From: stephen.hill@... (Steve Hill)
Date: 2001-08-16T00:12:28+09:00
Subject: [ruby-talk:19785] Re: "new" returning nil: how to report the failure of object creation

Hi,
surely an appropriate way to handle this problem would be using a
Factory design pattern.

so (writing off the top of my head - this is not a tested piece of
code) you would have something like this;

class DataFactory
   # initialize etc...

   def getData(file)
      ret_data=nil
      line=file.gets("data_end")
      md=/data_start(.*)data_end/.match(line)
      ret_data=Data.new(md[1]) unless md==nil
      return ret_data
   end
end

class Data
   def initialize(str)
   #...
   end
end

The basic idea is that Data is initialized from a string (in the
example above the data_start and data_end markups are not part of the
string, but it doesn't have to be this way. The DataFactory#getData
method reads in a line from file terminated by the end of the next
data object (or the end of the file), then uses a regexp to extract
the data fields in that line, which it uses to build a Data object.

Best regards

Steve  



> Eli Green <eli.green@codedogs.ca> wrote in message news:<20010814235952.A8247@codedogs.ca>...
> > Why not raise an exception?
> >
> > from Ryo Furue on 2001-08-15 at 12:51:00:
> > > Hi there,
> > > 
> > > How do you report the failure of object creation, that is, if you find
> > > it impossible to properly initialize the object in the "initialize"
> > > method?  Raising an exception is clearly an option.  But, what if the
> > > "failure" is a normal part of the flow of the program?[...]
> 
> Why not raise an exception?  Because "Whenever reasonable, one should
> stick to the `exception handling is an error handling' view"
> (Stroustrup, B., 1997, _The C++ Programming Language_, 3rd ed.,
> p.375).  I agree to this opinion. (Though the quoted statement is
> about C++, I think it almost equally applies to Ruby.)
> 
> Why doesn't IO#gets raise an exception when end of file is reached? 
> Because we usually expect an end of file so that handling it is a
> normal part of the program flow:
> 
>    while line = gets
>       ....
>    end
> 
> You wouldn't write
> 
>    begin
>       loop {
>          line = gets
>          ....
>       }
>    rescue  EndOfFileError # Assume gets raises this exceptsion
>                           # when EOF is reached.
>    end
> 
> My current problem is similar to this case.  My class constructs an
> object depending on data in a file.  The datafile is in this form:
> 
>    non-data
>    ...
>    data begin
>       data
>       data
>       ....
>    data end
>    non-data
>    ...
>    data begin
>     ...
> 
> Only the class knows what constitutes data and which part of the file
> to ignore.  So, only by calling "new" of that class can we know
> whether the remaining part of the file contains enough data.  But, we
> also know that the file is bound to reach its end, so that handling
> that condition is truly a normal part of our program:
> 
>    while s = MyClass.new(file)
>       arr.push(s)
>    end
>    # use arr
> 
> The following
> 
>    begin
>       loop {
>          s = MyClass.new(file)
>          arr.push(s)
>       }
>    rescue  NoMoreData
>    end
> 
> doesn't make sense to me because I prepare to see the condition
> NoMoreData to occur (as end of file for IO#gets): it's not an error.
> 
> Cheers,
> Ryo