Abbreviate multiple column names in linux, keeping the last field

Question

I have a file where all the column headers are the path names. I want to abbreviate each column header from something that looks like:

/mydir/cat/dog/hen/test/block/sample1.so.rg.mk.bam /mydir/cat/dog/hen/test/block/sample2.so.rg.mk.bam

to:

sample1 sample2

How do I do this in linux? My files have anywhere from 46 to 100+ columns so manually editing column names is not an option. My desired file names are each 7 characters in length, as above.

Thanks

The header has the filename. Each column header/ name is

/mydir/cat/dog/hen/test/block/sample1.so.rg.mk.bam 

where I just want it to be

sample1

To clarify, this is one text file with 46 columns. Each column header or name appears as the lengthy string above and I want to truncate each header to the 7 character version, e.g. 'sample1'...'sampl46'

Desired Example file (with data under each column header)

sample1 sample2 sample3 sample4 sample5 ...  

Answer 1

I would write a short program to copy the original file into a new file with the short names. Keeping the original file will give you a back up should something go wrong. Exactly what you write depends on the language you are comfortable with. This may be you shell such as Bash, or any of a number of languages such as java, c, pearl, python, etc.

Here is some psudo code: old is the original file and new is the new file create new

begin a loop to read each  line in old
   read line from old
   delete all characters from line up to and including the last "/"
   delete delete all characters from line after the first 7
//This is what you want to save unless it conflicts with a previously saved line
   determine if you have a conflict.
   if there is a conflict
      add a number to the end of line to make it unique
   save line to new
   end of loop

Answer 2

Let's say I have a file with 4 columns and two lines:

host:~ # cat file2
/mydir/cat/dog/hen/test/block/sample1.so.rg.mk.bam /mydir/cat/dog/hen/test/block/sample2.so.rg.mk.bam /mydir/cat/dog/hen/test/block/sample3.so.rg.mk.bam /mydir/cat/dog/hen/test/block/sample4.so.rg.mk.bam
abc def ghi jkl

This command worked for me (not very handy, but still):

host:~ # sed -i -e 's/^\///g' -e 's/[[:alnum:]]\+\///g' -e 's/\.[[:alnum:]]\+//g' -e 's/\///g' file2
host:~ # cat file2
sample1 sample2 sample3 sample4
abc def ghi jkl

I'm sure there's a more efficient way, but you could give it a try.

Answer 3

Assuming the unwanted suffix is always ".so.rg.mk.bam", then GNU sed's evaluate command can be used to run basename on just the first line of filename, replacing it with the required output:

sed -i '1s/.*/basename -as .so.rg.mk.bam -a &/e' filename

For non-GNU seds, head can be used instead:

sed -i '1s/.*/'"$(basename -as .so.rg.mk.bam -a $(head -1 filename))"'/' filename

--

Note: To see the results without changing the file, try it without the -i first.

Answer 4

You could use awk to process the headers. The following awk script works only on the first line (NR==1). It loops though all of the fields in that line, one at a time. For each field, it performs the following steps:

1. Find the first instance of the text /sample and trim the text up to that (and through the /).
2. Find the first instance in the remainder of a period and trim off the portion from the period onwards.
3. If the remainder is too long, then trim the sample text down as much as needed. The equation how much of it to keep turns out to be "6 plus the position of the first digit minus the overall length".
4. Once we're done processing this field, print it out with a trailing space.
5. Once we're done looping through all of the fields, print a newline.

Note that this leaves you with a trailing space at the end of the line.

The awk script:

NR == 1 {
  for(i=1; i <= NF; i++) {
    tail=substr($i, 1 + match($i, "/sample"))   # delete up to the first instance of "/sample"
    tail=substr(tail, 1, index(tail, ".") - 1)  # find, then stop short of, the first period
    if (length(tail) > 7) {                     # if it's too long
        match(tail, "[0-9]")                    # find the first digit
                                                # trim the beginning down, then append the number
        tail=substr(tail, 1, 6 + RSTART - length(tail))substr(tail, RSTART)
    }
    printf tail" "
  }
  print ""
}

On a sample input of:

/mydir/cat/dog/hen/test/block/sample1.so.rg.mk.bam /mydir/cat/dog/hen/test/block/sample47.so.rg.mk.bam /mydir/cat/dog/hen/test/block/sample4631.so.rg.mk.bam /mydir/cat/dog/hen/test/block/sample1234567.so.rg.mk.bam 

The sample output is:

sample1 sampl47 sam4631 1234567