12

-L is a useful feature of wc, or so I thought. It prints the length of the longest line. For some reason it expands a single-byte tab-char to a length of 8.
Is there some way to set this to not "expand" the tab? and what might be the rationale behind this expansion?

echo -n $'\t' | wc -L

outputs 8

wc (GNU coreutils) 7.4
GNU bash, version 4.1.5

Improve this question
1
  • The -n is orthogonal to the question.
    – user unknown
    Commented Sep 12, 2011 at 21:45

3 Answers 3

Reset to default
11

I find no bug report related to this, and the following lines in the source file wc.c

    case '\t':
        linepos += 8 - (linepos % 8);

seem to deliberately choose to behave in this way, probably to give an hint for width needed to display the file on screen.

A quick alternative could be

echo -n $'\t' | tr '\t' ' ' | wc -L
Improve this answer
2
  • 3
    Thanks enzo, I've now found that although man wc makes no mention of this issue, it is stated in info coreutils 'wc invocation' (which 'man' refers too)... Also, after trawling the google-sphere a bit more, I found this as an alternative echo -n $'\t' | expand -t1 | wc -L, which is pretty much the same as your alternative, but I've thrown it in for good measure.. And although the following link is a recompile wc hack*, it may be of interest to some: wc support for different tab widths
    – Peter.O
    Commented Sep 12, 2011 at 22:36
  • Counting tabs as 8 characters was added to GNU Coreutils in 1997 in the original patch that implemented the -L/--max-line-length feature. Before that they were counted as 1 character.
    – user114651
    Commented Dec 8, 2020 at 21:53
1

The wc -L description was ambiguous. It returns the widest display width. To control tab expansion you can filter through expand first.

Improve this answer
0

Normally a tab is expanded to the next position, (divisible by 8)+1 [1, 9, 17, 25, ...], so if you ask for it, you get it.

Note, that the -n is irrelevant for the question, but the $ is not.

echo foo$'\t' | wc -L

will return 8 too, because 

echo foo$'\t'bar 
foo     bar

You can omit the $, if you use -e for echo:

echo -e '\t' | wc -L
8

So if you want to count the '\t' as a single byte, just omit -e and $:

echo '\t' | wc -L
2
Improve this answer
11
  • Yes, expanding tabs is common enough for a printed/displayed output, but I found it odd that a program which counts bytes and words would count 1 character as anything other than 1 character... btw echo '\t' does not output a tab-char (\x09). It outputs a line whose length is 2, ie. a '\' and a 't'. A newline is not parte of a line's length... (I had a -n in my example to check whether wc would properly process a file which has no trailing newline-char...)
    – Peter.O
    Commented Sep 12, 2011 at 22:12
  • wc --help says: -L, --max-line-length print the length of the longest line?. It doesn't talk about bytes, but line lengths.
    – user unknown
    Commented Sep 12, 2011 at 23:05
  • 1
    Yes, it does say "print the length of the longest line"...` but it does not say "We assume that you want tabs expanded (not the usual character count, like most other length functions).. Oh, by the way, we will expand tabs to 8 spaces, regardless of what your specific tab stops are set to." ... That is the trap.. It is not properly documented.
    – Peter.O
    Commented Nov 1, 2011 at 21:35
  • How do you set the tab with? In Bash? Furthermore: Tabs are not expanded to 8 spaces, but to positions, see echo -e foo'\t'bar | wc -L which results in 11, not in 14.
    – user unknown
    Commented Nov 2, 2011 at 11:17
  • In the above foo\tbar example, wc has assumed tab-stops at a nominal spacing of 8... The following example shows how wc ignores the currently active tab-stop settings. It outputs a line to the terminal which is 8 terminal-columns wide/long, yet wc reports it to be 11. This example sets tab-stops to every 6th column... tabs -6; echo 12345678; echo -e "foo\tbar"|tee >(wc -L)
    – Peter.O
    Commented Nov 13, 2011 at 19:54

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .