coding horror keyword analysis

I have subscribed to quite a few programming blogs, one of them being Coding Horror.

Coding Horror is written by a guy named Jeff Atwood (you probably knew that already), and his blog has received massive attention, bringing 93 thousand (woah!) feed subscribers (as of April, 2008).

One thing that caught my attention on CodingHorror blog is that its traffic stats are publicly available!

coding horror traffic statistics

The statistics are hosted by, which keeps only the last 500 entries of any traffic activity.

recent keyword activity on codinghorror

I wanted to see a clearer picture of the most popular keywords people searched for and ended up in Coding Horror blog.

Thirty minutes later I had written a Perl program, which accessed the statistics, parsed the "Recent Keyword Activity" page, extracted the keywords, and inserted them in an SQLite database.

I always love to describe how my programs work. I'll make it short this time, as we are concentrating on the statistics and not on programming.

The Perl Program

The Perl program uses (or reuses) a few CPAN modules:

The program takes two optional arguments

  • -nodb not to insert the keywords in database (just print them out)
  • number - number of pages to extract keywords from

Here is the source code of the program:

# Peteris Krumins (, 2008
#  --  good coders code, great reuse
# Access traffic statistics and extract a few pages of latest search queries
# Released under GNU GPL
# 2008.04.08: Version 1.0

# run it as 'perl [-nodb] [number of pages to extract]'
# -nodb specifies not to insert keywords in database, just print them to stdout

use strict;
use warnings;

use DBI;
use WWW::Mechanize;
use HTML::TreeBuilder;
use Date::Parse;

# URL to publicly available codinghorror's statcounter stats
my $login_url = '';

# Query used to INSERT a new keyword in the database
my $insert_query = 'INSERT OR IGNORE INTO queries (query, unix_date, human_date) VALUES (?, ?, ?)';

# Path to SQLite database
my $db_path = 'codinghorror.db';

# Insert queries in database or not? Default, yes.
my $do_db = 1;

# Number of pages of keywords to extract. Default 1.
my $pages = 1;

for (@ARGV) {
    $pages = $_ if /^\d+$/;
    $do_db = 0 if /-nodb/;

my $dbh;
$dbh = DBI->connect("dbi:SQLite:$db_path", '', '', { RaiseError => 1 }) if $do_db;

my $mech = WWW::Mechanize->new();
my $login_req = $mech->get($login_url);

unless ($mech->success) {
    print STDERR "Failed getting $login_url:\n";
    print $login_req->message, "\n";
    exit 1;

unless ($mech->content =~ /Coding Horror/i) {
    # Could not access Coding Horror's stats
    print STDERR "Failed accessing Coding Horror stats\n";
    exit 1;

my $kw_req = $mech->follow_link(text => 'Recent Keyword Activity');
unless ($mech->success) {
    print STDERR "Couldn't find 'Recent Keyword Activity' link";
    print $kw_req->message, "\n";
    exit 1;

for my $page (1..$pages) {
    my $tree = HTML::TreeBuilder->new_from_content($mech->content);
    my $td_main_panel = $tree->look_down('_tag' => 'td', 'class' => 'mainPanel');
    unless ($td_main_panel) {
        print STDERR "Unable to find '<td class=mainPanel>'";
        exit 1;
    my $table = $td_main_panel->look_down('_tag' => 'table', 'class' => 'standard');
    unless ($table) {
        print STDERR "Unable to find 'table' tag";
        exit 1;
    my @trs = $table->look_down('_tag' => 'tr');
    my $idx = 0;
    for my $tr (@trs) {
        next unless $idx++;
        my @tds = $tr->look_down('_tag' => 'td');
        unless (@tds == 6) {
            print STDERR "<td> count was not 6!\n";
        my ($date, $time, $query) = map { $_->as_text } (@tds[1..2], $tds[4]);
        next unless $query;
        my $year = (localtime)[5] + 1900;
        my $ydt = "$date $year $time";
        my $unix_date = str2time($ydt);
        print "$date $year $time: $query\n";
        $dbh->do($insert_query, undef, $query, $unix_date, $ydt) if $do_db;
    if ($page != $pages) {
        my $page_req = $mech->follow_link(text => $page + 1);
        unless ($page_req) {
            print STDERR "Couldn't find page ", $page + 1, " of keywords", "\n";
            exit 1;

Download: coding horror keyword scraper (downloaded: 4521 times)

Here is an example run of the program:

$ ./ -nodb 2
8 Apr 2008 03:50:54: media player
8 Apr 2008 03:50:53: physical working environment programmers
8 Apr 2008 03:50:26: nano itx case
8 Apr 2008 03:50:23: how to clean some internet spyware or adware infection
8 Apr 2008 03:50:23: mercurial install tutorial windows
8 Apr 2008 03:50:22: iis 5.1 multiple websites
8 Apr 2008 03:50:17: javascript integer manipulation comparision
8 Apr 2008 03:50:16: build machines pc
8 Apr 2008 03:50:14: manage remote desktop connections
8 Apr 2008 03:50:07: check that all variables are initialized
8 Apr 2008 03:50:00: powergrep older version
8 Apr 2008 03:49:43: software counterfeiting
8 Apr 2008 03:48:59: floppy emulator windows xp
8 Apr 2008 03:48:35: safari rendering cleartype
8 Apr 2008 03:48:18: captchas goole broken
8 Apr 2008 03:48:11: vs2005 ide color
8 Apr 2008 03:47:55: optimising dual core for cubase sx3
8 Apr 2008 03:47:44: micosoft project scheduling
8 Apr 2008 03:47:36: dont buy from craig at australian computer resellers
8 Apr 2008 03:47:32: large scale stored procedures
8 Apr 2008 03:47:31: free diff tool
8 Apr 2008 03:46:58: games that support 3 monitors
8 Apr 2008 03:46:56: firefox multiple times same stylesheet
8 Apr 2008 03:46:48:
8 Apr 2008 03:46:37: apple software serial code blocker
8 Apr 2008 03:46:31: beautiful code jon bentley
8 Apr 2008 03:46:28: system.web.httpparseexception
8 Apr 2008 03:46:23: round in
8 Apr 2008 03:46:15: project postmortem software
8 Apr 2008 03:45:43: programming fun
8 Apr 2008 03:45:33: sending messages over ip using command prompt
8 Apr 2008 03:45:26: where did horror develop?

The SQLite Database

The database has just one table called 'queries' which contains a 'query', 'unix_date' and 'human_date' columns. The 'unix_date' column is used for sorting the entries chronologically, and 'human_date' is there just so I could easily see the date.

Here is the schema of the database:

CREATE TABLE queries (id INTEGER PRIMARY KEY, query TEXT, unix_date INTEGER, human_date TEXT);
CREATE UNIQUE INDEX unique_query_date ON queries (query, unix_date);

As the Perl program is run periodically, it might extract the same keywords several times. I created a UNIQUE index on 'query' and 'unix_date' fields, and left the job to drop the duplicate records to SQLite.

The Perl program uses the following SQL query to insert the data in database:

INSERT OR IGNORE INTO queries (query, unix_date, human_date) VALUES (?, ?, ?)

The 'OR IGNORE' makes sure the duplicate records get silently discarded.

Simple Statistics

I have been collecting keywords since March 31, and the database has now grown to a size of 73'336 records and 7MB (3MB compressed).

Download: coding horror keyword database (.zip) (downloaded: 810 times)

I ran a few simple SQL queries against the data using the GUI SQLite Database Browser to find the most popular keywords. I recommend downloading it, if you want to play around with the database.

The first query selected the 15 most popular keywords, along with their count, and percentage of all keywords.

The following SQL query did it:

 count(query) c,
 (round(count(query)/(1.0*(select count(*) from queries)),3)*100) || '%',
FROM queries
GROUP BY query

most popular coding horror’s keywords (sql query in sqlite database browser)

I also made a bar chart using the public Google Charts API:

This chart would look much better if it had vertical bars. I couldn't figure out how to add keywords nicely below each bar, though.

Here is how the messy query to Google Charts API looks like:'s%20Top%2015%20Keywords&cht=bhs&chd=t:100,77,12.07,10.18,9.09,8.74,8.64,8.49,7.05,6.51,5.91,5.71,5.66,5.61,5.22&chs=400x450&chxt=x,y&chxl=0:|0|2013|1:|command%20prompt%20commands|registration%20keys|cmd%20tricks|vista%20media%20center|sql%20joins|command%20prompt|you%20may%20be%20a%20victim...|codinghorror|dual%20core%20vs%20quad%20core|quad%20core%20vs%20dual%20core|cmd%20commands|command%20prompt%20tricks|system%20idle%20processes|coding%20horror|system%20idea%20process

Just to illustrate various ways to work with SQLite database, I did the same query from command line, and queried top 50 popular keywords, here they are:

$ sqlite3 ./codinghorror.db
sqlite> .header ON
sqlite> .explain ON
sqlite> SELECT count(query) c, query FROM queries GROUP BY query ORDER BY c DESC LIMIT 50;
c     query
----  -------------
2013  system idle process
1550  coding horror
243   system idle processes
205   command prompt tricks
183   cmd commands
176   quad core vs dual core
174   dual core vs quad core
171   codinghorror
142   you may be a victim of software counterfeiting
131   command prompt
119   sql joins
115   vista media center
114   cmd tricks
113   registration keys
105   command prompt commands
105   jeff atwood
99    quad core
96    dell xps m1330 review
89    rainbow tables
84    what is system idle process
82    software counterfeiting
80    fizzbuzz
78    laptop power consumption
77    quad core vs duo core
75    sql join
74    dell xps m1330
74    hard drive temperature
74    vista memory usage
73    source control
70    linked in
69    pontiac aztec
66    pontiac aztek
64    m1330 review
63    cracking
61    consolas
60    captcha
56    hyperterminal
56    ikea jerker
55    code horror
55    polling rate
55    source safe
54    coding horrors
54    dual core or quad core
54    programming quotes
54    visual source safe
53    logparser
51    sourcesafe
51    superfetch
51    three monitors
50    windows experience index

Knowing the most popular keywords can give you some hints what topics to write about on your blog. For example, an article named 'Windows Command Prompt Tricks' would start bringing good traffic from search engines instantly!

I did another bunch of queries to find the most popular programming languages on Coding Horror. I put the languages I could think of in langs.txt file, and ran the following Perl one-liner:

$ perl -MDBI -wlne 'BEGIN { $, = q/ /; $dbh = DBI->connect(q/dbi:SQLite:codinghorror.db/); } print +($dbh->selectrow_array(qq/SELECT count(query) FROM queries WHERE query LIKE "$_" OR query LIKE "$_ %" OR query LIKE "% $_" OR query LIKE "% $_ %"/))[0], $_' langs.txt | sort -n -r

It produced the following output:

1127 visual studio
1087 c#
407 c
287 javascript
239 java
139 asp
104 visual basic
59 php
44 ruby
42 python
26 perl
22 lisp
19 erlang
3 pascal
1 tcl
1 prolog
0 ml
0 haskell

I added 'visual studio' to the list of programming languages, as every beginner thinks it actually is a programming language. There were no keywords matching 'C++' because most search engines think of '+' as an operator rather than a valid search string.

I must say that Python is the answer to life, the universe and everything, as it was searched for 42 times! :)

Here is the same data put on a chart:

Here are some of the most popular search queries among programming languages:

I suggest that you download the keyword database and analyze the data that interests you the most yourself!


Download Perl program: coding horror keyword scraper
Downloaded: 4521 times

Download SQLite database(3 MB): coding horror keyword database (.zip)
Downloaded: 810

If you liked the post, why not vote for it?


latortuga Permalink
April 08, 2008, 21:01

You should consider submitting your dataset to seems like it would have at least some interest decent use. Maybe let it grow a bit more. Just a thought.

April 08, 2008, 21:26

latortuga, thanks for your idea! I'll let it grow bigger and submit it after some time!

Dan Permalink
April 09, 2008, 02:31

In your charts, "system idea" should be "system idle".

Nice post though.

April 09, 2008, 17:22

Great post. That is interesting, its almost like the ability to use keywords for making new blog posts. heh.

April 09, 2008, 17:33

Peter, I'm impressed. You've written a small but perfectly formed application which connects a database to a website and adds a user interface (google charts). All done in a few lines of Perl.
Thanks for linking to Word Aligned!

Roman Permalink
April 09, 2008, 20:32

You appear to have messed up HTML tags: everything is in fixed-width font, starting with "It produced the following output". Aside from that, nice post!

April 09, 2008, 23:51

Scott, yeap, something like that :)

April 09, 2008, 23:52

Thomas, thank you for your comment!

I have been reading your blog for more than a month now. Your article on "Essential Python Reading" got me hooked :)
I been programming Python just for a month, and I really like it. I'm still learning it, and your articles are really helpful.
But, I'd like to say that Perl is still my favorite language because of the massive, centralized CPAN library.

I'm glad that you noticed that I linked to you. :)

April 09, 2008, 23:53

Roman, it's fixed :) It was </cod> tag. Changed it to </code> and it's all fixed :)

April 20, 2008, 04:38

Wonderful post, do you really have much free time to do all that analysis? Where's my requested API 5 lines file? :P

Luci Permalink
December 01, 2008, 15:43

Very interesting blog! Thank you very much for the beautiful articles. I will surely read some more when I get to organize a blog feed.

Leave a new comment

(why do I need your e-mail?)

(Your twitter handle, if you have one.)

Type the word "computer_52": (just to make sure you're a human)

Please preview the comment before submitting to make sure it's OK.