Tool for viewing top N items in stream

Question

I'm looking for an existing Linux command-line tools that can accomplish the following:

Consider there is an input stream /dev/inputstream which spews an infinite stream of data that looks like this:

A
A
B
A
C
Z
A
.
.
.

NOTE: (Each line is one character from the alphabet).

The tool is used in the following fashion:

cat /dev/inputstream | tool

The output of the tool should be like the one from "top" command. It should be a continuously updated list of top N frequently occurring items.

Eg:

A -- 10 times
B -- 8 times
Z -- 7 times
C -- 2 times
D -- 1 time

If the stream were not infinite I could have accomplished this using sort and uniq.

Answer 1

I found this tool that matches my need exactly, called logtop. According to the documentation:

logtop displays real-time count of strings received in standard input.

Examples

Is useful for some cases, like getting the IP flooding your server:

$ tail -f /var/log/apache2/access.log | cut -d' ' -f1 | logtop

Or the top buzzing article of your blog:

$ tail -f /var/log/apache2/access.log | cut -d' ' -f7 | grep article | logtop

Fedora/CentOS/RHEL

There is no RPM that I could find for this application but it's pretty trivial to download, compile and install manually. It does require 2 dependencies to be installed to be able to compile it.

$ sudo yum install uthash ncurses-devel

You can download the software from Github:

$ git clone https://github.com/JulienPalard/logtop.git

To build it:

$ cd logtop
$ make

You can then install it or run it from this directory, ./logtop.

Example

$ tr -dc 'A-Z' </dev/urandom | fold -w 1 | head -n 10MB | ./logtop
10000 lines, 10000.00 lines/s
RANK    CNT   LINE/S LINE                                                                                                                             
   1    428   428.00 T
   2    417   417.00 G
   3    416   416.00 S
   4    414   414.00 K
   5    413   413.00 Q
   6    412   412.00 E
   7    401   401.00 I

Answer 2

In Perl:

perl -nle '
          $N = 5;      # Change as needed
          $freq{$_}++;
          if ( keys %freq >= $N ){
              @sorted = sort { $freq{$b} <=> $freq{$a} } keys %freq;
              for ( 0..$N-1){
                  $key = $sorted[$_];
                  print "$key -- $freq{$key} times";
              }
          }
          sleep 5;
          system(clear)
          ' < /dev/inputstream