MySQL get latest consecutive rows set by date - mysql

Question

I am retrieving following rows with query,
SELECT id, entry_id, DATE(entry_date)
FROM entries
WHERE entry_id = 51
ORDER BY entry_date DESC
+-----+----------+---------------------+
| id | entry_id | entry_date |
+-----+----------+---------------------+
| 84 | 51 | 2021-02-27 xx:xx:xx |<---
| 81 | 51 | 2021-02-26 xx:xx:xx | |
| 76 | 51 | 2021-02-25 xx:xx:xx | |-- consecutive set
| 74 | 51 | 2021-02-25 xx:xx:xx | |
| 73 | 51 | 2021-02-24 xx:xx:xx |<---
| 52 | 51 | 2021-02-20 xx:xx:xx |
| 44 | 51 | 2021-02-19 xx:xx:xx |
| 32 | 51 | 2021-02-18 xx:xx:xx |
| . | .. | ... |
| . | .. | ... |
+-----+----------+---------------------+
entry_date's data type is timestamp. The time does not matter here in entry_date. I am only concerned with the dates without time.
I want to get rows only with "latest consecutive dates" OR first and last date of the latest consecutive set for an "entry_id".
for example, for entry_id = 51, I want only rows,
+-----+----------+------------+
| id | entry_id | entry_date |
+-----+----------+------------+
| 84 | 51 | 2021-02-27 |
| 81 | 51 | 2021-02-26 |
| 76 | 51 | 2021-02-25 |
| 74 | 51 | 2021-02-25 |
| 73 | 51 | 2021-02-24 |
+-----+----------+------------+
OR I want to get first and last date of "latest consecutive dates" set for entry_id = 51
eg. in this case entry_date 2021-02-24 and 2021-02-27.
I don't have any experience with writing such queries. I can get all the records order by DESC for entry_id = 51 and write a script to get latest consecutive rows but since there are hundreds of thousands of rows which can be sometimes inefficient to process just to get latest consecutive rows.
Please note that there can be some entries with the same date (in this case: 2021-02-25) which are also considered in the result.
Edit: I am using MySQL 5.6.

Answer 1

This is a type of gaps-and-islands problem solved using lead() to determine where there is a gap of more than one day.
select entry_id, min(entry_date), max(entry_date)
from (select e.*,
sum(case when entry_date < next_entry_date - interval 1 day then 1 else 0 end) over (partition by entry_id order by entry_date desc) as grp
from (select e.*,
lead(entry_date) over (partition by entry_id order by entry_date) as next_entry_date
from entries e
) e
) e
where grp = 0
group by entry_id;
Then the cumulative sum is done in reverse order. So the last group has a cumulative sum of 0.
Here is a db<>fiddle.

Answer 2

You might try finding the gap dates with a self join, using group_concat to get them together, then using substring_index to split out the first two, then use that to get all rows bounded by the dates.
SELECT q.*
FROM entries q
JOIN (
SELECT entry_id,
SUBSTRING_INDEX(dts, ',', 1) AS upperDate,
SUBSTRING_INDEX(SUBSTRING_INDEX(dts, ',', 2), ',', -1) AS lowerDate
FROM (
SELECT x.entry_id,
GROUP_CONCAT(if(y.entry_date IS NULL, x.entry_date, NULL) ORDER BY x.entry_date DESC) AS dts
FROM entries x
LEFT JOIN entries y ON (
x.entry_id = y.entry_id AND
x.entry_date = y.entry_date - interval 1 DAY
)
GROUP BY 1
) z
) bounds ON (
q.entry_id = bounds.entry_id AND
q.entry_date <= bounds.upperDate AND
q.entry_date > bounds.lowerDate
)
You can avoid the self join with some variables, but that adds a certain level of complexity of its own and makes the logic a little harder to read and maintain.
SELECT entry_id, entry_date FROM (
SELECT entry_id, entry_date, elapsed,
NOT #latch AS keep,
#latch:= if(elapsed > 1, TRUE, #latch),
#latch := if(#currID <> entry_id, FALSE, #latch),
#currID := entry_id
FROM (
SELECT entry_id, entry_date,
TIMESTAMPDIFF(DAY, #prevDate, entry_date) AS elapsed,
#prevDate := entry_date
FROM (
SELECT entry_id, entry_date
FROM entries
JOIN (SELECT #currID := 0, #prevDate := null, #latch:=false) v
ORDER BY entry_id, entry_date ASC
) z
) y ORDER BY entry_id, entry_date DESC
) x WHERE keep

Answer 3

In MySQL 8, it can be done like this:
WITH e2 AS (
SELECT entry_id, entry_date
, COALESCE(LAG(DATE(entry_date))
OVER (PARTITION BY entry_id
ORDER BY entry_date DESC)
- DATE(entry_date), 0)
AS date_diff
FROM entries
), e3 AS (
SELECT e2.*
, MAX(date_diff)
OVER (PARTITION BY e2.entry_id
ORDER BY e2.entry_date DESC
ROWS UNBOUNDED PRECEDING)
AS max_diff
FROM e2
)
SELECT entry_id
, DATE(MIN(entry_date)) AS min_date
, DATE(MAX(entry_date)) AS max_date
FROM e3
WHERE max_diff <= 1
GROUP BY entry_id;
Result
entry_id
min_date
max_date
51
2021-02-24
2021-02-27
See DB Fiddle for demo.
If you want the WHERE entry_id = 51 condition, it should be added to the first WITH query.