I was trying to do masking of file with command 'tr' and 'awk' but failing with error fatal: cannot open pipe ( Too many open pipes) error. FILE has approx 1000000 records quite a huge number.
Below is the code I am trying :-
awk - F "|" - v OFS="|" '{ "echo \""$1"\" | tr \" 0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ\" \" QWERTYUIOPASDFGHJKLZXCVBNM9876543210mnbvcxzlkjhgfdsapoiuytrewq\"" | get line $1}1' FILE.CSV > test.CSV
It is showing error :-
awk: (FILENAME=- FNR=1019) fatal: cannot open pipe `echo ""TTP_123"" | tr "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ" "QWERTYUIOPASDFGHJKLZXCVBNM9876543210mnbvcxzlkjhgfdsapoiuytrewq"' (Too many open pipes)
Please let me know what I am doing wrong here
Also a Note any number of columns could be used for masking and can be at any positions in this example I have taken 1 and 2 column positions but it could be 3 and 10 or 5,7,25 columns
Thanks
AJ
First things first, you can't have a space between - and F or v.
I was going to suggest sed, but as you only want to translate the first column, that's not as easy.
Unfortunately, awk doesn't have built-in tr functionality, so you'd have to use the shell like you are and just close the pipe:
awk -F "|" -v OFS="|" '{
command="echo \"\\"$1"\\\" | tr \" 0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ\" \" QWERTYUIOPASDFGHJKLZXCVBNM9876543210mnbvcxzlkjhgfdsapoiuytrewq\""
command | getline $1
close(command)
}1' FILE.CSV > test.CSV
However, I suggest using perl, which can do field splitting and character translation:
perl -F'\|' -lane '$F[0] =~ tr/0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ/QWERTYUIOPASDFGHJKLZXCVBNM9876543210mnbvcxzlkjhgfdsapoiuytrewq/; print join("|", #F)' FILE.CSV > test.CSV
Or, for a shorter command line, just put the program into a file, drop the e in -lane and use the file name instead of the '...' command.
you can do the mapping in awk instead of making a system call for each line, or perhaps simply
paste -d'|' <(cut -d'|' -f1 file | tr '0-9' 'a-z') <(cut -d'|' -f2- file)
replace the tr arguments with yours.
This does not answer your question, but you can implement tr as an awk function that would save having to spawn lots of external processes
$ cat tr.awk
function tr(str, from, to, s,i,c,idx) {
s = ""
for (i=1; i<=length($str); i++) {
c = substr(str, i, 1)
idx = index(from, c)
s = s (idx == 0 ? c : substr(to, idx, 1))
}
return s
}
{
print $1, tr($1,
" 0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ",
" QWERTYUIOPASDFGHJKLZXCVBNM9876543210mnbvcxzlkjhgfdsapoiuytrewq")
}
Example:
$ printf "%s\n" hello wor-ld | awk -f tr.awk
hello KGCCN
wor-ld 3N8-CF
Related
Let's say I have a file like so:
test.txt
one
two
three
I'd like to get the following output: one|two|three
And am currently using this command: gawk -v ORS='|' '{ print $0 }' test.txt
Which gives: one|two|three|
How can I print it so that the last | isn't there?
Here's one way to do it:
$ seq 1 | awk -v ORS= 'NR>1{print "|"} 1; END{print "\n"}'
1
$ seq 3 | awk -v ORS= 'NR>1{print "|"} 1; END{print "\n"}'
1|2|3
With paste:
$ seq 1 | paste -sd'|'
1
$ seq 3 | paste -sd'|'
1|2|3
Convert one column to one row with field separator:
awk '{$1=$1} 1' FS='\n' OFS='|' RS='' file
Or in another notation:
awk -v FS='\n' -v OFS='|' -v RS='' '{$1=$1} 1' file
Output:
one|two|three
See: 8 Powerful Awk Built-in Variables – FS, OFS, RS, ORS, NR, NF, FILENAME, FNR
awk solutions work great. Here is tr + sed solution:
tr '\n' '|' < file | sed 's/\|$//'
1|2|3
just flatten it :
gawk/mawk 'BEGIN { FS = ORS; RS = "^[\n]*$"; OFS = "|"
} NF && ( $NF ? NF=NF : —-NF )'
ascii | = octal \174 = hex 0x7C. The reason for —-NF is that more often than not, the input includes a trailing new line, which makes field count 1 too many and result in
1|2|3|
Both NF=NF and --NF are similar concepts to $1=$1. Empty inputs, regardless of whether trailing new lines exist or not, would result in nothing printed.
At the OFS spot, you can delimit it with any string combo you like instead of being constrained by tr, which has inconsistent behavior. For instance :
gtr '\012' '高' # UTF8 高 = \351\253\230 = xE9 xAB x98
on bsd-tr, \n will get replaced by the unicode properly 1高2高3高 , but if you're on gnu-tr, it would only keep the leading byte of the unicode, and result in
1 \351 2 \351 . . .
For unicode equiv-classes, bsd-tr works as expected while gtr '[=高=]' '\v' results in
gtr: ?\230: equivalence class operand must be a single character
and if u attempt equiv-classes with an arbitrary non-ASCII byte, bsd-tr does nothing while gnu-tr would gladly oblige, even if it means slicing straight through UTF8-compliant characters :
g3bn 77138 | (g)tr '[=\224=]' '\v'
bsd-tr : 77138=Koyote 코요태 KYT✜ 高耀太
gnu-tr : 77138=Koyote ?
?
태 KYT✜ 高耀太
I would do it following way, using GNU AWK, let test.txt content be
one
two
three
then
awk '{printf NR==1?"%s":"|%s", $0}' test.txt
output
one|two|three
Explanation: If it is first line print that line content sans trailing newline, otherwise | followed by line content sans trailing newline. Note that I assumed that test.txt has not trailing newline, if this is not case test this solution before applying it.
(tested in gawk 5.0.1)
Also you can try this with awk:
awk '{ORS = (NR%3 ? "|" : RS)} 1' file
one|two|three
% is the modulo operator and NR%3 ? "|" : RS is a ternary expression.
See Ed Morton's explanation here: https://stackoverflow.com/a/55998710/14259465
With a GNU sed, you can pass -z option to match line breaks, and thus all you need is replace each newline but the last one at the end of string:
sed -z 's/\n\(.\)/|\1/g' test.txt
perl -0pe 's/\n(?!\z)/|/g' test.txt
perl -pe 's/\n/|/g if !eof' test.txt
See the online demo.
Details:
s - substitution command
\n\(.\) - an LF char followed with any one char captured into Group 1 (so \n at the end of string won't get matched)
|\1 - a | char and the captured char
g - all occurrences.
The first perl command matches any LF char (\n) not at the end of string ((?!\z)) after slurping the whole file into a single string input (again, to make \n visible to the regex engine).
The second perl command replaces an LF char at the end of each line except the one at the end of file (eof).
To make the changes inline add -i option (mind this is a GNU sed example):
sed -i -z 's/\n\(.\)/|\1/g' test.txt
perl -i -0pe 's/\n(?!\z)/|/g' test.txt
perl -i -pe 's/\n/|/g if !eof' test.txt
In a Linux script program, I've got the following awk command for other purposes and to rename the file.
cat $edifile | awk -F\| '
{ OFS = "|"
print $0
} ' | tr -d "\012" > $newname.hl7
While this is happening, I'd like to grab the 5th field of the MSH segment and save it for later use in the script. Is this possible?
If no, how could I do it later or earlier on?
Example of the segment.
MSH|^~\&|business1|business2|/u/tmp/TR0049-GE-1.b64|routing|201811302126||ORU^R01|20181130212105810|D|2.3
What I want to do is retrieve the path and file name in MSH 5 and concatenate it to the end of the new file.
I've used this to capture the data but no luck. If fpth is getting set, there is no evidence of it and I don't have the right syntax for an echo within the awk phrase.
cat $edifile | awk -F\| '
{ OFS = "|"
{fpth=$(5)}
print $0
} ' | tr -d "\012" > $newname.hl7
any suggestions?
Thank you!
Try
filename=`awk -F'|' '{print $5}' $edifile | head -1`
You can skip the piping through head if the file is a single line
First of all, it must be mentioned that the awk line in your first piece of code, has zero use:
$ cat $edifile | awk -F\| ' { OFS = "|"; print $0 }' | tr -d "\012" > $newname.hl7
This is totally equivalent to
$ cat $edifile | tr -d "\012" > $newname.hl7
because OFS is only used to redefine $0 if you redefine a field.
Example:
$ echo "a|b|c" | awk -F\| '{OFS="/"; print $0}'
a|b|c
$ echo "a|b|c" | awk -F\| '{OFS="/"; $1=$1; print $0}'
a/b/c
I understand that you have a hl7 file in which you have a single line starting with the string "MSH". From this line you want to store the 5th field: this is achieved in the following way:
fpth=$(awk -v outputfile="${newname}.hl7" '
BEGIN{FS="|"; ORS="" }
($1 == "MSH"){ print $5 }
{ print $0 > outputfile }' $edifile)
I have replaced ORS to an empty character set, as it is equivalent to tr -d "\012". The above will work very nicely if you only have a single MSH in your file.
I have a file with 100 columns of data. I want to print the first column and i-th column in 99 separate files, I am trying to use
for i in {2..99}; do awk '{print $1" " $i }' input.txt > data${i}; done
But I am getting errors
awk: illegal field $(), name "i"
input record number 1, file input.txt
source line number 1
How to correctly use $i inside the {print }?
Following single awk may help you too here:
awk -v start=2 -v end=99 '{for(i=start;i<=end;i++){print $1,$i > "file"i;close("file"i)}}' Input_file
An all awk solution. First test data:
$ cat foo
11 12 13
21 22 23
Then the awk:
$ awk '{for(i=2;i<=NF;i++) print $1,$i > ("data" i)}' foo
and results:
$ ls data*
data2 data3
$ cat data2
11 12
21 22
The for iterates from 2 to the last field. If there are more fields that you desire to process, change the NF to the number you'd like. If, for some reason, a hundred open files would be a problem in your system, you'd need to put the print into a block and add a close call:
$ awk '{for(i=2;i<=NF;i++){f=("data" i); print $1,$i >> f; close(f)}}' foo
If you want to do what you try to accomplish :
for i in {2..99}; do
awk -v x=$i '{print $1" " $x }' input.txt > data${i}
done
Note
the -v switch of awk to pass variables
$x is the nth column defined in your variable x
Note2 : this is not the fastest solution, one awk call is fastest, but I just try to correct your logic. Ideally, take time to understand awk, it's never a wasted time
I was trying get the total number of "??", " M", "A" and "D" from this:
?? this is a sentence
M this is another one
A more text here
D more and more text
I have this sample line of code but doesn't work:
awk -v pattern="\?\?" '{$1 == pattern} END{print " "FNR}'
$ awk '{ print $1 }' file | sort | uniq -c
1 ??
1 A
1 D
1 M
If for some reason you want an awk-only solution:
awk '{ ++cnt[$1] } END { for (i in cnt) print cnt[i], i }' file
but I think that's needlessly complicated compared to using the built-in unix tools that already do most of the work.
If you just want to count one particular value:
awk -v value='??' '$1 == value' file | wc -l
If you want to count only a subset of values, you can use a regex:
$ awk -v pattern='A|D|(\\?\\?)' '$1 ~ pattern { print $1 }' file | sort | uniq -c
1 ??
1 A
1 D
Here you do need to send a \ in order that the ?s are escaped within the regular expression. And because the \ is itself a special character within the string being passed to awk, you need to escape it first (hence the double backslash).
I have two files, one file is my data, and the other file is a list of line numbers that I want to extract from my data file. Can I use awk to read in my lines file, and then extract the lines that match the line numbers?
Example:
Data file:
This is the first line of my data
This is the second line of my data
This is the third line of my data
This is the fourth line of my data
This is the fifth line of my data
Line numbers file
1
4
5
Output:
This is the first line of my data
This is the fourth line of my data
This is the fifth line of my data
I've only ever used command line awk and sed for really simple stuff. This is way beyond me and I have been googling for an hour without an answer.
awk 'NR == FNR {nums[$1]; next} FNR in nums' numberfile datafile
simply referring to an array subscript creates the entry. Looping over the first file, while NR (record number) is equal to FNR (file record number) using the next statement stores all the line numbers in the array. After that when FNR of the second file is present in the array (true) the line is printed (which is the default action for "true").
One way with sed:
sed 's/$/p/' linesfile | sed -n -f - datafile
You can use the same trick with awk:
sed 's/^/NR==/' linesfile | awk -f - datafile
Edit - Huge files alternative
With regards to huge number of lines it is not prudent to keep whole files in memory. The solution in that case can be to sort the numbers-file and read one line at a time. The following has been tested with GNU awk:
extract.awk
BEGIN {
getline n < linesfile
if(length(ERRNO)) {
print "Unable to open linesfile '" linesfile "': " ERRNO > "/dev/stderr"
exit
}
}
NR == n {
print
if(!(getline n < linesfile)) {
if(length(ERRNO))
print "Unable to open linesfile '" linesfile "': " ERRNO > "/dev/stderr"
exit
}
}
Run it like this:
awk -v linesfile=$linesfile -f extract.awk infile
Testing:
echo "2
4
7
8
10
13" | awk -v linesfile=/dev/stdin -f extract.awk <(paste <(seq 50e3) <(seq 50e3 | tac))
Output:
2 49999
4 49997
7 49994
8 49993
10 49991
13 49988
Here is an awk example. inputfile is loaded up front, then matching records of datafile are output.
awk \
-v RS="[\r]*[\n]" \
-v FILE="inputfile" \
'BEGIN \
{
LINES = ","
while ((getline Line < FILE))
{
LINES = LINES Line ","
}
}
LINES ~ "," NR "," \
{
print
}
' datafile
I had the same problem. This is the solution already posted by Thor:
cat datafile \
| awk 'BEGIN{getline n<"numbers"} n==NR{print; getline n<"numbers"}'
If like me you don't have a numbers file, but it is instead passed on from stdin and you don't want to generate a temporary numbers file, then this is an alternative solution:
cat numbers \
| awk '{while((getline line<"datafile")>0) {n++; if(n==$0) {print line;next}}}'
This solution...
awk 'NR == FNR {nums[$1]; next} FNR in nums' numberfile datafile
...only prints unique numbers in the numberfile. What if the numberfile contains repeated entries? Then sed is a better (but much slower) alternative:
sed -nf <(sed 's/.*/&p/' numberfile) datafile
while read line; do echo $(sed -n '$(echo $line)p' Datafile.txt); done < numbersfile.txt