From: dblack@...
Date: 2003-08-24T14:37:21+09:00
Subject: Re: Wierd problem reading fields in a tab-delimitted row...

Hi --

On Sun, 24 Aug 2003, Kurt Euler wrote:

> Everyone--
>

> A wierd problem occurs in Ruby 1.67, 1.68 and 1.8 (and probably
> other versions?): If I step through rows in a tab-delimited text file
> using code like this:
>
> 	IO.foreach("wierd.txt") { |x|
> 		field = x.chop.split("\t")
> 		if field[4] != ""
> 			puts "some_string"
> 			puts field[4]
> 		else
> 			puts "no_string"
> 			puts field[4]
> 		end
> 		}
>
> to read each line of a text file (wierd.txt) as in this example:
>
> 	string1<tab>string2<tab>string3<tab><tab><tab><tab>string4<tab><tab>
> 	string4<tab>string5<tab>string6<tab><tab><tab><tab>string8<tab>
> 	string9<tab>string10<tab>string11<tab><tab><tab><tab>
> 	string12<tab>string13<tab>string14<tab><tab><tab><tab>string15
>
>
> I get this output:
>
> 	===================
> 	no_string
>
> 	no_string
>
> 	some_string
> 	nil
> 	no_string
> 	===================
>
> The question is, why does field[4] only get assigned nil in the 3rd
> line but some phantom non-"nil" value in the 1st, 2nd, and 4th
> lines? The general case I've found is that a nil value is read from
> a 'column' if there are no strings in any 'columns' to the right.

Technically, I think, it's that split stops adding values to the
array when it stop encountering anything other than delimiters on
the right:

  irb(main):002:0> "aaaba".split(/a/)
  => ["", "", "", "b"]

That last "a" in the string does not result in a final "" in the return
array.  And if the string has no non-delimiters:

  irb(main):003:0> "aaa".split(/a/)
  => []

Unless you give a negative second argument to split, which indicates
you want the empty strings returned:

  irb(main):001:0> "aaaba".split(/a/,-1)
  => ["", "", "", "b", ""]
  irb(main):002:0> "aaa".split(/a/,-1)
  => ["", "", "", ""]

So, in your third line, which ends with a bunch of tabs, you don't get
the empty strings in the return array.

(I think your test output might be clouding the results a little,
since "" gets called "no_string" while nil gets called "some_string"
:-) But anyway -- it's split's handling of delimiters on the right of
the string that's behind what's happening.)


David

-- 
David Alan Black
home: dblack@superlink.net
work: blackdav@shu.edu
Web:  http://pirate.shu.edu/~blackdav