pl - Swiss Army Knife of Perl One-Liners


Just one small script extends perl -E with many bells & whistles: Various one-letter commands & magic variables (with meaningful aliases too) and more nifty loop options take Perl programming to the command line. List::Util is fully imported. If you pass no program on the command line, starts a simple Perl Shell.

How to e(cho) values, including from @A(RGV), with single $q(uote) & double $Q(uote). Same for hard-to-print values:

    pl 'echo "${quote}Perl$quote", "$Quote@ARGV$Quote"' one liner
    pl 'e "${q}Perl$q", "$Q@A$Q"' one liner

    pl 'echo \"Perl", \@ARGV, undef' one liner
    pl 'e \"Perl", \@A, undef' one liner

Print up to 3 matching lines, resetting count (and $.) for each file:

    pl -rP3 '/Perl.*one.*liner/' file1 file2 file3

Loop over args, printing each with line ending. And same, SHOUTING:

    pl -opl '' Perl one liner
    pl -opl '$_ = uc' Perl one liner

Count hits in magic statistics hash %n(umber):

    pl -n '++$n{$1} while /(Perl|one|liner)/g' file1 file2 file3

Even though the point here is to make things even easier, most Perl one-liners from the internet work, just by omitting -e or -E. Known minor differences are: don't goto LINE, but next LINE is fine. In -n last goes straight to the next file instead of behaving like exit. Using pop, etc. to implicitly modify @A(RGV) works in -b begin code, but not in your main program (which gets compiled to a function. And shenanigans with unbalanced braces won't work.


Don't believe everything you read on the internet^H^H^H SourceForge! -- Aristotle ;-)

Pl follows Perl's philosophy for one-liners: the one variable solely used in one-liners, @F, is single-lettered. Because not everyone may like that, Pl has it both ways. Everything is aliased both as a word and as a single letter, including Perl's own @F & *ARGV.

-b doesn't do a BEGIN block. Rather it is in the same scope as your main PERLCODE. So you can use it to initialise my variables. Whereas, if you define a my variable in a -n, -p, -o or -O loop, it's a new variable each time. This echoes "a c" because -e emulates an END block, as a closure of the first $inner variable:

    pl -Ob 'my $outer'  -e 'echo $inner, $outer'  'my $inner = $outer = $ARGV' a b c
    pl -Ob 'my $outer'  -e 'e $inner, $outer'  'my $inner = $outer = $A' a b c


To steal ideas from one person is plagiarism. To steal from many is research. ;-)

Only some of these are original. Many have been adapted from the various Perl one-liner pages on the internet. This is no attempt to appropriate ownership, just to show how things are even easier and more concise with pl.

All examples use the long names and are repeated for short names, where applicable.

Dealing with Files

Delete matching files, except last one

If you have many files, which sort chronologically by name, and you want to keep only the last one, it can be quite painful to formulate Shell patterns. So check on each iteration of the -o loop, if the index $ARGIND (or $I) is less than the last, before unlinking (deleting). If you want to test it first, replace unlink with echo or e:

    pl -o 'unlink if $ARGIND < $#ARGV' file*
    pl -o 'unlink if $I < $#A' file*

If your resulting list is too long for the Shell, let Perl do it. Beware that the Shell has a clever ordering of files, while Perl does it purely lexically! In the -b begin code the result is assigned to @A(RGV), as though it had come from the command line. This list is then popped (shortened), instead of checking each time. Since the program doesn't contain special characters, you don't even need to quote it:

    pl -ob '@ARGV = <file*>; pop' unlink
    pl -ob '@A = <file*>; pop' unlink

You can exclude files by any other criterion as well:

    pl -ob '@ARGV = grep !/keep-me/, <file*>' unlink
    pl -ob '@A = grep !/keep-me/, <file*>' unlink

File statistics

42% of statistics are made up! :-)

Count files per suffix

Find and pl both use the -0 option to allow funny filenames, including newlines. Sum up encountered suffixes in sort-value-numerically-at-end hash %n(umber):

    find -print0 |
        pl -0ln '++$number{/(\.[^\/.]+)$/ ? $1 : "none"}'
    find -print0 |
        pl -0ln '++$n{/(\.[^\/.]+)$/ ? $1 : "none"}'
Count files per directory per suffix

Match to last / and after a dot following something, i.e. not just a dot-file. "" is the suffix for suffixless files. Stores in sort-by-key-and-stringify-at-end %s(tring). So count in a nested hash of directory & suffix:

    find -type f -print0 |
        pl -0ln '/^(.+)\/.+?(?:\.([^.]*))?$/; ++$string{$1}{$2}'
    find -type f -print0 |
        pl -0ln '/^(.+)\/.+?(?:\.([^.]*))?$/; ++$s{$1}{$2}'

This is the same, but groups by suffix and counts per directory:

    find -type f -print0 |
        pl -0ln '/^(.+)\/.+?(?:\.([^.]*))?$/; ++$string{$2}{$1}'
    find -type f -print0 |
        pl -0ln '/^(.+)\/.+?(?:\.([^.]*))?$/; ++$s{$2}{$1}'

This is similar, but stores in sort-by-number-at-end %n(umber). Since this matches suffixes optionally, a lone dot indicates no suffix. The downside is that it is neither sorted by directory, nor by suffix:

    find -type f -print0 |
        pl -0ln '/^(.+)\/.+?(?:\.([^.]*))?$/; ++$number{"$1 .$2"}'
    find -type f -print0 |
        pl -0ln '/^(.+)\/.+?(?:\.([^.]*))?$/; ++$n{"$1 .$2"}'

This avoids the lone dot:

    find -type f -print0 |
        pl -0ln '/^(.+)\/.+?(?:\.([^.]*))?$/; ++$number{length($2) ? "$1 .$2" : "$1 none"}'
    find -type f -print0 |
        pl -0ln '/^(.+)\/.+?(?:\.([^.]*))?$/; ++$n{length($2) ? "$1 .$2" : "$1 none"}'
Sum up file-sizes per suffix.

Find separates output with a dot and -F splits on that. The \\ is to escape one backslash from the Shell. No matter how many dots the filename contains, 1st element is the size and last is the suffix. Sum it in %n(umber), which gets sorted numerically at the end:

    find -name '*.*' -type f -printf "%s.%f\0" |
        pl -0lanF\\. '$number{".$FIELD[-1]"} += $FIELD[0]'
    find -name '*.*' -type f -printf "%s.%f\0" |
        pl -0lanF\\. '$n{".$F[-1]"} += $F[0]'

This is similar, but also deals with suffixless files:

    find -type f -printf "%s.%f\0" |
        pl -0lanF\\. '$number{@FIELD == 2 ? "none" : ".$FIELD[-1]"} += $FIELD[0]'
    find -type f -printf "%s.%f\0" |
        pl -0lanF\\. '$n{@F == 2 ? "none" : ".$F[-1]"} += $F[0]'
Count files per date

Incredibly, find has no ready-made ISO date, so specify the 3 parts. If you don't want days, just leave out -%Td. Sum up encountered dates in sort-value-numerically-at-end hash %n(umber):

    find -printf '%TY-%Tm-%Td\n' |
        pl -ln '++$number{$_}'
    find -printf '%TY-%Tm-%Td\n' |
        pl -ln '++$n{$_}'
Count files per date with rollup

Learn sign language! It's very handy. :-)

Rollup means, additionally to the previous case, sum up dates with the same prefix. The trick here is to count both for the actual year, month and day, as well as replacing once only the day, once also the month with "__",and once also the year with "____". This sorts after numbers and gives a sum for all with the same leading numbers. Use the sort-by-key-at-end hash %s(tring):

    find -printf '%TY-%Tm-%Td\n' |
        pl -ln '++$string{$_}; ++$string{$_} while s/[0-9]+(?=[-_]*$)/"_" x length $&/e'
    find -printf '%TY-%Tm-%Td\n' |
        pl -ln '++$s{$_}; ++$s{$_} while s/[0-9]+(?=[-_]*$)/"_" x length $&/e'

Diff several inputs by a unique key

Always remember you're unique, just like everyone else. :-)

The function k(eydiff) stores the 2nd arg or chomped $_ in %k(eydiff) keyed by 1st arg or $1 and the arg counter $ARGIND (or $I). Its sibling K(eydiff) does the same using 1st arg or 0 as an index into @F(IELD) for the 1st part of the key. At the end only the rows differing between files are shown. If you specify --color and have Algorithm::Diff the exact difference gets color-highlighted.

Diff several csv, tsv or passwd files by 1st field

This assumes no comma in key field and no newline in any field. Else you need a csv-parser package. -F implies -a, which implies -n (even using older than Perl 5.20, which introduced this idea):

    pl -F, Keydiff *.csv
    pl -F, K *.csv

This is similar, but removes the key from the stored value, so it doesn't get repeated for each file:

    pl -n 'keydiff if s/(.+?),//' *.csv
    pl -n 'k if s/(.+?),//' *.csv

A variant of csv is tsv, with tab as separator. Tab is \t, which must be escaped from the Shell as \\t:

    pl -F\\t Keydiff *.tsv
    pl -F\\t K *.tsv

    pl -n 'keydiff if s/(.+?)\t//' *.tsv
    pl -n 'k if s/(.+?)\t//' *.tsv

The same, with a colon as separator, if you want to compare passwd files from several hosts:

    pl -F: Keydiff /etc/passwd passwd*
    pl -F: K /etc/passwd passwd*

    pl -n 'keydiff if s/(.+?)://' /etc/passwd passwd*
    pl -n 'k if s/(.+?)://' /etc/passwd passwd*
Diff several zip archives by member name

This uses the same mechanism as the csv example. Addidionally it reads the output of unzip -vql for each archive through the pipe or p block. That has a fixed format, except for tiny members, which can report -200%, screwing the column by one:

    pl -o 'piped { keydiff if / Defl:/ && s/^.{56,57}\K  (.+)// } "unzip", "-vql", $_' *.zip
    pl -o 'p { k if / Defl:/ && s/^.{56,57}\K  (.+)// } "unzip", "-vql", $_' *.zip

If you do a clean build of java, many class files will have the identical crc, but still differ by date. This excludes the date:

    pl -o 'piped { keydiff $2 if / Defl:/ && s/^.{31,32}\K.{16} ([\da-f]{8})  (.+)/$1/ } "unzip", "-vql", $_' *.jar
    pl -o 'p { k $2 if / Defl:/ && s/^.{31,32}\K.{16} ([\da-f]{8})  (.+)/$1/ } "unzip", "-vql", $_' *.jar
Diff several tarballs by member name

This is like the zip example. But tar gives no checksum, so this is not very reliable. Each time a wider file size was seen, columns shift right. Reformat the columns, so this doesn't show up as a difference:

    pl -o 'piped { s/^\S+ \K(.+?) +(\d+) (.{16}) (.+)/sprintf "%-20s %10d %s", $1, $2, $3/e; keydiff $4 } "tar", "-tvf", $_' *.tar *.tgz *.txz
    pl -o 'p { s/^\S+ \K(.+?) +(\d+) (.{16}) (.+)/sprintf "%-20s %10d %s", $1, $2, $3/e; k $4 } "tar", "-tvf", $_' *.tar *.tgz *.txz

Again without the date:

    pl -o 'piped { s/^\S+ \K(.+?) +(\d+) .{16} (.+)/sprintf "%-20s %10d", $1, $2/e; keydiff $3 } "tar", "-tvf", $_' *.tar *.tgz *.txz
    pl -o 'p { s/^\S+ \K(.+?) +(\d+) .{16} (.+)/sprintf "%-20s %10d", $1, $2/e; k $3 } "tar", "-tvf", $_' *.tar *.tgz *.txz
Diff ELF executables or libraries by loaded dependencies

You get the idea: you can do this for any command that outputs records with a unique key. This one looks at the required libraries and which file they came from. For a change loop with -O and @A(RGV) to avoid the previous examples' confusion between outer $_ which are the cli args, and the inner one, which are the read lines:

    pl -O 'piped { keydiff if s/^\t(.+\.so.*) => (.*) \(\w+\)/$2/ } ldd => $ARGV' exe1 exe2 exe3
    pl -O 'p { k if s/^\t(.+\.so.*) => (.*) \(\w+\)/$2/ } ldd => $A' exe1 exe2 exe3

It's even more useful if you use just the basename as a key, because version numbers may change:

    pl -O 'piped { keydiff $2 if s/^\t((.+)\.so.* => .*) \(\w+\)/$1/ } ldd => $ARGV' exe1 exe2 exe3
    pl -O 'p { k $2 if s/^\t((.+)\.so.* => .*) \(\w+\)/$1/ } ldd => $A' exe1 exe2 exe3

Looking at Perl

A pig looking at an electric socket: "Oh no, who put you into that wall?" :)

VERSION of a File

Print the first line (-P1) where the substitution was successful. To avoid the hassle of protecting them from (sometimes multiple levels of) Shell quoting, there are variables for single $q(uote) & double $Q(uote):

    pl -P1 's/.+\bVERSION\s*=\s*[v$Quote$quote]([0-9.]+).+/$1/' pl
    pl -P1 's/.+\bVERSION\s*=\s*[v$Q$q]([0-9.]+).+/$1/' pl

For multple files, add the filename, and reset (-r) the -P count for each file:

    pl -rP1 's/.+\bVERSION\s*=\s*[v$Quote$quote]([0-9.]+).+/$ARGV: $1/' *.pm
    pl -rP1 's/.+\bVERSION\s*=\s*[v$Q$q]([0-9.]+).+/$A: $1/' *.pm
Only POD or non-POD

You can extract either parts of a Perl file, with these commands. Note that they don't take the empty line before into account. If you want that, and you're sure the files adheres strictly to this convention, use the option -00P instead (not exactly as desired, the empty line comes after things, but still, before next thing). If you want only the 1st POD (e.g. NAME & SYNOPSIS) use the option -P1 or -00P1:

    pl -P '/^=\w/../^=cut/' file

    pl -P 'not /^=\w/../^=cut/' file
Count Perl Code

This makes __DATA__ or __END__ the last inspected line of (unlike in perl -n!) each file. It strips any comment (not quite reliably, also inside a string). Then it strips leading whitespace and adds the remaining length to print-at-end $r(esult):

    pl -ln 'last if /^__(?:DATA|END)__/; s/(?:^|\s+)#.*//s; s/^\s+//; $result += length' *.pm
    pl -ln 'last if /^__(?:DATA|END)__/; s/(?:^|\s+)#.*//s; s/^\s+//; $r += length' *.pm

If you want the count per file, instead of $r(esult) use either sort-lexically $string{$ARGV} (or $s{$A}) or sort-numerically $number{$ARGV} (or $n{$A}).

Content of a Package

Pl's e(cho) can print any item. Packages are funny hashes, with two colons at the end. Backslashing the variable passes it as a unit to Data::Dumper, which gets loaded on demand in this case. Otherwise all elements would come out just separated by spaces:

    pl 'echo \%List::Util::'
    pl 'e \%List::Util::'
Library Loading

Where does perl load from, and what exactly has it loaded?

    pl 'echo \@INC, \%INC'
    pl 'e \@INC, \%INC'

Same, for a different Perl version, e.g. if you have perl5.20.0 in your path:

    pl -V5.20.0 'echo \@INC, \%INC'
    pl -V5.20.0 'e \@INC, \%INC'

You get %Config::Config loaded on demand and returned by c(onfig):

    pl 'echo config'
    pl 'e c'

It returns a hash reference, from which you can lookup an entry:

    pl 'echo config->{sitelib}'
    pl 'e c->{sitelib}'

You can also return a sub-hash, of only the keys matching any regexps you pass:

    pl 'echo config "random", qr/stream/'
    pl 'e c "random", qr/stream/'


ISO paper sizes

ISO replaced 8 standards by one. Now we have 9 standards. :-(

Can't put the A into the format, because 10 is wider. Uses Perl's lovely list assignment to swap and alternately halve the numbers. Because halving happens before echoing, start with double size:

    pl '($w, $h) = (1189, 1682); echof "%3s  %4dmm x %4dmm", "A$_", ($w, $h) = ($h / 2, $w) for 0..10'
    pl '($w, $h) = (1189, 1682); f "%3s  %4dmm x %4dmm", "A$_", ($w, $h) = ($h / 2, $w) for 0..10'

The table could easily be widened to cover B- & C-formats, by extending each list of 2, to an appropriate list of 6, e.g. ($Aw, $Ah, $Bw, ...).

ANSI foreground;background color table

If at first you don't succeed, destroy all evidence that you tried! ;-)

What a table, hardly a one-liner... You get numbers to fill into "\e[FGm", "\e[BGm" or "\e[FG;BGm" to get a color and close it with "\e[m". There are twice twice 8 different colors for dim & bright and for foreground & background. Hence the multiplication of escape codes and of values to fill them.

This fills @A(RGV) in -b, as though it had been given on the command line. It maps it to the 16fold number format to print the header. Then the main PERLCODE loops over it with $A(RGV), thanks to -O, to print the body. All numbers are duplicated with (N)x2, once to go into the escape sequence, once to be displayed:

    pl -Ob '@ARGV = map +($_, $_+8), 1..8; f "co:  fg;bg"."%5d"x16, @ARGV' \
        'echof "%2d:  \e[%dm%d;   ".("\e[%dm%4d "x16)."\e[m", $A, ($A + ($A > 8 ? 81 : 29))x2, map +(($_)x2, ($_+60)x2), 40..47'
    pl -Ob '@A = map +($_, $_+8), 1..8; f "co:  fg;bg"."%5d"x16, @A' \
        'f "%2d:  \e[%dm%d;   ".("\e[%dm%4d "x16)."\e[m", $A, ($A + ($A > 8 ? 81 : 29))x2, map +(($_)x2, ($_+60)x2), 40..47'

This does exactly the same, but explicitly loops over lists @co & @bg:

    pl '@co = map +($_, $_+8), 1..8; @bg = map +(($_)x2, ($_+60)x2), 40..47;
        echof "co:  fg;bg"."%5d"x16, @co;
        echof "%2d:  \e[%dm%d;   ".("\e[%dm%4d "x16)."\e[m", $_, ($_ + ($_ > 8 ? 81 : 29))x2, @bg for @co'
    pl '@co = map +($_, $_+8), 1..8; @bg = map +(($_)x2, ($_+60)x2), 40..47;
        f "co:  fg;bg"."%5d"x16, @co;
        f "%2d:  \e[%dm%d;   ".("\e[%dm%4d "x16)."\e[m", $_, ($_ + ($_ > 8 ? 81 : 29))x2, @bg for @co'



2 + 2 = 5 for extremely large values of 2. :-)

With the big* modules you can do arbitrary precision math:

    pl -Mbignum 'echo 123456789012345678901234567890 * 123456789012345678901234567890'
    pl -Mbignum 'e 123456789012345678901234567890 * 123456789012345678901234567890'

    pl -Mbignum 'echo 1.23456789012345678901234567890 * 1.23456789012345678901234567890'
    pl -Mbignum 'e 1.23456789012345678901234567890 * 1.23456789012345678901234567890'

    pl -Mbigrat 'echo 1/23456789012345678901234567890 * 1/23456789012345678901234567890'
    pl -Mbigrat 'e 1/23456789012345678901234567890 * 1/23456789012345678901234567890'
Separate big numbers with commas, dots or underscores

Loop and print with line-end (-opl) over remaining args in $_. If reading from stdin or files, instead of arguments, use only -pl. After a decimal dot, insert a comma before each 4th comma-less digit. Then do the same backwards from end or decimal dot:

    pl -opl '1 while s/[,.]\d{3}\K(?=\d)/,/; 1 while s/\d\K(?=\d{3}(?:$|[.,]))/,/' \
        12345678 123456789 1234567890 1234.5678 3.141 3.14159265358

The same for languages with a decimal comma, using either a dot or a space as spacer:

    pl -opl '1 while s/[,.]\d{3}\K(?=\d)/./; 1 while s/\d\K(?=\d{3}(?:$|[.,]))/./' \
        12345678 12345678 1234567890 1234,5678 3,141 3,141592653589

    pl -opl '1 while s/[, ]\d{3}\K(?=\d)/ /; 1 while s/\d\K(?=\d{3}(?:$|[ ,]))/ /' \
        12345678 12345678 1234567890 1234,5678 3,141 3,141592653589

The same for Perl style output:

    pl -opl '1 while s/[._]\d{3}\K(?=\d)/_/; 1 while s/\d\K(?=\d{3}(?:$|[._]))/_/' \
        12345678 123456789 1234567890 1234.5678 3.141 3.14159265358
Generate a random UUID

Lottery: a tax on people who are bad at math. :-)

This gives a hex number with the characteristic pattern of dashes. The hex format takes only the integral parts of the random numbers:

    pl '$x = "%04x"; echof "$x$x-$x-$x-$x-$x$x$x", map rand 0x10000, 0..7'
    pl '$x = "%04x"; f "$x$x-$x-$x-$x-$x$x$x", map rand 0x10000, 0..7'

To be RFC 4122 conformant, the 4 version & 2 variant bits need to have standard values. Note that Shell strings can span more than one line:

    pl '@u = map rand 0x10000, 0..7; ($u[3] /= 16) |= 0x4000; ($u[4] /= 4) |= 0x8000;
        $x = "%04x"; echof "$x$x-$x-$x-$x-$x$x$x", @u'
    pl '@u = map rand 0x10000, 0..7; ($u[3] /= 16) |= 0x4000; ($u[4] /= 4) |= 0x8000;
        $x = "%04x"; f "$x$x-$x-$x-$x-$x$x$x", @u'
DNS lookup

What do you call a sheep with no legs? A cloud. *,=,

The h(osts) function deals with the nerdy gethost... etc. and outputs as a hosts file. The file is sorted by address type (localhost, link local, private, public), version (IPv4, IPv6) and address. You tack on any number of IP-addresses or hostnames, either as Perl arguments or on the command-line via @A(RGV):

    pl 'hosts qw('
    pl 'h qw('

    pl 'hosts @ARGV'
    pl 'h @A'

If you don't want it to be sorted, call h(osts) for individual addresses:

    pl 'hosts for qw('
    pl 'h for qw('

    pl -o hosts
    pl -o h

If your input comes from a file, collect it in a list and perform at end (-e):

    pl -lne 'hosts @list' 'push @list, $_' file
    pl -lne 'h @list' 'push @list, $_' file


Even if it is rarer nowadays, Perl 5.10 is still found out in the wild. Pl tries to accomodate it gracefully, falling back to what works. Printed data-structures will be formatted with a wide margin and h(osts) will find the less IPv6 resolutions, the older your Perl.

Windows Notes

Do yourself a favour and get a real Shell, e.g. from Cygwin or MSYS! If you can't avoid or cmd.exe, you will have to first convert all inner quotes to qq. Then convert the outer single quotes to double quotes:

    pl "echo qq{${quote}Perl$quote}, qq{$Quote@A$Quote}" one liner
    pl "e qq{${q}Perl$q}, qq{$Q@A$Q}" one liner

Any help for getting this to work in PowerShell is welcome! While the old Windows 10 terminal understands Ansi escape sequences, it makes it horribly hard to activate them. So they are off by default, requiring --color to override that choice.