I have a regexp with a non-greedy modifier which does not seem to work. I have tried so many variations of the regexp and various other ways I could think of, without success, that I am losing my head
I want to remove all the empty strings embedded in the string s below. With my regexp I was expecting to remove all the things that matched something=""
s = 'a,b="cde",f="",g="hi",j=""'
puts s; puts s.gsub( /,.+?="",?/ , "," ).chomp(','); nil
Expected:
a,b="cde",g="hi"
What I get:
a,g="hi"
Why isn't the .+? non greedy in the gsub regexp below?
It works if I constrain the . to a set of characters [\w\d_-], but that forces me to do assumptions:
puts s; puts s.gsub( /,[\w\d_-]+?=""/ , "" ).chomp(','); nil
# outputs:
a,b="cde",f="",g="hi",j=""
a,g="hi"
It also works if I do some sort of negative lookup like:
puts s; puts s.gsub( /,.+?="",?/ , "," ).chomp(','); nil
# outputs:
a,b="cde",f="",g="hi",j=""
a,g="hi"
But still I do not understand why it did not work in the first case.
Regex matches from left to right. Your regex ,.+?="",? matches the first comma in the string a,b="cde",f="",g="hi",j="", the one between a and b. Then it tries to find ="" that exists after the ,g so you get the actual result.
What you want is: ,[^=]+?="",? that matches 1 or more any character that is not an equal sign before ="" and you'll get a,b="cde",g="hi" as result.
I have string formulas like this:
?{a,b,c,d}
It can be can be embedded like this:
?{a,b,c,?{x,y,z}}
or this is the same:
?{a,b,c,
?{x,y,z}
}
So I have to find those commas, what are in the second and greather "level" brackets.
In the example below I marked the "levels" where I have to find all commas:
?{a,b,c,
?{x,y, <--Those
?{1,2,3} <--Those
}
}
I've tried with lookahead and lookbehind, but I'm totally confused now :/
Here is my latest working try, but it is not good at all:
OnlineRegex
Update:
To avoid misunderstanding, I don't want to count the commas.
I'd like to get groups of commas to replace them.
The condition is find the commas where more than one "open tags" before it like this: ?{
.. without closing tag like this: }
Examlpe.:
In this case I have not replace any commas:
?{1,2,3} ?{a,b,c}
But in this case I have to replace commas between a b c
?{1,2,3,?{a,b,c}}
For the examples which you have provided, the following regex works(gives the desired output as mentioned by you):
(?<!^\?{[^{}]*),(?=[\s\S]*(?:\s*}){2,})
For String ?{a,b,c,d}, see Demo1 No Match
For String, ?{a,b,c,?{x,y,z}}, see Demo2 Match successful
For String,
?{a,b,c,
?{x,y,z}
}
see Demo3 Match Successful
For String,
?{a,b,c,
?{x,y,
?{1,2,3}
}
}
see Demo4 Match Successful
For String ?{1,2,3} ?{a,b,c} ?{1,2,3} ?{a,b,c}, see Demo5 No Match
Explanation:
(?<!^\?{[^{}]*), - negative lookbehind to discard the 1st level commas. The logic applied here is it should not match the comma which is preceded by start of the string followed by ?{ followed by 0+ occurrences of any character except { or }
(?=[\s\S]*(?:\s*}){2,}) - The comma matched above must be followed by atleast 2 occurrences of }(consecutive or having only whitespaces between them)
Your question is rather unclear #norbre, but I presume you'd like to extract (i.e. "count") the number of commas.
You can't do this with a regex. Regexps can't count number of occurences. However, you can use this to extract the "internal part" and then use a spreadsheet formula to count number of commas:
^(?:\?{[a-zA-Z0-9,]+?,\n??\s*?\?{)([a-zA-Z0-9,?{}\n\s]+?(?:\n*?\s*?|})+)(?:[a-zA-Z0-9,\n\s]*})$
Try: https://regex101.com/r/Rr0eFo/5
Examples
1.
Input:
?{a,b,c,?{e,f},1,2,3}
Output:
e,f}
2.
Input:
?{a,b,c,
?{x,y,z,e,
?{1,2,3,?{f,g,3},4,5,6}
}
,d,e,f}
Output:
x,y,z,e,
?{1,2,3,?{f,g,3},4,5,6}
}
3.
Input:
?{a,b,c,?{e},1,2,3}
Output:
e}
(note that there are no commas here!)
One caveat however. As I have said, regexps can't count number of occurences.
Hence, the following sample (don't know if it's valid or not for your case) would return wrong match:
?{a,b,c,?{e,f}
,1,2,3,?{a,b}
}
Output:
e,f}
,1,2,3,?{a,b}
OK replacing commas is another story so I'll add another answer.
Your regexp engine would need to support recursion.
Still I don't see a way to do it with one regex - one match would either contain the first comma or contain everything between the braces!
What I suggest is to use one regexp to get "what is inside the inner braces", run a replace (, => "") and assemble the whole line again using submatches from the regexp.
Here it is: (\?{[^?{}]*)((?>[^?{}]|(?R))+?)([^?{}]*?\})
Try: https://regex101.com/r/IzTeY0/3
Example 1:
Input:
?{a,b,c,
?{x,y,z,e,
?{1,2,3,?{f,g,3},4,5,6}
}
,d,e,f}
Submatches:
1. ?{a,b,c,
2. ?{x,y,z,e,
?{1,2,3,?{f,g,3},4,5,6}
}
3.
,d,e,f}
Replace all commas in submatch 2 with anything you want, then reassamble the whole string using submatches 1 and 3.
Again, this would break the regexp:
?{a,b,c,?{e,f}
,1,2,3,?{a,b}
}
Submatch 2 would look like this:
?{e,f}
,1,2,3,?{a,b}
Learning regex in bash, i am trying to fetch all lines which ends with .com
Initially i did :
cat patternNpara.txt | egrep "^[[:alnum:]]+(.com)$"
why : +matches one or more occurrences, so placing it after alnum should fetch the occurrence of any digit,word or signs but apparently, this logic is failing....
Then i did this : (purely hit-and-try, not applying any logic really...) and it worked
cat patternNpara.txt | egrep "^[[:alnum:]].+(.com)$"
whats confusing me : . matches only single occurrence, then, how am i getting the output...i mean how is it really matching the pattern???
Question : whats the difference between [[:alnum:]]+ and [[:alnum:]].+ (this one has . in it) in the above matching pattern and how its working???
PS : i am looking for a possible explanation...not, try it this way thing... :)
Some test lines for the file patternNpara.txt which are fetched as output!
valid email = abc#abc.com
invalid email = ab#abccom
another invalid = abc#.com
1 : abc,s,11#gmail.com
2: abc.s.11#gmail.com
Looking at your screenshot it seems you're trying to match email address that has # character also which is not included in your regex. You can use this regex:
egrep "[#[:alnum:]]+(\.com)" patternNpara.txt
DIfference between 2 regex:
[[:alnum:]] matches only [a-zA-Z0-9]. If you have # or , then you need to include them in character class as well.
Your 2nd case is including .+ pattern which means 1 or more matches of ANY CHARACTER
If you want to match any lines that end with '.com', you should use
egrep ".*\.com$" file.txt
To match all the following lines
valid email = abc#abc.com
invalid email = ab#abccom
another invalid = abc#.com
1 : abc,s,11#gmail.com
2: abc.s.11#gmail.com
^[[:alnum:]].+(.com)$ will work, but ^[[:alnum:]]+(.com)$ will not. Here is the reasons:
^[[:alnum:]].+(.com)$ means to match strings that start with a a-zA-Z or 0-9, flows two or more any characters, and end with a 'com' (not '.com').
^[[:alnum:]]+(.com)$ means to match strings that start with one or more a-zA-Z or 0-9, flows one character that could be anything, and end with a 'com' (not '.com').
Try this (with "positive-lookahead") :
.+(?=\.com)
Demo :
http://regexr.com?38bo0
I try to extract the name1 (first-row), name2 (second-row), name3 (third-row) and the street-name (last-row) with regex:
Company Inc.
JohnDoe
Foobar
Industrieterrein 13
The very last row is the street name and this part is already working (the text is stored in the variable "S2").
REGEXREPLACE(S2, "(.*\n)+(?!(.*\n))", "")
This expression will return me the very last line. I am also able the extract the first row:
REGEXREPLACE(S2, "(\n.*)", "")
My problem is, that I do not know how to extract the second and third row....
Also how do I test if the text contains one, two, three or more rows?
Update:
The regex is used in the context of Scribe (a ETL tool). The problem is I can not execute sourcecode, I only have the following functions:
REGEXMATCH(input, pattern)
REGEXREPLACE(input, pattern, replacement)
If the regex language provides support for lookaheads you may count rows backwards and thus get (assuming . does not match newline)
(.*)$ # matching the last line
(.*)(?=(\n.*){1}$) # matching the second last line (excl. newline)
(.*)(?=(\n.*){2}$) # matching the third last line (excl. newline)
just use this regex:
(.+)+
explain:
.
Wildcard: Matches any single character except \n.
+
Matches the previous element one or more times.
As for a regular expression that will match each of four rows, how about this:
(.*?)\n(.*?)\n(.*?)\n(.*)
The parentheses will match, and the \n will match a new line. Note: you may have to use \r\n instead of just \n depending; try both.
You can try the following:
((.*?)\n){3}