[SOLVED] Seeking an improved RegEx

danielbmartin · 08-21-2016, 12:10 PM

I have a file of English words, one word on each line.

As a learning exercise I coded this awk ...

Code:

echo; echo "Find four-letter words containing V, using AWK."
awk -F "" '(NF==4 && index($0,"v"))' $CleanWords >$OutFile

... and it works.

I coded this grep couplet ...

Code:

echo; echo "Find four-letter words containing V, using GREP."
grep "^....$" $CleanWords |grep "v" >$OutFile

... and it works but maybe it can be done in one line.

I coded this grep ...

Code:

echo; echo "Find four-letter words containing V, using GREP."
egrep '(^v...$)|(^.v..$)|(^..v.$)|(^...v$)' $CleanWords >$OutFile

... and it works but is an unappealing brute-force solution.

Is there a clever RegEx to handle this task?

Daniel B. Martin

astrogeek · 08-21-2016, 12:53 PM

This sed comes quickly to mind.

Code:

sed -n /^....$/s/v/v/p $CleanWords

You might change the address to /^.{4}$/ but it is the same number of characters and harder to type than ...., and you will need to escape the curly braces I think.

grail · 08-21-2016, 01:14 PM

The awk can be a little shorter:

Code:

awk -F '' 'NF == 4 && /v/' file

Your last grep can be shorter just by omission of repeating terms:

Code:

grep -E '^(v...|.v..|..v.|...v)$' file

danielbmartin · 08-21-2016, 01:53 PM

Quote:

Originally Posted by astrogeek

You might change the address to /^.{4}$/ but it is the same number of characters and harder to type than ...., and you will need to escape the curly braces I think.

Your suggestion works nicely ...

Code:

echo "Find four-letter words containing V, using SED with curly braces."
echo "Method of LQ Senior Member astrogeek."
sed -n '/^.\{4\}$/s/v/v/p' $CleanWords >$OutFile

... and lends itself to being a more general solution such as 9-character words containing "w."

Nitpick: the sed solution makes a v-for-v substitution which is not needed for the awk or grep solutions.

Counter-argument to the Nitpick: depending on the overall purpose of the code it might be desirable to make that "v" stand out by coding it this way ...

Code:

echo "Find four-letter words containing V, using SED with curly braces."
echo "Method of LQ Senior Member astrogeek."
sed -n '/^.\{4\}$/s/v/V/p' $CleanWords >$OutFile

Daniel B. Martin

grail · 08-21-2016, 02:17 PM

Without the substitution the sed does not work ... remember, it is how the tool functions that is the key and not how it compares to another commands process.

You can add -r to the sed switches and the braces should not need to be escaped.

keefaz · 08-21-2016, 02:40 PM

With grep -P

Code:

grep -P '(?=^....$).*v' $CleanWords

astrogeek · 08-21-2016, 03:29 PM

Quote:

Originally Posted by danielbmartin

... and lends itself to being a more general solution such as 9-character words containing "w."

Nitpick: the sed solution makes a v-for-v substitution which is not needed for the awk or grep solutions.

Counter-argument to the Nitpick: depending on the overall purpose of the code it might be desirable to make that "v" stand out...

I had played with it for words of various length, but the question was for four letters, so I let the moment pass...

With s/v/&/p it makes one less edit for substituting alternate characters as well.

But I had not thought of the caps replacement, nice touch!

I also seem to never remember -r for sed, it just doesn't stick in my remaining brain cell - thanks grail!

ntubski · 08-21-2016, 03:46 PM

Quote:

Originally Posted by grail

Without the substitution the sed does not work ... remember, it is how the tool functions that is the key and not how it compares to another commands process.

Actually, sed can use the same approach as awk here:

Code:

sed -n '/^....$/{/v/p}' input

danielbmartin · 08-21-2016, 08:37 PM

I am impressed by the ingenuity of the various solutions and improvements-on-solutions offered by all the contributors to this thread. Thanks to astrogeek, grail, keefaz, ntubski. Kudos and reps all around. This thread is marked SOLVED!

Daniel B. Martin

grail · 08-22-2016, 04:50 AM

Quote:

Originally Posted by ntubski

Actually, sed can use the same approach as awk here:

Code:

sed -n '/^....$/{/v/p}' input

I did think of that later

Good pick up.

pan64 · 08-22-2016, 07:12 AM

hm. really nice. But using awk the "natural" way would be:

Code:

awk ' length == 4 && /v/ ' input

# and obviously perl can be used too

Code:

perl -ne '/(?=^....$).*v/ && print' input