I name this script railstrunk on my machine. So, usage would be:

railstrunk app_name

That creates a rails application under app_name.

If app_name is a subversion working copy, rails is set as an external and generated files are committed, and some ignore properties are set.

Here's how I'd go about creating a versioned rails app:

mkdir -p happy_app happy_app/trunk happy_app/tags happy_app/branches 
svn import happy_app http://myserver/myrepos/happy_app -m "Layout for happy_app"
rm -rf happy_app
svn co http://myserver/myrepos/happy_app/trunk happy_app
railstrunk happy_app

$rails_dir is set at the beginning of script, and it should be a path to a working copy of rails trunk on your machine. It's merely used to speed things up, it's not necessary.

Notice I remove the silly public/index.html from the generated app.

And now the code:

#!/bin/bash

rails_dir=~/code/ruby/rails

if [ $# != 1 ]; then
  echo "Usage: $0 app_name"
  echo
  echo "Creates a rails application under app_name."
  echo
  echo "If app_name is a subversion working copy, rails is set as an external"
  echo "and generated files are committed."
  echo
  echo "If $rails_dir exists, things are sped up by either symlinking it"
  echo "(for non-versioned apps) or copying it to the vendor dir."
  echo
  echo "$rails_dir should be a rails trunk working copy."
  exit
fi

dir=$1
mkdir -p $dir
cd $dir

if [ -n "`ls`" ]; then
  echo "Can't create app: $dir is not empty." >&2
  exit 1
fi

if [ -d $rails_dir ]; then
  if [ "`svn info $rails_dir 2>/dev/null | grep URL:`" != "URL: http://dev.rubyonrails.org/svn/rails/trunk" ]; then
    echo "$rails_dir is not a rails trunk working copy. Not going to use it." >&2
  elif [ -n "`svn st $rails_dir`" ]; then
    echo "$rails_dir is modified. Not going to use it." >&2
  else
    use_rails_dir=1
  fi
fi

if [ -z "`svn info 2>/dev/null`" ]; then
  mkdir vendor
  if [ -n "$use_rails_dir" ]; then
    ln -s $rails_dir vendor/rails
    cd vendor/rails
    svn up
    cd ../..
  else
    svn co http://dev.rubyonrails.org/svn/rails/trunk vendor/rails
  fi
  ruby vendor/rails/railties/bin/rails .
  rm public/index.html
else
  svn mkdir vendor
  svn ps svn:externals 'rails http://dev.rubyonrails.org/svn/rails/trunk' vendor
  svn ci -m 'set rails external to rails trunk'
  [ -n "$use_rails_dir" ] && cp -r $rails_dir vendor/rails
  svn up
  ruby vendor/rails/railties/bin/rails .
  rm public/index.html
  rm log/*
  mv config/database.yml config/database.yml.sample
  svn add . --force
  svn ps svn:ignore '*' tmp/cache tmp/pids tmp/sessions tmp/sockets
  svn ps svn:ignore 'database.yml' config
  svn ps svn:ignore '*.log' log
  svn ci -m "- created rails app
- moved database.yml to database.yml.sample
- deleted public/index.html
- ignored logs and tmp"
  cp config/database.yml.sample config/database.yml
fi

perl -pi -e 's|^#!/.*|#!/usr/bin/env ruby|;' script/* script/*/* public/dispatch.*

Often in rails you'll do things like

<%= @post.author.name %>

but it may be that you have no post, or the post has no author, and even if that's ok, the above would raise. I came up with a solution that lets you do the following:

<%= ifnil("no author"){ @post.author.name } %>

If at any point the receiver is nil (either @post or #author), execution of the block is stopped and the default value is returned ("no author" in this case, but defaults to nil)

NoMethodError is still raised if you send an invalid message to a non-nil receiver.

Here's the code for ifnil

def ifnil(value=nil)
  yield
  rescue NoMethodError
  raise unless $!.message =~ /:NilClass$/
  value
end