It happens sometimes in scripts that you need to perform a mathematical operation. And bash itself and other standard commands can be used as a quick programmer's calculator and number base converter.
Bash works only with integers, its data type is signed int. My 64bit Fedora installation, has a range from –263 to 263–1. Numbers are stored in two's complement format, so in hexadecimal notation it is from 0x8000000000000000 to 0x7fffffffffffffff, so beware of number overflow.
You can use the following syntax to perform math operations:
In the first case, an expression is evaluated and interpreted as a command or a parameter. In the second, an expression is evaluated and a return value $? is set to false (1) if its evaluated value is zero and or true (0) if it is non-zero.
Examples:
# echo $[1+1] 2 # $[2+2] bash: 4: command not found... # ((0*1024)) ; echo $? 1 # ((2+2)) ; echo $? 0 # echo $((355/113)) $[ 355 % 133 ] 3 89 # i=10; ((i+=20, i=i-31)); echo $i -1
I like to use a syntax with square brackets instead of double parenthesis, because it is shorter although the square bracket syntax is not mentioned in documentation, but I've never had a problem with doing so in any bash versions.
We can use variables with or without a prefix dollar symbol, so the syntax $((x)) and $(($x)) will return the same value if $x contains a valid numeric value.
Arithmetic operators are the same as in the C language and its priorities are same. The only exception is in the power operator ** which comes from Fortran, but you might also know it from other programming languages. The operators are listed in order of descending priority , starting with the operator with the highest priority.
id++ id-- | postincrementation, postdecrementation |
++id --id | preincrementation, predecrementation |
- + | unary minus and plus |
! ~ | logic and bit negation |
** | power, a**b=ab |
* / % | multiply, division, remainder after division |
+ - | addition, subtraction |
<< >> | left and right bitwise shift |
<= >= < > | comparison |
== != | equality and inequality |
& | bitwise AND |
^ | bitwise XOR |
| | bitwise OR |
&& | logical AND |
|| | logical OR |
expr?expr:expr | conditional operátor |
= *= /= %= += -= <<= >>= &= ^= |= | assignment |
expr1 , expr2 | comma |
Few examples:
# echo $((1<<4 + 1<<5))
1024
# a=1; b=2; echo $[++a+++b]
5
# let "i = 2 * 3 ** 5"; echo $i
486
Bitwise shifts are often used for multiplication or division by the power of 2 but its priority is lower than that for addition, I also write it in this sneaky form so it is looks like the wanted evaluation/ result is 48, but that's wrong. Please note that an expression that accompanies a command must be written in quotes, because the symbol * can be replaced by shell to list all files in the current directory and other unwanted behaviours.
Bash uses decimal numbers in default, but in mathematical regimen you can use also octal and hexadecimal numbers in the usual way. The octal number base uses the prefix zero and hexadecimal uses 0x. Other number bases need to be written with a base before the # symbol. It is possible go up to base 32.
# echo $[0xffff] 65535 # echo $[2#101010101010] 2730 # echo $echo $[018] bash: 018: value too great for base (error token is "018") # echo $[0777] 511
It is possible to convert any number to decimal thanks to this syntax, but a conversion in the opposite direction is much more difficult and we need to use external utilities. The first of them is printf from the C standard library. Its control sequences uses %o for octal numbers and %x for decimal. And with %#o or %#x it will also print a number base prefix.
# printf "%#x\n" 65535 0xffff # printf "%o\n" $[2#101111111] 577
If we need any conversion with a base from 2 to 16 we need to use bc, but note that it uses capitals as a digit!
# echo "obase=2; 255" | bc 11111111 # echo "ibase=16; FFFF" | bc 65535
For binary conversion perl users can use the following one-liner:
# perl -e 'printf "%#b\n", 255' 0b11111111
Greek small letter alpha (α) is represented in UTF-8 as 0xCE 0xB1 and in UTF-16 as 0x03B1, with these values you can work in shell in the following way:
Use hexdump for conversion of UTF-8 characters to hexadecimal:
# echo -n "α" | hexdump -C 00000000 ce b1 |..| 00000002
The echo's option -n disables the print of line feed char, so its value 0x0A is not found in the output.
There is a built-in variable $RANDOM, it generates integers in range 0 to 32767. If you store a value into RANDOM it will be used as initial seed. I can't recall using it in a serious script, but you can have a lot of fun with it, for example:
P=(' ' █ ░ ▒ ▓);while :;do printf "\e[$[RANDOM%LINES+1];$[RANDOM%COLUMNS+1]f${P[$RANDOM%5]}";done
It is possible to generate numbers in lower ranges by using the remainder (modulo operation). For example, for the range 0 to 99 you can get its remainder after division by 100: $((RANDOM % 100)), etc.
It is possible to convert between decimal floats and its IEEE 754 representation, it is useful for the exploration of memory or binary file hexdumps with floating point numbers stored.
IEEE 754-1985 uses three precisions: single (32bit.), double (64bit.) and extended (80bit.), it is float, double and long double in C language.
# gdb --batch -ex "print/x (float *) ( (float) 1234.567 )" $1 = 0x449a5224 # gdb --batch -ex "print/f (float *) 0x449a5224" $1 = 1234.56689
I was unable to use gdb for conversion of other types so it was necessary to use other tools. For example in python you can use:
# python -c 'import struct; print "%#x" % struct.unpack("I", struct.pack("f", 1234.567))'
0x449a5225
# python -c 'import struct; print "%#x" % struct.unpack("Q", struct.pack("d", 1234.567))'
0x40934a449ba5e354
# python -c 'import struct; print "%f" % struct.unpack("d", struct.pack("Q", 0x4094b2449ba5e354))'
1324.567000
# python -c 'import struct; print "%f" % struct.unpack("f", struct.pack("I", 0x449a5225 ))'
1234.566895
Structures are used in both python and perl for easy conversion of binary data. It can pack data in one format and unpack in another: f float, d double, Q unsigned long int (8 Bytes), I unsigned int (4 Bytes).
The arbitrary precision calculator bc is used for computation with decimal point. The bc is sometimes just a converter of syntax to reverse or polish notation and the computation is done by dc, but sometimes in GNU/Linux the bc runs the calculation itself. The bc is an interpreter of a language similar to C, so it can be used for a lot of things. The number of digits after a decimal point can be set by variable scale. CLI option bc -l imports a library of mathematical functions: s(x) sin(x), c(x) cos(x), a(x)arctan(x), where x is in radians; l(x) natural logarithms ln(x), e(x) exp(x), j(n,x) Bessel funkce of integer order n of x.
# echo "scale=10; 355 / 113" | bc 3.1415929203 # echo "scale=10; (1+1/10000)^10000" | bc 2.7181459268 # echo "3.5 * 785.498" | bc 2749.243 #bc <<< "2340 * 1.22"2854.80 #bc -l <<< "scale=10; 4*a(1)"3.1415926532