Subquery table based on sys.tables - sql-server

Question

I'm trying to retrieve all primary keys iin a fact table and then count the number of records in that fact table grouped by that id, however so far i can only seem to get all the primary keys and its talbe. i guess i need to make some kind of subquery
SELECT
tab1.name AS [FactTable],
col1.name AS [PrimaryKey]
FROM sys.indexes ind1
INNER JOIN sys.tables tab1
ON tab1.object_id = ind1.object_id
INNER JOIN sys.schemas sch1
ON tab1.schema_id = sch1.schema_id
INNER JOIN sys.columns col1
ON col1.object_id = tab1.object_id AND col1.name like '%Id'
WHERE tab1.name like 'Fact%' AND ind1.is_primary_key = 1
sample output
primaryKey countRecordsGroupedByPrimaryKey
2 4000
3 8343
4 203
1 4023

Answer 1

I going out on a limb and guessing that you want to count the number of each dimensionPK used in your fact table. If a fact table has references to two different dimensions, you need two different statements to count the usage by that dimension.
For the query below, provide your fact table name and schema and it should generate a count statement joining your fact and dim and grouping by the join keys.
- If a fact table has two different FK relationships, you'll get two different statements.
- If a table uses a composite PK, both key columns will be included in the join
This is complicated and I don't have any tables with multiple FKs to test it on, so please let me know if it does what you want.
DECLARE #NameOfTableWithFKs sysname = 'your fact table name',
#SchemaOfTableWithFKs sysname = 'dbo';
WITH JoinColumns
AS (SELECT QUOTENAME(OBJECT_SCHEMA_NAME(parent.object_id)) + '.' + QUOTENAME(OBJECT_NAME(parent.object_id)) AS ParentTableName,
QUOTENAME(OBJECT_SCHEMA_NAME(referenced.object_id)) + '.' + QUOTENAME(OBJECT_NAME(referenced.object_id)) AS ReferencedTableName,
QUOTENAME(OBJECT_NAME(parent.object_id)) + '.' + QUOTENAME(parent.name) + ' = ' + QUOTENAME(OBJECT_NAME(referenced.object_id)) + '.'
+ QUOTENAME(referenced.name) AS JoinColumn,
QUOTENAME(OBJECT_NAME(referenced.object_id)) + '.' + QUOTENAME(referenced.name) AS GroupingColumn
FROM sys.foreign_key_columns AS fkc
INNER JOIN sys.columns AS parent
ON parent.object_id = fkc.parent_object_id
AND parent.column_id = fkc.parent_column_id
INNER JOIN sys.columns AS referenced
ON referenced.object_id = fkc.referenced_object_id
AND referenced.column_id = fkc.referenced_column_id
WHERE OBJECT_NAME(parent.object_id) = #NameOfTableWithFKs
AND OBJECT_SCHEMA_NAME(parent.object_id) = #SchemaOfTableWithFKs
),
JoinTables
AS (SELECT QUOTENAME(OBJECT_SCHEMA_NAME(tbl.object_id)) + '.' + QUOTENAME(OBJECT_NAME(tbl.object_id)) AS ParentTableName,
QUOTENAME(OBJECT_SCHEMA_NAME(rtbl.object_id)) + '.' + QUOTENAME(OBJECT_NAME(rtbl.object_id)) AS ReferencedTableName
FROM sys.tables AS tbl
INNER JOIN sys.foreign_keys AS cstr
ON cstr.parent_object_id = tbl.object_id
INNER JOIN sys.tables AS rtbl
ON rtbl.object_id = cstr.referenced_object_id
WHERE OBJECT_NAME(tbl.object_id) = #NameOfTableWithFKs
AND OBJECT_SCHEMA_NAME(tbl.object_id) = #SchemaOfTableWithFKs
)
SELECT 'SELECT Count(*)' + ( SELECT ', ' + JC.GroupingColumn
FROM JoinColumns AS JC
WHERE JC.ParentTableName = jt.ParentTableName
AND JC.ReferencedTableName = jt.ReferencedTableName
FOR XML PATH('')
) + ' FROM ' + JT.ParentTableName + ' INNER JOIN ' + JT.ReferencedTableName + ' ON'
+ SUBSTRING(( SELECT ' AND ' + JC.JoinColumn
FROM JoinColumns AS JC
WHERE JC.ParentTableName = JT.ParentTableName
AND JC.ReferencedTableName = JT.ReferencedTableName
FOR XML PATH('')
), 5, 8000
) + ' GROUP BY ' + SUBSTRING(( SELECT ', ' + JC.GroupingColumn
FROM JoinColumns AS JC
WHERE JC.ParentTableName = JT.ParentTableName
AND JC.ReferencedTableName = JT.ReferencedTableName
FOR XML PATH('')
), 2, 8000
)
FROM JoinTables AS JT;

Answer 2

If I understand the question correctly you are wanting to count all the rows in some tables based on some criteria. Not really sure why you care about the primary key portion since by definition a primary key must be unique so this could still be simplified to not check for primary key but whatever.
I did remove the join to sys.columns because why does it matter the name of the column unless you want only those table named Fact% and has a column named %Id.
This should get you pretty close as I understand it.
declare #SQL nvarchar(max) = ''
select #SQL = #SQL + 'select TableName = ''' + tab1.name + ''', NumRows = count(*) from ' + QUOTENAME(sch1.name) + '.' + QUOTENAME(tab1.name) + ' UNION ALL '
FROM sys.indexes ind1
INNER JOIN sys.tables tab1
ON tab1.object_id = ind1.object_id
INNER JOIN sys.schemas sch1
ON tab1.schema_id = sch1.schema_id
--INNER JOIN sys.columns col1
-- ON col1.object_id = tab1.object_id AND col1.name like '%Id'
WHERE tab1.name like 'Fact%'
AND ind1.is_primary_key = 1
select #SQL = LEFT(#SQL, LEN(#SQL) - 10) + ' ORDER BY TableName'
select #SQL --uncomment the exec line below once you are comfortable that the dynamic sql is what you want.
--exec sp_executesql #SQL