Behavior of substitutions with sub

ruby-talk
1

Hello,
I have seen a somehow strange behavior with Strings and the sub-method.
Maybe you cen give me an idea if it is a bug and where it comes from.
It have seen the problem when I wanted to make an substition with a
sub(/<regexp>/,"") instead of (/<regexp/,'').

To get the idea a example in irb. I tried it also with version 1.8.5 but
have seen it first with 1.8.4:

···

====================================================
irb(main):012:0> tst = "Das war ja nix"
=> "Das war ja nix"
irb(main):013:0> tst2 =tst
=> "Das war ja nix"
irb(main):014:0> tst2.sub(/nix/,"was")
=> "Das war ja was"
irb(main):015:0> tst2.sub(/(n)+ix/,"$1")
=> "Das war ja $1"
irb(main):016:0> tst2.sub(/(n)+ix/,"#{$1}")
=> "Das war ja n"
irb(main):017:0> tst2.sub(/(n)+(i)+(x)+/,"#{$3}#{$2}#{$1}")
=> "Das war ja n"
irb(main):018:0> tst2.sub(/(n)+(i)+(x)+/,'\3\2\1')
=> "Das war ja xin"
irb(main):019:0> tst2.sub(/(n)+(i)+(x)+/,"#{$3}#{$2}#{$1}")
=> "Das war ja xin"

==================================================================

What I do not understand is, that the substitution in the ''-expression with
\1, \2 and \3 work without trouble but when I use the normal ""-String and
the #{$1}, #{$2} and #{$3}-notation he seems alway to refer to the last
regular expression and not the one in sub-method.
It is only an example. In the script I have seen it I cicrcumvent the
problem by adding an extra regular expression before without using the
result. But I think it is a strange behavior.

Any comments about that.

Regards,
Tammo

2

Tammo Tjarks wrote:

Hello,
I have seen a somehow strange behavior with Strings and the sub-method.
Maybe you cen give me an idea if it is a bug and where it comes from. It have seen the problem when I wanted to make an substition with a sub(/<regexp>/,"") instead of (/<regexp/,'').

To get the idea a example in irb. I tried it also with version 1.8.5 but
have seen it first with 1.8.4:

====================================================
irb(main):012:0> tst = "Das war ja nix"
=> "Das war ja nix"
irb(main):013:0> tst2 =tst
=> "Das war ja nix"
irb(main):014:0> tst2.sub(/nix/,"was")
=> "Das war ja was"
irb(main):015:0> tst2.sub(/(n)+ix/,"$1")
=> "Das war ja $1"
irb(main):016:0> tst2.sub(/(n)+ix/,"#{$1}")
=> "Das war ja n"
irb(main):017:0> tst2.sub(/(n)+(i)+(x)+/,"#{$3}#{$2}#{$1}")
=> "Das war ja n"
irb(main):018:0> tst2.sub(/(n)+(i)+(x)+/,'\3\2\1')
=> "Das war ja xin"
irb(main):019:0> tst2.sub(/(n)+(i)+(x)+/,"#{$3}#{$2}#{$1}")
=> "Das war ja xin"

==================================================================

What I do not understand is, that the substitution in the ''-expression with
\1, \2 and \3 work without trouble but when I use the normal ""-String and
the #{$1}, #{$2} and #{$3}-notation he seems alway to refer to the last
regular expression and not the one in sub-method.

That's easily explained: the string you hand off to sub is interpreted before the call happens. But the grouping variables are only set after that, i.e. when sub executes.

The version with \1 etc. works, because method sub does the interpretation of the string and replaces \1 with contents of the first group etc:

irb(main):005:0> puts( "\\1" )
\1
=> nil
irb(main):006:0> puts( "#{$1}" )

=> nil

It is only an example. In the script I have seen it I cicrcumvent the
problem by adding an extra regular expression before without using the
result. But I think it is a strange behavior.

Not at all once you understand how this works. It is the general mechanism of method arguments being evaluated before the method call starts. It's basically the same as:

irb(main):001:0> def foo(x) puts x end
=> nil
irb(main):002:0> foo( 1 + 2 )
3
=> nil

Kind regards

  robert

3

Thanks for you answer. That makes sense.

Regards,
Tammo

Robert Klemme wrote:

···

Tammo Tjarks wrote:

Hello,
I have seen a somehow strange behavior with Strings and the sub-method.
Maybe you cen give me an idea if it is a bug and where it comes from.
It have seen the problem when I wanted to make an substition with a
sub(/<regexp>/,"") instead of (/<regexp/,'').

To get the idea a example in irb. I tried it also with version 1.8.5 but
have seen it first with 1.8.4:

====================================================
irb(main):012:0> tst = "Das war ja nix"
=> "Das war ja nix"
irb(main):013:0> tst2 =tst
=> "Das war ja nix"
irb(main):014:0> tst2.sub(/nix/,"was")
=> "Das war ja was"
irb(main):015:0> tst2.sub(/(n)+ix/,"$1")
=> "Das war ja $1"
irb(main):016:0> tst2.sub(/(n)+ix/,"#{$1}")
=> "Das war ja n"
irb(main):017:0> tst2.sub(/(n)+(i)+(x)+/,"#{$3}#{$2}#{$1}")
=> "Das war ja n"
irb(main):018:0> tst2.sub(/(n)+(i)+(x)+/,'\3\2\1')
=> "Das war ja xin"
irb(main):019:0> tst2.sub(/(n)+(i)+(x)+/,"#{$3}#{$2}#{$1}")
=> "Das war ja xin"

==================================================================

What I do not understand is, that the substitution in the ''-expression
with \1, \2 and \3 work without trouble but when I use the normal
""-String and the #{$1}, #{$2} and #{$3}-notation he seems alway to refer
to the last regular expression and not the one in sub-method.

That's easily explained: the string you hand off to sub is interpreted
before the call happens. But the grouping variables are only set after
that, i.e. when sub executes.

The version with \1 etc. works, because method sub does the
interpretation of the string and replaces \1 with contents of the first
group etc:

irb(main):005:0> puts( "\\1" )
\1
=> nil
irb(main):006:0> puts( "#{$1}" )

=> nil

It is only an example. In the script I have seen it I cicrcumvent the
problem by adding an extra regular expression before without using the
result. But I think it is a strange behavior.

Not at all once you understand how this works. It is the general
mechanism of method arguments being evaluated before the method call
starts. It's basically the same as:

irb(main):001:0> def foo(x) puts x end
=> nil
irb(main):002:0> foo( 1 + 2 )
3
=> nil

Kind regards

robert