I want to remove all non-english alphanumerals from a file.
tr -sc '[:alnum:][:punct:]' ' ' <file
gawk and sed have the -i flag but I cannot find anything similar in the tr man pages.
How can I do this inplace, that is, store the output in the input file itself?
No. Very few commands have this functionality, GNU awk (gawk) only added it recently, and even for those commands that have an -i or equivalent, all it does is create a temp file in the background.
So, the way to do this is:
tr -sc '[:alnum:][:punct:]' ' ' <file >newfile && mv newfile file
You could whip up a little function if you need this often:
tri(){
tmpFile=$(mktemp)
echo "$@"
case $# in
## You've given tr options
4)
trOpts="$1"
set1="$2"
set2="$3"
inputFile="$4"
;;
## No options, only set1 and set2 and the input file
3)
set1="$1"
set2="$2"
inputFile="$3"
;;
## Something wrong
*)
echo "Whaaaa?"
exit 1
;;
esac
tr "$trOpts" "$set1" "$set2" < "$inputFile" > "$tmpFile" &&
mv "$tmpFile" "$inputFile"
}
You'd then run this as:
tri -sc '[:alnum:][:punct:]' ' ' file
Note how, unlike the real tr, this expects the file name as an argument and not as redirected input (<file), and that options be specified together as shown above (and not like -s -c).
As that tr command always writes at most as much data as it reads, it should be safe for it to write its output in place:
With the ksh93 shell:
tr -sc '\n[:alnum:][:punct:]' '[ *]' < file 1<>; file
(here converted to standard/POSIX syntax [ *] and adding newline to the list of characters to be preserved to avoid creating non-text output).
The 1<>; ksh93-specific operator is like the standard 1<> read+write-without-truncation one except that upon successful completion of the command being redirected, the file is truncated at the current position.
