Remove hard line breaks from text with Ruby - ruby-on-rails

Question

I have some text with hard line breaks in it like this:
This should all be on one line
since it's one sentence.
This is a new paragraph that
should be separate.
I want to remove the single newlines but keep the double newlines so it looks like this:
This should all be on one line since it's one sentence.
This is a new paragraph that should be separate.
Is there a single regular expression to do this? (or some easy way)
So far this is my only solution which works but feels hackish.
txt = txt.gsub(/(\r\n|\n|\r)/,'[[[NEWLINE]]]')
txt = txt.gsub('[[[NEWLINE]]][[[NEWLINE]]]', "\n\n")
txt = txt.gsub('[[[NEWLINE]]]', " ")

Answer 1

Replace all newlines that are not followed by or preceded by a newline:
text = <<END
This should all be on one line
since it's one sentence.
This is a new paragraph that
should be separate.
END
p text.gsub /(?<!\n)\n(?!\n)/, ' '
#=> "This should all be on one line since it's one sentence.\n\nThis is a new paragraph that should be separate. "
Or, for Ruby 1.8 without lookarounds:
txt.gsub! /([^\n])\n([^\n])/, '\1 \2'

Answer 2

text.gsub!(/(\S)[^\S\n]*\n[^\S\n]*(\S)/, '\1 \2')
The two (\S) groups serve the same purposes as the lookarounds ((?<!\s)(?<!^) and(?!\s)(?!$)) in #sln's regexes:
they confirm that the linefeed really is in the middle of a sentence, and
they ensure that the [^\S\n]*\n[^\S\n]* part consumes any other whitespace surrounding the linefeed, making it possible for us to normalize it to a single space.
They also make the regex easier to read, and (perhaps most importantly) they work in pre-1.9 versions of Ruby that don't support lookbehinds.

Answer 3

There is more to formatting (turning off word wrap) than you think.
If the output is a result of a formatting operation, then you should go by
those rules to reverse engineer the original.
For instance, the test you have there is
This should all be on one line
since it's one sentence.
This is a new paragraph that
should be separate.
If you removed just the single newlines only, it would look like this:
This should all be on one line since it's one sentence.
This is a new paragraph thatshould be separate.
Also, other formatting such as intentional newlines will be lost, so something like:
This is Chapter 1
Section a
Section b
Turns into
This is Chapter 1 Section a Section b
Finding the newline in question is easy /(?<!\n)\n(?!\n)/
but, what do you replace it with.
Edit: Actually, its not that easy even to find standalone newlines, because visually they sit amongst hidden from view (horizontal) whitespaces.
There are 4 ways to go.
Remove newline, keep the surrounding formatting
$text =~ s/(?<!\s)([^\S\n]*)\n([^\S\n]*)(?!\s)/$1$2/g;
Remove newline and formatting, substitute a space
$text =~ s/(?<!\s)[^\S\n]*\n[^\S\n]*(?!\s)/ /g;
Same as above but ignore newline at beginning or end of string
$text =~ s/(?<!\s)(?<!^)[^\S\n]*\n[^\S\n]*(?!$|\s)/ /g;
$text =~ s/(?<!\s)(?<!^)([^\S\n]*)\n([^\S\n]*)(?!$|\s)/$1$2/g;
Example breakdown of regex (this is the minimum required just to isolate a single newline):
(?<!\s) # Not a whitespace behind us (text,number,punct, etc..)
[^\S\n]* # 0 or more whitespaces, but no newlines
\n # a newline we want to remove
[^\S\n]* # 0 or more whitespaces, but no newlines
(?!\s)/ # Not a whitespace in front of us (text,number,punct, etc..)

Answer 4

Well, there is this:
s.gsub /([^\n])\n([^\n])/, '\1 \2'
It won't do anything to leading or trailing newlines. If you don't need leading or trailing white space at all, then you will win with this variation:
s.gsub(/([^\n])\n([^\n])/, '\1 \2').strip

Answer 5

$ ruby -00 -pne 'BEGIN{$\="\n\n"};$_.gsub!(/\n+/,"\0")' file
This should all be on one line since it's one sentence.
This is a new paragraph thatshould be separate.