Consider this small example:
printf "Loading data..."; \
sleep 5; \
echo -e "\rThis is my cool data point."
This brings, of course, Loading data..., and after 5 seconds, that will be overwritten by This is my cool data point..
But what if the text printed out after the \r is shorter as the first line?
printf "Loading data..."; \
sleep 5; \
echo -e "\rNo data."
...brings No data.data... after the waiting time.
Do I have to keep track of the longest possible line and print "\rNo data. " or is there any "magic character" that fills the line until its end in a normal terminal?
You can delete to end of line with tput el. So you can do:
eol=$(tput el)
printf "Loading data..."
sleep 5
printf "\rNo data.${eol}\n"
It's not really a "magic character" so much as a "magic sequence", and the actual values that are used may vary with the terminal. tput will (should) do the right thing and give you a reasonably portable method. Attempting to determine precisely which sequence to use is a futile effort.
Another solution would be:
#!/bin/bash
msg1='Loading data...'
msg2='No data.'
printf '%s' "$msg1"
sleep 5
printf '\r%*s\n' $(( -(${#msg1} > ${#msg2} ? ${#msg1} : ${#msg2}) )) "$msg2"
This appends trailing spaces to the line if the length of msg1 is greater than that of msg2.
Related
Essentially for a set number of systems logs pulled and exported I need to indicate the scripts progress by printing a character "#". This should eventually create a progress bar with a width of 60. Something like what's presented below: ############################################# ,additionally I need the characters to build from left to right indicating the progression of the script.
The Question/Problem that this code was based off of goes as follows: "Use a separate invocation of wevtutil el to get the count of the number of logs and scale this to,say, a width of 60."
SYSNAM=$(hostname)
LOGDIR=${1:-/tmp/${SYSNAM}_logs}
i=0
LOGCOUNT=$( wevtutil el | wc -l )
x=$(( LOGCOUNT/60 ))
wevtutil el | while read ALOG
do
ALOG="${ALOG%$'\r'}"
printf "${ALOG}:\r"
SAFNAM="${ALOG// /_}"
SAFNAM="${SAFNAM//\//-}"
wevtutil epl "$ALOG" "${SYSNAM}_${SAFNAM}.evtx"
done
I've attempted methods such as using echo -ne "#", and printf "#%0.s" however the issue that I encounter is that the "#" characters gets printed with each instance of the name of the log file being retrieved; also the pattern is printed vertically rather than horizontally.
LOGCOUNT=$( wevtutil el | wc -l )
x=$(( LOGCOUNT/60 ))
echo -ne "["
for i in {1..60}
do
if [[ $(( x*i )) != $LOGCOUNT ]]
then
echo -ne "#"
#printf '#%0.s'
fi
done
echo "]"
printf "\n"
echo "Transfer Complete."
echo "Total Log Files Transferred: $LOGCOUNT"
I tried previously integrating this code into the first block but no luck. But something tells me that I don't need to establish a whole new loop, I keep thinking that the first block of code only needs a few lines of modification. Anyhow sorry for the lengthy explanation, please let me know if anything additional is needed for assistance--Thank you.
For the sake of this answer I'm going to assume the desired output is a 2-liner that looks something like:
$ statbar
file: /bin/cygdbusmenu-qt5-2.dll
[######## ]
The following may not work for everyone as it comes down to individual terminal attributes and how they can(not) be manipulated by tput (ie, ymmv) ...
For my sample script I'm going to loop through the contents of /bin, printing the name of each file as I process it, while updating the status bar with a new '#' after each 20 files:
there are 719 files under my /bin so there should be 35 #'s in my status bar (I add an extra # at the end once processing has completed)
we'll use a few tput commands to handle cursor/line movement, plus erasing previous output from a line
for printing the status bar I've pre-calculated the number of #'s and then use 2 variables ... $barspace for spaces, $barhash for #'s; for each 20 files I strip a space off $barspace and add a single # to $barhash; by (re)printing these 2x variables every 20x files I get the appearance of a moving status bar
Putting this all together:
$ cat statbar
clear # make sure we have plenty of room to display our status bar;
# if we're at the bottom of the console/window and we cause the
# windows to 'scroll up' then 'tput sc/rc' will not work
tput sc # save pointer/reference to current terminal line
erase=$(tput el) # save control code for 'erase (rest of) line'
# init some variables; get a count of the number of files so we can pre-calculate the total length of our status bar
modcount=20
filecount=$(find /bin -type f | wc -l)
# generate a string of filecount/20+1 spaces (35+1 for my particular /bin)
barspace=
for ((i=1; i<=(filecount/modcount+1); i++))
do
barspace="${barspace} "
done
barhash= # start with no #'s for this variable
filecount=0 # we'll re-use this variable to keep track of # of files processed so need to reset
while read -r filename
do
filecount=$((filecount+1))
tput rc # return cursor to previously saved terminal line (tput sc)
# print filename (1st line of output); if shorter than previous filename we need to erase rest of line
printf "file: ${filename}${erase}\n"
# print our status bar (2nd line of output) on the first and every ${modcount} pass through loop;
if [ ${filecount} -eq 1 ]
then
printf "[${barhash}${barspace}]\n"
elif [[ $((filecount % ${modcount} )) -eq 0 ]]
then
# for every ${modcount}th file we ...
barspace=${barspace:1:100000} # strip a space from barspace
barhash="${barhash}#" # add a '#' to barhash
printf "[${barhash}${barspace}]\n" # print our new status bar
fi
done < <(find /bin -type f | sort -V)
# finish up the status bar (should only be 1 space left to 'convert' to a '#')
tput rc
printf "file: -- DONE --\n"
if [ ${#barspace} -gt 0 ]
then
barspace=${barspace:1:100000}
barhash="${barhash}#"
fi
printf "[${barhash}${barspace}]\n"
NOTE: While testing I had to periodically reset my terminal in order for the tput commands to function properly, eg:
$ reset
$ statbar
I couldn't get the above to work on any of the (internet) fiddle sites (basically having problems getting tput to work with the web-based 'terminals').
Here's a gif displaying the behavior ...
NOTES:
the script does print every filename to stdout but since this script isn't actually doing anything with the files in question a) the printfs occur quite rapidly and b) the video/gif only captures a (relatively) few fleeting images ("Duh, Mark!" ?)
the last printf "file: -- DONE --\n" was added after I created the gif, and I'm being lazy by not generating and uploading a new gif
I'm working with an existing script which was written a bit messily. Setting up a loop with all of the spaghetti code could make a bigger headache than I want to deal with in the near term. Maybe when I have more time I can clean it up but for now, I'm just looking for a simple fix.
The script deals with virtual disks on a xen server. It reads multipath output and asks if particular LUNs should be formatted in any way based on specific criteria. However, rather than taking that disk path and inserting it, already formatted, into a configuration file, it simply presents every line in the format
'phy:/dev/mapper/UUID,xvd?,w',
UUID, of course, is an actual UUID.
The script actually presents each of the found LUNs in this format expecting the user to copy and paste them into the config file replacing each ? with a letter in sequence. This is tedious at best.
There are several ways to increment a number in bash. Among others:
var=$((var+1))
((var+=1))
((var++))
Is there a way to do the same with characters which doesn't involve looping over the entire alphabet such that I could easily "increment" the disk assignment from xvda to xvdb, etc?
To do an "increment" on a letter, define the function:
incr() { LC_CTYPE=C printf "\\$(printf '%03o' "$(($(printf '%d' "'$1")+1))")"; }
Now, observe:
$ echo $(incr a)
b
$ echo $(incr b)
c
$ echo $(incr c)
d
Because, this increments up through ASCII, incr z becomes {.
How it works
The first step is to convert a letter to its ASCII numeric value. For example, a is 97:
$ printf '%d' "'a"
97
The next step is to increment that:
$ echo "$((97+1))"
98
Or:
$ echo "$(($(printf '%d' "'a")+1))"
98
The last step is convert the new incremented number back to a letter:
$ LC_CTYPE=C printf "\\$(printf '%03o' "98")"
b
Or:
$ LC_CTYPE=C printf "\\$(printf '%03o' "$(($(printf '%d' "'a")+1))")"
b
Alternative
With bash, we can define an associative array to hold the next character:
$ declare -A Incr; last=a; for next in {b..z}; do Incr[$last]=$next; last=$next; done; Incr[z]=a
Or, if you prefer code spread out over multiple lines:
declare -A Incr
last=a
for next in {b..z}
do
Incr[$last]=$next
last=$next
done
Incr[z]=a
With this array, characters can be incremented via:
$ echo "${Incr[a]}"
b
$ echo "${Incr[b]}"
c
$ echo "${Incr[c]}"
d
In this version, the increment of z loops back to a:
$ echo "${Incr[z]}"
a
How about an array with entries A-Z assigned to indexes 1-26?
IFS=':' read -r -a alpharray <<< ":A:B:C:D:E:F:G:H:I:J:K:L:M:N:O:P:Q:R:S:T:U:V:W:X:Y:Z"
This has 1=A, 2=B, etc. If you want 0=A, 1=B, and so on, remove the first colon.
IFS=':' read -r -a alpharray <<< "A:B:C:D:E:F:G:H:I:J:K:L:M:N:O:P:Q:R:S:T:U:V:W:X:Y:Z"
Then later, where you actually need the letter;
var=$((var+1))
'phy:/dev/mapper/UUID,xvd${alpharray[$var]},w',
The only problem is that if you end up running past 26 letters, you'll start getting blanks returned from the array.
Use a Bash 4 Range
You can use a Bash 4 feature that lets you specify a range within a sequence expression. For example:
for letter in {a..z}; do
echo "phy:/dev/mapper/UUID,xvd${letter},w"
done
See also Ranges in the Bash Wiki.
Here's a function that will return the next letter in the range a-z. An input of 'z' returns 'a'.
nextl(){
((num=(36#$(printf '%c' $1)-9) % 26+97));
printf '%b\n' '\x'$(printf "%x" $num);
}
It treats the first letter of the input as a base 36 integer, subtracts 9, and returns the character whose ordinal number is 'a' plus that value mod 26.
Use Jot
While the Bash range option uses built-ins, you can also use a utility like the BSD jot utility. This is available on macOS by default, but your mileage may vary on Linux systems. For example, you'll need to install athena-jot on Debian.
More Loops
One trick here is to pre-populate a Bash array and then use an index variable to grab your desired output from the array. For example:
letters=( "" $(jot -w %c 26 a) )
for idx in 1 26; do
echo ${letters[$idx]}
done
A Loop-Free Alternative
Note that you don't have to increment the counter in a loop. You can do it other ways, too. Consider the following, which will increment any letter passed to the function without having to prepopulate an array:
increment_var () {
local new_var=$(jot -nw %c 2 "$1" | tail -1)
if [[ "$new_var" == "{" ]]; then
echo "Error: You can't increment past 'z'" >&2
exit 1
fi
echo -n "$new_var"
}
var="c"
var=$(increment_var "$var")
echo "$var"
This is probably closer to what the OP wants, but it certainly seems more complex and less elegant than the original loop recommended elsewhere. However, your mileage may vary, and it's good to have options!
I have a tab-delimited text file which I send to column to "pretty print" a table.
Original file:
1<TAB>blablablabla<TAB>aaaa bbb ccc
2<TAB>blabla<TAB>xxxxxx
34<TAB>okokokok<TAB>zzz yyy
Using column -s$'\t' -t <original file>, I get
1 blablablabla aaaa bbb xxx
2 blabla xxxxxx
34 okokokok zzz yyy
as desired. Now I want to add colors to the columns. I tried to add the escape codes around each tab-delimited field in the original file. column successfully prints in color, but the columns are no longer aligned. Instead, it just prints the TAB separators verbatim.
The question is: how can I get the columns aligned, but also with unique colors?
I've thought of two ways to achieve this:
Adjust the column parameters to make the alignment work with color codes
Redirect the output of column to another file, and do a search+replace on the first two whitespace-delimited fields (the first two columns are guaranteed to not contain spaces; the third column most likely will contain spaces, but no TAB characters)
Problem is, I'm not sure how to do either of those two...
For reference, here is what I'm passing to column:
Note that the fields are indeed separated by TAB characters. I've confirmed this with od.
edit:
There doesn't seem to be an issue with the colorization. I already have the file shown above with the color codes working. The issue is column won't align once I send it input with escape codes. I am thinking of passing the fields without color codes to column, then copying the exact number of spaces column output between each field, and using that in a pretty print scheme.
I wrote a bash version of column (similar to the one from util-linux) which works with color codes:
#!/bin/bash
which sed >> /dev/null || exit 1
version=1.0b
editor="Norman Geist"
last="04 Jul 2016"
# NOTE: Brilliant pipeable tool to format input text into a table by
# NOTE: an configurable seperation string, similar to column
# NOTE: from util-linux, but we are smart enough to ignore
# NOTE: ANSI escape codes in our column width computation
# NOTE: means we handle colors properly ;-)
# BUG : none
addspace=1
seperator=$(echo -e " ")
columnW=()
columnT=()
while getopts "s:hp:v" opt; do
case $opt in
s ) seperator=$OPTARG;;
p ) addspace=$OPTARG;;
v ) echo "Version $version last edited by $editor ($last)"; exit 0;;
h ) echo "column2 [-s seperator] [-p padding] [-v]"; exit 0;;
* ) echo "Unknow comandline switch \"$opt\""; exit 1
esac
done
shift $(($OPTIND-1))
if [ ${#seperator} -lt 1 ]; then
echo "Error) Please enter valid seperation string!"
exit 1
fi
if [ ${#addspace} -lt 1 ]; then
echo "Error) Please enter number of addional padding spaces!"
exit 1
fi
#args: string
function trimANSI()
{
TRIM=$1
TRIM=$(sed 's/\x1b\[[0-9;]*m//g' <<< $TRIM); #trim color codes
TRIM=$(sed 's/\x1b(B//g' <<< $TRIM); #trim sgr0 directive
echo $TRIM
}
#args: len
function pad()
{
for ((i=0; i<$1; i++))
do
echo -n " "
done
}
#read and measure cols
while read ROW
do
while IFS=$seperator read -ra COLS; do
ITEMC=0
for ITEM in "${COLS[#]}"; do
SITEM=$(trimANSI "$ITEM"); #quotes matter O_o
[ ${#columnW[$ITEMC]} -gt 0 ] || columnW[$ITEMC]=0
[ ${columnW[$ITEMC]} -lt ${#SITEM} ] && columnW[$ITEMC]=${#SITEM}
((ITEMC++))
done
columnT[${#columnT[#]}]="$ROW"
done <<< "$ROW"
done
#print formatted output
for ROW in "${columnT[#]}"
do
while IFS=$seperator read -ra COLS; do
ITEMC=0
for ITEM in "${COLS[#]}"; do
WIDTH=$(( ${columnW[$ITEMC]} + $addspace ))
SITEM=$(trimANSI "$ITEM"); #quotes matter O_o
PAD=$(($WIDTH-${#SITEM}))
if [ $ITEMC -ne 0 ]; then
pad $PAD
fi
echo -n "$ITEM"
if [ $ITEMC -eq 0 ]; then
pad $PAD
fi
((ITEMC++))
done
done <<< "$ROW"
echo ""
done
Example usage:
bold=$(tput bold)
normal=$(tput sgr0)
green=$(tput setaf 2)
column2 -s § << END
${bold}First Name§Last Name§City${normal}
${green}John§Wick${normal}§New York
${green}Max§Pattern${normal}§Denver
END
Output example:
I would use awk for the colorization (sed can be used as well):
awk '{printf "\033[1;32m%s\t\033[00m\033[1;33m%s\t\033[00m\033[1;34m%s\033[00m\n", $1, $2, $3;}' a.txt
and pipe it to column for the alignment:
... | column -s$'\t' -t
Output:
A solution using printf to format the ouput as well :
while IFS=$'\t' read -r c1 c2 c3; do
tput setaf 1; printf '%-10s' "$c1"
tput setaf 2; printf '%-30s' "$c2"
tput setaf 3; printf '%-30s' "$c3"
tput sgr0; echo
done < file
In my case, I wanted to selectively colorise values in a column depending on its value. Let's say I want okokokok to be green and blabla to be red.
I can do it such way (the idea is to colorise values of columns after columnisation):
GREEN_SED='\\033[0;32m'
RED_SED='\\033[0;31m'
NC_SED='\\033[0m' # No Color
column -s$'\t' -t <original file> | echo -e "$(sed -e "s/okokokok/${GREEN_SED}okokokok${NC_SED}/g" -e "s/blabla/${RED_SED}blabla${NC_SED}/g")"
Alternatively, with a variable:
DATA=$(column -s$'\t' -t <original file>)
GREEN_SED='\\033[0;32m'
RED_SED='\\033[0;31m'
NC_SED='\\033[0m' # No Color
echo -e "$(sed -e "s/okokokok/${GREEN_SED}okokokok${NC_SED}/g" -e "s/blabla/${RED_SED}blabla${NC_SED}/g" <<< "$DATA")"
Take a note of that additional backslash in values of color definitions. It is made for sed to not interpret an origingal backsash.
This is a result:
2021 Updated BASH Answer
TL;DR
I really liked #NORMAN GEIST's answer but was way too slow for what i needed... So i coded my own version of his script, this time written in Perl (stdin looping and formatting) + Bash (only for presentation/help).
You can find the full code here with an explanation on how to use it.
It is comprehensive of:
A Bash column-like command interface (same parameters like -t, -s, -o)
Exaustive help with column_ansi --help or column_ansi -h
Option to horizontally center.
The actual "core" code can broken down to only the Perl part.
Background and differences
I needed to format a very long awk-generated colored output (more than 300 lines) into a nice table.
I first thought of using column, but as i discovered it didn't take into consideration ANSI characters, since the output would come out not aligned.
After searching a bit on Google i found #NORMAN GEIST's interesting answer on SO which dynamically calculated the width of every single column in the output after removing the ANSI characters and THEN it built the table.
It was all good, but it was taking way too long to load (as someone pointed in the comments)...
So i tried to convert #NORMAN GEIST's column2 from bash to perl and my god if there was a change!
After trying out this version in my production script the time used to display data dropped from 30s to <1s!!
Enjoy!
Im trying to create a list of file-paths from a file but I don't seem to be able to get around the spaces in the file paths.
# Show current series list
PS3="Type a number or 'q' to quit: "
# Create a list of files to display
Current_list=`cat Current_series_list.txt`
select fileName in $Current_list; do
if [ -n "$fileName" ]; then
Selected_series=${fileName}
fi
break
done
The file path in the Current_series list is: /Volumes/Lara's Hard Drive/LARA HARD DRIVE/Series/The Big Bang Theory 3/The.Big.Bang.Theory S03E11.avi
and
/Volumes/Lara's Hard Drive/LARA HARD DRIVE/Series/nakitaS03E11.avi
So i would like them two be 1 and 2 respectively in my list but I get the following result.
1) /Volumes/Lara's 6) Big
2) Hard 7) Bang
3) Drive/LARA 8) Theory
4) HARD 9) 3/The.Big.Bang.Theory
5) DRIVE/Series/The 10) S03E11.avi
Type a number or 'q' to quit:
You need to trick it a little bit:
# Show current series list
PS3="Type a number or 'q' to quit: "
# Create a list of files to display
Current_list=$(tr '\n' ',' < Current_series_list.txt)
IFS=, read -a list <<< "$Current_list"
select fileName in "${list[#]}"; do
if [ -n "$fileName" ]; then
Selected_series="${fileName}"
fi
break
done
echo "you selected $fileName"
Executing:
$ ./a
1) /Volumes/Lara's Hard Drive/LARA HARD DRIVE/Series/The Big Bang Theory3/The.Big.Bang.Theory S03E11.avi
2) /Volumes/Lara's Hard Drive/LARA HARD DRIVE/Series/nakitaS03E11.avi
Type a number or 'q' to quit: 2
you selected /Volumes/Lara's Hard Drive/LARA HARD DRIVE/Series/nakitaS03E11.avi
The key point is that you have to convert a file into an array.
This part converts it into a "string one", "string two" format:
$ tr '\n' ',' < Current_series_list.txt
/Volumes/Lara's Hard Drive/LARA HARD DRIVE/Series/The Big Bang Theory 3/The.Big.Bang.Theory S03E11.avi,/Volumes/Lara's Hard Drive/LARA HARD DRIVE/Series/nakitaS03E11.avi,
While this one creates an array in the variable list based on the comma delimiter that was set in the previous step:
IFS=, read -a list <<< "$Current_list"
You could try to read each line of Current_series_list.txt separately into an array element and select from the expanded array "${Current_array[#]}":
# Show current series list
PS3="Type a number or 'q' to quit: "
# Create an array of files to display
Current_array=()
while read line; do Current_array+=("$line"); done < Current_series_list.txt
select fileName in "${Current_array[#]}"; do
if [ -n "$fileName" ]; then
Selected_series=${fileName}
fi
break
done
I have a homework assignment that is asking to shift a decimal number by a specified amount of digits. More clearly this bash script will take two input arguments, the first is the number(maximum 9 digits) that the shift will be performed on and the second is the number(-9 to 9) of digits to shift. Another requirement is that when a digit is shifted off the end, it should be attached to the other end of the number. One headache of a requirement is that we cannot use control statements of any kind: no loops, no if, and switch cases.
Example: 12345 3 should come out to 345000012 and 12345 -3 should be 12345000
I know that if I mod 12345 by 10^3 I get 345 and then if I divide 12345 by 10^3 I get 12 and then I can just concatenate those two variables together to get 34512. I am not quite sure if that is exactly correct but that is the closest I can get as of now. As far as the -3 shift, I know that 10^-3 is .001 and would work however when I try using 10^-3 in bash I get an error.
I am just lost at this point, any tips would be greatly appreciated.
EDIT: After several hours of bashing (pun intended) my head against this problem, I finally came up with a script that for the most part works. I would post the code right now but I fear another student hopelessly lost might stumble upon it. I will check back and post what I came up with in a week or two. I was able to do it with mods and division. Thank you all for the responses, it really helped me to open up and think about the problem from different angles.
Here's a hint:
echo ${string:0:3}
echo ${#string}
Edit (2011-02-11):
Here's my solution. I added some additional parameters with defaults.
rotate-string ()
{
local s=${1:-1} p=${2:--1} w=${3:-8} c=${4:-0} r l
printf -vr '%0*d' $w 0 # save $w zeros in $r
r=${r//0/$c}$s # change the zeros to the character in $c, append the string
r=${r: -w} # save the last $w characters of $r
l=${r: -p%w} # get the last part of $r ($p mod %w characters)
echo "$l${r::w-${#l}}" # output the Last part on the Left and the Right part which starts at the beginning and goes for ($w minus the_length_of_the_Left_part) characters
}
usage: rotate-string string positions-to-rotate width fill-character
example: rotate-string abc -4 9 =
result: ==abc====
Arguments can be omitted starting from the end and these defaults will be used:
fill-character: "0"
width: 8
positions-to-rotate: -1
string: "1"
More examples:
$ rotate-string
00000010
$ rotate-string 123 4
01230000
Fun stuff:
$ for i in {126..6}; do printf '%s\r' "$(rotate-string Dennis $i 20 .)"; sleep .05; done; printf '\n'
$ while true; do for i in {10..1} {1..10}; do printf '%s\r' "$(rotate-string : $i 10 .)"; sleep .1; done; done
$ while true; do for i in {40..2} {2..40}; do printf '%s\r' "$(rotate-string '/\' $i 40 '_')"; sleep .02; done; done
$ d=0; while true; do for i in {1..10} {10..1}; do printf '%s\r' "$(rotate-string $d $i 10 '_')"; sleep .02; done; ((d=++d%10)); done
$ d=0; while true; do for i in {1..10}; do printf '%s\r' "$(rotate-string $d $i 10 '_')"; sleep .2; ((d=++d%10)); done; done
$ shape='▁▂▃▄▅▆▇█▇▆▅▄▃▂▁'; while true; do for ((i=1; i<=COLUMNS; i++)); do printf '%s\r' "$(rotate-string "$shape" $i $COLUMNS ' ')"; done; done
In the absence of control structures, you need to use recursion, with index values as "choice selections", which is how functional programming often works.
#!/bin/sh
#
# cshift NUMBER N
cshift() {
let num=10#$1
num=`printf '%09d' $num`
lshift="${num:1:8}${num:0:1}"
rshift="${num:8:1}${num:0:8}"
next=( "cshift $lshift $(($2 + 1))" "echo $num" "cshift $rshift $(( $2 - 1 ))" )
x=$(( $2 == 0 ? 1 : $2 < 0 ? 0 : 2 ))
eval "${next[x]}"
}
cshift $1 $2
and, the testing:
$ for ((i=-9;i<=9;i++)); do cshift 12345 $i ; done
000012345
500001234
450000123
345000012
234500001
123450000
012345000
001234500
000123450
000012345
500001234
450000123
345000012
234500001
123450000
012345000
001234500
000123450
000012345
You can also do some math on the indexes and avoid the recursion, but I don't mind making the computer work harder so I don't have to. It's easy to think of how to do the shift by one in either direction, and then I use an evaluated choice that is selected by the signum of the shift value, outputting a value and stopping when the shift value is zero.