List files with line count and date

Question

I want to be able to list files showing the number of lines the each file has and the date. I can happily get the line count using wc -l *. Not a problem. I can get the date using ls -l.

Is there a way to combine the two commands to give me a single output in columns?

Solution 1

Here is something with find + wc + date.

find . -maxdepth 1 -exec sh -c '[ -f "$0" ] && \
  printf "%6s\t\t%s\t%s\n" "$(wc -l<"$0")" "$(date -r "$0")" "$0"' {} \;

Instead of date -r one can also use for example stat -c%y.

The output looks like this:

   394      Thu Oct 16 22:38:14 UTC 2014    ./.zshrc
     7      Thu Oct 30 11:19:01 UTC 2014    ./tmp.txt
     2      Thu Oct 30 06:02:00 UTC 2014    ./tmp2.txt
    40      Thu Oct 30 04:16:30 UTC 2014    ./pp.txt

Using this as starting point one can create a function which accepts directory and pattern as parameters:

myls () { find "$1" -maxdepth 1 -name "$2" -exec sh -c '[ -f "$0" ] && \
  printf "%6s \t\t%s\t%s\n" "$(wc -l<"$0")" "$(date -r "$0")" "$0"' {} \;; }

After that myls /tmp '*.png' will list only images from /tmp (notice single quotes around pattern to prevent shell from expanding a glob operator *).

Solution 2

The find-based solutions look the most elegant, but just for fun here are a couple of other ways to attack this.

This one uses sed & head to clean up the output of wc, then uses join to combine that to the output of stat, using the file name as the join field.

pat="A*";join -1 2 -2 1 -t ' ' <(wc -l $pat|head -n-1|sed 's/^[ ]*//') <(stat -c '%n %y' $pat) | awk '{printf "%-20s %6s %s %s %s\n",$1,$2,$3,$4,$5}'

And a simpler one that uses paste, with awk again tidying up the columns and also checking that the file names match.

pat="A*";paste -d ' ' <(wc -l $pat) <(stat -c '%n %y' $pat) | awk '$2==$3{printf "%6s %s %s %s\n", $1, $4, $5, $2}'

However, if the names don't match (IOW the file lists change while this command is running) then the errors are invisible. But this version reports the errors:

pat="A*";paste -d ' ' <(wc -l $pat) <(stat -c '%n %y' $pat) | awk '$2==$3{printf "%6s %s %s %s\n", $1, $4, $5, $2;next};{print "Error:" $0}'

Of course, that one will print an error message for the final line, but that's easy enough to fix. Or ignore. :)

Comments

[slm]

Explain what this is doing and how it works!

[John Reid]

Unfortunately wc dumps a new line which is being left behind.

[user8973106]

I added tr to remove the "\n"

[slm]

FYI, pay special attn to Jimmij's comment about using stat -c%y. You're running date on every file using the former command, date -r. Otherwise +1, this is a good solution to the Q.

[John Reid]

Ah I missed it, sorry. It's not working for a list of files now though as it removes all of the line breaks.

[John Reid]

Great answer. I changed date to (date -r "$0" -%x) to give me something to import into a spreadsheet. Exactly what I needed and later I'll have a look at this more in depth.

[Jakob Bennemann]

This seems so unnecessary given the find solution proposed above. This solution not only requires three actual commands, but it also uses a file on the system.

[jimmij]

@JohnReid I've changed original post a little, removed unnecessary variables, and added function for patterns. You may want to take a look. @slm you are right that one can do stat and wc on the whole list find ... {} + (if this is what you are saying), yes that would be for sure faster, but on the other hand more complicated since one would need to deal with loop over arrays later on (at least as I see it).

[G-Man Says 'Reinstate Monica']

Why would you choose to use a temporary file when all you need to do is pipe the output of the awk into tr?

[G-Man Says 'Reinstate Monica']

(1) I realize that the question says "list files", but, if there's only one file in the directory, wc will not output a "total" line, and so the head will discard the line count of the one existing file. (2) This breaks if filenames contain newlines. (3) This breaks if filenames contain spaces, because you say $file instead of "$file". You should always quote all shell variables references unless you have a good reason not to, and you’re sure you know what you’re doing.

[G-Man Says 'Reinstate Monica']

@jimmij: Why not say find -type f instead of making the shell do the test?

[jimmij]

@G-Man You are right, -type f would be simpler.