I have a file named list.txt containing a (supplier,product) pair and I must show the number of products from every supplier and their names using Linux terminal
Sample input:
stationery:paper
grocery:apples
grocery:pears
dairy:milk
stationery:pen
dairy:cheese
stationery:rubber
And the result should be something like:
stationery: 3
stationery: paper pen rubber
grocery: 2
grocery: apples pears
dairy: 2
dairy: milk cheese
Save the input to file, and remove the empty lines. Then use GNU datamash:
datamash -s -t ':' groupby 1 count 2 unique 2 < file
Output:
dairy:2:cheese,milk
grocery:2:apples,pears
stationery:3:paper,pen,rubber
The following pipeline shoud do the job
< your_input_file sort -t: -k1,1r | sort -t: -k1,1r | sed -E -n ':a;$p;N;s/([^:]*): *(.*)\n\1:/\1: \2 /;ta;P;D' | awk -F' ' '{ print $1, NF-1; print $0 }'
where
sort sorts the lines according to what's before the colon, in order to ease the successive processing
the cryptic sed joins the lines with common supplier
awk counts the items for supplier and prints everything appropriately.
Doing it with awk only, as suggested by KamilCuk in a comment, would be a much easier job; doing it with sed only would be (for me) a nightmare. Using both is maybe silly, but I enjoyed doing it.
If you need a detailed explanation, please comment, and I'll find time to provide one.
Here's the sed script written one command per line:
:a
$p
N
s/([^:]*): *(.*)\n\1:/\1: \2 /
ta
P
D
and here's how it works:
:a is just a label where we can jump back through a test or branch command;
$p is the print command applied only to the address $ (the last line); note that all other commands are applied to every line, since no address is specified;
N read one more line and appends it to the current pattern space, putting a \newline in between; this creates a multiline in the pattern space
s/([^:]*): *(.*)\n\1:/\1: \2 / captures what's before the first colon on the line, ([^:]*), as well as what follows it, (.*), getting rid of eccessive spaces, *;
ta tests if the previous s command was successful, and, if this is the case, transfers the control to the line labelled by a (i.e. go to step 1);
P prints the leading part of the multiline up to and including the embedded \newline;
D deletes the leading part of the multiline up to and including the embedded \newline.
This should be close to the only awk code I was referring to:
< os awk -F: '{ count[$1] += 1; items[$1] = items[$1] " " $2 } END { for (supp in items) print supp": " count[supp], "\n"supp":" items[supp]}'
The awk script is more readable if written on several lines:
awk -F: '{ # for each line
# we use the word before the : as the key of an associative array
count[$1] += 1 # increment the count for the given supplier
items[$1] = items[$1] " " $2 # concatenate the current item to the previous ones
}
END { # after processing the whole file
for (supp in items) # iterate on the suppliers and print the result
print supp": " count[supp], "\n"supp":" items[supp]
}
Related
I am trying to analyze this IDS log file from MIT, found here.
Summarized attack: 41.084031
IDnum Date StartTime Duration Destination Attackname insider? manual? console?success? aDump? oDump iDumpBSM? SysLogs FSListing StealthyNew? Category OS
41.08403103/29/1999 08:18:35 00:04:07 172.016.112.050ps out rem succ aDmp oDmp iDmp BSM SysLg FSLst Stlth Old llU2R
41.08403103/29/1999 08:19:37 00:01:56 209.154.098.104ps out rem succ aDmp oDmp iDmp BSM SysLg FSLst Stlth Old llU2R
41.08403103/29/1999 08:29:27 00:00:43 172.016.112.050ps out rem succ aDmp oDmp iDmp BSM SysLg FSLst Stlth Old llU2R
41.08403103/29/1999 08:40:14 00:24:26 172.016.112.050ps out rem succ aDmp oDmp iDmp BSM SysLg FSLst Stlth Old llU2R
I am trying to write commands that do two things:
First, parse through the entire file and determine the amount of distinct "summarized attacks" that begin with 4x.xxxxx. I have accomplished this with:
grep -o -E "Summarized attack: 4". It returns 80.
Second, for each of the "Summarized Attacks" found by the above command, parse the table and determine the amount of IDnum rows, and return the total amount of rows (i.e., attacks) across all "Summarized attack" finds. I would imagine that number is somewhere around 200.
However, I am struggling to get the individual number of IDs, i.e., that are in the IDnum column of this text file.
Since it is a text file with technically no structure, how can I parse this .txt file as if it had a tabular structure to retrieve the total entries in the IDnum column, for each Summarized attack that follows the above grep command's search text?
Desired output would be a count of all IDnum's for the Summarized attacks found by the above command. I don't know the count, but I would imagine an integer output, similar to the return of 80 for grep -o -E "Summarized attack: 4". The output would be <int> where <int> is the # of "attacks" as defined by rows in the IDnum column across all 80 of the found "Summarized attacks" by the above grep command.
If another command other than grep is better suited, that is OK.
to count matches you can use grep -c
grep -cE '(^Summarized.attack:.4[0-9]\.[0-9]+$)'
you can use colon as delimiter for cut -d
(if you loop over results the leading whitespace does not care)
grep -oE '(^Summarized.attack:.4[0-9]\.[0-9]+$)' | cut -d: -f2
example loop
file="path/to/master-listfile-condensed.txt"
for var in $(grep -oE '(^Summarized.attack:.4[0-9]\.[0-9]+$)' "$file" | cut -d: -f2)
do
printf "Summarized attacks: %s: %s\n" $var \
$(grep -cE "(^.${var}[0-9]+/[0-9]{2}/[0-9]{4})" "$file")
done
^ start of line
$ end of line
. any byte (in this case single whitespace)
\. single dot (escaped)
[0-9] single digit
+ one (or more) occurrence
{4} four occurrence
Assuming you have more than one "Summarized attack:" in your input file this may be what you're looking for:
$ cat tst.awk
/^Summarized attack:/ {
prt()
atk = ($3 ~ /^4/ ? $3 : 0)
cnt = 0
}
atk { cnt++ }
END {
prt()
print "TOTAL", tot
}
function prt() {
if ( atk ) {
cnt -= 2
print atk, cnt
}
tot += cnt
}
.
$ awk -f tst.awk file
For your first part, fgrep -c "Summarized attacks: 4" or fgrep -F "Summarized attacks: 4" is sufficient.
If I understand your second part, for each of those blocks, you want to add up the attack rows and print a grand total. You can do that with
gawk '/^Summarized attack: 4/ { on=1; next} /^ 4[0-9.]*/ { if (on) ++ids; next} /^ IDnum/ {next} /^ */ {next} { on=0} END {print ids;}'< master-listfile-condensed.txt
The first statement says, search (/.../) for every line that begins with (^) "Summarized attack: 4", and upon finding it, turn on the "on" flag, and go to the next line. The second statement says, if this is an attack record (i.e. begins with 4 followed by a string [*] of digits), then check the flag; if it is on, count it. Basically, we want the flag to be on when we are in a stanza of target attack records. The next two statements say for every line that starts with " IDnum" or are all whitespace (sometimes blank lines are inserted), go to the next line; this is needed to counteract the next statement, which says that if this is not a line that matches any of the previous statements, turn off the "on" flag. This prevents us from counting attacks outside the target. Finally, END means at the end, print the grand total. I get 757 which is pretty far out of your range. But I think it is correct.
But a far easier way, assuming the Summarized timestamp is always repeated in the IDnum at least to the first significant digit, would be to use
grep -Ec '^ 4' master-listfile-condensed.txt
That means count all the lines that begin with space-4. In this case it gives us the correct result.
Im new to bash and trying to extract a list of patterns from file:
File1.txt
ABC
BDF
GHJ
base.csv (tried comma separated and tab delimited)
line 1,,,,"hfhf,ferf,ju,ABC"
line 2 ,,,,,"ewy,trggt,gtg,ABC,RFR"
line 3 .."himk,n,hn.ujj., BDF"
etc
Suggested output is smth like
ABC
line 1..
line 2..(whole lines)
BDF
line 3..
and so on for each pattern from file 1
the code i tried was:
#!/bin/bash
for i in *.txt -# cycle through all files containing pattern lists
do
for q in "$i"; # # cycle through list
do
echo $q >>output.${i};
grep -f "${q}" base.csv >>output.${i};
echo "\n";
done
done
But output is only filename and then some list of strings without pattern names, e.g.
File1.txt
line 1...
line 2...
line 3..
so i don`t know to what pattern belongs each string and have to check and assign manually. Can you please point out my errors? Thanks!
grep can process multiple files in one go, and then has the attractive added bonus of indicating which file it found a match in.
grep -f File1.txt base.csv >output.txt
It's not clear what you hope for the inner loop to do; it will just loop over a single token at a time, so it's not really a loop at all.
If you want the output to be grouped per pattern, here's a for loop which looks for one pattern at a time:
while read -r pat; do
echo "$pat"
grep "$pat" *.txt
done <File1.txt >output.txt
But the most efficient way to tackle this is to write a simple Awk script which processes all the input files at once, and groups the matches before printing them.
An additional concern is anchoring. grep "ABC" will find a match in 123DEABCXYZ; is this something you want to avoid? You can improve the regex, or, again, turn to Awk which gives you more control over where exactly to look for a match in a structured line.
awk '# Read patterns into memory
NR==FNR { a[++i] = $1; next }
# Loop across patterns
{ for(j=1; j<=i; ++j)
if($0 ~ a[j]) {
print FILENAME ":" FNR ":" $0 >>output.a[j]
next }
}' File1.txt base.csv
You're not actually reading the files, you're just handling the filenames. Try this:
#!/bin/bash
for i in *.txt # cycle through all files containing pattern lists
do
while read -r q # read file line by line
do
echo "$q" >>"output.${i}"
grep -f "${q}" base.csv >>"output.${i}"
echo "\n"
done < "${i}"
done
Here is one that separates (with split, comma-separatd with quotes and spaces stripped off) words from file2 to an array (word[]) and stores the record names (line 1 etc.) to it comma-separated:
awk '
NR==FNR {
n=split($0,tmp,/[" ]*(,|$)[" ]*/) # split words
for(i=2;i<=n;i++) # after first
if(tmp[i]!="") # non-empties
word[tmp[i]]=word[tmp[i]] (word[tmp[i]]==""?"":",") tmp[1] # hash rownames
record[tmp[1]]=$0 # store records
next
}
($1 in word) { # word found
n=split(word[$1],tmp,",") # get record names
print $1 ":" # output word
for(i=1;i<=n;i++) # and records
print record[tmp[i]]
}' file2 file1
Output:
ABC:
line 1,,,,"hfhf,ferf,ju,ABC"
line 2 ,,,,,"ewy,trggt,gtg,ABC,RFR"
BDF:
line 3 .."himk,n,hn.ujj., BDF"
Thank you for your kind help, my friends.
Tried both variants above but kept getting various errors ( "do" expected) or misbehavior ( gets names of pattern blocks, eg ABC, BDF, but no lines.
Gave up for a while and then eventually tried another way
While base goal were to cycle through pattern list files, search for patterns in huge file and write out specific columns from lines found - i simply wrote
for *i in *txt # cycle throughfiles w/ patterns
do
grep -F -f "$i" bigfile.csv >> ${i}.out1 #greps all patterns from current file
cut -f 2,3,4,7 ${i}.out1>> ${i}.out2 # cuts columns of interest and writes them out to another file
done
I'm aware that this code should be improved using some fancy pipeline features, but it works perfectly as is, hope it`ll help somebody in similar situation. You can easily add some echoes to write out pattern list names as i initially requested
XYZNA0000778800Z
16123000012300321000000008000000000000000
16124000012300322000000007000000000000000
17234000012300323000000005000000000000000
17345000012300324000000004000000000000000
17456000012300325000000003000000000000000
9
XYZNA0000778900Z
16123000012300321000000008000000000000000
16124000012300322000000007000000000000000
17234000012300323000000005000000000000000
17345000012300324000000004000000000000000
17456000012300325000000003000000000000000
9
I have above file format from which I want to find a matching record. For example, match a number(7789) on line starting with XYZ and once matched look for a matching number (7345) in lines below starting with 1 until it reaches to line starting with 9. retrieve the entire line record. How can I accomplish this using shell script, awk, sed or any combination.
Expected Output:
XYZNA0000778900Z
17345000012300324000000004000000000000000
With sed one can do:
$ sed -n '/^XYZ.*7789/,/^9$/{/^1.*7345/p}' file
17345000012300324000000004000000000000000
Breakdown:
sed -n ' ' # -n disabled automatic printing
/^XYZ.*7789/, # Match line starting with XYZ, and
# containing 7789
/^1.*7345/p # Print line starting with 1 and
# containing 7345, which is coming
# after the previous match
/^9$/ { } # Match line that is 9
range { stuff } will execute stuff when it's inside range, in this case the range is starting at /^XYZ.*7789/ and ending with /^9$/.
.* will match anything but newlines zero or more times.
If you want to print the whole block matching the conditions, one can use:
$ sed -n '/^XYZ.*7789/{:s;N;/\n9$/!bs;/\n1.*7345/p}' file
XYZNA0000778900Z
16123000012300321000000008000000000000000
16124000012300322000000007000000000000000
17234000012300323000000005000000000000000
17345000012300324000000004000000000000000
17456000012300325000000003000000000000000
9
This works by reading lines between ^XYZ.*7779 and ^9$ into the pattern
space. And then printing the whole thing if ^1.*7345 can be matches:
sed -n ' ' # -n disables printing
/^XYZ.*7789/{ } # Match line starting
# with XYZ that also contains 7789
:s; # Define label s
N; # Append next line to pattern space
/\n9$/!bs; # Goto s unless \n9$ matches
/\n1.*7345/p # Print whole pattern space
# if \n1.*7345 matches
I'd use awk:
awk -v rid=7789 -v fid=7345 -v RS='\n9\n' -F '\n' 'index($1, rid) { for(i = 2; i < $NF; ++i) { if(index($i, fid)) { print $i; next } } }' filename
This works as follows:
-v RS='\n9\n' is the meat of the whole thing. Awk separates its input into records (by default lines). This sets the record separator to \n9\n, which means that records are separated by lines with a single 9 on them. These records are further separated into fields, and
-F '\n' tells awk that fields in a record are separated by newlines, so that each line in a record becomes a field.
-v rid=7789 -v fid=7345 sets two awk variables rid and fid (meant by me as record identifier and field identifier, respectively. The names are arbitrary.) to your search strings. You could encode these in the awk script directly, but this way makes it easier and safer to replace the values with those of a shell variables (which I expect you'll want to do).
Then the code:
index($1, rid) { # In records whose first field contains rid
for(i = 2; i < $NF; ++i) { # Walk through the fields from the second
if(index($i, fid)) { # When you find one that contains fid
print $i # Print it,
next # and continue with the next record.
} # Remove the "next" line if you want all matching
} # fields.
}
Note that multi-character record separators are not strictly required by POSIX awk, and I'm not certain if BSD awk accepts it. Both GNU awk and mawk do, though.
EDIT: Misread question the first time around.
an extendable awk script can be
$ awk '/^9$/{s=0} s&&/7345/; /^XYZ/&&/7789/{s=1} ' file
set flag s when line starts with XYZ and contains 7789; reset when line is just 9, and print when flag is set and contains pattern 7345.
This might work for you (GNU sed):
sed -n '/^XYZ/h;//!H;/^9/!b;x;/^XYZ[^\n]*7789/!b;/7345/p' file
Use the option -n for the grep-like nature of sed. Gather up records beginning with XYZ and ending in 9. Reject any records which do not have 7789 in the header. Print any remaining records that contain 7345.
If the 7345 will always follow the header,this could be shortened to:
sed -n '/^XYZ/h;//!H;/^9/!b;x;/^XYZ[^\n]*7789.*7345/p' file
If all records are well-formed (begin XYZ and end in 9) then use:
sed -n '/^XYZ/h;//!H;/^9/!b;x;/^[^\n]*7789.*7345/p' file
I am having this example of ini file. I need to extract the names between 2 patterns Name_Z1 and OBJ=Name_Z1 and put them each on a line.
The problem is that there are more than one occurences with Name_Z1 and OBJ=Name_Z1 and i only need first occurence.
[Name_Z5]
random;text
Names;Jesus;Tom;Miguel
random;text
OBJ=Name_Z5
[Name_Z1]
random;text
Names;Jhon;Alex;Smith
random;text
OBJ=Name_Z1
[Name_Z2]
random;text
Names;Chris;Mara;Iordana
random;text
OBJ=Name_Z2
[Name_Z1_Phone]
random;text
Names;Bill;Stan;Mike
random;text
OBJ=Name_Z1_Phone
My desired output would be:
Jhon
Alex
Smith
I am currently writing a more ample script in bash and i am stuck on this. I prefer awk to do the job.
My greatly appreciation for who can help me. Thank you!
For Wintermute solution: The [Name_Z1] part looks like this:
[CAB_Z1]
READ_ONLY=false
FilterAttr=CeaseTime;blank|ObjectOfReference;contains;511047;512044;513008;593026;598326;CL5518;CL5521;CL5538;CL5612;CL5620|PerceivedSeverity;=;Critical;Major;Minor|ProbableCause;!=;HOUSE ALARM;IO DEVICE|ProblemText;contains;AIRE;ALIMENTA;BATER;CONVERTIDOR;DISTRIBUCION;FUEGO;HURTO;MAINS;MALLO;MAYOR;MENOR;PANEL;TEMP
NAME=CAB_Z1
And the [Name_Z1_Phone] part looks like this:
[CAB_Z1_FUEGO]
READ_ONLY=false
FilterAttr=CeaseTime;blank|ObjectOfReference;contains;511047;512044;513008;593026;598326;CL5518;CL5521;CL5538;CL5612;CL5620|PerceivedSeverity;=;Critical;Major;Minor|ProbableCause;!=;HOUSE ALARM;IO DEVICE|ProblemText;contains;FUEGO
NAME=CAB_Z1_FUEGO
The fix should be somewhere around the "|PerceivedSeverity"
Expected Output:
511047
512044
513008
593026
598326
CL5518
CL5521
CL5538
CL5612
CL5620
This should work:
sed -n '/^\[Name_Z1/,/^OBJ=Name_Z1/ { /^Names/ { s/^Names;//; s/;/\n/g; p; q } }' foo.txt
Explanation: Written readably, the code is
/^\[Name_Z1/,/^OBJ=Name_Z1/ {
/^Names/ {
s/^Names;//
s/;/\n/g
p
q
}
}
This means: In the pattern range /^\[Name_Z1/,/^OBJ=Name_Z1/, for all lines that match the pattern /^Names/, remove the Names; in the beginning, then replace all remaining ; with newlines, print the whole thing, and then quit. Since it immediately quits, it will only handle the first such line in the first such pattern range.
EDIT: The update made things a bit more complicated. I suggest
sed -n '/^\[CAB_Z1/,/^NAME=CAB_Z1/ { /^FilterAttr=/ { s/^.*contains;\(.*\)|PerceivedSeverity.*$/\1/; s/;/\n/g; p; q } }' foo.txt
The main difference is that instead of removing ^Names from a line, the substitution
s/^.*contains;\(.*\)|PerceivedSeverity.*$/\1/;
is applied. This isolates the part between contains; and |PerceivedSeverity before continuing as before. It assumes that there is only one such part in the line. If the match is ambiguous, it will pick the one that appears last in the line.
An (g)awk way that doesn't need a set number of fields(although i have assumed that contains; will always be on the line you need the names from.
(g)awk '(x+=/Z1/)&&match($0,/contains;([^|]+)/,a)&&gsub(";","\n",a[1]){print a[1];exit}' f
Explanation
(x+=/Z1/) - Increments x when Z1 is found. Also part of a
condition so x must exist to continue.
match($0,/contains;([^|]+)/,a) - Matches contains; and then captures everything after
up to the |. Stores the capture in a. Again a
condition so must succeed to continue.
gsub(";","\n",a[1]) - Substitutes all the ; for newlines in the capture
group a[1].
{print a[1];exit}' - If all conditions are met then print a[1] and exit.
This way should work in (m)awk
awk '(x+=/Z1/)&&/contains/{split($0,a,"|");y=split(a[2],b,";");for(i=3;i<=y;i++)
print b[i];exit}' file
sed -n '/\[Name_Z1\]/,/OBJ=Name_Z1$/ s/Names;//p' file.txt | tr ';' '\n'
That is sed -n to avoid printing anything not explicitly requested. Start from Name_Z1 and finish at OBJ=Name_Z1. Remove Names; and print the rest of the line where it occurs. Finally, replace semicolons with newlines.
Awk solution would be
$ awk -F";" '/Name_Z1/{f=1} f && /Names/{print $2,$3,$4} /OBJ=Name_Z1/{exit}' OFS="\n" input
Jhon
Alex
Smith
OR
$ awk -F";" '/Name_Z1/{f++} f==1 && /Names/{print $2,$3,$4}' OFS="\n" input
Jhon
Alex
Smith
-F";" sets the field seperator as ;
/Name_Z1/{f++} matches the line with pattern /Name_Z1/ If matched increment {f++}
f==1 && /Names/{print $2,$3,$4} is same as if f == 1 and maches pattern Name with line if true, then print the the columns 2 3 and 4 (delimted by ;)
OFS="\n" sets the output filed seperator as \n new line
EDIT
$ awk -F"[;|]" '/Z1/{f++} f==1 && NF>1{for (i=5; i<15; i++)print $i}' input
511047
512044
513008
593026
598326
CL5518
CL5521
CL5538
CL5612
CL5620
Here is a more generic solution for data in group of blocks.
This awk does not need the end tag, just the start.
awk -vRS= -F"\n" '/^\[Name_Z1\]/ {n=split($3,a,";");for (i=2;i<=n;i++) print a[i];exit}' file
Jhon
Alex
Smith
How it works:
awk -vRS= -F"\n" ' # By setting RS to nothing, one record equals one block. Then FS is set to one line as a field
/^\[Name_Z1\]/ { # Search for block with [Name_Z1]
n=split($3,a,";") # Split field 3, the names and store number of fields in variable n
for (i=2;i<=n;i++) # Loop from second to last field
print a[i] # Print the fields
exit # Exits after first find
' file
With updated data
cat file
data
[CAB_Z1_FUEGO]
READ_ONLY=false
FilterAttr=CeaseTime;blank|ObjectOfReference;contains;511047;512044;513008;593026;598326;CL5518;CL5521;CL5538;CL5612;CL5620|PerceivedSeverity;=;Critical;Major;Minor|ProbableCause;!=;HOUSE ALARM;IO DEVICE|ProblemText;contains;FUEGO
NAME=CAB_Z1_FUEGO
data
awk -vRS= -F"\n" '/^\[CAB_Z1_FUEGO\]/ {split($3,a,"|");n=split(a[2],b,";");for (i=3;i<=n;i++) print b[i]}' file
511047
512044
513008
593026
598326
CL5518
CL5521
CL5538
CL5612
CL5620
The following awk script will do what you want:
awk 's==1&&/^Names/{gsub("Names;","",$0);gsub(";","\n",$0);print}/^\[Name_Z1\]$/||/^OBJ=Name_Z1$/{s++}' inputFileName
In more detail:
s==1 && /^Names;/ {
gsub ("Names;","",$0);
gsub(";","\n",$0);
print
}
/^\[Name_Z1\]$/ || /^OBJ=Name_Z1$/ {
s++
}
The state s starts with a value of zero and is incremented whenever you find one of the two lines:
[Name_Z1]
OBJ=Name_Z1
That means, between the first set of those lines, s will be equal to one. That's where the other condition comes in. When s is one and you find a line starting with Names;, you do two substitutions.
The first is to get rid of the Names; at the front, the second is to replace all ; semi-colon characters with a newline. Then you print it out.
The output for your given test data is, as expected:
Jhon
Alex
Smith
I'm trying to delete two lines either side of a pattern match from a file full of transactions. Ie. find the match then delete two lines before it, then delete two lines after it and then delete the match. The write this back to the original file.
So the input data is
D28/10/2011
T-3.48
PINITIAL BALANCE
M
^
and my pattern is
sed -i '/PINITIAL BALANCE/,+2d' test.txt
However this is only deleting two lines after the pattern match and then deleting the pattern match. I can't work out any logical way to delete all 5 lines of data from the original file using sed.
an awk one-liner may do the job:
awk '/PINITIAL BALANCE/{for(x=NR-2;x<=NR+2;x++)d[x];}{a[NR]=$0}END{for(i=1;i<=NR;i++)if(!(i in d))print a[i]}' file
test:
kent$ cat file
######
foo
D28/10/2011
T-3.48
PINITIAL BALANCE
M
x
bar
######
this line will be kept
here
comes
PINITIAL BALANCE
again
blah
this line will be kept too
########
kent$ awk '/PINITIAL BALANCE/{for(x=NR-2;x<=NR+2;x++)d[x];}{a[NR]=$0}END{for(i=1;i<=NR;i++)if(!(i in d))print a[i]}' file
######
foo
bar
######
this line will be kept
this line will be kept too
########
add some explanation
awk '/PINITIAL BALANCE/{for(x=NR-2;x<=NR+2;x++)d[x];} #if match found, add the line and +- 2 lines' line number in an array "d"
{a[NR]=$0} # save all lines in an array with line number as index
END{for(i=1;i<=NR;i++)if(!(i in d))print a[i]}' #finally print only those index not in array "d"
file # your input file
sed will do it:
sed '/\n/!N;/\n.*\n/!N;/\n.*\n.*PINITIAL BALANCE/{$d;N;N;d};P;D'
It works this way:
if sed has only one string in pattern space it joins another one
if there are only two it joins the third one
if it does natch to pattern LINE + LINE + LINE with BALANCE it joins two following strings, deletes them and goes at the beginning
if not, it prints the first string from pattern and deletes it and goes at the beginning without swiping the pattern space
To prevent the appearance of pattern on the first string you should modify the script:
sed '1{/PINITIAL BALANCE/{N;N;d}};/\n/!N;/\n.*\n/!N;/\n.*\n.*PINITIAL BALANCE/{$d;N;N;d};P;D'
However, it fails in case you have another PINITIAL BALANCE in string which are going to be deleted. However, other solutions fails too =)
For such a task, I would probably reach for a more advanced tool like Perl:
perl -ne 'push #x, $_;
if (#x > 4) {
if ($x[2] =~ /PINITIAL BALANCE/) { undef #x }
else { print shift #x }
}
END { print #x }' input-file > output-file
This will remove 5 lines from the input file. These lines will be the 2 lines before the match, the matched line, and the two lines afterwards. You can change the total number of lines being removed modifying #x > 4 (this removes 5 lines) and the line being matched modifying $x[2] (this makes the match on the third line to be removed and so removes the two lines before the match).
A more simple and easy to understand solution might be:
awk '/PINITIAL BALANCE/ {print NR-2 "," NR+2 "d"}' input_filename \
| sed -f - input_filename > output_filename
awk is used to make a sed-script that deletes the lines in question and the result is written on the output_filename.
This uses two processes which might be less efficient than the other answers.
This might work for you (GNU sed):
sed ':a;$q;N;s/\n/&/2;Ta;/\nPINITIAL BALANCE$/!{P;D};$q;N;$q;N;d' file
save this code into a file grep.sed
H
s:.*::
x
s:^\n::
:r
/PINITIAL BALANCE/ {
N
N
d
}
/.*\n.*\n/ {
P
D
}
x
d
and run a command like this:
`sed -i -f grep.sed FILE`
You can use it so either:
sed -i 'H;s:.*::;x;s:^\n::;:r;/PINITIAL BALANCE/{N;N;d;};/.*\n.*\n/{P;D;};x;d' FILE