[SOLVED] Replace a string of blanks with a "tailored" string of equal length

danielbmartin · 10-15-2015, 02:42 PM

It is desired to replace strings of 3 or more blanks with "tailored" strings of equal length, such that the replacement string has one leading blank, one trailing blank, and all the interior characters are "%".

InFile ...

Code:

cat dog
red   blue  green
   apple     pear
 honda toyota         nissan

Desired OutFile...

Code:

cat dog
red % blue  green
 % apple %%% pear
 honda toyota %%%%%%% nissan

A sed or awk solution is preferred.

Daniel B. Martin

danielbmartin · 10-15-2015, 05:53 PM

I hammered out a sed solution. It looks clumsy (3 steps) but works.

Step 1) Replace all blanks with %.
Step 2) Replace the leading character in a string of %s with a blank.
Step 3) Replace the trailing character in a string of %s with a blank.

With this InFile ...

Code:

cat dog
red   blue  green
   apple     pear
 honda toyota         nissan

... this sed ...

Code:

sed 's/ /%/g;s/\([^%]\)%%/\1 %/g;s/%\([^%]\)/ \1/g' $InFile >$OutFile

... produced this OutFile ...

Code:

cat dog
red % blue  green
%% apple %%% pear
 honda toyota %%%%%%% nissan

Shorter and cleaner solutions are welcomed!

Daniel B. Martin

millgates · 10-15-2015, 07:12 PM

Isn't your OutFile missing a space at the beginning of the third line?

Anyway, my first attempt would be

Code:

sed 's/\B \B//g'

In this case, the initial whitespace is also missing. That can be fixed with an aditional substitution:

Code:

sed 's/\B \B/%/g;s/^%/ /'

And a (not very nice) gawk solution:

Code:

awk -F'(^ | \\<)|(\\> | $)' '{for (i=1;i<NF;i++)gsub(/ /,"%",$i);print}'

danielbmartin · 10-15-2015, 09:50 PM

Quote:

Originally Posted by millgates

Isn't your OutFile missing a space at the beginning of the third line?

Yes. My proposed solution has a flaw.

Code:

sed 's/\B \B//g'

This plugs in no %s at all. Presumably you meant ...

Code:

sed 's/\B \B/%/g' $InFile >$OutFile

This sed is lovely!

Code:

sed 's/\B \B/%/g;s/^%/ /'

After making the original post I devised a debugging aid. It prints one line from InFile, the corresponding line from OutFile, and then a blank line, repeat, repeat.

Code:

paste -d "\n" $InFile $OutFile /dev/null

With this InFile ...

Code:

cat dog
red   blue  green
   apple     pear
 honda toyota         nissan

... this code ...

Code:

echo; echo "Method #2 of LQ Member millgates."
sed 's/\B \B/%/g;s/^%/ /' $InFile >$OutFile

# This PASTE interleaves three files.
# The third file (/dev/null) contains empty lines.
echo
paste -d "\n" $InFile $OutFile /dev/null

... produces this display ...

Code:

Method #2 of LQ Member millgates.

cat dog
cat dog

red   blue  green
red % blue  green

   apple     pear
 % apple %%% pear

 honda toyota         nissan
 honda toyota %%%%%%% nissan

Looks good! Thanks!

Daniel B. Martin

grail · 10-16-2015, 06:41 AM

For the awk I would probably follow the same solution as millgates has for sed:

Code:

awk 'gsub(/\B \B/,"") && sub(/^%/," ") || 1'

I did knock up a ruby solution, just to do it a different way from the regex solution:

Code:

ruby -pe '$_.chomp.scan(/ {3,}/){|n| $_.sub!(n, " #{n[1..-2].gsub(/ /,"%")} ")}'

danielbmartin · 10-16-2015, 07:30 AM

[QUOTE=grail;5435496]

Code:

awk 'gsub(/\B \B/,"") && sub(/^/," ") || 1'

This didn't work until I remembered something from a recent unrelated post. Single percent signs sometimes vanish from posts. That explains why a sed solution posted by millgates seemed to fail, and also why your awk seemed to fail. Thanks to both of you for constructive and instructive solutions.

Daniel B. Martin

grail · 10-16-2015, 11:07 AM

Yes sorry about that, I will remember one day