# Accessors and anonymous classes

**URL:** <https://www.ruby-forum.com/t/accessors-and-anonymous-classes/102969>\
**Category:** Ruby\
**Created:** [July 8, 2006, 11:59pm UTC](https://www.ruby-forum.com/t/accessors-and-anonymous-classes/102969 "2006-07-08T23:59:13Z")\
**Posts on this page:** 7\
**Page:** 1

<div class="post-metadata">

**Author:** ![russellf](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/r/7bcc69/32.png) [@russellf](https://www.ruby-forum.com/u/russellf)\
**Post date:** [July 8, 2006, 11:59pm UTC](https://www.ruby-forum.com/t/accessors-and-anonymous-classes/102969/1 "2006-07-08T23:59:13Z")

</div>

Hmmm… odd I wrote a version of this last night but it would appear  
that I failed to post it to the forum, doh! so here it goes again. If  
it did get out and this is a duplicate then please accept my apologies  
for the waste of bandwidth.

I am writing a recursive descent parser to parse a configuration file  
that consists of nested ‘Sections’ which all look like this:  
[] {  
  
  
.  
.  
}

where may be another section.

I wrote a clase Section to parse the basic structure. Section.initialise  
calls method specificItems to parse non section items (see below). For  
each type of section I have an anonymous class tied to a Class variable  
– I have included one such class below. I had to make the classes  
anonymous because I needed to include references to the classes in a  
data structure (see subSections ) which tells the parser which section  
types are valid in this section and what to do with them.

[ aside: someone asked why I wanted reference to scalar variable in my  
previous post – this is why. I needed to have a refernce to a variable  
in that structure into which I could store the parsed class]

Now to the problem: As you see I have defined a bunch of accessor  
methods using attr\_reader and attr\_writer, but these don’t work ( I get  
an undefined method error ). Adding the accessor methods explicitly  
works fine. Odd!

I’m not sure if this is a bug, a feature or simply some lack in my  
understanding of how things should work.

Any ideas?

Also this is my first real OO Ruby program and I would appreciate  
comments – it is too large (currently 1000 lines) to post here but if  
anyone would be willing to have a look at it I’d be happy to email it to  
them. My email is removed\_email\_address@domain.invalid.

I have run into various problems in this project and have managed to  
solve all of them but I am far from confident that my solution was the  
best way of doing it. The use of anonymous classes is one example. I  
wrote a generic text parser for this project and because Ruby does not  
support multiple inheiritance I ended up passing an instance of the  
parser as a parameter to all the methods – I figured that since the  
parser needed to maintain state that I could not make it a mixin…

Cheers, Russell

```
@@hostService = Class.new( Section ) {

  attr_reader = :services, :converted, :actions, :patterns, 

```

:realTime,  
:periodic, :files  
attr\_writer = :converted

```
  def initialize( head, parser )
    @services = []
    @actions = []
    @patterns = []
    @realTime = []
    @periodic = []
    @files = []
    @converted = false
    super( head, parser )
  end

```

# accessor methods — since the shorthand way did not work

```
  def actions
    @actions
  end

```

[boring repeditive stuff snipped]

```
  def converted=(value)
    @converted = value
  end

```

# this handles the non section items in the section

```
  def specificItem( firstToken, parser )

    case firstToken
    when 'service'
      if token = parser.expect(/^(\w+)/, "service name") then
        @services.push( token )
      else
        @errors = true
        parser.restofLine # ignore the rest of the line
      end
    else
      @errors = true
      parser.error( "#{firstToken} not valid in host section" )
      parser.restofLine # ignore the rest of the line
    end
 end

```

# defines which sections may be nested within this section

# Key is the section type the vailue is [ ,

# class to parse section and where to store the result ]

```
  def subSections( kind )
    {'actions' => [nil, @@actionList, @actions],
     'files' => [nil, @@commaList, @files],
     'patterns' => [nil, @@patternList, @patterns],
     'realtime' => [nil, @@matchList, @realTime],
     'periodic' => [nil, @@matchList, @periodic]
    }[kind]
  end
} # end of hostSection
```

---

<div class="post-metadata">

**Author:** ![russellf](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/r/7bcc69/32.png) [@russellf](https://www.ruby-forum.com/u/russellf)\
**Post date:** [July 9, 2006, 2:04am UTC](https://www.ruby-forum.com/t/accessors-and-anonymous-classes/102969/2 "2006-07-09T02:04:06Z")

</div>

Russell F. wrote:

> Now to the problem: As you see I have defined a bunch of accessor  
> methods using attr\_reader and attr\_writer, but these don’t work ( I get  
> an undefined method error ). Adding the accessor methods explicitly  
> works fine. Odd!

Hmmmm… I have now done what I should have done before posting this –  
written a little test program that illustrates the problem. Of course  
it works as expected so something more complex is going on. I’ll go  
back and double check everything yet again. If I can reporduce the  
‘problem’ in a small program I’ll post again.

If anyone feels like looking at the full program email me – address in  
original post.

Russell

---

<div class="post-metadata">

**Author:** ![russellf](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/r/7bcc69/32.png) [@russellf](https://www.ruby-forum.com/u/russellf)\
**Post date:** [July 9, 2006, 2:31am UTC](https://www.ruby-forum.com/t/accessors-and-anonymous-classes/102969/3 "2006-07-09T02:31:16Z")

</div>

Hi –

On Sun, 9 Jul 2006, Russell F. wrote:

> @@hostService = Class.new( Section ) {
> 
> ```
> attr_reader = :services, :converted, :actions, :patterns,
> 
> ```
> 
> :realTime,  
> :periodic, :files  
> attr\_writer = :converted

Just looking at this quickly: you’re assigning to local variables  
called attr\_reader and attr\_writer, when what you want to do is call  
the methods of those names:

attr\_reader :services, :converted, :actions, …

David

---

<div class="post-metadata">

**Author:** ![russellf](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/r/7bcc69/32.png) [@russellf](https://www.ruby-forum.com/u/russellf)\
**Post date:** [July 9, 2006, 4:24am UTC](https://www.ruby-forum.com/t/accessors-and-anonymous-classes/102969/4 "2006-07-09T04:24:53Z")

</div>

unknown wrote:

> Hi –
> 
> On Sun, 9 Jul 2006, Russell F. wrote:
> 
> > @@hostService = Class.new( Section ) {
> > 
> > ```
> > attr_reader = :services, :converted, :actions, :patterns,
> > 
> > ```
> > 
> > :realTime,  
> > :periodic, :files  
> > attr\_writer = :converted
> 
> Just looking at this quickly: you’re assigning to local variables  
> called attr\_reader and attr\_writer, when what you want to do is call  
> the methods of those names:
> 
> attr\_reader :services, :converted, :actions, …

Thanks David! Doh! it had to be something simple and obvious.

And that’s why my orginal test program worked fine - it was correct.

I see ruby suffers from the same problem as perl – just about anything  
is valid syntax, so it compiles but the semantics are not quite what you  
intended 😉

In a past life I spent half each day for several years as a programming  
consultant in a large university (in the days of mainframes) I lost  
count of the number of times that I had people come in with a listing of  
their program, lay it out in front of me and start explaining the latest  
incomprehensible behaviour then suddenly stop, go bright red and quickly  
gather everything up muttering that they now knew what the problem was.  
I know how they feel 🙂

Russell (who has just spent an hour or so carefully paring down his  
program to 10 lines that illustrates the problem 🙂

---

<div class="post-metadata">

**Author:** ![russellf](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/r/7bcc69/32.png) [@russellf](https://www.ruby-forum.com/u/russellf)\
**Post date:** [July 10, 2006, 7:59am UTC](https://www.ruby-forum.com/t/accessors-and-anonymous-classes/102969/5 "2006-07-10T07:59:06Z")

</div>

2006/7/9, Russell F. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid):

> I see ruby suffers from the same problem as perl – just about anything  
> is valid syntax, so it compiles but the semantics are not quite what you  
> intended 😉

Although Ruby’s syntax is indeed quite flexible I’m rarely bitten by  
this. I definitively struggled more with Perl in the old days I used  
it rather frequently. So, the good news is: although you stepped into  
this once you likely won’t too often in the future. Ah, and btw, the  
flag “-w” is also a good way to learn about some of these errors.

Kind regards

robert

---

<div class="post-metadata">

**Author:** ![russellf](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/r/7bcc69/32.png) [@russellf](https://www.ruby-forum.com/u/russellf)\
**Post date:** [July 10, 2006, 8:17pm UTC](https://www.ruby-forum.com/t/accessors-and-anonymous-classes/102969/6 "2006-07-10T20:17:26Z")

</div>

Russell, some of your questions haven’t been answered yet:

> (…)  
> I had to make the classes  
> anonymous because I needed to include references to the classes in a  
> data structure (see subSections ) which tells the parser which section  
> types are valid in this section and what to do with them.

I think you can use named classes just as well as anonymous classes:

class HostService \< Section  
…  
def sub\_sections( kind )  
{‘actions’ =\> [nil, ActionList, @actions],  
‘files’ =\> [nil, CommaList, @files],  
…  
end  
end

(Note that I renamed the method. See  
[http://wiki.rubygarden.org/Ruby/page/show/RubyStyleGuide](http://wiki.rubygarden.org/Ruby/page/show/RubyStyleGuide))

If you need to reference the classes before they are created, there are  
other possibilities besides class variables: you could use the names of  
the classes in the references and then call Module#const\_get to get to  
the class object:

def sub\_sections( kind )  
{‘actions’ =\> [nil, “ActionList”, @actions],  
‘files’ =\> [nil, “CommaList”, @files],  
…  
end

def class\_for\_name( name )  
Object.const\_get( name )  
end

Or you could pre-create the classes and re-open them later:

class HostService \< Section; end  
class ActionList \< Section; end  
class CommaList \< Section; end  
…

class HostService  
def sub\_sections( kind )  
{‘actions’ =\> [nil, ActionList, @actions],  
‘files’ =\> [nil, CommaList, @files],  
…  
end  
end

You could also add a convenience method to Section:

class Section

```
 def self.define_sub_sections( *names )
   names.each do |name|
     const_set( name, Class.new( Section ) )
   end
 end

 define_sub_sections "HostService", "ActionList", "CommaList"

 class HostService
   ...
 end

```

end

> [ aside: someone asked why I wanted reference to scalar variable in my  
> previous post – this is why. I needed to have a refernce to a variable  
> in that structure into which I could store the parsed class]

I still don’t understand the need to reference a variable. It seems all  
you want is to store a value in a variable, which you did with the class  
variables.

> (…)  
> I have run into various problems in this project and have managed to  
> solve all of them but I am far from confident that my solution was the  
> best way of doing it. The use of anonymous classes is one example.

See above.

> I  
> wrote a generic text parser for this project and because Ruby does not  
> support multiple inheiritance I ended up passing an instance of the  
> parser as a parameter to all the methods – I figured that since the  
> parser needed to maintain state that I could not make it a mixin…

Modules in Ruby can maintain state. The only caveat is initializing the  
state. There are different ways to do it, and it has been discussed  
often on ruby-talk, but I don’t have a reference.

Regards,  
Pit

---

<div class="post-metadata">

**Author:** ![russellf](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/r/7bcc69/32.png) [@russellf](https://www.ruby-forum.com/u/russellf)\
**Post date:** [July 15, 2006, 8:36am UTC](https://www.ruby-forum.com/t/accessors-and-anonymous-classes/102969/7 "2006-07-15T08:36:21Z")

</div>

Pit, thanks very my for your comments and insights!

Pit C. wrote:

> I think you can use named classes just as well as anonymous classes:
> 
> class HostService \< Section  
> …  
> def sub\_sections( kind )  
> {‘actions’ =\> [nil, ActionList, @actions],  
> ‘files’ =\> [nil, CommaList, @files],  
> …  
> end  
> end

Yup, that works. I tried this initially but had problems (clearly  
something else was wrong). That’s a useful simplication.

> (Note that I renamed the method. See  
> [http://wiki.rubygarden.org/Ruby/page/show/RubyStyleGuide](http://wiki.rubygarden.org/Ruby/page/show/RubyStyleGuide))

Ah, right… time for some global sustitutions 🙂

[good stuff snipped for brevity]

> > wrote a generic text parser for this project and because Ruby does not  
> > support multiple inheiritance I ended up passing an instance of the  
> > parser as a parameter to all the methods – I figured that since the  
> > parser needed to maintain state that I could not make it a mixin…
> 
> Modules in Ruby can maintain state.

Yep, I worked that out eventually and the parser now works much better  
as a mixin. I’ve renamed initialize to setup and call it explictly…

Seems to work fine. 🙂

I have redesigned things so I am not having to resort to reference to  
get values out of sub classes. So everything is looking good – thanks  
to the good folk on this list/forum!

Russell