mjhjmtu

I have a file and the format is as follows:

cat dog AHF123432 | 123432 | dhfshfjdh
lion AFG23412 |23412 | dfshjhfjdhj

I need to remove those words from first column which contains only alphabets i.e., cat(followed by space) and dog from first row
and lion from the second row

My output will look something like:

AHF123432 | 123432 | dhfshfjdh 
AFG23412 | 23412 | dhfshfjdh

This removes all pure-alphabetic words from the beginning of the line:

$ sed -r 's/^([[:alpha:]]* )*//' filename.tsv
AHF123432 | 123432 | dhfshfjdh
AFG23412 |23412 | dfshjhfjdhj

Or, to save the output in a new file:

sed -r 's/^([[:alpha:]]* )*//' filename.tsv > final.tsv

How it works

• 
  

  [[:alpha:]]* matches any number of alphabetic characters followed by a space. In other words, it matches a word followed by a space.

  
  
  

  In more detail, [[:alpha:]] matches a single alphabetic character. ``[[:alpha:]]*` matches any number of such characters.

  
  



• ([[:alpha:]]* )* matches any number of such words followed by spaces.




• ^([[:alpha:]]* )* matches any any number of words followed by spaces but, because of the ^, only starting from the beginning of the line.




• The substitute command removes all those matching words. The substitute command has the form s/old/new/ where, here, old is our expression above which matches any number of words starting from the beginning of the line. We want to replace these words with nothing so, here, we use the empty string for new.

In olden times, the regular expression to match an alphabetic characters was [a-zA-Z]. With modern unicode fonts, that is now longer reliable. Because our regular expression above uses [:alpha:], it is unicode-safe.

If your grep supports perl compatible regular expressions:

grep -Po '^([a-z]* +)*K.*'

or

pcregrep -o '^([a-z]* +)*K.*'

Using awk

awk -F'|' 'for (i=1; i<=NF;i++) [A-Z])+ /,"",$i); printf "%s",$i; if(i<NF) printf " else printf "n"' foo

Example

% cat foo
cat dog AHF123432 | 123432 | dhfshfjdh
lion AFG23412 | 23412 | dfshjhfjdh

% awk -F'|' 'for (i=1; i<=NF;i++) [A-Z])+ /,"",$i); printf "%s",$i; if(i<NF) printf " else printf "n"' foo
AHF123432 | 123432 | dhfshfjdh
AFG23412 | 23412 | dfshjhfjdh