regex fails to catch all matches

Question

Here is an example:

The two (Senior Officer Stuart & Officer Jess) were intercepted by Officer George.

Now, let's say I have two ranks "Officer" and "Senior Officer" and want to replace the name after them with a general token "PERSON". As you can see there are three names that come after a rank Stuart, Jess, George. I don't know why my regex solution fails to capture all of them. Here is my code:

    public static void main(String[] args) {
    String input = "The two (Senior Officer Stuart & Officer Jess) were intercepted by Officer George.";
    ArrayList<String> ranks = new ArrayList<String>();
    ranks.add("Senior Officer");
    ranks.add("Officer");
    for (String rank : ranks) {
        Pattern pattern = Pattern.compile(".*" + rank + " ([a-zA-Z]*?) .*");
        Matcher m = pattern.matcher(input);
        if (m.find()) {
            System.out.println(rank);
            System.out.println(m.group(1));
        }
    }
}

and here is its output:

Senior Officer
Stuart
Officer
Stuart

which captures Stuart twice (via Senior Officer and Officer), but ignores Jess and George. I am expecting to get this as the output:

Senior Officer
Stuart
Officer
Stuart
Officer
Jess
Officer
George

No regex expert, but that space looks suspicious: `([a-zA-Z]*?) .*` — Hovercraft Full Of Eels, Jun 26 '16 at 12:47
Something like,,..... `Pattern.compile(".*?" + rank + " ([a-zA-Z]*)");` — Hovercraft Full Of Eels, Jun 26 '16 at 12:49
or `Pattern pattern = Pattern.compile(".*?" + rank + " (\\w*)")` — Hovercraft Full Of Eels, Jun 26 '16 at 12:49

rock321987 · Accepted Answer · 2016-06-26T14:17:39.717

2

This will be sufficient

for (String rank : ranks) {
    Pattern pattern = Pattern.compile("\\b" + rank + "\\s+([a-zA-Z]*)");
    Matcher m = pattern.matcher(input);
    while (m.find()) {
        System.out.println(rank);
        System.out.println(m.group(1));
    }
}

Ideone Demo

Regex Breakdown (as per comments)

Officer #Match Officer literally
 ( #Capturing group
  (?: #Non-capturing group
    \s #Match space
     (?!(?:Senior\s+)?Officer) #Negative lookahead assures that its impossible to match the word Senior(which is optional) and Officer literally
    [A-Z][a-zA-Z]* #Match capital letter followed by combination of capital and small letter
  )* #Repeat the previous step any number of time till one of the condition of first letter being capital fails or word Officer is found
 )

edited Jun 26 '16 at 14:17

answered Jun 26 '16 at 12:56

rock321987

10,942
1
30
43

Is there a way to include catching two/three word nouns? For example in "Senior Officer Eduardo Garcia Lopez", to catch the entire "Eduardo Garcia Lopez" instead of the first word "Eduardo"? – user3639557 Jun 26 '16 at 13:52
@user3639557 you can do that ..but there is a problem because you can't determine beforehand how many words to capture – rock321987 Jun 26 '16 at 13:55
how about, as long as words are upper initial? – user3639557 Jun 26 '16 at 13:56
@user3639557 that can be done..but there is another problem, your rank also contains upper initial..so that will be merged which is incorrect – rock321987 Jun 26 '16 at 13:58
@user3639557 see if **[this](https://regex101.com/r/gH6aJ3/1)** does help you – rock321987 Jun 26 '16 at 14:02
why? assume that ranks do not share words (i.e. in this example ignore "Senior Officer" and only consider "Officer"). Doesn't the matcher work by matching the rank, and then catching the followup words? – user3639557 Jun 26 '16 at 14:02
@user3639557 i have pointed above a **[link](https://regex101.com/r/gH6aJ3/2)**..see if its good – rock321987 Jun 26 '16 at 14:03
this seems to work, but looks terribly complicated! ""((?:\s(?!(?:\s+)?Officer)[A-Z][a-zA-Z]*)*)"" – user3639557 Jun 26 '16 at 14:11
@user3639557 added an explanation – rock321987 Jun 26 '16 at 14:17
for some reason it doesn't work as it is promised to work in that link. – user3639557 Jun 26 '16 at 15:00
@user3639557 i have hardcoded the regex..see it working **[here](http://ideone.com/V1ZoLK)** – rock321987 Jun 26 '16 at 15:09
yeah - but hardcoded ranks don't work for me. I need something that is more generic – user3639557 Jun 26 '16 at 15:12
@user3639557 ok give me some time.. – rock321987 Jun 26 '16 at 15:16
I am not familiar with regex, but wonder why this (".*"+rank + "\\s(([A-Z][a-zA-Z]*)+?)") doesn't work? – user3639557 Jun 26 '16 at 15:17
@user3639557 check **[here](http://ideone.com/V1ZoLK)** – rock321987 Jun 26 '16 at 15:18
1

@user3639557 `.*` is greedy in nature..so it will consume whole string and match the last part..see **[here](https://regex101.com/r/gH6aJ3/4)** – rock321987 Jun 26 '16 at 15:20
I am learning a lot :) ... the solution you referred to involves hardcoded patterns still. What happens if I have 10 new ranks, etc. There must be a simpler solution to this damn problem! – user3639557 Jun 26 '16 at 15:32
@user3639557 in that case use alternation in negative lookahead part like `(?!(?:rank1|rank2|rank3|....))` – rock321987 Jun 26 '16 at 15:34
yeah I was trying it just now - something like this, it still doesn't catch the consecutive words with upperinitial case, but does something "(?!"+rank+") "+rank + "\\s(([A-Z][a-zA-Z]*)+?)" – user3639557 Jun 26 '16 at 15:38
@user3639557 i am doing a project which I have to show tomorrow. I will update you as soon as I am free.. – rock321987 Jun 26 '16 at 15:40
thanks man - appreciate it a lot :) – user3639557 Jun 26 '16 at 15:41
1

@user3639557 I created a string `tmp` with alternation for `rank`.this string can be used in negative lookahead..check **[here](http://ideone.com/V1ZoLK)** – rock321987 Jun 26 '16 at 16:47
it works, but is awfully slow for longer texts. – user3639557 Jun 26 '16 at 17:29
@user3639557 certainly there will be some side effect.. – rock321987 Jun 27 '16 at 02:04

Vlad Călin Buzea · Answer 2 · 2016-06-26T13:10:24.910

The for you are using finds ONLY the first match of each rank. First of all, you need a while clause inside the for.

 for (String rank : ranks) {
        Pattern pattern = Pattern.compile(rank + " [A-z]+");
        Matcher m = pattern.matcher(input);
        while (m.find()) {
            System.out.println(rank);
            System.out.println(m.group(1));
        }
    }

However, this does not solve the problem of finding the "senior officer" rank twice: once when you search for "senior officer" and once when you search for "officer". I am not sure how you want to handle this issue. If you want Stuart to appear twice, then this code is good enough. If you want Stuart to be detected only once, you need to work on your regEx.

P.S. use an online tool to test the regex before coding it. It saves up a lot of time.

@user3639557: he's right about using while and not if. 1+ – Hovercraft Full Of Eels Jun 26 '16 at 12:53 — Hovercraft Full Of Eels, Jun 26 '16 at 12:53
@user3639557 i have updated the regex – Vlad Călin Buzea Jun 26 '16 at 13:02 — Vlad Călin Buzea, Jun 26 '16 at 13:02

regex fails to catch all matches

2 Answers2