Simple one-liners: last line missing

ruby-talk
1

Hi

Let's say I'd want to modify each line of a file. Here's the dummy
data:

  $ cat data.txt
  first foo
  second foo
  third foo
  $

This works:

  $ ruby -ne '$_ =~ /(\S+)\s+(\S+)/;puts "#{$2} #{$1}"' data.txt
  foo first
  foo second
  foo third
  $

I don't really need to achieve that modification. Instead I'd like to
understand why the following version one doesn't work:

  $ ruby -pe 'sub(/(\S+)\s+(\S+)/,"#{$2} #{$1}")' data.txt

  foo first
  foo second
  $

TIA,
Tobi

2

  $ ruby -pe 'sub(/(\S+)\s+(\S+)/,"#{$2} #{$1}")' data.txt

The replacement string is build when #sub is called

for the first call $1 = $2 = nil ==> the replacement string is ' '

for the second call, $1 = 'first', $2 = 'foo' ==> the replacement string
is 'foo first'

etc,

Guy Decoux

3

[snip]

I don't really need to achieve that modification. Instead I'd like to
understand why the following version one doesn't work:

  $ ruby -pe 'sub(/(\S+)\s+(\S+)/,"#{$2} #{$1}")' data.txt

  foo first
  foo second
  $

When doing substitutions with a string replacement, you do not get what
you might expect because the interpolation of #{$1} in the replacement
string happens when the arguments are evaluated and passed into the sub()
or gsub() methods, not during the match. That means they will have whatever
value they had from the last successful match (or nil, as in your case with
the first line printed).

To access backreferences within a replacement string, you may use the same
notation as you do for backreferences within the pattern itself: \1, \2,
etc. For your example:

  ruby -pe 'sub(/(\S+)\s+(\S+)/,%q(\2 \1))' data.txt

Alternatively, you may use the block form of replacement, in which case the
block is evaluated at match-time and $1, $2, etc are available and refer to
their respective backreferences within the current match:

  ruby -pe 'sub(/(\S+)\s+(\S+)/){"#{$2} #{$1}"}' data.txt

cheer's
andrew

···

On Tue, 14 Mar 2006 01:46:50 +0900, Tobi Reif <tobiasreif@pinkjuice.com> wrote:

--
Andrew L. Johnson http://www.siaris.net/
      Reality is that which, when you stop believing in
      it, doesn't go away.
          -- Philip K. Dick

4

I think you want,

>ruby -pe 'sub(/(\S+)\s+(\S+)/,"\\2 \\1")' data.txt

Tobi Reif wrote:

···

Hi

Let's say I'd want to modify each line of a file. Here's the dummy
data:

  $ cat data.txt
  first foo
  second foo
  third foo
  $

This works:

  $ ruby -ne '$_ =~ /(\S+)\s+(\S+)/;puts "#{$2} #{$1}"' data.txt
  foo first
  foo second
  foo third
  $

I don't really need to achieve that modification. Instead I'd like to
understand why the following version one doesn't work:

  $ ruby -pe 'sub(/(\S+)\s+(\S+)/,"#{$2} #{$1}")' data.txt

  foo first
  foo second
  $

TIA,
Tobi

5

Thanks. I guess my (current) opinion is that it should work.

Tobi

···

On Tue 2006-03-14 ts wrote:

> $ ruby -pe 'sub(/(\S+)\s+(\S+)/,"#{$2} #{$1}")' data.txt

The replacement string is build when #sub is called

for the first call $1 = $2 = nil ==> the replacement string is ' '

for the second call, $1 = 'first', $2 = 'foo' ==> the replacement string
is 'foo first'

etc,

6

Now I understand that it can't work.

Thanks again,
Tobi

···

On Tue 2006-03-14 Tobi Reif wrote:

On Tue 2006-03-14 ts wrote:

>
> > $ ruby -pe 'sub(/(\S+)\s+(\S+)/,"#{$2} #{$1}")' data.txt
>
> The replacement string is build when #sub is called
>
> for the first call $1 = $2 = nil ==> the replacement string is ' '
>
> for the second call, $1 = 'first', $2 = 'foo' ==> the replacement string
> is 'foo first'
>
> etc,

Thanks. I guess my (current) opinion is that it should work.

Tobi