I have a file with many "start token lines" that starts with the word "MODEL", and many lines that end with an "end token line" that starts with the word "ENDMDL".
I would like to parse the file so that it grabs all lines starting from the "start token line" and ending at the "end token line" into a new output file.
In other words, if I ran this on a file with 100 of these "start token line" and "end token line", I would like to produce 100 files.
I have an awk command working:
awk '/MODEL/ {flag=1;next} /ENDMDL/{flag=0} flag {print}' 1KZS.pdb > TEST
However, this command just prints all lines between MODEL-->ENDMDL into the same output file. But I would like each MODEL-->ENDMDL to output in a new output file.
How can my awk command be tweaked to accomplish this?
This will do the trick:
$ awk '/MODEL/{f=1;s="FILE"++i;next}/ENDMDL/{f=0;close(s)}f{print > s}' 1KZS.pdb
Related
I am using the awk command in terminal on my Mac.
I want to print the contents of an already existing file and give a title to each column which i'll separate using a tab then I want to send the output to another file. What line of code would I use to give titles to the columns? Im hoping to use simple awk commands and preferably if I can complete the task in as little lines as possible.
So far I have tried to use the BEGIN command. (The titles I want to give are first name, second name and score)
BEGIN { print "First Name\tSecond Name\tScore}**
then I want to print the entire contents of the file.
{print} filename.txt
Finally I want to save the output to another file
End{print} filename.txt > output.txt
to do this all all together
awk 'BEGIN {print "First Name\tSecond Name\tScore";}
{print}
End{print}' filename.txt > output.txt
However, this only saved the titles to the output file and not the contents of the original file under the columns.
awk processes files line by line. Before it starts processing the file you can have it do something. We use the BEGIN keyword to note that the following block of code is to be executed before processing. Same with END running after the processing of each line of the file is complete.
While your code has some superfluous bits in it, like the unnecessary END block, it still should do exactly what you are wanting to do, assuming you have data in your filename.txt.
A more succinct awk code would be:
awk 'BEGIN {print "First Name\tSecond Name\tScore";}1' filename.txt > output.txt
In action (using commas instead of tabs because it's easier and I'm lazy):
$ echo "1,2,3" > filename.txt
$ awk 'BEGIN {print "c1,c2,c3"}1' filename.txt > output.txt
$ cat output.txt
c1,c2,c3
1,2,3
I'm trying to remove newline character in the first column in csv file using awk but it doesn't seems to work
Sample File:
"This
is a test
","Something","Something"
"This is
another
test","something","something"
"One
more
test","something","something"
The command i'm using is
awk -F, '{gsub("\n","",$1); print}' sample
The output doesn't remove the new line character
I'm looking for a solution using awk not sed or perl
Can someone please help?
The output required is,
"This is a test","something","something"
"This is another test","something","something"
"One more test","something","something"
Assuming what you have is a CSV exported from Excel or some other windows tool (since that's what it looks like) and so it has \r\n line endings, all you need is this with GNU awk for multi-char RS:
$ awk -v RS='\r\n' -F'\n' '{$1=$1}1' file
"This is a test ","Something","Something"
"This is another test","something","something"
"One more test","something","something"
Otherwise with GNU awk for multi-char RS this would work for the sample you posted:
$ awk -v RS='"\\s+("|$)' -F'\n' '{$1=$1; gsub(/^"?|"?$/,"\"")}1' file
"This is a test ","Something","Something"
"This is another test","something","something"
"One more test","something","something"
I am very new to using linux and I am trying to find/replace some of the text in my file.
I have successfully been able to find and replace "0/0" using gsub:
awk '{gsub(/0\/0/,"0")}; 1' filename
However, if I try to replace "./." using the same idea
awk '{gsub(/\.\/\./,"U")}; 1' filename
the output is truncated and stops at the location of the first "./." in the file. I know that "." is a special wildcard character, but I thought that having the "\" in front of it would neutralize it. I have searched but have been unable to find an explanation why the formula I used would truncate the file.
Any thoughts would be much appreciated. Thank you.
Recall that the basic outline of an awk is:
awk 'pattern { action }'
The most common patterns are regexes or tests against line counts:
awk '/FOO/ { do_something_with_a_line_with_FOO_in_it }'
awk 'FNR==10'
The last one has no action so the default is to print the line.
But functions that return a value are also useable as patterns. gsub is a function and returns the number of substitutions.
So given:
$ echo "$txt"
abc./.def line 1
ghk/lmn won't get printed
abc./.def abc./.def printed
To print only lines that have a successful substitution you can do:
$ echo "$txt" | awk 'gsub(/\.\/\./,"U")'
abcUdef line 1
abcUdef abcUdef printed
You do not need to put gsub into an action block since you want to run it on every line and the return tells you something about what happened. The lines that successfully are matched are printed since gsub returns the number of substitutions.
If you want every line printed regardless if there is a match:
$ echo "$txt" | awk 'gsub(/\.\/\./,"U") || 1'
abcUdef line 1
ghk/lmn won't get printed
abcUdef abcUdef printed
Or, you can use the function as an action with an empty pattern and then a 1 with an empty action:
$ echo "$txt" | awk '{gsub(/\.\/\./,"U")} 1'
abcUdef line 1
ghk/lmn won't get printed
abcUdef abcUdef printed
In either case, 1 as a pattern with no action prints the line regardless if there is a match and the gsub makes the substitution if any.
The second awk is what you have. Why it is not working on your input data is probably related to you input data.
Your awk script is fine, your input contains control-Ms, probably from being created by a Windows program. You can see them with cat -v file and use dos2unix or similar to remove them.
I have a long ASCII log-file from a simulation and need to extract some data from it.
The lines I want have the structure:
Main step= 1 a= 0.00E+00 b=-6.85E-08 c= 4.58E-08
The phrase "Main step" is only used in the lines I want. This is easy to grep for, but I also want to include the next line following the line above, which has the structure:
Fine step= 1 t=-1.31854E+01
Note that "Fine step" is used other places in the log-file.
My question boils down to this: How can I extract the lines containing a keyword/phrase (here "Main step") and also make sure that I get the next following line using grep or AWK or some other standard Linux program?
You can use sed
sed -n '/Main step/,/./p' inputFile
This prints only the lines in a range starting from Main step and ending with . (the wildcard). Effectively, every line which reads Main step and the following are printed.
Posted according to the tag awk. And the one through awk's getline function,
awk '/Main step/{print; getline; print}' file
It would print the Main step line and also the next line.
Because you tagged "grep", and since this is the most obvious solution to me:
grep -A1 'Main step' file
...although this will add "--" between matches. So to get the same output as the awk and sed answer:
grep -A1 'Main step' file | grep -v '^--$'
I am trying to work with an AWK script that was posted earlier on this forum. I am trying to split a large FASTA file containing multiple DNA sequences, into separate FASTA files. I need to separate each sequence into its own FASTA file, and the name of each of the new FASTA files needs to be the name of the DNA sequence from the original, large multifasta file (all the characters after the >).
I tried this script that I found here at stackoverflow:
awk '/^>chr/ {OUT=substr($0,2) ".fa"}; OUT {print >OUT}' your_input
It works well, but the DNA sequence begins directly after the name of the file- with no space. The DNA sequence needs to begin on a new line (regular FASTA format).
I would appreciate any help to solve this.
Thank you!!
Do you mean something like this?
awk '/^>chr/ {OUT=substr($0,2) ".fa";print " ">OUT}; OUT{print >OUT}' your_input
where the new file that is created for each "chromosome/sequence/thing" gets a blank line at the start?
I think this should work.
awk '/^>chr/ {OUT=substr($0,2) ".fa"}; {print >> OUT; close(OUT)}' Input_File
Hope this perl script could help.
#!/usr/bin/perl
open (INFILE, "< your_input.fa")
or die "Can't open file";
while (<INFILE>) {
$line = $_;
chomp $line;
if ($line =~ /\>/) { #if has fasta >
close OUTFILE;
$new_file = substr($line,1);
$new_file .= ".fa";
open (OUTFILE, ">$new_file")
or die "Can't open: $new_file $!";
}
print OUTFILE "$line\n";
}
close OUTFILE;
The .fa (or .fasta) format looks like:
>ID1
SEQUENCE
>ID2
SEQUENCE
When splitting a fasta file it is actually not desired inserting a new line character at its top. Therefore the answer of Pramod is more appropriate. Additionally, the ID can be defined more generally to match only the > character. Consequently, the complete line would be:
awk '/^>/ {OUT=substr($0,2) ".fa"}; {print >> OUT; close(OUT)}' Input_File
If you don't want to mess up your current directory with all the split files you can also output into a subdirectory (subdir):
awk '/^>/ {OUT="subdir/" substr($0,2) ".fa"}; {print >> OUT; close(OUT)}' Input_File
awk to split multi-sequence fasta file into separate sequence files
This problem is best approached by considering each sequence (complete with header) a single record and changing awk's default record separator RS (usually a line break) to be the unique (one per record) > symbol used to define the start of a header. As we will want to use the header text as a file name, and as fasta headers cannot contain line-breaks, it is also convenient to reset awk's default field separaor FS (usually white space) to be line breaks.
Both of these are done in an awk BEGIN block:
BEGIN{RS=">";FS="\n"}
Since the file begins with >, the first record will be empty and therefore must be ignored to prevent an error caused by trying to write to a file name extracted from an empty record. Thus, the main awk action block is filtered to only process records beginning with record number (NR) 2. This is achieved by placing a condition before the action block as follows:
NR>1{ ... }
with the record separator set to > each record is a whole sequence including its header, and each is split into fields at line breaks (because we set the field separator to "\n"). Thus, field 1 ($1) of each record contains the text we wish to use as filenames. Note the record separator (>) is no longer part of any field and so the entire first field can be used to build the filename. In this example, ".fasta" has been appended as a file extension:
fnme=$1 ".fasta";
Next, the fasta header marker ">" is printed, followed by the entire record ($0) to the filename fnme just formed, using awk's > redirect:
print ">" $0 > fnme;
lastly, the file is closed to prevent awk exceeding the system limit for the number of open files allowed, if many files are to be written (see footnote):
close(fnme);
whole procedure
awk command
awk 'BEGIN{RS=">";FS="\n"} NR>1{fnme=$1".fasta"; print ">" $0 > fnme; close(fnme);}' example.fasta
Tested on the following mock file named example.fasta:
>DNA sequence 1
GCAAAAGAACCGCCGCCACTGGTCGTGAAAGTGGTCGATCCAGTGACATCCCAGGTGTTGTTAAATTGAT
CATGGGCAGTGGCGGTGTAGGCTTGAGTACTGGCTACAACAACACTCGCACTACCCGGAGTGATAGTAAT
GCCGGTGGCGGTACCATGTACGGTGGTGAAGT
>DNA sequence 2
TCCCAGCCAGCAGGTAGGGTCAAAACATGCAAGCCGGTGGCGATTCCGCCGACAGCATTCTCTGTAATTA
ATTGCTACCAGCGCGATTGGCGCCGCGACCAGGATCCTTTTTAACCATTTCAGAAAACCATTTGAGTCCA
TTTGAACCTCCATCTTTGTTC
>DNA sequence 3
AACAAAAGAATTAGAGATATTTAACTCCACATTATTAAACTTGTCAATAACTATTTTTAACTTACCAGAA
AATTTCAGAATCGTTGCGAAAAATCTTGGGTATATTCAACACTGCCTGTATAACGAAACACAATAGTACT
TTAGGCTAACTAAGAAAAAACTTT
results (terminal commands and output)
$ ls
'DNA sequence 1.fasta' 'DNA sequence 3.fasta'
'DNA sequence 2.fasta' example.fasta
$ cat DNA\ sequence\ 1.fasta
>DNA sequence 1
GCAAAAGAACCGCCGCCACTGGTCGTGAAAGTGGTCGATCCAGTGACATCCCAGGTGTTGTTAAATTGAT
CATGGGCAGTGGCGGTGTAGGCTTGAGTACTGGCTACAACAACACTCGCACTACCCGGAGTGATAGTAAT
GCCGGTGGCGGTACCATGTACGGTGGTGAAGT
$ cat DNA\ sequence\ 2.fasta
>DNA sequence 2
TCCCAGCCAGCAGGTAGGGTCAAAACATGCAAGCCGGTGGCGATTCCGCCGACAGCATTCTCTGTAATTA
ATTGCTACCAGCGCGATTGGCGCCGCGACCAGGATCCTTTTTAACCATTTCAGAAAACCATTTGAGTCCA
TTTGAACCTCCATCTTTGTTC
$ cat DNA\ sequence\ 3.fasta
>DNA sequence 3
AACAAAAGAATTAGAGATATTTAACTCCACATTATTAAACTTGTCAATAACTATTTTTAACTTACCAGAA
AATTTCAGAATCGTTGCGAAAAATCTTGGGTATATTCAACACTGCCTGTATAACGAAACACAATAGTACT
TTAGGCTAACTAAGAAAAAACTTT
footnote
"To write numerous files, successively, in the same awk program. If the files aren’t closed, eventually awk may exceed a system limit on the number of open files in one process. It is best to close each one when the program has finished writing it."
quoted from https://www.gnu.org/software/gawk/manual/html_node/Close-Files-And-Pipes.html