Weird behaviour escaping special characters in a string

This instance method added to the String class returns a copy of the
receiver with occurrences of \ replaced with \, and occurrences of ’
replaced with ':

class String
def to_source_string
gsub(/(\|’)/, ‘\\\1’)
end
end

The idea is that it will give you a string that you can write out a
Ruby file that will later print the string. For, example, let’s take
the string, foo (3 characters):

“puts '” + “foo”.to_source_string + “’” # puts ‘foo’

Or a string with special characters in it like ‘foo’ (5 characters,
including enclosing single quotes):

“puts '” + “‘foo’”.to_source_string + “’” # puts ‘‘foo’’

My RSpec specs and experimentation in irb confirm that the method
works but I am at a loss to explain one thing:

Why do I need so many backslashes in my replacement expression?

There are five slashes in the replacement expression:

gsub(/(\\|')/, '\\\\\1')

But I would have thought that three would work:

gsub(/(\\|')/, '\\\1')

I basically want to replace “whatever is found in the pattern” with a
backslash (\) followed by “whatever was found” (\1); so that’s three
slashes. But with only three slashes Ruby gives me \1foo\1 instead of
‘foo’. Four slashes produces the same result. Five slashes and
suddenly everything works (funnily enough, six slashes also works).
Two slashes and one slash have no effect (no escaping is performed).

I’ve got working code so it’s not a huge problem, but my curiosity is
piqued. What’s going on here that I don’t understand?

Cheers,
Greg

On 2/21/07, Greg H. [email protected] wrote:

This instance method added to the String class returns a copy of the
receiver with occurrences of \ replaced with \, and occurrences of ’
replaced with ':

class String
def to_source_string
gsub(/(\|')/, ‘\\\1’)
end
end

class String
def to_source_string
gsub(/(\|')/) { “\#$1” }
end
end

-austin

On Feb 21, 2007, at 12:36 PM, Austin Z. wrote:

class String
def to_source_string
gsub(/(\|’)/) { “\#$1” }
end
end

It’s probably better to use a character class [\’] instead of
alternation (\|’).

James Edward G. II

On Thu, Feb 22, 2007 at 02:55:09AM +0900, Greg H. wrote:

Why do I need so many backslashes in my replacement expression?

There are five slashes in the replacement expression:

gsub(/(\\|')/, '\\\\\1')

But I would have thought that three would work:

gsub(/(\\|')/, '\\\1')

Because even in single quotes, blackslashes must be doubled; this in
turn is
because ’ is the way that you insert a single quote within a
single-quoted
string.

irb(main):001:0> a=’\’
=> “\”
irb(main):002:0> a.size
=> 1
irb(main):003:0> b=’’’
=> “’”
irb(main):004:0> b.size
=> 1
irb(main):005:0> c=’\x’
=> “\x”
irb(main):006:0> c.size
=> 2

I basically want to replace “whatever is found in the pattern” with a
backslash (\) followed by “whatever was found” (\1); so that’s three
slashes. But with only three slashes Ruby gives me \1foo\1 instead of
‘foo’. Four slashes produces the same result. Five slashes and
suddenly everything works (funnily enough, six slashes also works).
Two slashes and one slash have no effect (no escaping is performed).

I’ve got working code so it’s not a huge problem, but my curiosity is
piqued. What’s going on here that I don’t understand?

irb(main):009:0> a=’\\1’
=> “\\1”
irb(main):010:0> a.size
=> 3
irb(main):011:0> a=’\\\1’
=> “\\\1”
irb(main):012:0> a.size
=> 4
irb(main):013:0> a=’\\\1’
=> “\\\1”
irb(main):014:0> a.size
=> 4

In a single-quoted string:
’ => ’
\ =>
\x => \x for all other x

So ‘…\1’ and ‘…\1’ are identical.

HTH,

Brian.

On 21 feb, 19:45, James Edward G. II [email protected]
wrote:

On Feb 21, 2007, at 12:36 PM, Austin Z. wrote:

It’s probably better to use a character class [\‘] instead of
alternation (\|’).

James Edward G. II

I did some quick and dirty benchmarks and using a character class is a
little bit quicker. Interpolation (“\#$1”) is slower but more
readable. I guess I’ll stick with the character class and no
interpolation though.

require ‘benchmark’
include Benchmark

bm(6) do |x|
x.report(‘alternation’) { 100_000.times { “‘foo’”.gsub(/(\|‘)/, ‘\
\1’) } }
x.report(‘char class’) { 100_000.times { “‘foo’”.gsub(/[\’]/, ‘\\
&’) } }
x.report(‘interpolation’) { 100_000.times { “‘foo’”.gsub(/(\|‘)/, "
#$1") } }
x.report(‘interpolation with char class’) { 100_000.times
{ “‘foo’”.gsub(/[\’]/, “\#$&”) } }
end
user system total real
alternation 0.450000 0.000000 0.450000 ( 0.452661)
char class 0.390000 0.000000 0.390000 ( 0.396193)
interpolation 0.540000 0.010000 0.550000 ( 0.532106)
interpolation with char class 0.480000 0.000000 0.480000
( 0.485922)

On Thu, 22 Feb 2007 13:55:06 +0100, Greg H.
[email protected]
wrote:

4 slashes, and the same for 5 and 6 slashes.

%q{…} is your friend.

David V.

On 21 feb, 20:50, Brian C. [email protected] wrote:

In a single-quoted string:
' => ’
\ =>
\x => \x for all other x

So ‘…\1’ and ‘…\1’ are identical.

Excellent, that explains why I was getting the same results for 3 and
4 slashes, and the same for 5 and 6 slashes.

Cheers,
Greg

2007/2/21, Greg H. [email protected]:

The idea is that it will give you a string that you can write out a
Ruby file that will later print the string. For, example, let’s take
the string, foo (3 characters):

“puts '” + “foo”.to_source_string + “'” # puts ‘foo’

Or a string with special characters in it like ‘foo’ (5 characters,
including enclosing single quotes):

“puts '” + “‘foo’”.to_source_string + “'” # puts ‘'foo'’

Why don’t you just use #inspect?

Kind regards

robert