Remove rows that do not differ from the previous row in MySQL - mysql

Question

Suppose I have a table that records changes to my database over time:
TimeOfChange FieldA FieldB FieldC
-------------------------------------
2019-01-01 A1 B1 C1 /*(R1)*/
2019-01-02 A2 B2 C1 /*(R2)*/
2019-01-03 A2 B2 C1 /*(R3)*/
2019-01-05 A1 B1 C2 /*(R4)*/
2019-01-07 A1 B1 C1 /*(R5)*/
My database has many rows where nothing significant changed, eg row (R3) is the same as (R2).
I would like to remove these rows. I have found many references on how to use a common table expression to remove duplicate rows from the table. So it's possible to remove the duplicate (ignoring the TimeOfChange column) rows. But this will remove (R5) as well because it is the same as R1. I only want to remove the rows that have the same ABC values as the previous row, when ordered by the TimeOfChange column. How do I do that?
edit: You can assume that TimeOfChange values are all unique

Answer 1

Assuming the TimeOfChange is unique, you can do:
delete
from data
where TimeOfChange in (
select TimeOfChange
from (
select d2.TimeOfChange
from data d1
join data d2
where d2.TimeOfChange in (
select min(x.TimeOfChange)
from data x
where x.TimeOfChange>d1.TimeOfChange
) and d1.FieldA=d2.FieldA and d1.FieldB=d2.FieldB and d1.FieldC=d2.FieldC
) as q
);
So you first want to determine which rows are the "next" and then check if the "next" has the same values as the "current". For those the "next" would form a result set that you want to use in DELETE. The select * from data is there to circumvent the reuse of the table in DELETE and in the subquery.
You probably will get much better performance if you separate the logic into a stored procedure and store the id's for rows to be deleted into a temp table.
See DB Fiddle

Answer 2

Presuming, you really meant "when the same A, B, C occurred on the most recent day prior that had any data", this should be usable to identify the rows that need removed:
SELECT t2.TimeOfChange, t2.FieldA, t2.FieldB, t2.FieldC
FROM (
SELECT tMain.TimeOfChange, tMain.FieldA, tMain.FieldB, tMain.FieldC
, MAX(tPrev.TimeOfChange) AS prevTimeOfChange
FROM t AS tMain
LEFT JOIN t AS tPrev ON t.TimeOfChange> tPrev.TimeOfChange
GROUP BY tMain.TimeOfChange, tMain.FieldA, tMain.FieldB, tMain.FieldC
) AS t2
INNER JOIN t AS tPrev2
ON t2.prevTimeOfChange = tPrev2.TimeOfChange
AND t2.FieldA = tPrev2.FieldA
AND t2.FieldB = tPrev2.FieldB
AND t2.FieldC = tPrev2.FieldC
This can then be used in a DELETE with some indirection to force a temp table to be created.
DELETE td
FROM t AS td
WHERE (td.TimeOfChange, td.FieldA, td.FieldB, td.FieldC)
IN (SELECT * FROM ([the query above]) AS tt) -- Yes, you have to wrap the query from above in a select * so mysql will not reject it.
;
However, after getting this far, what happens when....
2019-01-01 A1 B1 C1
2019-01-02 A2 B2 C1
2019-01-03 A2 B2 C1
2019-01-04 A1 B1 C2
2019-01-05 A1 B1 C3
2019-01-05 A1 B1 C1
2019-01-06 A1 B1 C3
2019-01-07 A1 B1 C1
becomes
2019-01-01 A1 B1 C1
2019-01-02 A2 B2 C1
2019-01-04 A1 B1 C2
2019-01-05 A1 B1 C3
2019-01-05 A1 B1 C1
2019-01-07 A1 B1 C1
Does a second pass now need made to remove the 2019-01-07 entry?
Are you going to run the query repeatedly until no rows are affected?