Populating an array from a file should be basic, but I can't get it to work.
#!/bin/bash
declare -a a
i=0
cat "file.txt" | while read line; do
a[$i]="$line"
i=$(($i + 1))
done
echo "${a[0]}"
This prints an empty line. For a file that contains the lines "Foo" and "Bar", here's the output from bash -x:
+ declare -a a
+ i=0
+ cat file.txt
+ read line
+ a[$i]=Foo
+ i=1
+ read line
+ a[$i]=Bar
+ i=2
+ read line
+ echo ''
I can't even get the readarray builtin working:
#!/bin/bash
declare -a b
cat "file.txt" | readarray b
echo "${b[0]}"
+ declare -a b
+ cat file.txt
+ readarray b
+ echo ''
What am I doing wrong here?
This is a basic bash pitfall: all segments of a pipeline execute in subshells by default, which means that whatever variables they create will go out of scope when these subshells terminate (the current shell won't see them).
Depending on what version of bash you're using, you have two options:
[Bash v4.2+] Execute shopt -s lastpipe before you use the pipeline to ensure that the last pipeline segment runs in the current shell.
Use input redirection (<) instead of a pipeline (or, in more complex cases, process substitution (<(...)) to ensure that your while loop doesn't run in a subshell:
a=()
while IFS= read -r line; do
a+=( "$line" )
# ...
done < "file.txt"
In the simple case of reading lines from a file, you can read directly into an array ("${a[#]}", in this example):
Bash 4.x:
readarray -t a < "file.txt"
Earlier versions:
IFS=$'\n' read -r -d '' -a a < "file.txt"
Related
I'm taking over a bash script from a colleague that reads a file, process it and print another file based on the line in the while loop at the moment.
I now need to append some features to it. The one I'm having issues with right now is to read a file and put each line into an array, except the 2nd column of that line can be empty, e.g.:
For a text file with \t as separator:
A\tB\tC
A\t\tC
For a CSV file same but with , as separator:
A,B,C
A,,C
Which should then give
["A","B","C"] or ["A", "", "C"]
The code I took over is as follow:
while IFS=$'\t\r' read -r -a col; do
# Process the array, put that into a file
lp -d $printer $file_to_print
done < $input_file
Which works if B is filled, but B need to be empty now sometimes, so when the input files keeps it empty, the created array and thus the output file to print just skips this empty cell (array is then ["A","C"]).
I tried writing the whole bloc on awk but this brought it's own sets of problems, making it difficult to call the lp command to print.
So my question is, how can I preserve the empty cell from the line into my bash array, so that I can call on it later and use it?
Thank you very much. I know this might be quite confused so please ask and I'll specify.
Edit: After request, here's the awk code I've tried. The issue here is that it only prints the last print request, while I know it loops over the whole file, and the lp command is still in the loop.
awk 'BEGIN {
inputfile="'"${optfile}"'"
outputfile="'"${file_loc}"'"
printer="'"${printer}"'"
while (getline < inputfile){
print "'"${prefix}"'" > outputfile
split($0,ft,"'"${IFSseps}"'");
if (length(ft[2]) == 0){
print "CODEPAGE 1252\nTEXT 465,191,\"ROMAN.TTF\",180,7,7,\""ft[1]"\"" >> outputfile
size_changer = 0
} else {
print "CODEPAGE 1252\nTEXT 465,191,\"ROMAN.TTF\",180,7,7,\""ft[1]"_"ft[2]"\"" >> outputfile
size_changer = 1
}
if ( split($0,ft,"'"${IFSseps}"'") > 6)
maxcounter = 6;
else
maxcounter = split($0,ft,"'"${IFSseps}"'");
for (i = 3; i <= maxcounter; i++){
x=191-(i-2)*33
print "CODEPAGE 1252\nTEXT 465,"x",\"ROMAN.TTF\",180,7,7,\""ft[i]"\"" >> outputfile
}
print "PRINT ""'"${copies}"'"",1" >> outputfile
close(outputfile)
"'"`lp -d ${printer} ${file_loc}`"'"
}
close("'"${file_loc}"'");
}'
EDIT2: Continuing to try to find a solution to it, I tried following code without success. This is weird, as just doing printf without putting it in an array keeps the formatting intact.
$ cat testinput | tr '\t' '>'
A>B>C
A>>C
# Should normally be empty on the second ouput line
$ while read line; do IFS=$'\t' read -ra col < <(printf "$line"); echo ${col[1]}; done < testinput
B
C
For tab, it's complicated.
From 3.5.7 Word Splitting in the manual:
A sequence of IFS whitespace characters is also treated as a delimiter.
Since tab is an "IFS whitespace character", sequences of tabs are treated as a single delimiter
IFS=$'\t' read -ra ary <<<$'A\t\tC'
declare -p ary
declare -a ary=([0]="A" [1]="C")
What you can do is translate tabs to a non-whitespace character, assuming it does not clash with the actual data in the fields:
line=$'A\t\tC'
IFS=, read -ra ary <<<"${line//$'\t'/,}"
declare -p ary
declare -a ary=([0]="A" [1]="" [2]="C")
To avoid the risk of colliding with commas in the data, we can use an unusual ASCII character: FS, octal 034
line=$'A\t\tC'
printf -v FS '\034'
IFS="$FS" read -ra ary <<<"${line//$'\t'/"$FS"}"
# or, without the placeholder variable
IFS=$'\034' read -ra ary <<<"${line//$'\t'/$'\034'}"
declare -p ary
declare -a ary=([0]="A" [1]="" [2]="C")
One bash example using parameter expansion where we convert the delimiter into a \n and let mapfile read in each line as a new array entry ...
For tab-delimited data:
for line in $'A\tB\tC' $'A\t\tC'
do
mapfile -t array <<< "${line//$'\t'/$'\n'}"
echo "############# ${line}"
typeset -p array
done
############# A B C
declare -a array=([0]="A" [1]="B" [2]="C")
############# A C
declare -a array=([0]="A" [1]="" [2]="C")
NOTE: The $'...' construct insures the \t is treated as a single <tab> character as opposed to the two literal characters \ + t.
For comma-delimited data:
for line in 'A,B,C' 'A,,C'
do
mapfile -t array <<< "${line//,/$'\n'}"
echo "############# ${line}"
typeset -p array
done
############# A,B,C
declare -a array=([0]="A" [1]="B" [2]="C")
############# A,,C
declare -a array=([0]="A" [1]="" [2]="C")
NOTE: This obviously (?) assumes the desired data does not contain a comma (,).
It may just be your # Process the array, put that into a file part.
IFS=, read -ra ray <<< "A,,C"
for e in "${ray[#]}"; do o="$o\"$e\","; done
echo "[${o%,}]"
["A","","C"]
See #Glenn's excellent answer regarding tabs.
My simple data file:
$: cat x # tab delimited, empty field 2 of line 2
a b c
d f
My test:
while IFS=$'\001' read -r a b c; do
echo "a:[$a] b:[$b] c:[$c]"
done < <(tr "\t" "\001"<x)
a:[a] b:[b] c:[c]
a:[d] b:[] c:[f]
Note that I used ^A (a 001 byte) but you might be able to use something as simple as a comma or pipe (|) character. Choose based on your data.
I need to read file line by line, and every line split by ",", and store to array.
File source_file.
usl-coop,/root
usl-dev,/bin
Script.
i=1
while read -r line; do
IFS="," read -ra para_$i <<< $line
echo ${para_$i[#]}
((i++))
done < source_file
Expected output.
para_1[0]=usl-coop
para_1[1]=/root
para_2[0]=usl-dev
para_2[1]=/bin
Script will out error about echo.
./sofimon.sh: line 21: ${para_$i[#]}: bad substitution
When I echo array one by one field, for example
echo para_1[0]
it shows, that variables are stored.
But I need use it with variable within, something like this.
${para_$i[1]}
Is possible to do this?
Thanks.
S.
There is a trick to simulate 2D arrays using associative arrays. It works nice and I think is the most flexible and extensible:
declare -A para
i=1
while IFS=, read -r -a line; do
for j in ${!line[#]}; do
para[$i,$j]="${line[$j]}"
done
((i++)) ||:
done < source_file
declare -p para
will output:
declare -A para=([1,0]="usl-coop" [1,1]="/root" [2,1]="/bin" [2,0]="usl-dev" )
Without modifying your script that much you could use indirect variable expansion. It's sometimes used in simpler scripts:
i=1
while IFS="," read -r -a para_$i; do
n="para_$i[#]"
echo "${!n}"
((i++)) ||:
done < source_file
declare -p ${!para_*}
or basically the same with a nameref a named reference to another variable (side note: see how [#] needs to be part of the variable in indirect expansion, but not in named reference):
i=1
while IFS="," read -r -a para_$i; do
declare -n n
n="para_$i"
echo "${n[#]}"
((i++)) ||:
done < source_file
declare -p ${!para_*}
both scripts above will output the same:
usl-coop /root
usl-dev /bin
declare -a para_1=([0]="usl-coop" [1]="/root")
declare -a para_2=([0]="usl-dev" [1]="/bin")
That said, I think you shouldn't read your file into memory at all. It's just a bad design. Shell and bash is build around passing your files with pipes, streams, fifos, redirections, process substitutions, etc. without ever saving/copying/storing the file. If you have a file to parse, you should stream it to another process, parse and save the result, without ever storing the whole input in memory. If you want some data to find inside a file, use grep or awk.
Here is a short awk script that do the task.
awk 'BEGIN{FS=",";of="para_%d[%d]=%s\n"}{printf(of, NR, 0, $1);printf(of, NR, 1, $2)}' input.txt
Provide the desired output.
Explanation:
BEGIN{
FS=","; # set field seperator to `,`
of="para_%d[%d]=%s\n" # define common printf output format
}
{ # for each input line
printf(of, NR, 0, $1); # output for current line, [0], left field
printf(of, NR, 1, $2) # output for current line, [1], right field
}
How can I get just the filenames into an array using the cat command?
How I've been trying:
array=()
while IFS= read -r -d $'\0'; do
array+=("$REPLY")
done < <(cat /proc/swaps | grep "swap")
This either grabs all the information from the output into an array, or just doesn't work. How can I successfully get my expected output of [/swapfile, /dev/hda1, /some/other/swap] into an array form using the cat command?
readarray array < <(awk '/swap/{print $1}' /proc/swaps)
Bash introduced readarray in version 4 which can take the place of the while read loop. readarray is the solution you want.
here is the syntax
readarray variable < inputfile
echo "${variable[0]}" ' to print the first element in array
I'm trying to write a small script that will take the 4th columns of a file and store it in an array then do a little comparison. If the element in the array is greater than 0 and less than 500 I have to increment the counter. However when I run the script the counter always shows 0. Here's my script
#!/bin/bash
mapfile -t my_array < <(cat file1.txt | awk '{ print $4 }' > test.txt)
COUNTER=0
for i in ${my_array[#]}; do
if [["${my_array[$i]}" -gt 0 -a "${my_array[$i]}" -lt 500 ]]
then
COUNTER=$((COUNTER + 1))
fi
printf "%s\t%s\n" "%i" "${my_array[$i]}"//just to test if the mapfile command is working
done
echo $COUNTER
output:
./script1.bash
0
#!/bin/bash
mapfile -t my_array < <(awk '{ print $4 }' file1.txt | tee test.txt)
COUNTER=0
for idx in "${!my_array[#]}"; do
value=${my_array[$idx]}
if (( value > 0 )) && (( value < 500 )); then
COUNTER=$((COUNTER + 1))
fi
printf "%s\t%s\n" "$idx" "$value"
done
echo "$COUNTER"
The use of cat here is needless: It added nothing but inefficiency (requiring an extra process to be started, and forcing awk to read from a pipe rather than direct from a file).
mapfile had nothing to read because the output of awk was redirected to test.txt. If you want it to go to both a file and stdout, then you need to use tee.
-a is not valid in [[ ]]; use && instead there. However, since you're doing only arithmetic, (( )) is more appropriate. Incidentally, -a is officially marked obsolescent even for [ ] and test; see the current POSIX standard.
${my_array[#]} iterates over values. If you want to iterate over indexes, you need ${!my_array[#]} instead.
Whitespace is mandatory in separating command names. [["$foo" is a different command from [[, unless $foo is empty or starts with a character in $IFS.
If you redirect the output to a file: > test.txt then there is no output in "standard output" because it is consumed by the file. So, first, you need to remove that redirection. You may use:
mapfile -t my_array < <(cat file1.txt | awk '{ print $4 }' )
But since awk could perfectly well read a file, this is better:
mapfile -t my_array < <(awk '{ print $4 }' file1.txt)
And since you are using awk, it could do the comparison to 0 and 500 and output the whole count.
counter=$(awk '{if($4>0 && $4<500){c++}}END{print c}' file1.txt)
echo "$counter"
Simpler, faster.
That will also avoid some simple mistakes in your script, like missing an space in the […] construct:
if [[ "${my … # NOT "if [["${my …"
And some missing quotes:
for i in "${my_array[#]}" # NOT for i in ${my_array[#]}
In general, it is a good idea to check your script with ShellCheck.net to remove some simple mistakes.
I am very new to Unix shell script and trying to get some knowledge in shell scripting. Please check my requirement and my approach.
I have a input file having data
ABC = A:3 E:3 PS:6
PQR = B:5 S:5 AS:2 N:2
I am trying to parse the data and get the result as
ABC
A=3
E=3
PS=6
PQR
B=5
S=5
AS=2
N=2
The values can be added horizontally and vertically so I am trying to use an array. I am trying something like this:
myarr=(main.conf | awk -F"=" 'NR!=1 {print $1}'))
echo ${myarr[1]}
# Or loop through every element in the array
for i in "${myarr[#]}"
do
:
echo $i
done
or
awk -F"=" 'NR!=1 {
print $1"\n"
STR=$2
IFS=':' read -r -a array <<< "$STR"
for i in "${!array[#]}"
do
echo "$i=>${array[i]}"
done
}' main.conf
But when I add this code to a .sh file and try to run it, I get syntax errors as
$ awk -F"=" 'NR!=1 {
> print $1"\n"
> STR=$2
> FS= read -r -a array <<< "$STR"
> for i in "${!array[#]}"
> do
> echo "$i=>${array[i]}"
> done
>
> }' main.conf
awk: cmd. line:4: FS= read -r -a array <<< "$STR"
awk: cmd. line:4: ^ syntax error
awk: cmd. line:5: for i in "${!array[#]}"
awk: cmd. line:5: ^ syntax error
awk: cmd. line:8: done
awk: cmd. line:8: ^ syntax error
How can I complete the above expectations?
This is the awk code to do what you want:
$ cat tst.awk
BEGIN { FS="[ =:]+"; OFS="=" }
{
print $1
for (i=2;i<NF;i+=2) {
print $i, $(i+1)
}
print ""
}
and this is the shell script (yes, all a shell script does to manipulate text is call awk):
$ awk -f tst.awk file
ABC
A=3
E=3
PS=6
PQR
B=5
S=5
AS=2
N=2
A UNIX shell is an environment from which to call UNIX tools (find, sort, sed, grep, awk, tr, cut, etc.). It has its own language for manipulating (e.g. creating/destroying) files and processes and sequencing calls to tools but it is NOT intended to be used to manipulate text. The guys who invented shell also invented awk for shell to call to manipulate text.
Read https://unix.stackexchange.com/questions/169716/why-is-using-a-shell-loop-to-process-text-considered-bad-practice and the book Effective Awk Programming, 4th Edition, by Arnold Robbins.
First off, a command that does what you want:
$ sed 's/ = /\n/;y/: /=\n/' main.conf
ABC
A=3
E=3
PS=6
PQR
B=5
S=5
AS=2
N=2
This replaces, on each line, the first (and only) occurrence of = with a newline (the s command), then turns all : into = and all spaces into newlines (the y command). Notice that
this works only because there is a space at the end of the first line (otherwise it would be a bit more involved to get the empty line between the blocks) and
this works only with GNU sed because it substitutes newlines; see this fantastic answer for all the details and how to get it to work with BSD sed.
As for what you tried, there is almost too much wrong with it to try and fix it piece by piece: from the wild mixing of awk and Bash to syntax errors all over the place. I recommend you read good tutorials for both, for example:
The BashGuide
Effective AWK Programming
A Bash solution
Here is a way to solve the same in Bash; I didn't use any arrays.
#!/bin/bash
# Read line by line into the 'line' variable. Setting 'IFS' to the empty string
# preserves leading and trailing whitespace; '-r' prevents interpretation of
# backslash escapes
while IFS= read -r line; do
# Three parameter expansions:
# Replace ' = ' by newline (escape backslash)
line="${line/ = /\\n}"
# Replace ':' by '='
line="${line//:/=}"
# Replace spaces by newlines (escape backslash)
line="${line// /\\n}"
# Print the modified input line; '%b' expands backslash escapes
printf "%b" "$line"
done < "$1"
Output:
$ ./SO.sh main.conf
ABC
A=3
E=3
PS=6
PQR
B=5
S=5
AS=2
N=2