Typed Dataset Design Question

[Garrek]

I'm curious as to opinions on the use of Typed Datasets with a database
larger than but a handful of tables as provided in most samples.

Say I have a database with 100 tables. I'll break these 100 tables
into logical groups such as User tables, Customer tables, Order tables,
Product tables, Vendor tables, and others.

Is there a recommended approach to any of the following:

1. Create one massive Typed Dataset that represents all tables in your
system. Whether you need it to just store a new customer or create a
complete chain of events from a customer ordering a new product from a
new vendors, you will always use this single Typed Dataset.

One of the possible issues with this scenario is that I may want to
read partial data. If I created a massive Typed Dataset with relations
between each table I'll have to worry about turning some 'off' to push
the data in correctly. Sounds like a nightmare.

2. Create multiple Typed Datasets broken into the logical groups as
stated above. For instance, if there are four tables to describe a
Customer then you would create a single Typed Dataset of those four
tables. One point of confusion, in my opinion with this method, is how
do you decide where to place a table linking many Customers to an Order
(in this scenario the groups of tables for Orders would be in a
separate Typed Dataset).

3. What about Typed Datasets for reporting? Or in this case is the
consensus to stick with an Untyped Dataset? For instance, I want a
report of all orders complete with some min/max values, averages, etc.
I don't believe it would be appropriate to use a Typed Dataset created
in method #1 or #2 mentioned above. In my opinion the Typed Datasets
defined in option #1 or #2 are for transactional purposes; not
reporting.

Any feedback on this topic would be appreciated. I understand there is
no magic bullet. I've used Typed Dataset for a project and haven't
been as excited about their use as I had hoped. I'm wondering if
perhaps I did it improperly or... there is a better method. =)

Thanks.

 

[David Browne]

I'm curious as to opinions on the use of Typed Datasets with a database
larger than but a handful of tables as provided in most samples.

Say I have a database with 100 tables. I'll break these 100 tables into
logical groups such as User tables, Customer tables, Order tables, Product
tables, Vendor tables, and others.

Is there a recommended approach to any of the following:

1. Create one massive Typed Dataset that represents all tables in your
system. Whether you need it to just store a new customer or create a
complete chain of events from a customer ordering a new product from a new
vendors, you will always use this single Typed Dataset.

One of the possible issues with this scenario is that I may want to read
partial data. If I created a massive Typed Dataset with relations between
each table I'll have to worry about turning some 'off' to push the data in
correctly. Sounds like a nightmare.

2. Create multiple Typed Datasets broken into the logical groups as stated
above. For instance, if there are four tables to describe a Customer then
you would create a single Typed Dataset of those four tables. One point
of confusion, in my opinion with this method, is how do you decide where
to place a table linking many Customers to an Order (in this scenario the
groups of tables for Orders would be in a separate Typed Dataset).

Close. Tables should be groped into DataSets by Use Case, and a table may
appear in more than one Typed Dataset.

The DataSet supporting order entry might contain
CUSTOMER,ORDER,ORDER_DETAIL.

The DataSet suporting customer profile maintence might contain CUSTOMER,
CUSTOMER_CONTACT, CUSTOMER_SHIPPING, etc.

The DataSet supporting fulfillment might contain ORDER, ORDER_DETAIL,
PRODUCT, PRODUCT_PART, etc.

For each Use Case, decide what the main table(s) you need to update, and
then add just enough other tables to make you life easier.

3. What about Typed Datasets for reporting? Or in this case is the
consensus to stick with an Untyped Dataset? For instance, I want a report
of all orders complete with some min/max values, averages, etc. I don't
believe it would be appropriate to use a Typed Dataset created in method
#1 or #2 mentioned above. In my opinion the Typed Datasets defined in
option #1 or #2 are for transactional purposes; not reporting.

The _other_ use for Typed Datasets is data binding. If you use a Typed
DataSet you can do visual layout using the column names and set column
formatting in the UI. This would be the only reason to use Typed Datasets
for reporting.

David