From: Phlip <phlip2005@...>
Date: 2009-03-16T16:32:38+09:00
Subject: Re: `&' interpreted as argument prefix

Arun Kumar wrote:
> Hi,
> I'm new to rubi and at present i'm trying to write a sample code for
> html extraction and store it in the database. The following is the code
> for that :
> 
> doc = Hpricot(open("http://www.google.com/"))

The offending line is:

       ele.instance_variable_set("@#{k}", v)

Apparently Google pages contain very complex HTML, to relieve strain on their 
servers. Then, Hpricot does not "sanitize" its input. That k variable might 
contain a &, which Ruby then warns about. instance_variable_set() creates an 
instance variable, like this:

    @foo = v

where 'foo' was in k. But if k contains '&foo', you get this:

    @&foo = v

You can't write that in raw Ruby, so instance_variable_set() is warning you that 
you should not write it in "meta-programming" Ruby either.

But none of this is your fault: It's a bug in Hpricot, which Google's advanced 
HTML uncovered.

The conclusion: Switch to Nokogiri. It has an Hpricot compatibility mode, but 
its internal engine is libxml, which is one of the industry's leading XML (and 
therefor HTML) implementations.

-- 
   Phlip