I have multiple count files that look like this:
File1.tab
6 10 0
49 0 53
15 0 15
0 0 0
0 0 0
0 0 0
Other file:
File2.tab
3 1 2
29 0 29
4 0 4
0 0 0
0 0 0
0 0 0
I have over 30 files and I want to combine the second column of each file into one big file.
I know this question have already been asked and I found a similar here How to combine column from multiple text files?
I used the answer from previous question for my problem:
paste *.tab | awk '{i=2;while($i); {printf("%d ",$i);i+=3}printf("\n")}'
The problem is that zero values are not printed, I get something like this:
10 1
and I want something like this:
10 1
0 0
0 0
0 0
0 0
0 0
I cheked the printf format specifiers, but none works. How can I solve this problem?
You picked a bad "answer" to build on. Try this:
paste *.tab |
awk '{for (i=2; i<=NF; i+=3) printf "%s%s", (i>2?OFS:""), $i; print ""}'
I have a log file with lots of unnecessary information. The only important part of that file is a table which describes some statistics. My goal is to have a script which will accept a column name as argument and return the sum of all the elements in the specified column.
Example log file:
.........
Skipped....
........
WARNING: [AA[409]: Some bad thing happened.
--- TOOL_A: READING COMPLETED. CPU TIME = 0 REAL TIME = 2
--------------------------------------------------------------------------------
----- TOOL_A statistics -----
--------------------------------------------------------------------------------
NAME Attr1 Attr2 Attr3 Attr4 Attr5
--------------------------------------------------------------------------------
AAA 885 0 0 0 0
AAAA2 1 0 2 0 0
AAAA4 0 0 2 0 0
AAAA8 0 0 2 0 0
AAAA16 0 0 2 0 0
AAAA1 0 0 2 0 0
AAAA8 0 0 23 0 0
AAAAAAA4 0 0 18 0 0
AAAA2 0 0 14 0 0
AAAAAA2 0 0 21 0 0
AAAAA4 0 0 23 0 0
AAAAA1 0 0 47 0 0
AAAAAA1 2 0 26 0
NOTE: Some notes
......
Skipped ......
The expected usage script.sh Attr1
Expected output:
888
I've tried to find something with sed/awk but failed to figure out a solution.
tldr;
$ cat myscript.sh
#!/bin/sh
logfile=${1}
attribute=${2}
field=$(grep -o "NAME.\+${attribute}" ${logfile} | wc -w)
sed -nre '/NAME/,/NOTE/{/NAME/d;/NOTE/d;s/\s+/\t/gp;}' ${logfile} | \
cut -f${field} | \
paste -sd+ | \
bc
$ ./myscript.sh mylog.log Attr3
182
Explanation:
assign command-line arguments ${1} and ${2} to the logfile and attribute variables, respectively.
with wc -w, count the quantity of words within the line that
contains both NAME and ${attribute} (the field index) and assign it to field
with sed
suppress automatic printing (-n) and enable extended regular expressions (-r)
find lines between the NAME and NOTE lines, inclusive
delete the lines that match NAME and NOTE
translate each contiguous run of whitespace to a single tab and print the result
cut using the field index
paste all numbers as an infix summation
evaluate the infix summation via bc
Quick and dirty (without any other spec)
awk -v CountCol=2 '/^[^[:blank:]]/ && NF == 6 { S += $( CountCol) } END{ print S + 0 }' YourFile
with column name
awk -v ColName='Attr1' '/^[[:blank:]]/ && NF == 6 { for(i=1;i<=NF;i++){if ( $i == ColName) CountCol = i } /^[^[:blank:]]/ && NF == 6 && CountCol{ S += $( CountCol) } END{ print S + 0 }' YourFile
you should add a header/trailer filter to avoid noisy line (a flag suit perfect for this) but lack of info about structure to set this flag, i use sthe simple field count (assuming text field have 0 as value so not changing the sum when taken in count)
$ awk -v col='Attr3' '/NAME/{for (i=1;i<=NF;i++) f[$i]=i} col in f{sum+=$(f[col]); if (!NF) {print sum+0; exit} }' file
182
I have a file which looks like follows:
ENSG00000197111:I12 0
ENSG00000197111:I12 1
ENSG00000197111:I13 0
ENSG00000197111:I18 0
ENSG00000197111:I2 0
ENSG00000197111:I3 0
ENSG00000197111:I4 0
ENSG00000197111:I5 0
ENSG00000197111:I5 1
I have some lines that are duplicated but I cannot remove by sort -u because the second column has different values for them (1 or 0). How do I remove such duplicates by keeping the lines with second column as 1 such that the file will be
ENSG00000197111:I12 1
ENSG00000197111:I13 0
ENSG00000197111:I18 0
ENSG00000197111:I2 0
ENSG00000197111:I3 0
ENSG00000197111:I4 0
ENSG00000197111:I5 1
you can use awk and or operator, if the order isn't mandatory
awk '{d[$1]=d[$1] || $2}END{for(k in d) print k, d[k]}' file
you get
ENSG00000197111:I2 0
ENSG00000197111:I3 0
ENSG00000197111:I4 0
ENSG00000197111:I5 1
ENSG00000197111:I12 1
ENSG00000197111:I13 0
ENSG00000197111:I18 0
Edit, only sort solution
You can use sort with a double pass, example
sort -k1,1 -k2,2r file | sort -u -k1,1
you get,
ENSG00000197111:I12 1
ENSG00000197111:I13 0
ENSG00000197111:I18 0
ENSG00000197111:I2 0
ENSG00000197111:I3 0
ENSG00000197111:I4 0
ENSG00000197111:I5 1
My data(tab separated):
1 0 0 1 0 1 1 0 1
1 1 0 1 0 1 0 1 1
1 1 1 1 1 1 1 1 1
0 0 0 0 0 0 0 0 0
...
how can i grep the lines with exact, for example, 5 '1's,
ideal output:
1 0 0 1 0 1 1 0 1
Also, how can i grep lines with equal or more than (>=) 5 '1's,
ideal output:
1 0 0 1 0 1 1 0 1
1 1 0 1 0 1 0 1 1
1 1 1 1 1 1 1 1 1
i tried,
grep 1$'\t'1$'\t'1$'\t'1$'\t'1
however this will only output consecutive '1's, which is not all i want.
i wonder if there will be any simple method to achieve this, thank you!
John Bollinger's helpful answer and anishane's answer show that it can be done with grep, but, as has been noted, that is quite cumbersome, given that regular expression aren't designed for counting.
awk, by contrast, is built for field-based parsing and counting (often combined with regular expressions to identify field separators, or, as below, the fields themselves).
Assuming you have GNU awk, you can use the following:
Exactly 5 1s:
awk -v FPAT='\\<1\\>' 'NF==5' file
5 or more 1s:
awk -v FPAT='\\<1\\>' 'NF>=5' file
Special variable FPAT is a GNU awk extension that allows you to identify fields via a regex that describes the fields themselves, in contrast with the standard approach of using a regex to define the separators between fields (via special variable FS or option -F):
'\\<1\\>' identifies any "isolated" 1 (surrounded by non-word characters) as a field, based on word-boundary assertions \< and \>; the \ must be doubled here so that the initial string parsing performed by awk doesn't "eat" single \s.
Standard variable NF contains the count of input fields in the line at hand, which allows easy numerical comparison. If the conditional evaluates to true, the input line at hand is implicitly printed (in other words: NF==5 is implicitly the same as NF==5 { print } and, more verbosely, NF==5 { print $0 }).
A POSIX-compliant awk solution is a little more complicated:
Exactly 5 1s:
awk '{ l=$0; gsub("[\t0]", "") }; length($0)==5 { print l }' file
5 or more 1s:
awk '{ l=$0; gsub("[\t0]", "") }; length($0)>=5 { print l }' file
l=$0 saves the input line ($0) in its original form in variable l.
gsub("[\t0]", "") replaces all \t and 0 chars. in the input line with the empty string, i.e., effectively removes them, and only leaves (directly concatenated) 1 instances (if any).
length($0)==5 { print l } then prints the original input line (l) only if the resulting string of 1s (i.e., the count of 1s now stored in the modified input line ($0)) matches the specified count.
You can use grep. But that would be an abuse of regex.
$ cat countme
1 0 0 1 0 1 1 0 1
1 1 0 1 0 1 0 1 1
1 1 1 1 1 1 1 1 1
0 0 0 0 0 0 0 0 0
$ grep -P '^[0\t]*(1[0\t]*){5}[0\t]*$' countme # Match exactly 5
1 0 0 1 0 1 1 0 1
$ grep -P '^[0\t]*(1[0\t]*){5,}[0\t]*$' countme # Match >=5
1 0 0 1 0 1 1 0 1
1 1 0 1 0 1 0 1 1
1 1 1 1 1 1 1 1 1
You can do this to get lines with exactly five '1's:
grep '^[^1]*\(1[^1]*\)\{5,5\}[^1]*$'
You can simplify that to this for at least five '1's:
grep '\(1[^1]*\)\{5,\}'
The enumerated quantifier (\{n,m\}) enables you to conveniently specify a particular number or range of numbers of consecutive matches to a sub-pattern. To avoid matching lines with extra matches to such a pattern, however, you must also anchor it to the beginning and end of the line.
The other other trick is to make sure the gaps previous to the first 1, between the 1s, and after the last 1 are matched. In your case, all of those gaps can be represented pretty simply as ranges of zero or more characters other than 1: [^1]*. Putting those pieces together gives you the above regular expressions.
Do
sed -nE '/^([^1]*1[^1]*){5}$/p' your_file
for exactly 5 matches and
sed -nE '/^([^1]*1[^1]*){5,}$/p' your_file
for 5 or more matches.
Note: In GNU sed you may not see the -E option in the manpage, but it is supported. Using -E is for portability to, say, Mac OSX.
with perl
$ perl -ane 'print if (grep {$_==1} #F) == 5' ip.txt
1 0 0 1 0 1 1 0 1
$ perl -ane 'print if (grep {$_==1} #F) >= 5' ip.txt
1 0 0 1 0 1 1 0 1
1 1 0 1 0 1 0 1 1
1 1 1 1 1 1 1 1 1
-a to automatically split input line on whitespaces and save to #F array
grep {$_==1} #F returns array with elements from #F array which are exactly equal to 1
(grep {$_==1} #F) == 5 in scalar context, comparison will be done based on number of elements of array
See http://perldoc.perl.org/perlrun.html#Command-Switches for details on -ane options
I want to extract some fields out of output of command xentop. It's like top command; provides an ongoing look at cpu usage,memory usage,...in real time.
If I run this command in batch mode, I will have its output as you see in a file:
NAME STATE CPU(sec) CPU(%) MEM(k) MEM(%) MAXMEM(k) MAXMEM(%) VCPUS NETS NETTX(k) NETRX(k) VBDS VBD_OO VBD_RD VBD_WR VBD_RSECT VBD_WSECT SSID
Domain-0 -----r 13700 33.0 7127040 85.9 no limit n/a 8 0 0 0 0 0 0 0 0 0 0
fed18 -----r 738 190.6 1052640 12.7 1052672 12.7 3 1 259919 8265 1 0 82432 22750 2740966 1071672 0
and running this
cat file| tr '\r' '\n' | sed 's/[0-9][;][0-9][0-9][a-Z]/ /g' | col -bx | awk '{print $1,$4,$6}'
on this file gives me what I want
NAME CPU(%) MEM(%)
Domain-0 33.0 85.9
fed18 190.6 12.7
but my script doesn't work on realtime output of xentop. I even tried to just run xentop one time by setting itteration option as 1(xentop -i 1) but It does not work!
How can I pipe output of xentop as "not" realtime to my script?
It may not be sending any output to the standard output stream. There are several ways of sending output to the screen without using stdout. A quick google search didn't provide much information about how it works internally.
I use xentop version 1.0 on xenserver 7.0 like :
[root#xen] xentop -V
xentop 1.0
[root#xen] cat /etc/centos-release
XenServer release 7.0.0-125380c (xenenterprise)
If you want to save the xentop output you can do it with '-b' (batch mode) and '-i' (number of iterations before exiting) options :
[root#xen] xentop -b -i 1
NAME STATE CPU(sec) CPU(%) MEM(k) MEM(%) MAXMEM(k) MAXMEM(%) VCPUS NETS NETTX(k) NETRX(k) VBDS VBD_OO VBD_RD VBD_WR VBD_RSECT VBD_WSECT SSID
Domain-0 -----r 132130 0.0 4194304 1.6 4194304 1.6 16 0 0 0 0 0 0 0 0 0 0
MY_VM --b--- 5652 0.0 16777208 6.3 16915456 6.3 4 0 0 0 1 - - - - - 0
[root#xen] xentop -b -i 1 > output.txt
[root#xen] cat output.txt
NAME STATE CPU(sec) CPU(%) MEM(k) MEM(%) MAXMEM(k) MAXMEM(%) VCPUS NETS NETTX(k) NETRX(k) VBDS VBD_OO VBD_RD VBD_WR VBD_RSECT VBD_WSECT SSID
Domain-0 -----r 132130 0.0 4194304 1.6 4194304 1.6 16 0 0 0 0 0 0 0 0 0 0
MY_VM --b--- 5652 0.0 16777208 6.3 16915456 6.3 4 0 0 0 1 - - - - - 0