grep use -v on one regexp and disable for another

Question

i want to grep for lines containing a OR lines not containing l

example text file:

ant
beetle
cow
deer
eel
flamingo
giraffe

desired output:

ant
cow
deer
flamingo
giraffe

it would be nice to use something like grep -e 'a' -ev 'l' but this doesnt work. is this a limitation of grep? either a grep solution (preferred) or other POSIX BRE (basic regular expression) processing tool is what i'm looking for, so something like awk '/a/ || !/l/' does not fit my use case

Answer 1

You could use sed to delete non-matching lines instead, so inverting the logic. You want A || !B, which when inverted is !A && B. So:

sed '/a/!{/l/d;}'

Answer 2

If your grep has perl-like regex support with -P, you can use look ahead operators for that:

grep -P -e a -e '^(?!.*l)'

Would match on lines that contain a or where the start of the line (^) is not followed ((?!...)) by any number of characters then l.

For the special case of single character search strings here, you can use:

grep -e a -e '^[^l]*$'

Which matches on lines that either contain a or contain only any number (*) of characters other than l ([^l]) from start (^) to finish ($).

perl, sed or awk (sed being the only one of those still supporting basic regexps) would make it easier in the general case. For lines that contain yes or don't contain no:

perl -ne 'print if /yes/ || ! /no/'

sed -e '/yes/b' -e '/no/!d'

(or @muru's variant)

awk '/yes/ || ! /no/'

---

Below is my original answer where I had misinterpreted the question as asking for lines that include a and not l. I'll leave it here in case that's useful to anyone as it's a connected question.

If your grep includes PCRE support with -P, you can use look ahead operators for that:

grep -P '^(?=.*a)(?!.*l)'

Would match on the start of the line (^) provided it's followed ((?=...)) by any number (*) of characters (.) then a and that it's not followed ((?!...)) by any number of characters then l.

For the special case of single character search strings here, you can use:

grep -x '[^l]*a[^l]*'

Which matches on lines made exactly (as a whole) of any number of characters other than l, one a and any number of characters other than l.

perl, sed or awk would make it easier in the general case. For lines that contain yes and not no:

perl -ne 'print if /yes/ && ! /no/'

sed -e '/yes/!d' -e '/no/d'

sed -ne /no/d -e /yes/p

awk '/yes/ && ! /no/'

Answer 3

To be completely honest, it looks like your question doesn't tell your whole goal. At least what you say in comments contradicts what it says in the question.

You said:

my actual use case involves large and complicated strings it needs to not match, and a single character it needs to match.

but there's no example of what that means, and also you did not mention why you want BRE in particular (and not ERE).

---

Now, I'll throw a wild guess that you don't actually want regexes at all, but only fixed strings, and that your logic is as follows:

• for each line
  1. print if the line contains the "include" character/string
  2. ignore the line if it contains any of the "exclude" strings
  3. otherwise print it

If that's the case, then using AWKs index() function might be helpful to avoid quoting special characters, as it matches against strings only.

This should print any lines in strings.txt that either contain the value assigned to include (i.e. i) or don't contain any of the strings in excludepatterns.txt (one per line).

% awk -v include=i '
   NR == FNR { excludes[NR] = $0; next };
   index($0, include) {print; next};
   { for (i in excludes) 
       if (index($0, excludes[i])) next;
       print; 
   }'  excludepatterns.txt strings.txt

Given the input strings.txt

ixy
zzi
xyz
iii
abc

and excludepatterns.txt:

xy
zz

That should print

ixy
zzi
iii
abc

i.e. lines that have an i or don't have xy or zz. (NR == FNR is AWK trickery for "is this the first file being processed")

Answer 4

Trying to piece together information from your comments, and especially:

my actual use case involves large and complicated strings it needs to not match (-v), and a single character it needs to match. grep -v '^complicatedRegex$\|[^l]' and grep -v '^complicatedRegex$\|.*[^l].*' didn't work as expected

I think what you really have is some string of any characters you want treated literally and skipped and some single character you also want treated literally and do want to match. If so, the implementation of what you were trying to do with the grep commands in your comment is this in any awk:

awk 'index($0,"l") && ($0 != "complicatedString")'

or if you wanted to do a partial-line instead of full-line match on complicatedString too then it'd be:

awk 'index($0,"l") && !index($0,"complicatedString")'

The above are just doing literal string comparisons, see how-do-i-find-the-text-that-matches-a-pattern for more information.