# [Copyright 1998 Peter Maydell <pmaydell@chiark.greenend.org.uk>]
# Not a shell script; the comments make the syntax invalid.
# Comments are on lines starting with '#'; if you strip them all out
# then the syntax is OK.

# Standard boilerplate to create a working directory /tmp/t
# Note that the last \ on each line is the shell line continuation marker.
cd /tmp && mkdir t && cd t && \
# More standard boilerplate. -noleaf is important so that find doesn't
# do an optimisation which causes it to not see directories created
# during the execution of commands relating to the parent directory.
find . -noleaf 
# This is the standard way to do something as an initialisation routine
# (similar concept to awk's BEGIN section): since /tmp/t is empty, the
# first filename to be processed is '.'
# We create 2 (ie /tmp/t/2) as a subdirectory to start the ball rolling,
# and print '2', since 2 is a prime.
-path . -exec mkdir 2 \; -printf '2\n' \
# This is the termination condition: if we're processing the directory
# '100n' then we should stop here. We terminate by simply not creating
# a subdirectory.
-o -name 100n \
# The rest of this handles the general case. Suppose we are in a directory
# named x-1. Then we create a subdirectory x if x is prime or xn if x
# is not prime. The result is to create a directory structure
# /tmp/t/2/3/4n/5/6n/7/8n/9n/10n/11/...
# First we need to find out whether x is prime or not. To do this we
# exec a process which exits successfully if x is not prime: 
-o -exec sh -c "find . -path . -o -name '*n' \
-o -exec expr \( \( \( {} : '.*/\([^n]*\)' \) + 1 \) / \{\} : '.*/\(.*\)' \
\) \* \{\} : '.*/\(.*\)' - \( {} : '.*/\([^n]*\)' \) - 1 ';' | grep '^0$' \
>/dev/null " \; \
# Phew. What this does is run a sub-find which executes expr for
# each directory which doesn't end 'n' (ie all the primes). 
# The expr uses  {} : '.*/\([^n]*\)'  to extract the final number 
# (eg given ./2/3/4n/5 it evaluates to 5) and thus the find essentially
# evaluates ((x+1)/n) * n - x - 1  for all prime n between 2 and x.
# This is zero for some n iff x+1 is not prime, so we grep for that, 
# throwing away the output of the grep and just using its exit status
# (true for if it matched a '0'). Note that we use {} to get the top
# level find to substitute its filename, and \{\} to get the sub-find
# to do the substitution. This works because the top find sees \{\}
# but the argument presented to the subfind has gone through an extra
# level of shell quoting so it gets {} (and gets a substituted file
# name where the top find saw {}).
#
# We now use the fact that -foo -bar is equivalent to -foo -a -bar,
# so if -foo fails we wouldn't do -bar. 
# If the subfind succeeds we execute this mkdir; otherwise we
# continue executing after the next -o.
# If we're currently in ./2/3/4n/5, this creates the directory
# ./2/3/4n/5/6n, using expr to do the addition and the shell for 
# quoting services (backticks).
-exec sh -c "mkdir {}/\`expr {} : '.*/\([^n]*\)' + 1\`n" \; \
# if the sub-find failed, we do a similar thing, but without the trailing 'n'
-o -exec sh -c "mkdir {}/\`expr {} : '.*/\([^n]*\)' + 1\`" \; \
# and we also print x+1, since it's prime. We do this by printing a 
# listing of the current directory (which for eg ./2/3/4n will have 
# a single entry '5'); this is both for variety and to avoid using a
# grotty expr pattern again.
-exec sh -c "echo \`ls {}\`" \; \
# This is the standard trailer to remove all the directories we created.
# We can't do that as part of the find command because it confuses it...
&& cd /tmp && rm -r t 
#
# See? Perfectly comprehensible.