Group By same or similar string sql - sql-server

Question

1) Suppose i have a table like this:-
| id | color_code | fruit |
|:------|--------------|----------------:|
| 1 | 000001 | apple |
| 2 | 000001 | apple |
| 3 | 000001 | apple |
| 4 | 000002 | lemon |
| 5 | 000002 | lemon |
| 6 | 000003 | grapes |
| 7 | 000003 | grapes |
How can i group by the fruit column according to the color_code column in sql server?
like this i suppose:-
| id | color_code | fruit | group_concat(id) |
|:------|--------------|-----------------|---------------------|
| 1 | 000001 | apple | 1,2,3 |
| 4 | 000002 | lemon | 2,5 |
| 6 | 000003 | grapes | 6,7 |
2) What if i have 3 tables (like shown below) which require join, how can i achieve this?
series_no table:
| id | desc_seriesno |
|:------|----------------:|
| 7040 | AU1011 |
| 7041 | AU1022 |
| 7042 | AU1033 |
| 7043 | AU1044 |
| 7044 | AU1055 |
| 7045 | AU1066 |
brand table:
| id | desc_brand |
|:------|----------------:|
| 1020 | Audi |
| 1021 | Bentley |
| 1022 | Ford |
| 1023 | BMW |
| 1024 | Mazda |
| 1025 | Toyota |
car_info table:
| seriesno_id | brand_id | color |
|:---------------|------------|--------:|
| 7040 | 1020 | white |
| 7040 | 1020 | black |
| 7040 | 1020 | pink |
| 7041 | 1021 | yellow |
| 7041 | 1021 | brown |
| 7042 | 1022 | purple |
| 7042 | 1022 | black |
| 7042 | 1022 | green |
| 7043 | 1023 | blue |
| 7044 | 1024 | red |
| 7045 | 1025 | maroon |
| 7045 | 1025 | white |
this is my current query with sql server 2014:-
SELECT SN.id AS seriesid, B.id AS brandid, B.desc_brand
FROM [db1].[dbo].[series_no] SN
LEFT JOIN [db1].[dbo].[car_info] CI
ON CI.seriesno_id = SN.id
RIGHT JOIN [db1].[dbo].[brand] B
ON B.id = CI.brand_id
GROUP BY SN.id, B.id
ORDER BY SN.id ASC
but unfortunately it gave me an error since i cannot group by similar string this way.
i want it to be like this:-
| seriesid | brandid | desc_brand | count |
|:-----------|------------|---------------|-------|
| 7040 | 1020 | Audi | 3 |
| 7041 | 1021 | Bentley | 2 |
| 7042 | 1022 | Ford | 3 |
| 7043 | 1023 | BMW | 1 |
| 7044 | 1024 | Mazda | 1 |
| 7045 | 1025 | Toyota | 2 |

Answer 1

1 Fruit Color
Assuming the table name is FruitColor, you can get the desired output by the following query -
SELECT MIN(id) AS id
, color_code
, fruit
, group_concat_id = STUFF((SELECT ',' + CAST(id AS VARCHAR)
FROM FruitColor AS fci
WHERE fci.fruit = fc.fruit AND fci.color_code = fc.color_code
FOR XML PATH(''), TYPE).value('.', 'NVARCHAR(MAX)'), 1, 1, '')
FROM FruitColor AS fc
GROUP BY color_code, fruit
ORDER BY id;
The MIN() selects the first id of the group.
Since there is no default GROUP_CONCAT function like in MySql in SQL Server, you have to use the STUFF function and FOR XML PATH. To learn more about group concat you can visit this link https://sqlperformance.com/2014/08/t-sql-queries/sql-server-grouped-concatenation
You can customize the WHERE clause to match only by color_code.
2. You can have several options for this -
Option (a): Show counts for series with brands
SELECT seriesno_id AS seriesid, ci.brand_id AS bandid, desc_brand, COUNT(*) AS [count]
FROM db1.dbo.car_info AS ci
LEFT JOIN db1.dbo.brand AS b ON (b.id = ci.brand_id)
GROUP BY seriesno_id, ci.brand_id, desc_brand;
Here you don't need to use the series table if you want to show counts for cars having brand(s).
You may not need to use the RIGHT JOIN on the brand table because if brand table contains a record which
is not in car_info table, then seriesno_id would be null.
Option (b): Show counts for all the series with or without a brand
SELECT sn.id AS seriesid, ci.brand_id AS bandid, desc_brand, COUNT(*) AS [count]
FROM db1.dbo.series_no AS sn
LEFT JOIN db1.dbo.car_info AS ci ON (ci.seriesno_id = sn.id)
LEFT JOIN db1.dbo.brand AS b ON (b.id = ci.brand_id)
GROUP BY sn.id, ci.brand_id, desc_brand;
Option (c): The work around for selecting a column which is not in a GROUP BY
SELECT seriesno_id AS seriesid, ci.brand_id AS bandid, MAX(desc_brand) AS desc_brand, COUNT(*) AS [count]
FROM db1.dbo.car_info AS ci
LEFT JOIN db1.dbo.brand AS b ON (b.id = ci.brand_id)
GROUP BY seriesno_id, ci.brand_id;
Here, if we are certain that each brand contains only one desc_brand, we can use an aggregate on it.
This is bcause applying aggregate only one value returns that value. I used MAX here.
Personally I would go with option (a) as it makes more sense.
Update on GROUP BY exception for desc_brand being NTEXT...
Cast desc_brand to NVARCHAR to avoid the exception.
CAST(desc_brand AS NVARCHAR(200))
Also I highly recommend using VARCHAR / NVARCHAR instead of any TEXT, CHAR etc. because they usually occupy more memory.

Answer 2

SELECT
id = SUBSTRING(group_concat,1,1),
color_code,
fruit,
group_concat
FROM(
SELECT distinct
m.color_code,
m.fruit,
group_concat = STUFF((SELECT ',' + CONVERT(varchar(10),md.id)
FROM [Test_1].[dbo].[Stuff] md
WHERE m.fruit = md.fruit
AND m.color_code = md.color_code
FOR XML PATH(''), TYPE).value('.', 'NVARCHAR(MAX)'), 1, 1, '')
FROM [Test_1].[dbo].[Stuff] m)x

Answer 3

use below code ..
SELECT distinct
m.color_code
, m.fruit
, group_concat = STUFF((
SELECT ',' + CONVERT(varchar(10),md.id)
FROM dbo.tablename md
WHERE m.fruit = md.fruit and m.color_code = md.color_code
FOR XML PATH(''), TYPE).value('.', 'NVARCHAR(MAX)'), 1, 1, '')
FROM dbo.tablename m
for second :
SELECT SN.id AS seriesid, B.id AS brandid, B.desc_brand ,count(*)
FROM [db1].[dbo].[series_no] SN
LEFT JOIN [db1].[dbo].[car_info] CI
ON CI.seriesno_id = SN.id
RIGHT JOIN [db1].[dbo].[brand] B
ON B.id = CI.brand_id
GROUP BY SN.id, B.id ,B.desc_brand
ORDER BY 4 ASC