String interpolation list item text #71

Activity

1. jdugan commented on Oct 12, 2017

   

   jdugan

   on Oct 12, 2017

   Last edited by jdugan

   Contributor

   More actions

   Hi, Andy.

   Scoping is an fairly gnarly issue for a gem like Caracal, but I suspect there is an existing way around your issue.

   What You Actually Asked Me About

   First, a small bit of background.

   The main difference between a Word document and an HTML document (for example) is that the Word document actually requires several output streams to be written and collected to form the "output". This complicates the rendering process for Caracal because it can't really just "render as it goes" like, say, Prawn can. For better or worse, Caracal was designed to model all DSL instructions first and only when all the instructions are stored does it attempt to produce the various output files and zip them into the final document. The reason this complicates things is because Caracal is fundamentally turning all your instructions into a series of nested model objects as it parses your commands, which by default rapidly destroys the idea of a single context in which the commands can be evaluated.

   Having written that, there is generally a way to preserve outer context. In an effort to give programmers some degree of control over how their instructions are interpreted, Caracal's BaseModel implements an arity strategy I discovered in similar gems when I was designing Caracal--it chooses a block evaluation strategy based on the arity of the block. So, if you call a Caracal method without a block argument (i.e., arity < 1), the gem assumes all the variables in the block are local and it uses instance _eval(&block); if you provide a block argument, Caracal assumes the variables are not all local and uses block[self] instead--self here is the caller rather than the receiver.

   Consider the following example, which removes Caracal from the equation:

   class InnerClass
     attr_reader :birth_year
   
     def initialize(age)
       @birth_year = 2017 - age
     end
   
     def format(&block)
       (block.arity < 1) ? instance_eval(&block) : block[self]
     end
   end
   
   class OuterClass
     attr_reader :name, :age, :inner
   
     def initialize(name, age)
       @name  = name
       @age   = age
       @inner = InnerClass.new(age)
     end
   
     # This evaluates self in the context of the caller 
     # so it explicitly prefixes the value coming from 
     # receiver.
     #
     def print_cool
       inner.format do |ic|
         puts "#{ name } - #{ age } - #{ ic.birth_year }"
       end
     end
   
     # This evaluates self in the context of the caller  
     # but it fails to identify the birth year's owner. 
     # So, ruby thinks the caller owns birth_year, which 
     # is not true, hence the error.
     #
     def print_notcool
       inner.format do |ic|
         puts "#{ name } - #{ age } - #{ birth_year }"
       end
     end
   
     # This evaluates self in the context of the receiver, 
     # so the birth year is cool but ruby has no idea what 
     # name and age mean now, hence the error.
     #
     def print_alsonotcool
       inner.format do
         puts "#{ name } - #{ age } - #{ birth_year }"
       end
     end
   end

   You Didn't Ask, But I'd Appreciate Your Opinion

   FWIW, I realise this feature of Caracal is not documented in any way. Probably it should be. But self is a sufficient advanced concept that you're the first person to ask me about scoping in a non-trivial way.

   I genuinely don't think there's a great, auto-magical way to spare advanced users from managing scope explicitly, but I'm 100% open to a comprehensively designed alternative to what I've implemented.

   I'm also not clear how best to even write up this discussion in a way that an ordinary ruby user would understand. From what I can tell, most Caracal users are building relatively simple, top-down Word documents through Rails, so I typically advise them to localise their values into simple strings/objects and reference those in their blocks. (The instance_eval calls pick these data structures fine.) My sense is if I started delving into scoping and self, I'd lose most of them immediately, you know?
2. jdugan

   added

   documentation

   invalid

   on Dec 31, 2017
3. jdugan commented on Jan 3, 2018

   

   jdugan

   on Jan 3, 2018

   Last edited by jdugan

   Contributor

   More actions

   A Practical Solution

   I'm going to close this issue, so I wanted to follow-up in case you were still interested. (If not, that's also 100% fine.)

   I added a link to this discussion in the README to help others with lexical scoping questions. I'm also going to explicitly answer your immediate question so that the issue will include a "solution" per se.

   The ruby issue in your example is that the variables are implicitly referencing self.
   But by the time the interpreter encounters name and age, self is no longer a instance of MyClass but rather an instance of Caracal ListItem. Caracal's ListItem model has no idea what name and age mean.

   To get the example to work, we need to tell the interpreter explicitly which object owns name and age. To do so, we can employ a strategy familiar to javascripters:

   class MyClass
       attr_reader :name, :age
   
       def initialize(name, age)
         @name = name
         @age = age
       end
   
       def create_doc
         obj = self
         Caracal::Document.save('my_file.docx') do |document|
           document.h1 'Example'
           document.ul do
             li "Name: #{obj.name}"
             li "Age: #{obj.age}"
           end
         end
       end
     end
   
     MyClass.new('john', 30).create_doc

   👍React with 👍1Reacted by Goulven Champenois
4. jdugan

   closed this as completedon Jan 3, 2018
5. Uelb commented on Jul 1, 2019

   

   Uelb

   on Jul 1, 2019

   More actions

   hey, I had the same problem and had a hard time finding this issue, I believe it would be nice to add that to the readme.
6. jdugan commented on Jul 4, 2019

   

   jdugan

   on Jul 4, 2019

   Contributor

   More actions

   Hi, @Uelb. There is a direct link to this issue in the README under a section named Using Variables. Were you recommending something other than that?
7. Uelb commented on Jul 5, 2019

   

   Uelb

   on Jul 5, 2019

   More actions

   Actually I missed that and it may be enough ! Thanks. Maybe a better way would be to directly include the bad and the good example in the readme.
8. koutzasnew commented on Dec 9, 2020

   

   koutzasnew

   on Dec 9, 2020

   More actions

   Hi, this could also work using

   @@name = name
   @@age = age
   

   inside the initialize method and then using @@name and @@age in your main code.