Hi.
I'm working on a small project of mine (a system to collect and easily
graph data via a "smart" wrapper around rrdtool), and I'm thinking about
the design (while the whole stuff is already in production, how
typical).
Overall, it goes like this : the system has a few types of plugins ; to
be classified as a plugin, an object must inherit from a pre-defined
class and implement a few methods. Some of these methods receive helper
objects and have to call methods on these objects to achieve something.
The trouble is I have a few "dangerous" methods defined in the various
objects, and a few which aren't useful for the user. With this design,
I don't see how I could protect those methods - and logically, I think
they belong to their objects.
So, my questions are : does this look like a sound design ? What are
the ways to have a better "protection" of the code ? Should I care ?
(It's a rhetoric question - I care, only if that gives me the occasion
to learn something... 
)
For instance (not actual code, but a very simplified example) :
$tests = []
class Tester
def Tester.inherited(c)
$tests.push(c.new())
end
def initdatastruct(datastruct)
raise "initdatastruct must be implemented"
end
end
class DataStruct
attr_reader :sources
def initialize()
@sources = []
end
def addsource(name = nil, min = nil, max = nil)
@sources.push(Source.new(name, min, max))
end
def doit()
cl = "program -c " + @sources.collect do |s|
s.to_cmd
end.join(' ')
system(cl)
# puts cl
end
end
class Source
attr_reader :name, :min, :max
def initialize(name = nil, min = nil, max = nil)
@name, @min, @max = name, min, max
end
def to_cmd()
[ @name, @min, @max ].join(':')
end
end
class PluginTester < Tester
def initdatastruct(ds)
ds.addsource("blah", 0, 100)
ds.addsource("blih", -100, 100)
end
end
$tests.each do |t|
ds = DataStruct.new()
t.initdatastruct(ds)
ds.doit
end
(The dangerous method would be the DataStruct#doit part ; the unuseful
one would be Source#to_cmd.)