Re: [Patch] FTP.new with block

On Sun, Jul 20, 2003 at 06:51:03PM +0900, ts wrote:
> >>>>> "R" == Richard Zidlicky <[e-mail]> writes:
> 
> R> something more systematic is needed. So far only 2 widely used 
> R> classes appear to evaluate blocks passed to new: Proc and Tk.
> 
> svg% ruby -e 'p Array.new(2) { 12 }'
> [12, 12]
> svg% 
> 
> svg% ruby -e 'a = Hash.new { 12 }; p a[1]'
> 12

I am still on ruby 1.6.7 so I missed that:

$ ruby -e 'p Array.new(2) { 12 }'
[nil, nil]


> R> Furthermore Tk is somehwat confusing for the beginner because
> R> it uses instance_eval or something completely different to eval
> R> the block.
> 
>  Why a method must eval the block when it's called ? See Hash::new

it would be less confusing for the beginner if the syntax could be
disambiguated somehow. Clearly you won't write

  Hash.new begin 12 end
so why should
  Hash.new { 12 }
work?

> R> How to get it consistent? Leave the 2 widely used exceptions
> R> to the rule (no block arg for new) some time to phase out and
> R> use either a clear naming convention (eg new_wb instead of new)
> 
>  and change the name of *all* methods to have _wb when they use a block ?

The top priority imho is to make ruby at least force an error when 
a block expecting method doesn岐 get a block or some method ignores
a given block. Example in 1.6.7:

  TCPSocket.new(host,43{|sock|
      sock.each{|str| print str}}

it is rather irritating if the block is plainly ignored.
Another example, 

   http=Net::HTTP.Proxy($proxy_addr,$proxy_port).start(host, port)

simply returns nil as it expects to yield a block but doesn't
complain if there isn't any.


Perhaps it would be possible to use a convention to append eg "&"
to all method names that expect a block? 

The idea with "_wb" was really only a last resort workaround for me..

Richard