I want to join tables in tableau but, when I am adding any table to join it is getting joined with the 1st table I added irrespective of keys. Even if tables do not have same keys table is getting joined.
Example: Sales to be joined with OnlineSales and Customer to be joined with OnlineSales but Customer is getting joined with Sales.
I tried using custom SQL as well but it is not working.
Click on the Venn diagram icon for the join and set the left and right keys as desired. You can choose different tables for each join key; you aren't limited to the ones listed first. After you save your join key choices, the diagram will update to show which tables are involved.
No need for custom SQL. In fact, save custom SQL for the rare situation where Tableau gives you no other choice because Custom SQL prevents useful query optimizations.
I want establish a one-to-one relationship between two columns (a program code and a test code) in the same table. I want all tests with the same test code to have the same program code.
My first thought was to use a UDF to find cases where the same test code corresponds to two different programs. I learned that this won't work because t-sql only checks UDFs in check constraints after INSERTS -- not after UPDATES
why is t-sql allowing me to violate a check constraint that uses a UDP?
My next thought was to move the logic from the UDF into the check constraint itself. But t-sql is saying that sub-queries are not allowed in the check constraint. This also means I can't use EXISTS syntax (which I think also uses a sub-query).
ALTER TABLE [dbo].[mytable] WITH CHECK ADD CONSTRAINT [oneProgramPerTest] CHECK
(
(select COUNT(*)
from mydb.dbo.mytable u1
inner join
mydb.dbo.mytable u2 on u1.testcode=u2.testcode and u1.progcode <> u2.progcode
)=0
)
Unless there is some way to enforce this logic without (1) a udf or (2) a subquery then it seems like I need to create a "dummy" table of program codes and then enforce a one-to-one relationship between test codes from myTable and the dummy table. This seems really ugly so there has got to be a better way. Right?
Have you read about normalization (and if you haven't why are you designing a datbase?). You should havea structure with
tableA
id (PK)
programcode
other fields
tableB
programcode (PK)
testcode
Add a formal foreign key between the two tables and define program code as the PK in tableb.
Then to get the data you want:
select <Name specific fields, never use select *>
from tableA a
join tableB b on a.programcode = b.programcode
I have two data buckets using a cryptic naming convention.
Am I able to update the data in the fields on the main table where the record entry is equal to the primary key on the other table?
Something like Table1 has 5 columns, t1A t1B t1C t1D t1E
and Table2 has 2 columns description, and Table1code.
Am I able to switch the data in Table1 with the description field in Table2?
I have tried doing a sql update/case statement but kept getting non-boolean errors when I would run it.
Any help would be appreciated.
You need to do an update with a join. Have a read of this http://www.bennadel.com/blog/938-Using-A-SQL-JOIN-In-A-SQL-UPDATE-Statement-Thanks-John-Eric-.htm
I am new to PostgreSQL. I have 2 databases in PostgreSQL 9.0, db1 and db2, and with db2 I have read only access. I want to create a stored function that would be otherwise easily accomplished with a JOIN or a nested query, something PostgreSQL can't do across databases.
In db1, I have table1 where I can query for a set of foreign keys keys that I can use to search for records in a table2 in db2, something like:
SELECT * from db2.table2 WHERE db2.table2.primary_key IN (
SELECT db1.table1.foreign_key FROM db1.table1 WHERE
db1.table1.primary_key="whatever");
What is the best practice for doing this in Postgres? I can't use a temporary tables in db2, and passing in the foreign keys as a parameter in a stored function running in db2 doesn't seem like a good solution.
Note: the keys are all VARCHAR(11)
You'll want to look into the db_link contrib.
As an aside if you're familiar with C, there also is a cute functionality called foreign data wrappers. It allows to manipulate pretty much any source using plain SQL. Example with Twitter:
SELECT from_user, created_at, text FROM twitter WHERE q = '#postgresql';
I need to create a report that shows, for each master row, two sets of child (detail) rows. In database terms you can think of this as table 'A' having a 1:M relationship to table 'B' and also a 1:M relationship to table 'C'. So, for each row from table 'A', I want to display a list of child rows from 'B' (preferably in one section) and a list of child rows from 'C' (preferably in another section). I would also prefer to avoid the use of sub-reports, if possible. How can I best accomplish this?
Thanks.
I think I understand your question correctly, ie for a given row in Table A, you want the details section(s) to show all connected rows in Table B and then all connected rows in Table C (for any number of rows in B or C, including zero). I only know of two solutions to this, neither of which is straightforward.
The first is, as you've guessed, the disliked subreport option.
The second involves some additional work in the database; specifically, creating a view that combines the entries in Table B and Table C into one table, which can then be used in the main report as a linkable object to report on and you can group on Table A as desired. How straightforward this is will depend on how similar the structures of B and C are. If they were effectively identical, the view could contain something simple like
SELECT 'B' AS DetailType, Field1, Field2, FieldLinkedToTableA
FROM TableB
UNION ALL
SELECT 'C' AS DetailType, Field1, Field2, FieldLinkedToTableA
FROM TAbleC
However, neither option will scale well to reports with lots of results (unless your server supports indexing the view).
This is exactly what Crystal was made for :)
Make a blank .rpt and connect to your data sources as you normally would.
Go to Report->Group Expert and choose your grouping field (aka Index field).
You will now see a Group Header section in your design view. This is your "Master row". The Details sections will be your "Child rows".
In the example image below, this file is grouped by {Client}. For client "ZZZZ", there are 2 records, so there are 2 details sections while all the other clients only have 1 details section.
Edit
Based on your response, how about this:
In your datasource (or perhaps using some kind of intermediary like MS Access), start SQLing as follows.
Make a subquery left joining the primary key of TblA and the foreign key of TblB. Add a third column containing a constant, e.g. "TblB"
Make a subquery left joining the primary key of TblA and the foreign key of TblC. Add a third column containing a different constant, e.g. "TblC"
Union those 2 queries together. That'll be your "index table" of your crystal report.
In Crystal, you can have multiple grouping levels. So group first by your constant column, then by the TblA primary key, then by the foreign key.
This way, all the results from TblB will be displayed first, then TblC. And with a little work, tables B & C won't even have to have the same field definitions.
You can use or create columns that are used for grouping, then group on the table A column, then the table B column, then C. (Crystal's group is not the same as t-sql "group by". In Crystal, it's more of a sorting than a grouping.)