SELECT group by twice - mysql

I'm not strong in DB at all and I need your help.
I need SQL request with GROUP by twice.
Example of my data in table
<table border="1" style="border-collapse:collapse">
<tr><th>id</th><th>market_id</th><th>price</th><th>low</th><th>high</th><th>symbol</th><th>created_at</th></tr>
<tr><td>1</td><td>1</td><td>5773.8</td><td>5685</td><td>6020</td><td>btcusd</td><td>2017-10-27 16:46:10</td></tr>
<tr><td>2</td><td>1</td><td>0.4274</td><td>0.39</td><td>0.43983</td><td>iotusd</td><td>2017-10-27 16:46:11</td></tr>
<tr><td>3</td><td>1</td><td>0.20026</td><td>0.1986</td><td>0.20352</td><td>xrpusd</td><td>2017-10-27 16:46:12</td></tr>
<tr><td>4</td><td>2</td><td>5771</td><td>5685</td><td>6020</td><td>btcusd</td><td>2017-10-27 16:46:18</td></tr>
<tr><td>5</td><td>2</td><td>0.4274</td><td>0.39</td><td>0.43983</td><td>iotusd</td><td>2017-10-27 16:46:18</td></tr>
<tr><td>6</td><td>2</td><td>0.20026</td><td>0.1986</td><td>0.20352</td><td>xrpusd</td><td>2017-10-27 16:46:19</td></tr>
<tr><td>7</td><td>1</td><td>5773.1</td><td>5685</td><td>6020</td><td>btcusd</td><td>2017-10-27 16:46:25</td></tr>
<tr><td>8</td><td>1</td><td>0.4274</td><td>0.39</td><td>0.43983</td><td>iotusd</td><td>2017-10-27 16:46:25</td></tr>
<tr><td>9</td><td>1</td><td>0.20026</td><td>0.1986</td><td>0.20352</td><td>xrpusd</td><td>2017-10-27 16:46:26</td></tr>
<tr><td>10</td><td>2</td><td>5773.1</td><td>5685</td><td>6020</td><td>btcusd</td><td>2017-10-27 16:46:32</td></tr>
<tr><td>11</td><td>2</td><td>0.42741</td><td>0.39</td><td>0.43983</td><td>iotusd</td><td>2017-10-27 16:46:32</td></tr>
<tr><td>12</td><td>2</td><td>0.20026</td><td>0.1986</td><td>0.20352</td><td>xrpusd</td><td>2017-10-27 16:46:33</td></tr></table>
I would like to get latest data for every market_id and symbol
That's mean I need somethind like that in the end :
- id market_id symbol
- 7 1 btcusd
- 8 1 iotusd
- 9 1 xrpusd
- 10 2 btcusd
- 11 2 iotusd
- 12 2 xrpusd
Really need help, a little bit blocked.

You are almost there. Try this
SELECT c.*
FROM CRYPTO as C
JOIN (
SELECT market_id, symbol, MAX(id) as maxid
FROM CRYPTO
GROUP BY market_id, symbol
) AS C2
ON C2.maxid = C.id and C.market_id = c2.market_id and c.symbol = c2.symbol

Along these lines...
SELECT MAX(id), market_id, symbol
FROM crypto
GROUP BY market_id, symbol

Here's my comment stated as SQL.
SELECT A.ID, A.MarketID, A.Symbol, A.Price, A.Low, A.High
FROM CRYPTO A
INNER JOIN (SELECT max(Created_at) MCA, Market_ID, Symbol
FROM crypto
GROUP BY Market_ID, Symbol) B
on A.Created_At = B.MCA
and A.market_ID = B.Market_ID
and A.Symbol = B.Symbol
What this does:
The derived table (aliased B) generates 1 line for each market_ID and symbol having the max created_at time. It then uses this derived table set to join back to the base set (aliased A) to limit the data to just those having the max created_at. this allows us to show the whole record from A for each unique market_Id and symbol; but only for records having the max created_at.
Other engines would allow you to use a cross apply or an analytic to obtain the desired results.

I tried these requests
SELECT * FROM CRYPTO as C3
JOIN (
SELECT MAX(id) as max
FROM CRYPTO as C1
GROUP BY symbol
) AS C2
ON C2.max = C3.id
SELECT M.id, M.name, R.symbol FROM MARKET AS M
JOIN (
SELECT DISTINCT C.symbol, C.market_id
FROM CRYPTO as C
) as R
ON M.id = R.market_id
But finally I did not find the good combination.

Related

How to determine equaivalent groupings (recursively) in SQL?

I have a list of products identified by their SKUs. To simplify it, I just name them as A, B, C, D,... here. Each of these SKUs has been assigned by default an already existing GroupID, for simplicity I just number them as 1, 2, 3,... here.
The same GroupID would mean "These SKUs are equivalent, so it is ok to use/buy either one of them, as it makes no difference".
The problem is, some SKUs show up more than once as they come from a different buying source, but as they come from a different source, they have a different grouping.
The goal is therefore to consolidate the grouping and make sure they have the same groupings.
I already apologize if my illustration may not be super pretty, but I'm trying. Here's a small data table sample on how the raw data looks like (first line is the column names):
Source SKU GroupID
Seller1 A 1
Seller1 B 1
Seller1 C 1
Seller2 B 2
Seller2 D 2
Seller2 E 2
Seller3 A 3
Seller3 B 3
Seller4 F 4
Seller4 G 4
Seller4 H 4
The result should be like:
Source SKU GroupID
Seller1 A 1
Seller1 B 1
Seller1 C 1
Seller2 B 1
Seller2 D 1
Seller2 E 1
Seller3 A 1
Seller3 B 1
Seller4 F 4
Seller4 G 4
Seller4 H 4
Basically, if Any SKU in GroupID X is a subset of GroupID Y, then GroupID Y = GroupID X. But that should be applied on all GroupIDs, so it appears to be recursive.
I wish I could show the code that I tried already and I tried already for a few days, but I literally only managed to produce garbage.
In C# I'd know how to deal with this, but I can't seem to wrap my head around SQL as I am not that experienced and unfortunately I would need this in SQL.
I would be thankful for any kind of help, even if it's just a hint or direction you guys would suggest I should try. Thanks a lot!
You want a correspondence between groups, which you can calculate with a recursive CTE:
with recursive tt as (
select distinct t1.groupid as groupid1, t2.groupid as groupid2
from t t1 join
t t2
on t1.sku = t2.sku
),
cte as (
select tt.groupid1, tt.groupid2, concat_ws(',', tt.groupid1, tt.groupid2) as visited
from tt
union all
select cte.groupid1, tt.groupid2, concat_ws(',', visited, tt.groupid2)
from cte join
tt
on cte.groupid2 = tt.groupid1
where find_in_set(tt.groupid2, cte.visited) = 0
)
select groupid1, min(groupid2) as overall_group
from cte
group by groupid1;
You can then join this back to the original table to get the "overall group":
with recursive tt as (
select distinct t1.groupid as groupid1, t2.groupid as groupid2
from t t1 join
t t2
on t1.sku = t2.sku
),
cte as (
select tt.groupid1, tt.groupid2, concat_ws(',', tt.groupid1, tt.groupid2) as visited
from tt
union all
select cte.groupid1, tt.groupid2, concat_ws(',', visited, tt.groupid2)
from cte join
tt
on cte.groupid2 = tt.groupid1
where find_in_set(tt.groupid2, cte.visited) = 0
)
select t.*, g.overall_group
from t join
(select groupid1, min(groupid2) as overall_group
from cte
group by groupid1
) g
on t.groupid = g.groupid1;
Here is a db<>fiddle.
Note: Your sample data is rather "complete" so you don't need a recursive CTE for that particular data. However, I am guessing that your real groups have a bit less overlap in which case recursion is necessary.
First is to get all those sellers with subsets based on count. then filter using Group By
select table1.Source, SKU, case when table1.Source = t6.Source and t6.cnt > 1 then 1 else 2 end as GroupID
from table1
left join
(select t5.Source, count(t5.cnt) as cnt from (
select distinct t4.Source, t4.cnt from (
select t3.Source, count(t3.SKU) as cnt from (
select t1.Source, t1.SKU from table1 t1
left join table1 t2 on t2.SKU = t1.SKU ) t3
group by t3.Source, t3.SKU
order by t3.Source) t4) as t5
group by t5.Source) t6 on t6.Source = table1.Source

MYSQL Select max value in multi column in a date

Suppose I have table called "bowlmark" and the summary of table as below link:
(this table is recording the bowling marks for each bowler)
Here is the sample table summary:
bowler_id---record_date---round1mark---round2mark
101---2018-06-02---100---164
102---2018-06-02---102---120
101---2018-06-03---150---124
103---2018-06-03---200---122
I want to output the highest marks in each record day and show the bowler id
For the above example, I want the output to be:
date:2018-06-02, bowler_id:101, highestmark: 164
date:2018-06-03, bowler_id:103, highestmark: 200 .
How to write this SQL query? Thanks.
It's a little ugly, but I was able to do it this way:
SELECT b.bowler_id, a.* FROM (
SELECT record_date, MAX(GREATEST(round1mark, round2mark)) AS greatest
FROM bowlers
GROUP BY record_date
) a
INNER JOIN bowlers b ON
a.record_date = b.record_date AND
a.greatest = GREATEST(b.round1mark, b.round2mark)
SQLFiddle here: http://sqlfiddle.com/#!9/3718ec/17
You could use greatest and an inner join on the subquery for max round mark
select m.bowler_id, m.record_date
, greatest(round, round2, round3, round4, round5, round6) highest_mark
from bowlmark m
inner join (
select record_date
, max(greatest(round, round2, round3, round4, round5, round6)) as max_round
from bowlmark
group by record_date
) t on t.record_date = m.record_date
and greatest(round, round2, round3, round4, round5, round6) = t.max_round
anyway a proper datababase normalization could help you to avoid this kind of problems

Using select from another select in MySQL

I have the following query where I will have finally a 205 patient IDs to work with:
select
patient_id
FROM
visit
WHERE
month(visit.date_of_visit)=3
AND
year(visit.date_of_visit)=2018
AND
visit.visit_status='Active'
GROUP BY patient_id
I want to get all the 205 IDs and run them into other query to see how many diseases we have as cardio-vascular and then as respiratory disease.
My database structure is as the following:
What I want is to get for each patient id, what they are diagnosed at ONLY their first visit to the hospital (so here we will work with min(visit.date_of_visit))
The desired result for `diagnosis_name LIKE '%Cardio%':
E.g>:
Patients: 150 (Or something)
And the query is changed to get the respiratory info.
I tried the following for the Cardio diseases where I use select from select:
SELECT count(*)
FROM
(
select
min(visit.date_of_visit), visit_id, patient_id, count(*) as patientId
FROM
visit
WHERE
month(visit.date_of_visit)=3
AND
year(visit.date_of_visit)=2018
AND
visit.visit_status='Active'
GROUP BY patient_id
) as vid
LEFT JOIN
consultation ON consultation.visit_id=vid.visit_id
LEFT JOIN
diagnosis ON diagnosis.diagnosis_id=consultation.diagnosis_id
WHERE diagnosis.diagnosis_name LIKE '%Cardio%'
The result was: 5 which is a wrong number.
This can be done easily with PHP and MYSQL together but this will exhaust the server by repeating the same query for 205 times and increment a counter. So the desired result should be only done with MySQL.
Data example:
Visit Table
visit_id= 1; date_of_visit=2018-03-03; visit_reason=Active; ...;
patient_id=1234;
visit_id= 2; date_of_visit=2018-03-04; visit_reason=Active; ...;
patient_id=1239;
visit_id= 3; date_of_visit=2018-03-07; visit_reason=Active; ...;
patient_id=1234;
Consultation Table
consultation_id=1; ...; diagnosis_id=12; visit_id=1;...;
consultation_id=2; ...; diagnosis_id=12; visit_id=2;...;
Diagnosis Table
diagnosis_id=12; diagnosis_name: hypertension (cardio disease);
diagnosis_id=13; diagnosis_name: renal disease
By running the query to see patients who came to hospital and that they were diagnosed as having cardio disease in their initial first visit, the result should be in the example as 2 as you can see from the example where patient_id=1234 had 2 visits but I need to know what he had in his first one.
You can use window functions in MySQL 8+. But in older versions you need to calculate the value some other way.
The question for you is what you are counting:
SELECT COUNT(*) as num_diagnoses, COUNT(DISTINCT patient_id) as num_patients
FROM visit v JOIN
(SELECT patient_id,
MIN(v.date_of_visit) as min_dov
FROM visit v
WHERE v.date_of_visit >= '2018-03-01' AND
v.date_of_visit < '2018-04-01' AND
v.visit_status = 'Active'
) vf
ON v.patient_id = vf.patient_id AND v.date_of_visit = vf.min_dov JOIN
consultation c
ON c.visit_id = v.visit_id JOIN
diagnosis d
ON d.diagnosis_id = c.diagnosis_id
WHERE d.diagnosis_name LIKE '%Cardio%';
When working with dates, it is best to compare column values directly to dates, rather than dissecting them.
BRO, it works fine. Test it now on the live scenario.
SELECT count(*)
FROM
(
select
min(visit.date_of_visit) first_date, patient_id, count(*) as patientId
FROM
visit
WHERE
month(visit.date_of_visit)=3
AND
year(visit.date_of_visit)=2018
AND
visit.visit_status='Active'
GROUP BY patient_id
) as vid
INNER JOIN visit b ON
B.patient_id = vid.patient_id AND
B.date_of_visit = vid.first_date and
month(B.date_of_visit)=3 AND
year(B.date_of_visit)=2018 AND
B.visit_reason='Active'
INNER JOIN consultation ON
consultation.visit_id = B.visit_id
INNER JOIN diagnosis ON
diagnosis.diagnosis_id = consultation.diagnosis_id AND
diagnosis.diagnosis_name LIKE '%Cardio%'

Inner query is difficult to write

I have two tables:
customer with schema_id
Schema table has: schema_id, period, amt, updated_date
I need to take join of customer and schema but only retrieve the latest record joined and not the others.
customer table
cust_id name schema_id
1 ABC 1
Schema table
schema_id period amt updated_date
1 1 100 2010-4-1
1 2 150 2011-4-1
If you need the max(updated_date) for each schema_id, then you can use an subquery:
select c.cust_id, c.name, c.schema_id, s.period, s.amt, s.updated_date
from customer c
inner join
(
select s1.schema_id, s1.period, s1.amt, s1.updated_date
from `schemas` s1
inner join
(
select schema_id, max(updated_date) MaxDate
from `schemas`
group by schema_id
) s2
on s1.schema_id = s2.schema_id
and s1.updated_date = s2.maxdate
) s
on c.schema_id = s.schema_id
See SQL Fiddle with Demo
The subquery is then used in a join back to your table to return the rows that have the matching date and schema_id.
If I understood your problem, you need to take lastest register of the "schema".
I think you need to use max() function. So, try the query below:
select *
from customer c,
schema s
where c.schema_id = s.schema_id
and s.updated_date = ( select max(s2.updated_date)
from schema s2
where s2.schema_id = s.schema_id
)
Regards!
Edmilton

MS Access - Aggregate functions and max

I have this query, where I am trying to get max age of a retail store seller(There's multiple towns), and show multiple if there's multiple people with the same (max)age. I am using Microsoft Access 2010. Here is the query:
SELECT Linnad.Linn, Myyjad.Nimi, Max(Myyjad.Vanus) As Vanus
FROM Linnad INNER JOIN Myyjad ON Linnad.LinnID = Myyjad.LinnID
GROUP BY Linnad.Linn, Myyjad.Nimi
ORDER BY Linnad.Linn;
The problem is, it seems to ignore the MAX, and just shows all of the values, and I can't remove the group by Myyjad.Nimi, because it gives me an error that aggregate function not included for Myyjad.Nimi.
And the output should be:
Town - Name - Max(Age)
Also, Linn = Town, Nimi = Name and the Vanus = Age.
I think this may be what your looking for:
SELECT L.Linn, M.Nimi, M.Vanus
FROM Linnad As L,
(
SELECT M2.LinnID, M2.Nimi, M2.Vanus
FROM Myyjad As M2
WHERE M2.Vanus = (SELECT Max(Z.Vanus) FROM Myyjad As Z WHERE Z.LinnID = M2.LinnID)
) As M
WHERE M.LinnID = L.LinnID
This performs a sub-select to get a list of the Linn ID's with all Nimi's showing the maximum Vanus, then we link this sub-select back to the Linnad table via the LinnID.
I think you want:
SELECT Linnad.Linn, Myyjad.Nimi, Myyjad.Vanus
FROM Linnad INNER JOIN Myyjad ON Linnad.LinnID = Myyjad.LinnID
WHERE DateValue(Myyjad.Vanus)
= (SELECT Max(DateValue(Myyjad.Vanus)) FROM Myyjad)
ORDER BY Linnad.Linn
Top N per group:
SELECT Linnad.Linn, Myyjad.Nimi, Myyjad.Vanus
FROM FROM Linnad INNER JOIN Myyjad ON Linnad.LinnID = Myyjad.LinnID
WHERE Myyjad.ID In (
SELECT Top 1 m.ID
FROM Myyjad m
WHERE m.LinnID=Linnad.ID
ORDER BY m.Vanus Desc, m.ID)
Grouping by Linn (town) and Nimi (name) tells the db engine to give you one row for each combination of town and name, and show you the maximum Vanus (age) for each of those combinations. And logically, that's not what you want. You want the name of each person whose age is the same as the maximum age in that town.
First verify you can retrieve the max age for each LinnID.
SELECT
LinnID,
Max(Vanus) As MaxOfVanus
FROM
Myyjad
GROUP BY LinnID;
If that works, you can save it as "qryTownAge", then use it in another query where you join it (on LinnID) with Linnad. That will allow you to retrieve the matching Linn.
SELECT l.LinnID, l.Linn, q.MaxOfVanus
FROM
Linnad AS l
INNER JOIN qryTownAge AS q
ON l.LinnID = q.LinnID
ORDER BY l.Linn;
If that works, save it as qryTownAge2. Then try this query.
SELECT q.Linn, q.MaxOfVanus, m.Nimi
FROM
qryTownAge2 AS q
INNER JOIN Myyjad AS m
ON (
m.LinnID = q.LinnID
AND m.Vanus = q.MaxOfVanus
)
ORDER BY q.Linn;
If that all works, you could create a single query which does it all. However, doing it step by step should help us pinpoint errors.