The #L Challenge http://twitter.com/stephenfry/status/1136005076

L_project.sh

== An elementary as hell solution to stephenfry's #L challenge: ==

-- Get the twitter scrape from infochimps.org 
--   http://blog.infochimps.org/2008/12/29/massive-scrape-of-twitters-friend-graph/
-- (not available yet, but sometime soon)
-- 
-- Then run this pig (http://wiki.apache.org/pig/) code 
-- to extract all twitter screen names with three or more L's
-- It's not properly case insensitive but who cares.
--
Ells_1 = FOREACH Users GENERATE screen_name;
Ells_2 = FILTER Ells_1 BY screen_name MATCHES '.*l.*l.*l.*';
STORE Ells_2 INTO 'foo/ells' ;

# ====== SHELL CODE ========

# Whip up a ruby-from-command-line oneoff to find long screen_names with a high fraction of L's
hadoop dfs -cat foo/ells/part\* | \
  ruby -ne '$_ = $_.downcase.chomp!; \
    ls = $_.count("l") ; \
    lf = ls.to_f/$_.length ; \
    puts "%7.3f\t%7d\t%7d\t%s" % [lf, ls, $_.length, $_]
  ' | sort -rn | head -n 50

# The output:
#
#  1.000       6       6 llllll
#  1.000       5       5 lllll
#  1.000      14      14 llllllllllllll
#  1.000      10      10 llllllllll
#  0.867      13      15 illllllllllllli
#  0.867      13      15 hulllllllllllll
#  0.857       6       7 lll_lll
#  0.833       5       6 dlllll
#  0.800       8      10 alllllllly
#  0.800       4       5 llull
#  0.800       4       5 lllel
#  0.800       4       5 llill
#  0.800      12      15 lllvlllclllvlll
#  0.800      12      15 lll8lll8lll8lll
#  0.778       7       9 halllllll
#  0.750       6       8 llilllil

# The third through sixth users' names add up to exactly 50 L's
#
# Compose a message with no other L's except the #L hashtag:
#
#  @stephenfry - @llllllllllllll have you met @llllllllll @IlllllllllllllI
#   and @hulllllllllllll? Your orthographic unity = easy #L project win.