Question

I have several tables in my ERD which which I would like to combine in a relational manner.

I have several use cases, but I completely lost track of what kind of relations to use between the tables.

• Every user can work on multiple projects.
• Every user has one specific role per project (Manager, Contributor, User)
• Every project has multiple datasets (currDataXXX columns in ‘projects’) which need to be linked to the table data.
• The application will keep track of datasets that have been added by users. Thus I assume I need a relation between tables ‘users’ and ‘data’ too?

I used a bridge model in table ‘roles’ with 2 PK’s to link the users and projects together and defining a role for that user and project at the same time (is this the correct way?).

Could somebody please help me assist the correct relations between the tables? and maybe suggest columns (which are missing) for tables (or tables as a whole of course).

Kind of lost sight of this.

With kind regards,

B.

REVISED ERD:

(Original image: http://i55.tinypic.com/2mq2ejs.jpg)

Answer 1

Every project has multiple datasets (currDataXXX columns in ‘projects’)

This means that there is a one-to-many relationship from project to dataset. As such, the project ID should be a foreign key on the dataset table, while the currDataXX columns should be removed from the project table. (The current design represents a denormalisation.)

Having done this, you now have two many-to-many relationships between project and user – one with a link entity of roles, and one with a link entity of dataset. Such relationships are normally redundant – in this case, I would assume that a) only users with a role on a project can add datasets, and b) that a single user can add many datasets for a single project.

If both of these assumptions are correct, then there is actually a one to many relationship from role to dataset, and therefore the role ID should be a foreign key on the dataset table, while the user and project IDs become redundant on the dataset table and can be removed. Like this:

+----------+    +----------+    +----------+
|   User   +---<|   Role   |>---+  Project |
+----------+    +----+-----+    +----------+
                     |
                    /|\
                +----------+
                |  Dataset |
                +----------+

[If assumption b) is incorrect, then role and dataset can be combined into a single table, while if assumption a) is incorrect then role and dataset remain two distinct linking entities, unrelated to each other.]

EDIT – updated proposed structure, following Rhizosis’ edits:

                +----------+
                |   Role   |
                +----+-----+
                     |
                    /|\
+----------+    +----------+    +----------+
|   User   +---<|  Users/  |>---+  Project |
|          |    | Projects |    |          |
+----------+    +----+-----+    +----------+
                     |
                    /|\
                +----------+
                |  Dataset |
                +----------+