10

I have a file ma.txt and it contains the output of ls -l; when I run the tr command (tr "nik-pc" "root") I get this output:

nik-pc@nik:~$ cat ma.txt 
total 52
drwxr-xr-x 2 nik-pc nik-pc 4096 Mar 11 11:33 Desktop
lrwxrwxrwx 1 nik-pc nik-pc    2 Mar  8 22:54 di -> hd
drwxr-xr-x 3 nik-pc nik-pc 4096 Mar 13 13:28 Documents
drwxr-xr-x 7 nik-pc nik-pc 4096 Mar 14 18:21 Downloads
drwxr-xr-x 2 nik-pc nik-pc 4096 Mar 11 09:39 dwhelper
-rw-r--r-- 1 nik-pc nik-pc 2134 Mar 13 17:40 hd
-rw-r--r-- 1 nik-pc nik-pc    3 Mar 13 15:34 m
-rw-r--r-- 1 nik-pc nik-pc    0 Mar 17 19:48 ma.txt
drwxr-xr-x 3 nik-pc nik-pc 4096 Mar 13 14:58 Music
drwxr-xr-x 2 nik-pc nik-pc 4096 Mar  5 12:30 Pictures
drwxr-xr-x 2 nik-pc nik-pc 4096 Mar  5 11:44 Public
drwxr-xr-x 2 nik-pc nik-pc 4096 Mar 13 15:58 sd
drwxr-xr-x 2 nik-pc nik-pc 4096 Mar  5 11:44 Templates
drwxr-xr-x 2 nik-pc nik-pc 4096 Mar  5 11:44 Videos
drwxr-xr-x 2 nik-pc nik-pc 4096 Mar 11 11:33 xdm-helper

nik-pc@nik:~$ tr "nik-pc" "root" < ma.txt 
tttat 52
drwxr-xr-x 2 too-tt too-tt 4096 Mar 11 11:33 Desottt
trwxrwxrwx 1 too-tt too-tt    2 Mar  8 22:54 do -> hd
drwxr-xr-x 3 too-tt too-tt 4096 Mar 13 13:28 Dttutetts
drwxr-xr-x 7 too-tt too-tt 4096 Mar 14 18:21 Dtwtttads
drwxr-xr-x 2 too-tt too-tt 4096 Mar 11 09:39 dwhetter
-rw-r--r-- 1 too-tt too-tt 2134 Mar 13 17:40 hd
-rw-r--r-- 1 too-tt too-tt    3 Mar 13 15:34 t
-rw-r--r-- 1 too-tt too-tt    0 Mar 17 19:48 ta.txt
drwxr-xr-x 3 too-tt too-tt 4096 Mar 13 14:58 Musot
drwxr-xr-x 2 too-tt too-tt 4096 Mar  5 12:30 Pottures
drwxr-xr-x 2 too-tt too-tt 4096 Mar  5 11:44 Pubtot
drwxr-xr-x 2 too-tt too-tt 4096 Mar 13 15:58 sd
drwxr-xr-x 2 too-tt too-tt 4096 Mar  5 11:44 Tetttates
drwxr-xr-x 2 too-tt too-tt 4096 Mar  5 11:44 Vodets
drwxr-xr-x 2 too-tt too-tt 4096 Mar 11 11:33 xdt-hetter

At line one it replaced "nik" with "too" and the spelling of "Desktop" became "Desottt".

Why is this? What is logic behind it?

muru
  • 207,970
Manish Bharti
  • 341
  • 3
    info coreutils 'tr invocation' tells you en detail what tr does. – Nephente Mar 17 '16 at 14:49
  • 4
    What are you trying to do by the way? The command did exactly what you told it to do, but I guess you actually want to replace nik-pc with root? – kos Mar 17 '16 at 14:51
  • 3
    First step is always to check man page of the command. – Mostafa Ahangarha Mar 17 '16 at 15:06
  • The output you posted is actually a bit inconsistent, because your ns are translated to ts instead of rs as they should be.. Just to be sure, is there any chance you ran tr "nik-pc" "toot" instead? – David Z Mar 17 '16 at 20:04
  • 1
    @DavidZ: Note that n is also in the k-p range. The result of a character appearing more than once on the first set is unspecified by POSIX. – hmakholm left over Monica Mar 17 '16 at 20:35
  • lol man so funny I so saw where this was going from the title – Joshua Mar 17 '16 at 22:01
  • 1
    I think what you really need, is the sed command... ls -l | sed 's/nik-pc/root/' would do what you want. tr is for converting single characters - eg. lower-case to upper-case, or dos-style lineshift (\r) to Unix-style (\n), or something like replacing all backslashes (Windows) with slashes. It can also delete "unwanted" characters, eg. all numbers or all upper-case letters. – Baard Kopperud Mar 17 '16 at 23:26
  • tr replaces the first set with the second, so "nik-pc" and "root" means that all n becomes r, and i & k both becomes o, - becomes t... and I think p & c also becomes t (not sure though) – Baard Kopperud Mar 17 '16 at 23:31
  • @HenningMakholm oh, silly me, I missed that n was in the range. – David Z Mar 18 '16 at 08:48

2 Answers2

19

tr is for translating characters, not for complete words. It can translate sets. In your example you have "nik-pc" as first collection chars, and "root" is another. In fact, k-p is a range, so it includes all chars from k to p. It will match chars one by one, so n will translate to r, i to o, k to o, and whatever else beyond 4th char will be t. That's why you have "Desktop" translated to "Desottt"

You can see it more clearly in this example:

$ echo "ABCDEF" | tr "ABCDEF"  "12"                            
122222

Here you can see tr set 1 have D in position 4. But set 2 has no position 4, so it will use last position set 2 has to translate.

What you are doing is translating one word into another. What you want to do is use more advanced tool like sed or awk.

For instance,

$ ls -l /etc/passwd | awk '{gsub(/root/,"TEST");print}'        
-rw-r--r-- 1 TEST TEST 2575 Feb 29 12:30 /etc/passwd
Sergiy Kolodyazhnyy
  • 107,880
18

tr translates a string character-wise. It searches the letters from the first set and replaces them by those form the second set.

You had nik-pc as first set. tr expands the k-p part in that to all letters in the range from "k" to "p", so the set is equal to niklmnopc.

Your second set was root.

What tr now does is to search all occurrences of the first character in the (evaluated) first set and replace them with the first character of the second set. When there is no more character in set 2, it simply repeates its last character. See the table below:

n --> r
i --> o
k --> o
l --> t
m --> t
n --> t
o --> t
p --> t
c --> t

So now it's clear why e.g. "Desktop" becomes "Desottt". The behaviour is fully correct and intended this way.

What you're looking for instead can be achieved using sed:

sed 's/nik-pc/root/g' ma.txt

The syntax is this:

sed 's/SEARCH_PATTERN/REPLACE_STRING/FLAGS' INPUT_FILE

So we let it search for the pattern "nik-pc" and replace the entire match with "root". We need to add the "g" flag to enable global replacing. Without that, it would only replace each first match per line.

Byte Commander
  • 110,523
  • That table with arrows makes the explanation even clearer, I like that. +1-ed . A user could probably make something similar with printf "A\nB\nC\n" | tr 'ABC' '12' – Sergiy Kolodyazhnyy Mar 17 '16 at 15:09