Compare fields from the query results in Oracle - oracle

Question

I am in a scenario to obtain all the records from a table where FIRSTNAME and LASTNAME of a particular record is the same but the BIRTHDATE is greater than or equal to 15 years.
Consider my table looks like:
_______________________________________________________________________________
| PRIMARY_ID | UNIQUE_ID | FIRSTNAME | LASTNAME | SUFFIX | BIRTHDATE |
_______________________________________________________________________________
| 12345 | abcd | john | collin | Mr | 1975-10-01 00:00:00|
| 12345 | cdef | john | collin | Mr | 1960-10-01 00:00:00|
| 12345 | efgh | john | collin | Mr | 1975-10-01 00:00:00|
| 12345 | ghij | john | collin | Mr | 1960-10-01 00:00:00|
| 12345 | aaaa | john | collin | Mr | 1975-10-01 00:00:00|
| 12345 | bdfs | john | collin | Mr | 1975-10-01 00:00:00|
| 12345 | asdf | john | collin | Mr | null |
| 12345 | dfgh | john | collin | Mr | null |
| 23456 | ghij | jeremy | lynch | Mr | 1982-10-15 00:00:00|
| 23456 | aaaa | jacob | lynch | Mr | 1945-10-12 00:00:00|
| 23456 | bdfs | jeremy | lynch | Mr | 1945-10-12 00:00:00|
| 23456 | asdf | jacob | lynch | Mr | null |
| 23456 | dfgh | jeremy | lynch | Mr | null |
_______________________________________________________________________________
In this table, for the PRIMARY_ID 12345, the FIRSTNAME and LASTNAME are all same but the BIRTHDATE difference between the UNIQUE_IDs if 15 years. So this PRIMARY_ID needs to be pulled out. Wherein for PRIMARY_ID 23456, the FIRSTNAME is not the same for all UNIQUE_ID records, so it must not be pulled out.
The table might contain NULL values for BIRTHDATE, which should be ignored.
This is what I have tried till now:
SELECT
/*PARALLEL(16)*/
PRIMARY_ID,
UNIQUE_ID,
FIRSTNAME,
LASTNAME,
SUFFIX,
BIRTHDATE,
RANK() OVER ( ORDER BY FIRSTNAME, LASTNAME, SUFFIX, BIRTHDATE) "GROUP"
FROM TABLE;
I have queried to form separate groups to distinguish by FIRSTNAME, LASTNAME and BIRTHDATE. I do not know on how to proceed further with this.
Can someone please help out?
NOTE: The BIRTHDATE field is in varchar datatype and I use Oracle 12C.

Answer 1

As I understand it, the goal is to return the distinct set of primary_id for which adjacent (alphabetically) unique_id that share the same firstname and lastname are separated by 15+ years. As I understand it, NULL should interrupt comparison (and be considered a non-match (otherwise, primary_id 23456 would also match here for pseudo-adjacent bdfs + ghij).
There are other ways to do this, but one way available in 12c is to use pattern-matching. An example is below. The example just uses a difference of 5478 days as to represent 15-years, but one could nuance that if greater exactitude was needed for intercalary days etc.
SELECT DISTINCT PRIMARY_ID
FROM THE_TABLE
MATCH_RECOGNIZE (
PARTITION BY PRIMARY_ID
ORDER BY UNIQUE_ID
ONE ROW PER MATCH
AFTER MATCH SKIP PAST LAST ROW
PATTERN(FIFTEEN_DIFF)
DEFINE FIFTEEN_DIFF AS
(FIFTEEN_DIFF.FIRSTNAME = PREV(FIFTEEN_DIFF.FIRSTNAME)
AND FIFTEEN_DIFF.LASTNAME = PREV(FIFTEEN_DIFF.LASTNAME)
AND (ABS(EXTRACT( DAY FROM (TO_TIMESTAMP(FIFTEEN_DIFF.BIRTHDATE,'YYYY-MM-DD HH24:MI:SS') - PREV(TO_TIMESTAMP(FIFTEEN_DIFF.BIRTHDATE,'YYYY-MM-DD HH24:MI:SS'))))) >= 5478)));
Result:
PRIMARY_ID
12345
1 row selected.
The above query does the following:
PARTITIONs to look at each PRIMARY_ID group individually,
then ORDERs by the UNIQUE_ID, so only alphabetically-adjacent records are compared.
Then each record is compared to the last, and if they share FIRSTNAME and LASTNAME, and their BIRTHDATEs differ by 15+ years, they are counted as a MATCH, and returns one record to indicate this.
After any match is found, it skips to the next row and resumes comparing.
Since only the distinct matches are desired, a DISTINCT is included in the select statement.
EDIT:
In response to follow-up questions, adding two additional examples.
Alternative 1: Pre-Filter NULL
This will bring different UNIQUE_ID into proximity, giving different matches.
SELECT DISTINCT PRIMARY_ID
FROM (SELECT PRIMARY_ID, UNIQUE_ID, FIRSTNAME, LASTNAME, SUFFIX, BIRTHDATE
FROM THE_TABLE
WHERE BIRTHDATE
IS NOT NULL)
MATCH_RECOGNIZE (
PARTITION BY PRIMARY_ID
ORDER BY UNIQUE_ID
ONE ROW PER MATCH
AFTER MATCH SKIP PAST LAST ROW
PATTERN (FIFTEEN_DIFF)
DEFINE FIFTEEN_DIFF AS
(FIFTEEN_DIFF.FIRSTNAME = PREV(FIFTEEN_DIFF.FIRSTNAME)
AND FIFTEEN_DIFF.LASTNAME = PREV(FIFTEEN_DIFF.LASTNAME)
AND (ABS(EXTRACT(DAY FROM (TO_TIMESTAMP(FIFTEEN_DIFF.BIRTHDATE , 'YYYY-MM-DD HH24:MI:SS') -
PREV(TO_TIMESTAMP(FIFTEEN_DIFF.BIRTHDATE , 'YYYY-MM-DD HH24:MI:SS'))))) >= 5478)));
Result (this now includes PRIMARY_ID 23456, as removing NULL brings two UNIQUE_IDs into order that ar 15+ years apart) :
PRIMARY_ID
12345
23456
2 rows selected.
Alternative 2: Count NULL as a match
SELECT DISTINCT PRIMARY_ID
FROM THE_TABLE
MATCH_RECOGNIZE (
PARTITION BY PRIMARY_ID
ORDER BY UNIQUE_ID
ONE ROW PER MATCH
AFTER MATCH SKIP PAST LAST ROW
PATTERN (FIFTEEN_DIFF)
DEFINE FIFTEEN_DIFF AS
(FIFTEEN_DIFF.FIRSTNAME = PREV(FIFTEEN_DIFF.FIRSTNAME)
AND FIFTEEN_DIFF.LASTNAME = PREV(FIFTEEN_DIFF.LASTNAME)
AND ((ABS(EXTRACT(DAY FROM (TO_TIMESTAMP(FIFTEEN_DIFF.BIRTHDATE , 'YYYY-MM-DD HH24:MI:SS') -
PREV(TO_TIMESTAMP(FIFTEEN_DIFF.BIRTHDATE , 'YYYY-MM-DD HH24:MI:SS'))))) >= 5478)
OR (LEAST(FIFTEEN_DIFF.BIRTHDATE,PREV(FIFTEEN_DIFF.BIRTHDATE)) IS NULL
AND COALESCE(FIFTEEN_DIFF.BIRTHDATE,PREV(FIFTEEN_DIFF.BIRTHDATE)) IS NOT NULL))));
Result (This also return both PRIMARY_ID, as NULL is now counted as a match):
PRIMARY_ID
12345
23456
2 rows selected.