C++11 Regex search - Exclude empty submatches

Question

From the following text I want to extract the number and the unit of measurement.

I have 2 possible cases:

This is some text 14.56 kg and some other text

or

This is some text kg 14.56 and some other text

I used | to match the both cases. My problem is that it produces empty submatches, and thus giving me an incorrect number of matches.

This is my code:

std::smatch m;
std::string myString = "This is some text kg 14.56 and some other text";

const std::regex myRegex(
        R"(([\d]{0,4}[\.,]*[\d]{1,6})\s+(kilograms?|kg|kilos?)|s+(kilograms?|kg|kilos?)(\s+[\d]{0,4}[\.,]*[\d]{1,6}))",
        std::regex_constants::icase
);

if( std::regex_search(myString, m, myRegex) ){
    std::cout << "Size: " << m.size() << endl;
    for(int i=0; i<m.size(); i++)
        std::cout << m[i].str() << std::endl;
}
else
    std::cout << "Not found!\n";

OUTPUT:

Size: 5
kg 14.56


kg
14.56

I want an easy way to extract those 2 values, so my guess is that I want the following output:

WANTED OUTPUT:

Size: 3
kg 14.56
kg
14.56

This way I can always directly extract 2nd and 3th, but in this case I would also need to check which one is the number. I know how to do it with 2 separate searches, but I want to do it the right way, with a single search without using c++ to check if a submatch is an empty string.

I've edited my answer, and it works - but you might want to wait for the explanation as well — Robo Mop, Jan 15 '19 at 05:26

score 2 · Accepted Answer · answered Jan 15 '19 at 06:31

2

Using this regex, you just need the contents of Group 1 and Group 2

((?:kilograms?|kilos?|kg)|(?:\d{0,4}(?:\.\d{1,6})))\s*((?:kilograms?|kilos?|kg)|(?:\d{0,4}(?:\.\d{1,6})))

Click for Demo

Explanation:

((?:kilograms?|kilos?|kg)|(?:\d{0,4}(?:\.\d{1,6})))
- (?:kilograms?|kilos?|kg) - matches kilograms or kilogram or kilos or kilo or kg
- | - OR
- (?:\d{0,4}(?:\.\d{1,6})) - matches 0 to 4 digits followed by 1 to 6 digits of decimal part
\s* - matches 0+ whitespaces

answered Jan 15 '19 at 06:31

Gurmanjot Singh

10,224
2
19
43

Hi, is there some way to return groups by the same order? I read that ECMAScript does no allow named groups: https://stackoverflow.com/questions/16886992/c11-regex-capture-groups-by-name – Boy Jan 16 '19 at 03:56
1

By "same order", do you mean that for both the examples, group 1 should contain the numeric value and group 2 should contain the unit? – Gurmanjot Singh Jan 16 '19 at 05:28
Yes, exactly that. But my bigger problem is that I forgot to mention that it should also allow integers, not just float :( I tried many things but it always catches some other numbers that aren't surrounded by unit. If you are willing to help I'll update question. Anyway, thanks a lot for helping man. – Boy Jan 16 '19 at 06:00
1

That Integer problem can be solved by doing a slight modification as shown [HERE](https://regex101.com/r/lYmx7l/3) – Gurmanjot Singh Jan 16 '19 at 06:15
1

Also, for the first problem, I think you can just programmatically check for the contents of each group. If the contents are non-numeric, that means you have captured the unit, otherwise you have captured the value. – Gurmanjot Singh Jan 16 '19 at 06:19
Omg, 0 or 1 for a group, can't believe I haven't tried that! thank you! Anyway, I encountered another problem, check this: https://regex101.com/r/kLnnOg/1 For the ordering problem, yes, I know I can fairly easy do it by code, I wanted to know if it is possible with regex. – Boy Jan 16 '19 at 06:33
I understand that it is happening because of OR, because it is not strict so that each alternative inside capturing groups can appear only once. But thank you anyway man, you taught me a lot. If I could, I would give you my more points :) – Boy Jan 16 '19 at 06:56
1

No worries brother. I will update the solution if I come up with a better solution to include your negative test cases too. Will give it another try after my office hours. Glad that the solution helped you :) – Gurmanjot Singh Jan 16 '19 at 07:13

score 1 · Answer 2 · edited Jun 20 '20 at 09:12

1

You can try this out:

((?:(?<!\d)(\d{1,4}(?:[\.,]\d{1,6})?)\s+((?:kilogram|kilos|kg)))|(?:((?:kilogram|kilos|kg))\s+(\d{1,4}(?:[\.,]\d{1,6})?)))

As shown here: https://regex101.com/r/9O99Fz/3

USAGE -

As I've shown in the 'substitution' section, to reference the numeral part of the quantity, you have to write $2$5, and for the unit, write: $3$4

Explanation -

There are two capturing groups we could possibly need: the first one here (?:(?<!\d)(\d{1,4}(?:[\.,]\d{1,6})?)\s+((?:kilogram|kilos|kg))) is to match the number followed by the unit,
and the other (?:((?:kilogram|kilos|kg))\s+(\d{1,4}(?:[\.,]\d{1,6})?)) to match the unit followed by the number

edited Jun 20 '20 at 09:12

Community

1
1

answered Jan 15 '19 at 05:11

Robo Mop

3,485
1
10
23

Hi, thank you for helping. Please note that I'm using default regex engine, which is ECMAScript. – Boy Jan 15 '19 at 05:18
I see that it produces following error: "? The preceding token is not quantifiable". – Boy Jan 15 '19 at 05:32
Nope, I use c++11 – Boy Jan 15 '19 at 05:34
@Boy Ok, let me check it out again – Robo Mop Jan 15 '19 at 05:34
C++ returned this error: "Invalid special open parenthesis." – Boy Jan 15 '19 at 05:36
@Boy Ok, this is a temporary fix, but can you try it out with this regex: `((?:(\d{1,4}(?:[\.,]\d{1,6})?)\s+((?:kilogram|kilos|kg)))|(?:((?:kilogram|kilos|kg))\s+(\d{1,4}(?:[\.,]\d{1,6})?)))` – Robo Mop Jan 15 '19 at 05:45
It produces empty submatches, similar to my output – Boy Jan 15 '19 at 05:58

C++11 Regex search - Exclude empty submatches

2 Answers2

USAGE -

Explanation -