问题
This is my previous question: How can I merge two strings of comma-separated numbers in MySQL?
I tried to use delimited list with these reasons:
- the data is more than 2000 rows.
- I have just one day.
- this is not production level
- I need to parse the data with my hands.
I thought I don't have any choice. But I found this: SQL split values to multiple rows
So, I got some hopes from it. But it's quite difficult to apply it to my table.
- Mine is different form. I have multiple columns and I'd like to avoid duplicate row.
- I rather UPDATE and INSERT than SELECT because changing the formation of table will be easier to manage with this reason:Is storing a delimited list in a database column really that bad?
- I need to find the non listed numbers. For example, if the time value is 1, 3, 8 then I need 2,4,5,6,7,9,10,11,12,13 (between 1 and 13)
My example table looks like this:
+-----+------+--------+------+-------+-----------+-------------+
| cid | sid | type | day | time | building | room_number |
+-----+------+--------+------+-------+-----------+-------------+
| 1 | 1 | daytime | mon | 6,7,8 | sky | 507 |
| 2 | 2 | daytime | thu | 2,3,4 | nuri | 906 |
| 3 | 3 | daytime | tue | 6,7,8 | nuri | 906 |
| 4 | 4 | daytime | thu | 6,7 | sky | 1003 |
| 5 | 5 | daytime | mon | 2,3,4 | sky | 507 |
| 6 | 6 | daytime | wed | 6,7,8 | belief | 1003 |
| 7 | 7 | daytime | mon | 2,3,4 | belief | 905 |
| 8 | 8 | daytime | fri | 6,7,8 | truth | 905 |
| 9 | 9 | daytime | tue | 6,7,8 | truth | 905 |
| 10 | 10 | daytime | fri | 2,3,4 | truth | 905 |
| 11 | 11 | daytime | wed | 6,7,8 | truth | 905 |
| 12 | 12 | daytime | fri | 2,3,4 | truth | 1003 |
| 13 | 13 | daytime | mon | 6,7,8 | truth | 905 |
| 14 | 14 | daytime | tue | 2,3,4 | truth | 905 |
| 15 | 15 | daytime | tue | 6,7,8 | sky | 208 |
| 16 | 16 | daytime | tue | 2,3,4 | sky | 208 |
| 17 | 17 | daytime | tue | 2,3,4 | truth | 1004 |
| 18 | 19 | daytime | mon | 2,3,4 | sky | 208 |
| 19 | 20 | daytime | thu | 2,3,4 | truth | 1003 |
| 20 | 21 | daytime | wed | 6,7,8 | sky | 208 |
| 21 | 22 | night | tue | 4,5,6 | nuri | 405 |
| 22 | 23 | night | tue | 1,2,3 | nuri | 405 |
| 23 | 24 | night | tue | 1,2,3 | nuri | 306 |
| 24 | 25 | night | thu | 1,2,3 | nuri | 205 |
| 25 | 26 | night | thu | 4,5,6 | sky | 306 |
| 26 | 27 | night | wed | 1,2,3 | nuri | 306 |
| 27 | 28 | night | wed | 4,5,6 | sky | 309 |
| 28 | 29 | night | wed | 4,5,6 | nuri | 407 |
| 29 | 30 | night | tue | 4,5,6 | nuri | 306 |
| 30 | 31 | night | thu | 1,2,3 | nuri | 307 |
| 31 | 0 | always | | | sky | 201 |
| 32 | 0 | always | | | sky | 202 |
| 33 | 0 | always | | | sky | 203 |
| 34 | 0 | always | | | sky | 204 |
| 35 | 0 | always | | | nuri | 205 |
| 36 | 0 | always | | | nuri | 206 |
| 37 | 0 | always | | | truth | 207 |
| ... | ... | ... | | | ... | ... |
| 2000 | 0 | always | | | belief | 1101 |
+-----+------+--------+------+-------+-----------+-------------+
- 507|sky building on monday: 2,3,4,6,7,8 are scheduled. So, 1,5,9,10,11,12,13 will be added.
- 906|nuri building has schedule on thu, tue. which is different day. So, they will not affect each other. So, 1,5,6,7,8,9,10,11,12,13 on thu & 1,2,3,4,5,9,10,11,12,13 on tue.
- If there's duplicate row, it won't affect.
And my expectation of changed table is this:
+-----+------+--------+------+-------+-----------+-------------+
| cid | sid | type | day | time | building | room_number |
+-----+------+--------+------+-------+-----------+-------------+
| 1 | 1 | daytime | mon | 1 | sky | 507 |
| 2 | 1 | daytime | mon | 5 | sky | 507 |
| 3 | 1 | daytime | mon | 9 | sky | 507 |
| 4 | 1 | daytime | mon | 10 | sky | 507 |
| 5 | 1 | daytime | mon | 11 | sky | 507 |
| 6 | 1 | daytime | mon | 12 | sky | 507 |
| 7 | 1 | daytime | mon | 13 | sky | 507 |
| 8 | 2 | daytime | thu | 1 | nuri | 906 |
| 9 | 2 | daytime | thu | 5 | nuri | 906 |
| 10 | 2 | daytime | thu | 6 | nuri | 906 |
| 11 | 2 | daytime | thu | 7 | nuri | 906 |
| 12 | 2 | daytime | thu | 8 | nuri | 906 |
| 13 | 2 | daytime | thu | 9 | nuri | 906 |
| 14 | 2 | daytime | thu | 10 | nuri | 906 |
| 15 | 2 | daytime | thu | 11 | nuri | 906 |
| 16 | 2 | daytime | thu | 12 | nuri | 906 |
| 17 | 2 | daytime | thu | 13 | nuri | 906 |
| 18 | 3 | daytime | tue | 1 | nuri | 906 |
| 19 | 3 | daytime | tue | 2 | nuri | 906 |
| 20 | 3 | daytime | tue | 3 | nuri | 906 |
| ... | ... | ... | | | ... | ... |
| 302 | 0 | always | | | nuri | 206 |
| 303 | 0 | always | | | truth | 207 |
| ... | ... | ... | | | ... | ... |
| 4020 | 0 | always | | | belief | 1101 |
+-----+------+--------+------+-------+-----------+-------------+
The reason why I do this is the origin data is the time schedule of classes. And I want to find empty time so that people can use the class room during the time there's any classes.
回答1:
Here is an approach using a recusive query to split the strings into rows. You can then use this resultset to identify the missing records: for this, you can cross join the original table with the list of times, and use an anti-left join to pull out missing records:
with recursive
cte as (
select cid, sid, type, day, room_number,
substring(time, 1, locate(',', time) - 1) time,
substring(concat(time, ','), locate(',', time) + 1) rest
from mytable
union all
select cid, sid, type, day, room_number,
substring(rest, 1, locate(',', rest) - 1),
substring(rest, locate(',', rest) + 1)
from cte
where locate(',', rest) > 0
),
all_times as (
select 1 time
union all select time + 1 from all_times where time < 13
)
select t.cid, t.sid, t.type, t.day, at.time, t.building, t.room_number
from all_times at
cross join mytable t
left join cte c on c.cid = t.cid and c.time = at.time
where c.cid is null
order by t.cid, at.time
Demo on DB Fiddle:
Sample data:
cid | sid | type | day | time | building | room_number --: | --: | :------ | :-- | :---- | :------- | ----------: 1 | 1 | daytime | mon | 6,7,8 | sky | 507 2 | 2 | daytime | thu | 2,3,4 | nuri | 906
Query results:
cid | sid | type | day | time | building | room_number --: | --: | :------ | :-- | ---: | :------- | ----------: 1 | 1 | daytime | mon | 1 | sky | 507 1 | 1 | daytime | mon | 2 | sky | 507 1 | 1 | daytime | mon | 3 | sky | 507 1 | 1 | daytime | mon | 4 | sky | 507 1 | 1 | daytime | mon | 5 | sky | 507 1 | 1 | daytime | mon | 9 | sky | 507 1 | 1 | daytime | mon | 10 | sky | 507 1 | 1 | daytime | mon | 11 | sky | 507 1 | 1 | daytime | mon | 12 | sky | 507 1 | 1 | daytime | mon | 13 | sky | 507 2 | 2 | daytime | thu | 1 | nuri | 906 2 | 2 | daytime | thu | 5 | nuri | 906 2 | 2 | daytime | thu | 6 | nuri | 906 2 | 2 | daytime | thu | 7 | nuri | 906 2 | 2 | daytime | thu | 8 | nuri | 906 2 | 2 | daytime | thu | 9 | nuri | 906 2 | 2 | daytime | thu | 10 | nuri | 906 2 | 2 | daytime | thu | 11 | nuri | 906 2 | 2 | daytime | thu | 12 | nuri | 906 2 | 2 | daytime | thu | 13 | nuri | 906
You can easily turn this to an insert query by just preceding it with with an insert instruction, like:
insert into sometable(cid, sid, type, day, time, building, room_number)
with recursive ...
select ...
from ...
where ...
Actually, since the list of values in the csv list is pre-defined, this can very well be done without recursion. You can join the list of days with the table and use find_in_set() to identify the gaps. The upside is that it would work on versions of MySQL / MariaDB that do not support recusive queries:
select t.cid, t.sid, t.type, t.day, at.time, t.building, t.room_number
from (
select 1 time union all select 2 time union all select 3
union all select 4 union all select 5 union all select 6
union all select 7 union all select 8 union all select 9
union all select 10 union all select 11 union all select 12 union all select 13
) at
cross join (
select distinct cid, sid, type, day, building, room_number
from mytable
) t
left join mytable t1 on t1.cid = t.cid and find_in_set(at.time, t1.time)
where t1.cid is null
order by t.cid, at.time
Demo
来源：https://stackoverflow.com/questions/62143110/how-can-i-parse-delimited-list-in-string-to-rows