Let's assume I have the following text:
file.txt 4 minutes 11 seconds 102 msec
file_1.txt 5 minutes 10 seconds 100 msec
file_3_2.txt 0 minutes 10 seconds 100 msec
How do I delete all the words while retaining numbers and the filenames using sed?
file.txt 4 11 102
file_1.txt 5 10 100
file_3_2.txt 0 10 100
There are multiple ways of extracting the data that you require.
If you see each line as a number of fields delimited by a space, then the most basic tool you could use to solve your issue would be cut:
$ cut -d ' ' -f 1,2,4,6 file
file.txt 4 11 102
file_1.txt 5 10 100
file_3_2.txt 0 10 100
This extracts the 1st, 2nd, 4th, and 6th space-delimited field from the file.
Similarly, with awk:
$ awk '{ print $1,$2,$4,$6 }' file
file.txt 4 11 102
file_1.txt 5 10 100
file_3_2.txt 0 10 100
This is doing the same thing, but there's slightly more going into it. The awk tool treats its input as a stream of records made up by fields. By default, a record is simply a line and a field is any non-blank text delimited by blanks (spaces or tabs, possibly more than one). The output is created, one record at a time (by default, a line at a time), and the fields are separated by a single space.
A slightly more involved approach would be to cut out the first field from the rest of the data, delete anything which is a letter in the rest of the data, and then paste the two together again.
$ paste -d ' ' \
<( cut -d ' ' -f 1 file ) \
<( cut -d ' ' -f 2- file | tr -s '[:alpha:]' '[ *]' )
file.txt 4 11 102
file_1.txt 5 10 100
file_3_2.txt 0 10 100
This is a bit inelegant though, because it reads the input twice. The tr command used will change all alphabetical characters to spaces, and then "squeeze" (with -s) any set of multiple consecutive spaces down into single spaces.
The paste command gets two input streams and concatenates these line by line with a space character in-between them. The first stream is provided via a process substitution (<( ... )) running a cut command that extract only the first column from the data. The second stream is provided by another process substitution running cut to extract all columns from column 2 on, and which then replaces alphabetical characters with spaces in these columns using tr as just previously described.
paste pipeline works?
It's unclear how specific your example is of your problem. Assuming it is literally what you are asking (which makes me worried it might be a class exercise) then you can take advantage of the fact that you have only 3 known strings to remove and that you want to end up with only single spaces between symbols to create a pretty short sed command. This example is using BSD sed supplied with MacOS, input is in the file input.txt and output is to stdout.
sed -E -e 's/ (minutes|seconds|msec)//g' < input.txt
This will match on a single space followed by one of your target strings and then replace it with nothing. The 'g' on the end says to apply repeatedly to the same line.
.out files with program execution times which look like this: CH3F_opt.out 0 days 0 hours 0 minutes 10 seconds 850 msec. As you see it is a bit longer, but still your suggestion works for me
Your problem can be tackled by the following GNU sed command:
$ sed -Ee ' /(\s+\S+)\s+\S+/\1/g' file
Where we keep alternating space-word and reject the next space-word combo.
One solution is to remove all "one space" followed by "non-numbers words":
$ sed -E 's/ \<[^0-9]*\>//g' file
file.txt 4 11 102
file_1.txt 5 10 100
file_3_2.txt 0 10 100