise-xor.
.TP
.BR ~ \ (1)
Bitwise-complement..
.TP
.BR _ \ (1)
Unary minus.
.TP
.BR << ,\  shl \ (2)
Shift left.
.TP
.BR >> ,\  shr \ (2)
Shift right.
.TP
.BR = ", " == ", " eq " (2)"
Equality.
.TP
.BR != ", " neq " (2)"
Inequality.
.TP
.BR > ", " gt " (2)"
Greater than.
.TP
.BR < ", " lt " (2)"
Less than.
.TP
.BR <= ", " le " (2)"
Less than or equal to.
.TP
.BR >= ", " ge " (2)"
Greater than or equal to.
.TP
.BR ! ", " not " (1)"
Logical negation.
.TP 
.BI rep \ \f1(\fPn\f1)\fP
.B Rep
repeats the last operation (which must
have been a two-operand operation other
than
.BR seq )
until the values in the stack are exhausted.
.TP
.BR seq \ (2)
.B Seq
pushes on the stack a sequence of numbers ranging
numerically from its first argument up to and including
its second argument. If its second argument is
less than its first, the sequence will descend.
.TP
.BR rand \ (1)
(\fImpexpr\fP only). Push a secure random number;
the argument value gives the size of the number, in bits.
.TP
.BR bits \ (1)
(\fImpexpr\fP only). Push the size, in bits, of the argument.
.TP
.BR expmod ", " invert " (2)"
(\fImpexpr\fP only). See
.IR keyring-ipint (2).
.TP
.BR exp ", " xx ", " **
(\fImpexpr\fP only). Exponentiation.
.SH SOURCE
.B /appl/cmd/sh/expr.b
.SH SEE ALSO
.IR sh (1),
.IR sh-std (1),
.IR sh-tk (1),
.IR keyring-ipint (2)
.SH BUGS
Postfix notation can be confusing.
Any operators that contain shell metacharacters (e.g. ``*'', ``>'')
must be quoted to avoid interpretation by the shell.
Base64 notation can contain # characters, which need
quoting to avoid interpretation by the shell.