I have a text file in the following format:
Model 1
Atom….
Atom….
Atom….
ENDMDL
Model 2
Atom….
Atom….
Atom….
ENDMDL
Model n
Atom….
Atom….
Atom….
ENDMDL
I need to split this file into the files corresponding to each Model. The names of new files according to the Model’s number.
This is easily done using a small awk script.
#!/usr/bin/awk -f
# Write sections of the input file to separate files
# Written by PM 2Ring 2016.06.14
BEGIN{outbase = "outfile"}
/^Model/{outname = outbase $2}
{print > outname}
outbase is the base file name. It gets the Model number appended to it, so for your sample file the output files outfile1, outfile2, etc will get created. With a minor change to the script you could set outbase from the command line, using awk's -v option.
The heart of this script is
/^Model/{outname = outbase $2}
It says: If the current line starts with "Model" append the contents of field #2 to the outbase string, assigning the result to outname.
By default, awk process a file line by line, splitting each line into fields using whitespace as the field separator.
{print > outname}
simply prints the current line to the file whose name is stored in outname.
This script is small enough to write the whole thing on the command line:
awk 'BEGIN{outbase = "outfile"}; /^Model/{outname = outbase $2}; {print > outname}' infile.txt
You can actually supply multiple input file arguments and they will be handled correctly, as long as you don't have duplicated Model numbers.
I would probably go for this using csplit. This will work for a file called file.txt:
csplit -ksz file.txt '/^Model/' '{*}'
for xx in xx*
do
newname=$(awk '{print $2; exit}' "$xx")
test ! -f "$newname" && mv -f "$xx" "$newname"
done
The csplit splits file.txt into multiple parts based on the RE. Filenames are (by default) named as xx and a monotonically increasing numeric suffix. We look at each of these in turn and rename them to the model number found inside the file.
Any files matching xx* at the end of the loop contain duplicate model numbers (the renaming is performed on a first come first served basis).
#!/bin/bash
while read -r line
do
case $line in
Model*)
f="${line//[[:space:]]/}"
touch "$f" # file name without white spces
;;
ENDMDL)
:
;;
*)
echo "$line" >> "$f"
;;
esac
done < "$1"
Something like this. You should run it providing models file as argument: ./script_name models.txt
Note that as mentioned by @PM 2Ring this approach is slow specially if you have large files.
Model.............N(dots are space).I'd assume OP would not want all that space in filenames. – 123 Jun 14 '16 at 11:10readis rather slow. It obtains its input character by character, with a system call required for each character. It's really designed for interactive use, not text processing. Please see Why is using a shell loop to process text considered bad practice? for further details. – PM 2Ring Jun 14 '16 at 11:21