MYSQL Alias join NOT EXISTS - mysql

Question

I have two tables. In TableNames the library of several songs is stored. In TableTimes the exact date and time on which the song will be played is stored.
In every new week I'm checking if there are new songs which were never played before. Easy. But my problem is the following: I need to know if a song will be played before or after 12:00:00 (24h). Therefore I'm running a CASE WHEN and output an alias TimeCase with value + or -, please see query below.
My query should output if there are songs in the actual week which has a new "TimeCase". You can find an example below with ID 101.
ID 101 is not new in Week 201908 but it plays the first time after 12:00:00 and therefore needs to be in my query output.
TableNames
+-------+-----+
| Name | id |
+-------+-----+
| Song1 | 100 |
+-------+-----+
| Song2 | 101 |
+-------+-----+
| Song3 | 102 |
+-------+-----+
| Song4 | 103 |
TableTimes:
+--------+--------+------+
| Week | Time | idFI |
+--------+--------+------+
| 201908 | 08:00 | 100 |
+--------+--------+------+
| 201908 | 19:00 | 101 |
+--------+--------+------+
| 201907 | 09:00 | 101 |
+--------+--------+------+
| 201906 | 22:00 | 103 |
My Query:
SELECT t2.idFI, t1.id, Name, TimeCase
FROM TableTimes t2
JOIN TableNames t1 ON t2.idFI = t1.id
JOIN (
SELECT idFI,
CASE WHEN Time < '12:00:00' THEN '-'
ELSE '+'
END AS TimeCase
FROM TableTimes GROUP BY idFI, TimeCase
) t3 ON t2.idFI = t3.idFI
WHERE '201908' = Week
AND deleteSZ = false
AND NOT EXISTS
(
SELECT null
FROM TableTimes t5
JOIN (
SELECT idFI,
CASE WHEN Time < '12:00:00' THEN '-'
ELSE '+'
END AS TimeCaseBefore
FROM TableTimes GROUP BY idFI, TimeCaseBefore
) t4 ON t5.idFI = t4.idFI
WHERE t5.idFI = t2.idFI
AND TimeCaseBefore = TimeCase
AND Week<'201908'
AND deleteSZ = false
GROUP BY t5.idFI, TimeCaseBefore)
GROUP BY TimeCase, t2.idFI
Which is giving me the following:
+-------+-----+------+----------+
| Name | id | idFI | TimeCase |
+-------+-----+------+----------+
| Song1 | 100 | 100 | - |
+-------+-----+------+----------+
What I actually want is:
+-------+-----+------+----------+
| Name | id | idFI | TimeCase |
+-------+-----+------+----------+
| Song1 | 100 | 100 | - |
+-------+-----+------+----------+
| Song2 | 101 | 101 | + |
+-------+-----+------+----------+
My understanding of my problem is, that I should be able to change t5.idFI = t2.idFI to t5.TimeCaseBefore = t2.TimeCase which is not possible due to the alias.
How can I solve this?
Final solutions based on P.Salmon:
select tn.name,
idfi,
case when thiswkafter12 > 0 and priorwkafter12 = 0 and thiswkbefore12 > 0 and priorwkbefore12 = 0 then '+ -'
when thiswkafter12 > 0 and priorwkafter12 = 0 and priorwkbefore12 = 0 then '+'
when thiswkbefore12 > 0 and priorwkbefore12 = 0 and priorwkafter12 = 0 then '-'
end as timecase
from
(
SELECT idfi,
sum(case when week = 201908 and time >= '12:00:00' then 1 else 0 end) as thiswkafter12,
sum(case when week = 201908 and time < '12:00:00' then 1 else 0 end) as thiswkbefore12,
sum(case when week < 201908 and time >= '12:00:00' then 1 else 0 end) as priorwkafter12,
sum(case when week < 201908 and time < '12:00:00' then 1 else 0 end) as priorwkbefore12
FROM TABLETIMES
group by idfi
having (thiswkafter12 > 0 and priorwkafter12 = 0 and thiswkbefore12 > 0 and priorwkbefore12 = 0) or
(thiswkafter12 > 0 and priorwkafter12 = 0 and priorwkbefore12 = 0) or
(thiswkbefore12 > 0 and priorwkbefore12 = 0 and priorwkafter12 = 0)
) s
join tablenames tn on tn.id = s.idfi ;

Answer 1

Maybe a simpler approach working out how often a song has been played this week and comparing to how often played in previous weeks, split into am and pm. I assume that the inverse of your example for 101 can also apply ie a case where a song has been played in the am this week but was played in the pm previously.
Note I have added a song played for the first time in the am AND pm this week
+--------+----------+------+
| Week | Time | idFI |
+--------+----------+------+
| 201908 | 08:00:00 | 100 |
| 201908 | 19:00:00 | 101 |
| 201907 | 09:00:00 | 101 |
| 201906 | 22:00:00 | 103 |
| 201908 | 08:00:00 | 104 |
| 201908 | 13:00:00 | 104 |
| 201908 | 08:00:00 | 104 |
| 201908 | 13:00:00 | 104 |
+--------+----------+------+
8 rows in set (0.00 sec)
select tn.name,
idfi,
case when thiswkafter12 > 0 and priorwkafter12 = 0 then '+'
when thiswkbefore12 > 0 and priorwkbefore12 = 0 then '-'
end as timecase
from
(
SELECT idfi,
sum(case when week = 201908 and time >= '12:00:00' then 1 else 0 end) as thiswkafter12,
sum(case when week = 201908 and time < '12:00:00' then 1 else 0 end) as thiswkbefore12,
sum(case when week < 201908 and time >= '12:00:00' then 1 else 0 end) as priorwkafter12,
sum(case when week < 201908 and time < '12:00:00' then 1 else 0 end) as priorwkbefore12
FROM TABLETIMES
group by idfi
having (thiswkafter12 > 0 and priorwkafter12 = 0) or
(thiswkbefore12 > 0 and priorwkbefore12 = 0)
) s
join tablenames tn on tn.id = s.idfi ;
+-------+------+----------+
| name | idfi | timecase |
+-------+------+----------+
| Song1 | 100 | - |
| Song2 | 101 | + |
| song5 | 104 | + |
+-------+------+----------+
3 rows in set (0.00 sec)
You haven't said what should happen in the case of 104 so I'm assumming pm wins.