I have a text file with lot of rows with letter spacing, i.e.
cat test.txt
Some word here: T h e Q u i c k B r o w n F o x J u m p s O v e r T h e L a z y D o g
Some doggerel: J a c k A n d J i l l W e n t U p T h e H i l l
I ask some regular expression to apply to this text file to delete spacing between characters, with some command line tool in Linux.
cat result.txt
Some word here: The Quick Brown Fox Jumps Over The Lazy Dog
Some doggerel: Jack And Jill Went Up The Hill
Thank you
If what you want is what was divined by TrebuchetMS in this comment, it's not difficult using awk:
$ awk -F: '{gsub(/ /,"",$2); gsub(/[A-Z]/," &",$2) ; print $1":"$2}' file.txt
The one-line program ① split a line on :, ② erases all the spaces after the :, ③ puts a space in front of each capital letter (also in front of the first one) and ④ prints the concatenation of $1 (what precedes the :), a : and
$2, that is, the modified second part.
I mentioned in the comments that you could use sed for this. After trying it out, I lost hope in sed since I couldn't get lookarounds to work in their regexes. Apparently, the perl command can parse regexes with lookarounds. If you have the perl command, you can try this
perl -pe 's/ ([a-z])(?= |$)/\1/g' file.txt
or
cat file.txt | perl -pe 's/ ([a-z])(?= |$)/\1/g'
The perl option -e tells the perl command to accept a script (that's the monstrous regex you see right after it) and -p will loop the script around the file. (I'm no perl expert, so I need someone to double check this, I only looked at perl -h for help.)
Now the regex.
The s/<match>/<replace>/g follows sed's syntax. It'll search globally for the <match> and substitute it with the <replace>.
Here, the match was ([a-z])(?= |$), which tells perl to match places with spaces followed by a lower-case letter (([a-z]), where [a-z] denotes the set of characters to match, and () denotes a capture group, used in the <replace> section).
And to make sure that what follows is either a space or the end of the line ((?= |$)), that's the [positive] lookahead I was referring to before. The vertical bar implies "or". Thus, the lookahead will search for a space () "or" the end of the line ($). The lookahead ensures the correct match while not including the space/end in the match.
The replace was \1, which will replace the match with the first capture group. In this case, the capture group is whatever lower-case letter was matched.
If you look at the first line of your text file:
Some word here: T h e Q u i c k B r o w n F o x J u m p s O v e r T h e L a z y D o g
We only want to match lower-case letters, which have a space after them, i.e. a-z. If we only match a-z, that will include Some, word, and here. So we match lower-case letters, with spaces at the front and back. We remove the first space by matching it, by only replacing the letter, dropping the space.
If your file had
Lol a word here: T h e Q u i c k B r o w n F o x J u m p s O v e r T h e L a z y D o g
then the output would include:
Lola word here: The Quick Brown Fox Jumps Over The Lazy Dog
not as accurate as gboffi's answer in that it matches after the colon, but still regexes are a short hack ¯\_(ツ)_/¯.
Further Reading: Reference: What does this regex mean?
This might work for you (GNU sed):
sed -r ':a;s/^(.*: .*) ([[:lower:]])/\1\2/;ta' file
Replace all cases of a space followed by a lower-case character by a the lower-case character following a : in the current line. This solution works its way back along the line until it fails when all cases have been catered for.
With GNU awk for gensub():
$ awk 'BEGIN{FS=OFS=":"} {$2=gensub(/ ([^[:upper:]])/,"\\1","g",$2)}1' file
Some word here: The Quick Brown Fox Jumps Over The Lazy Dog
Some doggerel: Jack And Jill Went Up The Hill
with any awk:
$ awk 'BEGIN{FS=OFS=":"} {gsub(/ /,"",$2); gsub(/[[:upper:]]/," &",$2)}1' file
Some word here: The Quick Brown Fox Jumps Over The Lazy Dog
Some doggerel: Jack And Jill Went Up The Hill
Here is one more variant using Perl
$ cat peter.txt
Some word here: T h e Q u i c k B r o w n F o x J u m p s O v e r T h e L a z y D o g
Some doggerel: J a c k A n d J i l l W e n t U p T h e H i l l
$ perl -F":" -lane ' $F[1]=~s/ //g; $F[1]=~s/([A-Z])/ \1/g; print "$F[0]:$F[1]" ' peter.txt
Some word here: The Quick Brown Fox Jumps Over The Lazy Dog
Some doggerel: Jack And Jill Went Up The Hill
This problem can be solve by many different way. The easiest way I can think of is just remove blank before lower case. I have tried using SED as TrebuchetMS mentioned "SED didn't have lookarounds in their regexes"
echo "T h e Q u i c k B r o w n F o x J u m p s O v e r T h e L a z y D o g" | sed 's/[[:blank:]]\([[:lower:]]\)/\1/g'
Output: The Quick Brown Fox Jumps Over The Lazy Dog
