[SOLVED] Seeking 7-letter words containing 2 Us

danielbmartin · 08-08-2021, 09:12 PM

This topic is only a learning exercise.

Wanted: a list of English words which...
- are 7 letters long
- contain exactly 2 Us in any sequence

Examples: fulcrum and surplus are good
but cumulus and unusual are not.

My solution...

Code:

WordList='/usr/share/dict/words'
 sed -rn '/^.{7}$/p' $WordList  \
|sed -rn '/u.*u/p'              \
|sed -r  '/u.*u.*u/d'

This works but it seems clumsy.

Brainteaser: is there an elegant RegEx which could do this with only one sed?

Daniel B. Martin

.

astrogeek · 08-08-2021, 10:41 PM

This comes quickly to mind...

Code:

sed -nr '/^.{7}$/{/^[^u]*u[^u]*u[^u]*$/p}' /usr/share/dict/words

...but not very pretty - probably not a candidate for "elegant"!

danielbmartin · 08-09-2021, 09:50 AM

Quote:

Originally Posted by astrogeek

Code:

sed -nr '/^.{7}$/{/^[^u]*u[^u]*u[^u]*$/p}' /usr/share/dict/words

...but not very pretty - probably not a candidate for "elegant"!

Ding ding ding ding ding! We have a winner!

Pretty? No.
Elegant? Not really.
Excellent? Yes!

A good definition of Technical Excellence:
Completeness of function coupled with economy of means.

Thank you for the education!

Daniel B. Martin

.

astrogeek · 08-09-2021, 01:35 PM

Thank U2!

grail · 08-10-2021, 09:34 AM

How about:

Code:

awk 'length == 7 && split($0,_,"u") > 2' /usr/share/dict/words
#OR
awk -Fu 'length == 7 && NF > 2' /usr/share/dict/words

danielbmartin · 08-10-2021, 10:10 AM

Quote:

Originally Posted by grail

Code:

awk 'length == 7 && split($0,_,"u") > 2' /usr/share/dict/words

I was looking for a sed solution but an awk solution is certainly welcome.

With this WordList ...

Code:

nothing
fulcrum
surplus
cumulus
unusual
abcdefg
ubudefg
abcuefu
ubudufg
abuuefu
abuuufg
uuudefg
uuuuefg
ubudufu
abcduuu
ubudufu
ubudef
ubudefgh

... this awk ...

Code:

awk 'length == 7 && split($0,_,"u") > 2' $WordList

... produced this result ...

Code:

fulcrum
surplus
cumulus
unusual
ubudefg
abcuefu
ubudufg
abuuefu
abuuufg
uuudefg
uuuuefg
ubudufu
abcduuu
ubudufu

It fails the criterion of finding words with exactly two Us.

The second awk has the same shortcoming.

Daniel B. Martin

.

grail · 08-10-2021, 10:16 AM

Quote:

It fails the criterion of finding words with exactly two Us.

That is my bad for the misread, obviously just change the sign from > to ==

danielbmartin · 08-10-2021, 01:56 PM

Quote:

Originally Posted by grail

That is my bad for the misread, obviously just change the sign from > to ==

This didn't produce the desired result ...

Code:

awk 'length == 7 && split($0,_,"u") == 2' $WordList

... but this did.

Code:

awk 'length == 7 && split($0,_,"u") == 3' $WordList

Daniel B. Martin

.

danielbmartin · 08-11-2021, 05:38 AM

grail's awk solution has an advantage: scalability. Suppose some industrial application dealt with electronic signals instead of English words. The problem might be to identify lines of length 112 with exactly 27 "U"s. The sed solution would be unwieldy but the awk solution would have two numbers change.

Daniel B. Martin

.

shruggy · 08-11-2021, 05:49 AM

There are grep-like tools that allow either matching against several patterns at once (ugrep, faster than awk, but slower than sed)

Code:

ug -x% '.{7} [^u]*u[^u]*u[^u]*' $WordList

or constraining the match to selected regions via an additional regex (greple, slow)

Code:

greple --inside '^.{7}$' --strict '^[^u]*u[^u]*u[^u]*$' $WordList

pan64 · 08-11-2021, 05:59 AM

I would suggest a different approach (although it is already solved).
You can always [try to] invert the requirement[s], sometimes that can be implemented easier (and you do not need to pipe sed[s] into each other).
I mean something like this:

Code:

sed -rn '/^.{7}$/!d;
         /u.*u.*u/d;
         /u.*u/p'

shruggy · 08-11-2021, 06:15 AM

^ -n is unnecessary in the latter case

Code:

sed '/^.\{7\}$/!d;/u.*u.*u/d;/u.*u/!d'

syg00 · 08-11-2021, 06:32 AM

Readability - and the aforementioned scalability - goes a long way to nailing awk as the solution. Its syntax is becoming accepted elsewhere as well - bpftrace for example. Worse ways to spend your time than becoming proficient in it. Thanks grail for getting me interested years ago.

rnturn · 08-31-2021, 12:17 PM

Quote:

Originally Posted by danielbmartin

This topic is only a learning exercise.

Wanted: a list of English words which...
- are 7 letters long
- contain exactly 2 Us in any sequence

Examples: fulcrum and surplus are good
but cumulus and unusual are not.

With a little work, this could be a helpful tool to help solve those word jumble puzzles in the Sunday paper's Books section. :^)

danielbmartin · 08-31-2021, 01:49 PM

Quote:

Originally Posted by rnturn

With a little work, this could be a helpful tool to help solve those word jumble puzzles in the Sunday paper's Books section. :^)

Quite right! The puzzles in the Sunday NYTimes magazine section sparked my interest in this subject.

Daniel B. Martin

.