Error TR command width :punct: param

Question

I have the following command:

cat original.txt | tr [:upper:] [:lower:] | tr -d [:digit:] | 
tr -d [:punct:] > preproces.txt

And the system returns me the following error.

when translating, the only character classes that may appear in replaceString are 'upper' and 'lower'

The problem is giving:

tr -d [:punct:]

What's the solution to this problem?

I have searched Google and StackOverflow and don't find an answer to this problem.

Tiny edit (but significant for cut&pasters): Changed '--d' to '-d' before [:punct:] the '--d' was being rendered as '\342\200\223d' and the rendered difference was nearly invisible to the eye.

Answer 1

Try the following command :

tr '[:upper:]' '[:lower:]' < original.txt  |
    tr -d '[:digit:]' |
    tr –d '[:punct:]' > preproces.txt

This avoid the infamous useless use of cat :

http://iki.fi/era/unix/award.html

Answer 2

You only get the message you said you're seeing when you try something like:

tr '[:alpha:]' '[:digit:]'

I was able to use

tr -d '[:punct:]'

to delete punctuation without a problem. For example:

$ echo "tr '[:alpha:]' [:digit:] <chkit.sh" | tr -d '[:punct:]'
tr alpha digit chkitsh
$

This was on RHEL 5 for x86/64 (tr version 'tr (GNU coreutils) 5.97').

You might note I enclosed the arguments to tr in quotes (single or double would do, though single are generally safer). If you have any single-letter file names lurking in your directory, you could run into problems with the shell interpreting square brackets as character classes, and passing those names to tr. Say you had a file p:

tr [:upper:] [:lower:]

That command is now tr p [:lower:]. The error you get is slightly different, but it might be that your problem is related to that. Be very careful with shell metacharacters; quotes are your friend.

Note that doubled square brackets are not part of the solution:

$ echo "tr -d '[[:alpha:]]' <chkit.sh" | tr -d '[[:alpha:]]'        
 - '::' <.
$

Answer 3

If your tr does not support the [:digit:] and [:punct:] character classes, you will have to specify them explicitly.

tr '[:upper:]' '[:lower:]' < original.txt  |
    tr -d '0-9!"#$%&'"'"'()*+,./:;<=>?@[\\]^_`{|}~-' > preproces.txt

(The refactoring to merge the two tr -d commands into one could have been done by tr -d '[:digit:][:punct:]' in the original command, too.)

_{The definition of [:punct:] was obtained from http://www.regular-expressions.info/posixbrackets.html -- I have not attempted any verification, though I refactored it slightly to make it compatible with tr and the shell context it occurs in here.}