[bash] How do you echo a 4-digit Unicode character in Bash?

I'd like to add the Unicode skull and crossbones to my shell prompt (specifically the 'SKULL AND CROSSBONES' (U+2620)), but I can't figure out the magic incantation to make echo spit it, or any other, 4-digit Unicode character. Two-digit one's are easy. For example, echo -e "\x55", .

In addition to the answers below it should be noted that, obviously, your terminal needs to support Unicode for the output to be what you expect. gnome-terminal does a good job of this, but it isn't necessarily turned on by default.

On macOS's Terminal app Go to Preferences-> Encodings and choose Unicode (UTF-8).

I'm using this:

$ echo -e '\u2620'
?

This is pretty easier than searching a hex representation... I'm using this in my shell scripts. That works on gnome-term and urxvt AFAIK.

If you don't mind a Perl one-liner:

$ perl -CS -E 'say "\x{2620}"'
?

-CS enables UTF-8 decoding on input and UTF-8 encoding on output. -E evaluates the next argument as Perl, with modern features like say enabled. If you don't want a newline at the end, use print instead of say.

Just put "?" in your shell script. In the correct locale and on a Unicode-enabled console it'll print just fine:

$ echo ?
?
$

An ugly "workaround" would be to output the UTF-8 sequence, but that also depends on the encoding used:

$ echo -e '\xE2\x98\xA0'
?
$

% echo -e '\u2620'     # \u takes four hexadecimal digits
?
% echo -e '\U0001f602' # \U takes eight hexadecimal digits

This works in Zsh (I've checked version 4.3) and in Bash 4.2 or newer.

In bash to print a Unicode character to output use \x,\u or \U (first for 2 digit hex, second for 4 digit hex, third for any length)

echo -e '\U1f602'

I you want to assign it to a variable use $'...' syntax

x=$'\U1f602'
echo $x

Quick one-liner to convert UTF-8 characters into their 3-byte format:

var="$(echo -n '?' | od -An -tx1)"; printf '\\x%s' ${var^^}; echo

Easy with a Python2/3 one-liner:

$ python -c 'print u"\u2620"'    # python2
$ python3 -c 'print(u"\u2620")'  # python3

Results in:

?

So long as your text-editors can cope with Unicode (presumably encoded in UTF-8) you can enter the Unicode code-point directly.

For instance, in the Vim text-editor you would enter insert mode and press Ctrl + V + U and then the code-point number as a 4-digit hexadecimal number (pad with zeros if necessary). So you would type Ctrl + V + U 2 6 2 0. See: What is the easiest way to insert Unicode characters into a document?

At a terminal running Bash you would type CTRL+SHIFT+U and type in the hexadecimal code-point of the character you want. During input your cursor should show an underlined u. The first non-digit you type ends input, and renders the character. So you could be able to print U+2620 in Bash using the following:

echo CTRL+SHIFT+U2620ENTERENTER

(The first enter ends Unicode input, and the second runs the echo command.)

Credit: Ask Ubuntu SE

Here's a fully internal Bash implementation, no forking, unlimited size of Unicode characters.

fast_chr() {
    local __octal
    local __char
    printf -v __octal '%03o' $1
    printf -v __char \\$__octal
    REPLY=$__char
}

function unichr {
    local c=$1    # Ordinal of char
    local l=0    # Byte ctr
    local o=63    # Ceiling
    local p=128    # Accum. bits
    local s=''    # Output string

    (( c < 0x80 )) && { fast_chr "$c"; echo -n "$REPLY"; return; }

    while (( c > o )); do
        fast_chr $(( t = 0x80 | c & 0x3f ))
        s="$REPLY$s"
        (( c >>= 6, l++, p += o+1, o>>=1 ))
    done

    fast_chr $(( t = p | c ))
    echo -n "$REPLY$s"
}

## test harness
for (( i=0x2500; i<0x2600; i++ )); do
    unichr $i
done

Output was:

-?¦?????????+???
+???+???+???+???
????¦???????-???
????-???????+???
????????????????
-¦++++++++++++¦¦
¦¦¦¦------+++???
????????????????
¯???_???¦???¦???
¦¦¦¦????????????
¦???????????????
????????????????
????????????????
????????????????
????????????????
????????????????

Any of these three commands will print the character you want in a console, provided the console do accept UTF-8 characters (most current ones do):

echo -e "SKULL AND CROSSBONES (U+2620) \U02620"
echo $'SKULL AND CROSSBONES (U+2620) \U02620'
printf "%b" "SKULL AND CROSSBONES (U+2620) \U02620\n"

SKULL AND CROSSBONES (U+2620) ?

After, you could copy and paste the actual glyph (image, character) to any (UTF-8 enabled) text editor.

If you need to see how such Unicode Code Point is encoded in UTF-8, use xxd (much better hex viewer than od):

echo $'(U+2620) \U02620' | xxd
0000000: 2855 2b32 3632 3029 20e2 98a0 0a         (U+2620) ....

That means that the UTF8 encoding is: e2 98 a0

Or, in HEX to avoid errors: 0xE2 0x98 0xA0. That is, the values between the space (HEX 20) and the Line-Feed (Hex 0A).

If you want a deep dive into converting numbers to chars: look here to see an article from Greg's wiki (BashFAQ) about ASCII encoding in Bash!

If hex value of unicode character is known

H="2620"
printf "%b" "\u$H"

If the decimal value of a unicode character is known

declare -i U=2*4096+6*256+2*16
printf -vH "%x" $U              # convert to hex
printf "%b" "\u$H"

You may need to encode the code point as octal in order for prompt expansion to correctly decode it.

U+2620 encoded as UTF-8 is E2 98 A0.

So in Bash,

export PS1="\342\230\240"

will make your shell prompt into skull and bones.

Based on Stack Overflow questions Unix cut, remove first token and https://stackoverflow.com/a/15903654/781312:

(octal=$(echo -n ? | od -t o1 | head -1 | cut -d' ' -f2- | sed -e 's#\([0-9]\+\) *#\\0\1#g')
echo Octal representation is following $octal
echo -e "$octal")

Output is the following.

Octal representation is following \0342\0230\0240
?

Here is a list of all unicode emoji's available:

https://en.wikipedia.org/wiki/Emoji#Unicode_blocks

Example:

echo -e "\U1F304"

For get the ASCII value of this character use hexdump

echo -e "" | hexdump -C

00000000  f0 9f 8c 84 0a                                    |.....|
00000005

And then use the values informed in hex format

echo -e "\xF0\x9F\x8C\x84\x0A"

The printf builtin (just as the coreutils' printf) knows the \u escape sequence which accepts 4-digit Unicode characters:

   \uHHHH Unicode (ISO/IEC 10646) character with hex value HHHH (4 digits)

Test with Bash 4.2.37(1):

$ printf '\u2620\n'
?

Sorry for reviving this old question. But when using bash there is a very easy approach to create Unicode codepoints from plain ASCII input, which even does not fork at all:

unicode() { local -n a="$1"; local c; printf -vc '\\U%08x' "$2"; printf -va "$c"; }
unicodes() { local a c; for a; do printf -vc '\\U%08x' "$a"; printf "$c"; done; };

Use it as follows to define certain codepoints

unicode crossbones 0x2620
echo "$crossbones"

or to dump the first 65536 unicode codepoints to stdout (takes less than 2s on my machine. The additional space is to prevent certain characters to flow into each other due to shell's monospace font):

for a in {0..65535}; do unicodes "$a"; printf ' '; done

or to tell a little very typical parent's story (this needs Unicode 2010):

unicodes 0x1F6BC 32 43 32 0x1F62D 32 32 43 32 0x1F37C 32 61 32 0x263A 32 32 43 32 0x1F4A9 10

Explanation:

• printf '\UXXXXXXXX' prints out any Unicode character
• printf '\\U%08x' number prints \UXXXXXXXX with the number converted to Hex, this then is fed to another printf to actually print out the Unicode character
• printf recognizes octal (0oct), hex (0xHEX) and decimal (0 or numbers starting with 1 to 9) as numbers, so you can choose whichever representation fits best
• printf -v var .. gathers the output of printf into a variable, without fork (which tremendously speeds up things)
• local variable is there to not pollute the global namespace
• local -n var=other aliases var to other, such that assignment to var alters other. One interesting part here is, that var is part of the local namespace, while other is part of the global namespace.
  • Please note that there is no such thing as local or global namespace in bash. Variables are kept in the environment, and such are always global. Local just puts away the current value and restores it when the function is left again. Other functions called from within the function with local will still see the "local" value. This is a fundamentally different concept than all the normal scoping rules found in other languages (and what bash does is very powerful but can lead to errors if you are a programmer who is not aware of that).

In Bash:

UnicodePointToUtf8()
{
    local x="$1"               # ok if '0x2620'
    x=${x/\\u/0x}              # '\u2620' -> '0x2620'
    x=${x/U+/0x}; x=${x/u+/0x} # 'U-2620' -> '0x2620'
    x=$((x)) # from hex to decimal
    local y=$x n=0
    [ $x -ge 0 ] || return 1
    while [ $y -gt 0 ]; do y=$((y>>1)); n=$((n+1)); done
    if [ $n -le 7 ]; then       # 7
        y=$x
    elif [ $n -le 11 ]; then    # 5+6
        y=" $(( ((x>> 6)&0x1F)+0xC0 )) \
            $(( (x&0x3F)+0x80 ))" 
    elif [ $n -le 16 ]; then    # 4+6+6
        y=" $(( ((x>>12)&0x0F)+0xE0 )) \
            $(( ((x>> 6)&0x3F)+0x80 )) \
            $(( (x&0x3F)+0x80 ))"
    else                        # 3+6+6+6
        y=" $(( ((x>>18)&0x07)+0xF0 )) \
            $(( ((x>>12)&0x3F)+0x80 )) \
            $(( ((x>> 6)&0x3F)+0x80 )) \
            $(( (x&0x3F)+0x80 ))"
    fi
    printf -v y '\\x%x' $y
    echo -n -e $y
}

# test
for (( i=0x2500; i<0x2600; i++ )); do
    UnicodePointToUtf8 $i
    [ "$(( i+1 & 0x1f ))" != 0 ] || echo ""
done
x='U+2620'
echo "$x -> $(UnicodePointToUtf8 $x)"

Output:

-?¦?????????+???+???+???+???+???
????¦???????-???????-???????+???
????????????????-¦++++++++++++¦¦
¦¦¦¦------+++???????????????????
¯???_???¦???¦???¦¦¦¦????????????
¦???????????????????????????????
????????????????????????????????
????????????????????????????????
U+2620 -> ?