I have a reviews table as follows:

I am trying to match 2 things. The TID and the weight (kg). As you can see there are a 3 things to keep in mind

• Weight is always in kg and is written case insensitively and in 2 ways, kg and k.g and written in 2 ways <weight> <kg or k.g> <weight><kg or k.g> (one with space, one without space)
• TID is written case insensitively and can be written in 2 ways TID<id> or TID <id> (one with a space, one without a space.
• Some comments have multiple TID and weights. I am making the assumption that the first appearance of the TID is tied to the first appearance of the weight and the the second appearance of the TID is for the second appearance of the weight. I've only gone up to 2 instances of the TID/weight but I would like it to dynamically work for any amount of instances.

So I am able to extract the TID and the weight if the comment only has 1 weight and 1 TID. However, if it has multiple, I fail to do so. So I want to separate the multiple into different rows.

This is my desired output

SQL to create table/dummy data:

CREATE TABLE reviews(
  r_id number(3) NOT NULL,
  my_comment VARCHAR(255) NOT NULL
);

INSERT INTO reviews (r_id, my_comment) VALUES (1, 'Boxes with the TID 823 cannot exceed 40 kg');
INSERT INTO reviews (r_id, my_comment) VALUES (2, 'Parcel with the marking tid 63157 must not make the weight go over 31 k.g');
INSERT INTO reviews (r_id, my_comment) VALUES (3, 'Envelopes with TID 104124 and TID 92341 cant excel above 94.477kg');
INSERT INTO reviews (r_id, my_comment) VALUES (4, 'TID38204 cannot go over 45.4 kg and TID 8242602 cannot go over 92kg');
INSERT INTO reviews (r_id, my_comment) VALUES (5, 'Box with the TID 94514 cannot go over 52kg but also cannot go over 51KG');

In my attempt, I am able to extract the tid and weight, but only the first instance and not able to split it into rows.

SELECT
    r_id,

   REGEXP_SUBSTR (
        REGEXP_SUBSTR (my_comment, '(tid).*?[0-9] ', 1, 1, 'i'),
        '[0-9] '
    ) as "tid",

    REGEXP_SUBSTR (
        REGEXP_SUBSTR (my_comment, '(cannot exceed|go over| excel above).*?[0-9]  ?(kg|k.g)', 1, 1, 'i'),
        '[0-9] '
    ) as "weight"

FROM reviews;

CodePudding user response：

I am able to extract the tid and weight, but only the first instance and not able to split it into rows.

Your query, modified:

• I didn't do much with what you already wrote as you seem to be satisfied with extracted tid and weight
  • what I did change, was regexp_substr's occurrence parameter (was 1, now it is column_value)
• in order to get split data, cross join was added which "loops" through my_comment as many times as there's greatest occurrence between tid and kg (in any form)
  • for example, if there are 2 tid and 1 kg, it'll "loop" 2 times
  • it is also used to avoid duplicates you'd get if you used only connect by level clause

You did tag the question as Oracle 10; I don't have it any more, but I know it doesn't support regexp_count function. If that's really the case (you never answered Koen's question), then it won't work and you'll have to calculate number of tid/weight occurrences using some other way. I hope you aren't on 10g, though.

I ran this code in SQL*Plus. BREAK is here just to nicely distinguish r_id and my_comment values, doesn't have any other purpose.

SQL> break on r_id on my_comment
SQL> SELECT r_id,
  2         my_comment,
  3         REGEXP_SUBSTR (REGEXP_SUBSTR (my_comment,
  4                                       '(tid).*?[0-9] ',
  5                                       1,
  6                                       COLUMN_VALUE,
  7                                       'i'),
  8                        '[0-9] ') AS "tid",
  9         REGEXP_SUBSTR (
 10            REGEXP_SUBSTR (
 11               my_comment,
 12               '(cannot exceed|go over| excel above).*?[0-9]  ?(kg|k.g)',
 13               1,
 14               COLUMN_VALUE,
 15               'i'),
 16            '[0-9] ') AS "weight"
 17    FROM reviews
 18         CROSS JOIN
 19         TABLE (
 20            CAST (
 21               MULTISET (
 22                      SELECT LEVEL
 23                        FROM DUAL
 24                  CONNECT BY LEVEL <= GREATEST (REGEXP_COUNT (my_comment, 'tid'     , 1, 'i'),
 25                                                REGEXP_COUNT (my_comment, '(kg|k.g)', 1, 'i')))
 26                  AS SYS.odcinumberlist));

which results in

 R_ID MY_COMMENT                                                                tid     weight
----- ------------------------------------------------------------------------- ------- -------
    1 Boxes with the TID 823 cannot exceed 40 kg                                823     40
    2 Parcel with the marking tid 63157 must not make the weight go over 31 k.g 63157   31
    3 Envelopes with TID 104124 and TID 92341 cant excel above 94.477kg         104124  94
                                                                                92341
    4 TID38204 cannot go over 45.4 kg and TID 8242602 cannot go over 92kg       38204   45
                                                                                8242602 92
    5 Box with the TID 94514 cannot go over 52kg but also cannot go over 51KG   94514   52
                                                                                        51

8 rows selected.

SQL>