grep two words near each other - regex

Question

Say I have a line in a file "This is perhaps the easiest place to add new functionality." and I want to grep two words close to each other. I do
grep -ERHn "\beasiest\W+(?:\w+\W+){1,6}?place\b" *
that works and gives me the line. But when I do
grep -ERHn "\beasiest\W+(?:\w+\W+){1,10}?new\b" *
it fails, defeating the whole point of the {1,10}?
This one is listed in the regular-expression.info site and also a couple of Regex books. Though they do not describe it with grep but that should not matter.
Update
I put the regex into a python script. Works, but doesn't have the nice grep -C thing ...
#!/usr/bin/python
import re
import sys
import os
word1 = sys.argv[1]
word2 = sys.argv[2]
dist = sys.argv[3]
regex_string = (r'\b(?:'
+ word1
+ r'\W+(?:\w+\W+){0,'
+ dist
+ '}?'
+ word2
+ r'|'
+ word2
+ r'\W+(?:\w+\W+){0,'
+ dist
+ '}?'
+ word1
+ r')\b')
regex = re.compile(regex_string)
def findmatches(PATH):
for root, dirs, files in os.walk(PATH):
for filename in files:
fullpath = os.path.join(root,filename)
with open(fullpath, 'r') as f:
matches = re.findall(regex, f.read())
for m in matches:
print "File:",fullpath,"\n\t",m
if __name__ == "__main__":
findmatches(sys.argv[4])
Calling it as
python near.py charlie winning 6 path/to/charlie/sheen
works for me.

Answer 1

Do you really need the look ahead structure?
Maybe this is enough:
grep -ERHn "\beasiest\W+(\w+\W+){1,10}new\b" *
Here is what I get:
echo "This is perhaps the easiest place to add new functionality." | grep -EHn "\beasiest\W+(\w+\W+){1,10}new\b"
(standard input):1:This is perhaps the easiest place to add new
functionality.
Edit
As Camille Goudeseune said:
To make it easily usable, this can be added in a .bashrc:
grepNear() {
grep -EHn "\b$1\W+(\w+\W+){1,10}$2\b"
}.
Then at a bash prompt: echo "..." | grepNear easiest new

Answer 2

grep does not support the non-capturing groups of Python regular expressions. When you write something like (?:\w+\W+), you are asking grep to match a question mark ? followed by a colon : followed by one or more word chars \w+ followed by one or more non-word chars \W+. ? is a special character for grep regexes, for sure, but since it is following the beginning of a group, it is automatically escaped (in the same way that the regex [?] matches the question mark).
Let us test it? I have the following file:
$ cat file
This is perhaps the easiest place to add new functionality.
grep does not match it with the expression you used:
$ grep -ERHn "\beasiest\W+(?:\w+\W+){1,10}?new\b" file
Then, I created the following file:
$ cat file2
This is perhaps the easiest ?:place ?:to ?:add new functionality.
Note that each word is preceded by ?:. In this case, your expression matches the file:
$ grep -ERHn "\beasiest\W+(?:\w+\W+){1,10}?new\b" file2
file2:1:This is perhaps the easiest ?:place ?:to ?:add new functionality.
The solution is to remove the ?: of the expression:
$ grep -ERHn "\beasiest\W+(\w+\W+){1,10}?new\b" file
file:1:This is perhaps the easiest place to add new functionality.
Since you do not even need a non-capturing group (at least as far as I've seen) it does not bear any problem.
Bonus point: you can simplify your expression changing {1,10} to {0,10} and removing the following ?:
$ grep -ERHn "\beasiest\W+(\w+\W+){0,10}new\b" file
file:1:This is perhaps the easiest place to add new functionality.