Scope problem (?) in implementing Design Patterns in Ruby

Ruby
1

Hi,

I’m hoping that employing “Design Patterns in Ruby” will lead to less
coding errors and more easily maintained code. I’m stuck with the Not
pattern in the File Finding pattern in Chapter 15.

I’ve posted my code and output in http://www.pastie.org/1889586 and
http://www.pastie.org/188968, respectively. Any idea about the cause
of the reported syntax error?

Thanks in Advance,
Richard

2

I think the second link points to the wrong pastie.

I can see a couple of problems with the code, first of which is that you
are
setting up Expression.new to require an argument but not passing it an
argument when you call All_basic.new, etc.

More than design patterns (which I think are a little much here), I
would
suggest looking into Ruby’s higher order functions like reduce and map.
They
make your code a lot simpler. This is an alternate implementation I did.
Let
me know what you think:

http://www.pastie.org/1889838

On Wed, May 11, 2011 at 1:20 PM, RichardOnRails <

3

Correction: My bad, your version of All_basic doesn’t inherit from
Expression like I thought it would from the pattern it follows.

4

One more suggestion: if you’re going to build a hierarchy, might as well
work the parent as much as possible. My final implementation would
probably
look something like this:

http://www.pastie.org/1889969

5

RichardOnRails wrote in post #998059:

  1. Your case statement syntax doesn’t work in ruby 1.9.2:

prog.rb:4: syntax error, unexpected ‘:’, expecting keyword_then or ‘,’
or ‘;’ or ‘\n’

You have to use ‘then’ in place of a colon.

  1. Next, I get this error:

prog.rb:67:in evaluate': uninitialized constant Not::All (NameError) from prog.rb:74:in’

which relates to this code:

class Not < Expression
def initialize(expression)
@expression = expression
end

def evaluate(dir)
all = All.new.evaluate(dir)
other = @expression.evaluate(dir)
all - other
end
end

In ruby, constants are looked up like directories and files (or if you
prefer constants are ‘lexically scoped’). When you
are inside the Not class (which is a module), the ‘directory’ you are in
for constant lookups is the ‘Not’ directory. When you write All.new,
because the name All is not preceded by a directory name, ruby looks in
the current ‘directory’ for the constant All. The current directory is
Not, so ruby is looking for Not::All, i.e. the ‘file’ All in the
‘directory’ Not. However, All is not defined inside Not, so you get an
error. In fact, there is no constant named All defined anywhere in
your program, so the error is more serious than a scope problem.

If you are inside a class/module and you need to access a class at the
top level, you do this:

class All
def greet
puts ‘hi’
end
end

class Dog
def do_stuff
::All.new.greet #<*****
end
end

Dog.new.do_stuff #=>hi

And, if you want to access a class contained inside another module, you
do this:

module A
class B
def greet
puts ‘hi’
end
end
end

class Cat
def do_stuff
A::B.new.greet #<******
end
end

Cat.new.do_stuff #=>hi

6

I think the lookup starts at the innermost scope but since All isn’t
defined
in Not, it reaches the top-level.

7

7stud – wrote in post #998081:

RichardOnRails wrote in post #998059:

  1. Your case statement syntax doesn’t work in ruby 1.9.2:

prog.rb:4: syntax error, unexpected ‘:’, expecting keyword_then or ‘,’
or ‘;’ or ‘\n’

You have to use ‘then’ in place of a colon.

  1. Next, I get this error:

prog.rb:67:in evaluate': uninitialized constant Not::All (NameError) from prog.rb:74:in’

which relates to this code:

class Not < Expression
def initialize(expression)
@expression = expression
end

def evaluate(dir)
all = All.new.evaluate(dir)
other = @expression.evaluate(dir)
all - other
end
end

In ruby, constants are looked up like directories and files (or if you
prefer constants are ‘lexically scoped’). When you
are inside the Not class (which is a module), the ‘directory’ you are in
for constant lookups is the ‘Not’ directory. When you write All.new,
because the name All is not preceded by a directory name, ruby looks in
the current ‘directory’ for the constant All. The current directory is
Not, so ruby is looking for Not::All, i.e. the ‘file’ All in the
‘directory’ Not. However, All is not defined inside Not, so you get an
error. In fact, there is no constant named All defined anywhere in
your program, so the error is more serious than a scope problem.

If you are inside a class/module and you need to access a class at the
top level, you do this:

class All
def greet
puts ‘hi’
end
end

class Dog
def do_stuff
::All.new.greet #<*****
end
end

Dog.new.do_stuff #=>hi

In fact, that is unnecessary:

class All
def greet
puts ‘hi’
end
end

class Dog
def do_stuff
All.new.greet #<*****
end
end

Dog.new.do_stuff #=>hi

I guess the lookup actually starts at the toplevel. So your error is a
result of not defining the constant All at the toplevel, and ruby
obfuscates the error by telling you that All is not defined inside the
Not module, giving you the error message: Not::All doesn’t exist.

8

David J. wrote in post #998093:

I think the lookup starts at the innermost scope but since All isn’t
defined
in Not, it reaches the top-level.

Yes, you are right:

===
Constants defined within a class or module may be accessed unadorned
anywhere within the class or module.

(Programming Ruby)

===
Constants declared outside of a class or module are assigned global
scope.

(http://www.techotopia.com/index.php/Ruby_Variable_Scope#Ruby_Constant_Scope)

So it’s a case of the inner All hiding the global All. And you can use
the :: prefix to leap over an inner scope constant that hides a toplevel
constant:

class All
def greet
puts ‘All#greet’
end
end

module C
class All
def greet
puts “C::All#greet”
end
end

class Dog
def do_stuff
::All.new.greet #<*****
end
end

end

C::Dog.new.do_stuff

–output:–
All#greet

9

…which doesn’t seem like that should work.

Why not?

10

7stud – wrote in post #998090:

I guess the lookup actually starts at the toplevel.

Well, that’s not true either. If I define two All classes: one at the
top level and one inside a module,

class All
def greet
puts ‘All#greet’
end
end

module C
class All
def greet
puts “C::All#greet”
end
end

class Dog
def do_stuff
All.new.greet #<*****
end
end

end

C::Dog.new.do_stuff

–output:–
C::All#greet

Then if I delete C::All:

class All
def greet
puts ‘All#greet’
end
end

module C
class Dog
def do_stuff
All.new.greet #<*****
end
end
end

C::Dog.new.do_stuff

–output:–
All#greet

…and I don’t think both should work. So, I guess I don’t understand
the scoping of constants yet.

11

On May 11, 1:52pm, David J. [email protected] wrote:

make your code a lot simpler. This is an alternate implementation I did. Let

of the reported syntax error?

Thanks in Advance,
Richard

I think the second link points to the wrong pastie.
It sure does. I apologize for that error. My results are at
http://www.pastie.org/1889681

This is an alternate implementation I did. Let
me know what you think:

http://www.pastie.org/1889838

I downloaded your example, removed a few spurious “end” statement and
added a “p” to get the Regexp class name.
It ran perfectly as far as I can see, but I’d have to format the
output to confirm that.
The code looks great, but I’ll await a real comparison until I get
the book’s code working

I can see a couple of problems with the code, first of which is that you are
setting up Expression.new to require an argument but not passing it an
argument when you call All_basic.new, etc.

Thanks for this possible solution. I’ll try to correct that after I
look at the other posts I’ve been lucky enough to get.

Best wishes,
Richard

12

On May 11, 2:37pm, David J. [email protected] wrote:

More than design patterns (which I think are a little much here), I would

I’m hoping that employing “Design Patterns in Ruby” will lead to less
coding errors and more easily maintained code. I’m stuck with the Not
pattern in the File Finding pattern in Chapter 15.

I’ve posted my code and output inhttp://www.pastie.org/1889586and
http://www.pastie.org/188968, respectively. Any idea about the cause
of the reported syntax error?

Thanks in Advance,
Richard

Hi David,

Assuming I’m responding to the correct you version works perfectly, as
evidenced by results below gotten after I decorated your code with
puts versions.

The Design Patterns author, Russ Olson, is aims to support a neat DSL
to support file search with arbitrary logical combinations of the
fundamental search elements … which is my adopted goal. So I’ve got
to master his approach to implementing this search language before I
can really weigh approaches.

7stud has responded with the key thing I saw but couldn’t figure out
how to address the problem, so I’ve got to follow up on that now.

Thanks for very detailed responses in intrinsic Ruby education.

Best wishes,
Richard

======== Output ==========

ruby DavesTempFilesDirs_2.rb

files.select &all

AlternativeFileSearch.rb
DavesTempFilesDirs_1.rb
DavesTempFilesDirs_2.rb
FindFilesDirs-01.rb
FindFilesDirs-02.rb
FindFilesDirs.rb
TempFindFilesDirs.rb
[email protected]
TestFolder
TestFolder/TestDoc.txt

files.select &filename(/-01/))

FindFilesDirs-01.rb

files.select &writable

AlternativeFileSearch.rb
DavesTempFilesDirs_1.rb
DavesTempFilesDirs_2.rb
FindFilesDirs-01.rb
FindFilesDirs-02.rb
FindFilesDirs.rb
TempFindFilesDirs.rb
TestFolder/TestDoc.txt

files.select &_not(writable)

[email protected]
TestFolder

13

Hi 7Zip and David,

Thanks very much to you both for hanging in there with me.

I hacked up a working version (agnostic for 1.8.6/1.9.2) here:
http://www.pastie.org/1893405,
with results here: http://www.pastie.org/1893420

I apologize for having screwed a Pastie URL last time.

I left comments in code for the hacks I made. When I first
encountered the problem, I made a feeble attempt at prefixing :: for
All but I failed to take note of the fact that the All definition was
subordinate to Expression. That was stupid on my part, but I blame it
on the fact that I’m 77, have been retired from programming for 7
years and have only got serious about learning Ruby/Rails about a year
ago. It’s been a struggle; I used to be faster on the uptake.

BTW, David Black has a very nice exposition of lookup for methods and
classes on page 98 et seq of his “The Well-Grounded Rubyist”, one of
my “bibles”.

I don’t know how my hacks will playout for the parsing scheme in
“Design Patterns in Ruby”. I going to follow it until this newsgroup
and/or I discover it’s a fool’s errand.

With my very best wishes to you both and thanks for your generous
insights,
Richard

14

Hi Richard,

Happy to help.

I would definitely encourage you to not only look at design patterns
(which are okay, but often overkill or already built in to Ruby) but
also to Ruby’s standard features to reduce code errors.

Ruby is a rich language. It has a lot of really well-crafted methods
that take care of a most of the tedium of coding, and for that reason
the standard library is well worth learning.

For example, in your code, instead of importing ‘find’ and using it to
look at directories, use Ruby’s built-in Dir module. Dir[’**/*’] will
get you all subdirectories and files recursively.

Second, look into higher order functions. They let you change code from
this:

def evaluate(dir, arg)
results = []
Dir[’**/*’].each do |p|
next unless File.file? p
name = File.basename§
case arg
when String
results << p if File.fnmatch(@arg, name)
when Regexp
results << p if @arg.match(name)
end
end
results
end

Into this:

def match?(str_or_re, file)
str_or_re.is_a?(String) ? File.fnmatch(str_or_re, file) : str_or_re =~
file
end

def evaluate(dir, arg)
Dir[’**/*’].select {|f| File.file?(f) and match?(arg, f) }
end

Any time you see this pattern …

my_temporary_variable = []

my_collection.each do |elem|
my_temporary_variable << elem if some_condition
end

return my_temporary_variable

… you know that higher order functions would probably be a better
solution. :slight_smile:

Hope that helps, and good luck learning this beautiful language.

Cheers,
David

Any time you see this pattern in your code, you should automatically
think

15

David J. wrote in post #998417:

Hi Richard,

Happy to help.

I would definitely encourage you to not only look at design patterns

Hard to do when you are reading a book on Design Patterns, and you are
trying to understand one of the examples. I’m also currently reading
the same book as RichardOnRails, but I’m not quite that far along.

The author does a good job of showing how some of the patterns
devolve into simpler constructs in ruby.

Second, look into higher order functions.

Does your example really involve higher order functions? It’s my
understanding a higher order function is a function that takes another
function as an argument. Of course, in ruby blocks are like passing one
function to another function, but your method neither takes a method as
an argument nor does it yield to a block.

16

Is it a higher order function because it calls a global method?

17

Hard to do when you are reading a book on Design Patterns, and you are

trying to understand one of the examples. I’m also currently reading
the same book as RichardOnRails, but I’m not quite that far along.
I’ve read the book. It’s good but I think it’s healthy to understand how
design patterns aren’t always necessary in Ruby like they are in Java.
I’m just pointing out other, potentially more effective ways to write
re-usable code.

Cheers,
David

18

No, sorry for the confusion, the higher order function transformation
was this:

results = []
Dir[’**/*’].each do |p|
next unless (… …)
results << p if (… …)
end
results

Into this:

Dir[’**/*’].select {|x| (… …) }

Other good examples are Array#map, Array#reduce and Array#grep.

PS I apologize for the indentation, my mail client is acting up.
On Thursday, 12 May 2011 at 10:20 pm, 7stud – wrote:
Is it a higher order function because it calls a global method?

19

On May 12, 9:56pm, David J. [email protected] wrote:

Second, look into higher order functions. They let you change code from this:
results << p if @arg.match(name)

end
Any time you see this pattern in your code, you should automatically think

Thanks very much to you both for hanging in there with me.
subordinate to Expression. That was stupid on my part, but I blame it
and/or I discover it’s a fool’s errand.

With my very best wishes to you both and thanks for your generous
insights,
Richard

Hi Dave,

I checked out your Version 2. Added puts stuff and restriction to
names of files, which should probably be another one of your methods
instead of my hard-coding.
Expanded code: http://www.pastie.org/1895917
Output: http://www.pastie.org/1895932

I have only one question: I put a space between the & and the method
following it; the code broke. Is the construct you used documented on-
line somewhere? I’ve peeked at lambda documentation so I’ve got the
drift but I haven’t used it in any of my code so far.

I copied your memo on the approach and plan to review it and use/
expand your code after I finish getting Russ Olson’s code working or
giving up on it. To his credit, he posted his code on-line so I can
avoid typos and he included an apparently thorough set of unit tests.
So I’ve got a lot of studying to do.

Again, thanks for your guidance and great ideas,
Richard

20

Hi Richard,

The & has to be adjacent to “all” or “filename” to act as expected. The
&
can mean one of two related things when you pass it to a method:

  1. It can mean “pass the lambda expression that this variable points to
    into
    a block”. That makes the following to snippets equivalent:

Snippet 1

filter_function = lambda {|x| x == 2 }
[1, 2, 3, 4, 2].select &filter_function

Snippet 2

[1, 2, 3, 4, 2].select {|x| x == 2 }

  1. It can mean “pass the instance method named by a symbol, and treat it
    as
    a block”.

That makes these two equivalent:

Snippet 1

[1, 2, 3, 4, 2].select {|x| x.even? }

Snippet 2

[1, 2, 3, 4, 2].select(&:even?)

So what my code is doing is calling the all method, for example, and
that
method’s sole job is to return a lambda that can then be passed in per
the
above. You don’t have to wrap the lambda in a method like I did. You
could
also just say:

all = lambda {|x| x }

I hope that makes things a little clearer!

Cheers,
David
On Fri, May 13, 2011 at 3:31 AM, RichardOnRails <