Regular Expression Pattern Help Needed

[Ricky Murphy]

I have been struggling to achive following by using regex in Java:

For example. if I have following 5 sentences:

1. A123456 some 2008 junk in my PC
2. A234567 another pile of 2008 junk
3. A345678 collected 2009 rocks
4. A456 got that in 2009.
5. A567890 sent me 3 letters.


What I want to achieve is: pick out the sentences that each should:
a). begins with A and followed by 6 digits
b). does NOT have the word 2008 in it.

This should give me sentence 3) and 5).

How do I construct a regex patten?

Greatly appreciate it

Thank you

-RickM

[Daniel Croft]

I believe this is what you want: "A\\d{6}((?!2008).)*$" I ran it on your examples and it worked, but I would try it with a few more because I'm not 100% sure about it, my regex skills are a bit rusty.

The A catches your first A, \d{6} stands for "any 6 digits" but you need the extra \ because java will recognize the first \ as an escape character. ((?!2008).)*$ means "any number of characters that don't contain 2008". Hope this works for you =)

[Henry Wong]

Daniel Croft wrote:I believe this is what you want: "A\\d{6}((?!2008).)*$" I ran it on your examples and it worked, but I would try it with a few more because I'm not 100% sure about it, my regex skills are a bit rusty.

The A catches your first A, \d{6} stands for "any 6 digits" but you need the extra \ because java will recognize the first \ as an escape character. ((?!2008).)*$ means "any number of characters that don't contain 2008". Hope this works for you =)

Have you tested what happens when A is followed by more than 6 digits?

Furthermore, while it does work, I don't like the way the negative look ahead works. It tests it with every character.

Henry

[Daniel Croft]

Good catch on the more than 6 digits part. You're right, the .* after it allows this thing to accept more than 6 digits, although his specs don't technically disallow this. As for the negative lookahead, I sort of chopped that out of a Google search, but it was the best I could find on short notice; I suspect there's probably a better way.

[Henry Wong]

Ricky Murphy wrote:I have been struggling to achive following by using regex in Java:

For example. if I have following 5 sentences:

1. A123456 some 2008 junk in my PC
2. A234567 another pile of 2008 junk
3. A345678 collected 2009 rocks
4. A456 got that in 2009.
5. A567890 sent me 3 letters.


What I want to achieve is: pick out the sentences that each should:
a). begins with A and followed by 6 digits
b). does NOT have the word 2008 in it.

This should give me sentence 3) and 5).

How do I construct a regex patten?

If you do *not* know (or just a beginner) with regexes. I recommend that this be done with two regexes. One that looks for the first case, and the other for the second. The first regex should succeed and the second should fail.

As already mentioned, it is possible to use a single regex -- but for a beginner, it may be a good idea to avoid that until you are more comfortable with regexes. At least, to the point where you know what "negative look ahead" means.

Henry

[Ricky Murphy]

Thank you Daniel and Henry.

I think that I know what I did wrong (but not why). I started with the "positive" comparison:
a). begins with A and followed by 6 digits
b). AND has the word 2008 in it.

So i ended up with regex: ^A\d{6}\s.*(2008).*$ and ( ^A\d{6}\s.*((2008).)*$ works too )

While above worked for me, I thought the (?!2008) would give what I originally wanted ( i.e. the regex: ^A\d{6}\s.*(?!2008).*$ ). Wrong, not the case, with added ?! I still got a match.

By comparing to your solution, you didn't have the "\s.*" in your regex. So i removed that from my version of the positive match. And the positive match stopped working. So, the conclusion, for a positive match, I need to have \s.* after{6}; while to do a negative match with ?!, I need to remove \s.* . Could you help me out why?

Not sure if I explained it clearly.

Thank you,

-RickM

[Henry Wong]

Ricky Murphy wrote:
I think that I know what I did wrong (but not why). I started with the "positive" comparison:
a). begins with A and followed by 6 digits
b). AND has the word 2008 in it.

So i ended up with regex: ^A\d{6}\s.*(2008).*$ and ( ^A\d{6}\s.*((2008).)*$ works too )

I still think that you should do it as two different regexes, instead of trying to do both with one regex. It isn't as easy as you think it is.

Henry

[Henry Wong]

Ricky Murphy wrote:
While above worked for me, I thought the (?!2008) would give what I originally wanted ( i.e. the regex: ^A\d{6}\s.*(?!2008).*$ ). Wrong, not the case, with added ?! I still got a match.

First of all, it is a negative look ahead. It is not a negative match..... ie. "(?!2008)" is not the opposite case of "(2008)".

But to answer why it always matches.... the first ".*" will basically greedily take the whole string, leaving the negative look ahead to always not match (meaning succeed), and the last ".*" to always have a zero length match.

Henry

[Ricky Murphy]

But to answer why it always matches.... the first ".*" will basically greedily take the whole string, leaving the negative look ahead to always not match (meaning succeed), and the last ".*" to always have a zero length match.

Thank you Henry. I will certainly give some thought on breaking it into two regexs. It sure will be more clearer for a later on revisit.

First of all, it is a negative look ahead. It is not a negative match..... ie. "(?!2008)" is not the opposite case of "(2008)".

You are right that the (2008) is not the opposite of (?!2008). Then, is the (?=2008) the opposite of (?!2008)? While ^A\d{6}((?!2008).)*$ gives me good negative filtering,
^A\d{6}((?=2008).)*$ does NOT give me any positive findings (I tried other combinations too, such as ((?=(2008)), etc.). Could you please help me on on this? I am not very clear on the look ahead mixed with the regular expression.

-RickM

[Henry Wong]

Ricky Murphy wrote:
While ^A\d{6}((?!2008).)*$ gives me good negative filtering,

^A\d{6}((?!2008).)*$ -- is looking for an "A" followed by six digits... followed by zero or more of any character, with the qualification that at the point of the character, it must not be a "2008" at that location. Of course, this means that, after the first 7 places, it will match any string, as long as there isn't a "2008" at any location.

Ricky Murphy wrote:
^A\d{6}((?=2008).)*$ does NOT give me any positive findings

^A\d{6}((?=2008).)*$ -- is looking for an "A" followed by six digits... followed by zero or more of any character, with the qualification that at the point of the character, it must be a "2008" at that location. Of course, this means that, after the first 7 places, it will match any string, as long as there is a "2008" at every location. And since it is not possible to have "2008" at every location, the only string that will work is when it is a blank string.

Ricky Murphy wrote:
I tried other combinations too, such as ((?=(2008)), etc.

The extra parens specifies another group -- which can be used to capture the string in that group.... but since the string is always "2008", it doesn't really make much sense to capture it. If it matches, you know what it is.

Henry

[Winston Gutkowski]

Ricky Murphy wrote:Thank you Daniel and Henry.

I think that I know what I did wrong (but not why). I started with the "positive" comparison:
a). begins with A and followed by 6 digits
b). AND has the word 2008 in it.

OK, first off I agree with Henry: it's far better to do it in two separate passes. For one thing, it's clearer, for the second, the search for "2008" probably doesn't warrant a regex anyway.

Secondly, the regex that will find A + at least 6 digits is "A\\d{6,}", and furthermore it will eliminate anything that appears in that initial 'code number' from the search (I presume you don't want to match '2008' in that).

Have you tried "^A\\d{6,}.*\\b2008\\b.*$"? That's what I'd do for a matches() check, but you can probably do a lot better if you use java.util.regex.Pattern and java.util.regex.Matcher (again, if you really think you need to).

Winston

[Ricky Murphy]

Thank you guys for all your valuable suggestions. I learned a lot indeed.

-RickM