From: Ben Giddings <bg-rubytalk@...>
Date: 2003-12-12T04:12:12+09:00
Subject: Re: expect, rexpect and others

John Carter wrote:
> Slightly unfair of me to comment since I wrote it....
> 
> But yes, I have used it to drive very many (100's) of real and virtual
> devices simultaneously. So yes, it is robust and does scale, which I
> found the standard one didn't.

Great!

> That's not quite the absolute latest version I have but should be good
> to try. (The API hasn't changed except I now through a TimeoutError
> instead of EOFError on timeout.)

Ok, a few questions then:


Is there some historical reason why the result of a match is "buffer", $1, 
$2, $3...?  To me, it would make more sense to do

matchdata = regexp.match(buffer)

return matchdata.to_a

I don't see why it is useful to return the whole buffer, and also this 
allows really complex regexps with more than 9 elements.

Also, the other efficiency thing that concerns me is that both versions 
attempt to match against the entire buffer.  The majority of regexps 
*could* match something huge, but are generally only used for something 
less than say 80 chars long.  I want to use an 'expect' type library 
against something that is updating a count of each byte of a 200k+ file 
written.  After that progress meter it prints "Done.", which is all I 
really want to match against.  It would be more efficient if there were an 
option to specify the maximum size of the buffer to use as a match, and to 
throw out parts of a buffer bigger than that.  Something like:

# Append the data to the match buffer, and shorten it if necessary
match_buff << read_buff
if match_buff.length > maxlength
   match_buff = match_buff[-maxlength .. -1]
end
debug("match buff is now <<#{match_buff}>>")

> RubyForge didn't exist when I submitted it, I guess I should create a
> RubyForge project for it. Except it is so small it is more appropriate
> to included it as part of some other project.

It really isn't all that small, I mean, look at most of the ruby libs, 
they're much smaller.  I think it's worth adding to the stdlib.

Ben