Hello,
The following SQL code produces the famous multiple cascading paths problem.
How should I design the tables to have the below functionality, but keep the
cascading paths? A Doc doesn't necessarily have to be related to a Folder,
but must be related to a Cust.
Changing ON UPDATE to NO ACTION would solve it partly, but it just doesn't
feel right.
Thanks for any help!
cheers,
Jonah
CREATE TABLE Cust (
usr_name varchar(20) NOT NULL,
usr_pwd varchar(40) NOT NULL,
customer_name nvarchar(50) NOT NULL,
created_date datetime default getdate() NOT NULL,
change_date datetime default getdate() NOT NULL,
deactivate_date datetime default getdate() NULL
) ON [PRIMARY]
GO
ALTER TABLE Cust ADD CONSTRAINT
PK_Cust PRIMARY KEY CLUSTERED
(
usr_name
) ON [PRIMARY]
GO
CREATE TABLE Folders (
folder_id int NOT NULL ,
folder_name nvarchar(20) NOT NULL ,
folder_description nvarchar(150) NULL ,
usr_name varchar(20) NOT NULL
) ON [PRIMARY]
GO
ALTER TABLE Folders ADD CONSTRAINT
PK_Folders PRIMARY KEY CLUSTERED
(
folder_id
) ON [PRIMARY]
GO
ALTER TABLE Folders ADD CONSTRAINT
FK_Folders_Cust FOREIGN KEY
(
usr_name
) REFERENCES Cust
(
usr_name
) ON UPDATE CASCADE
GO
CREATE TABLE Docs (
doc_id int NOT NULL,
header nvarchar(255) not null,
created_date datetime default getdate() NOT NULL,
updated_date datetime default getdate() NOT NULL,
usr_name varchar(20) NOT NULL
) ON [PRIMARY]
GO
ALTER TABLE Docs ADD CONSTRAINT
PK_Docs PRIMARY KEY CLUSTERED
(
doc_id
) ON [PRIMARY]
GO
ALTER TABLE Docs ADD CONSTRAINT
FK_Cust_Docs FOREIGN KEY
(
usr_name
) REFERENCES Cust
(
usr_name
) ON UPDATE CASCADE
ON DELETE NO ACTION
GO
CREATE TABLE DocsInFolders
(
doc_id int NOT NULL,
folder_id int NOT NULL
) ON [PRIMARY]
GO
ALTER TABLE DocsInFolders ADD CONSTRAINT
PK_DocsInFolders PRIMARY KEY CLUSTERED
(
doc_id,
folder_id
) ON [PRIMARY]
GO
ALTER TABLE DocsInFolders ADD CONSTRAINT
FK_DocsInFolders_Folders FOREIGN KEY
(
folder_id
) REFERENCES Folders
(
folder_id
) ON UPDATE CASCADE
ON DELETE CASCADE
GO
ALTER TABLE DocsInFolders ADD CONSTRAINT
FK_DocsInFolders_Docs FOREIGN KEY
(
doc_id
) REFERENCES Docs
(
doc_id
) ON UPDATE CASCADE
ON DELETE CASCADE
GOYou say 'A Doc *doesn't necessarily have to be* related to a Folder,
but *must be* related to a Cust.'
Keep cascading on FK_DocsInFolders_Docs, but manage the Cust<--Docs
relationship in procedures. You make it sound as if the importance of the
Cust<--Docs relationship supercedes the importance of the Folders<--Docs
relationship.
Vital relationships should not be cascading.
ML|||Correct. The Cust<--Docs relationship is more importance, but that's why I
have only ON UPDATE CASCADE (to make possible usr_name changes up to date)
and not ON DELETE because I don't want Custs having Docs deleted by mistake.
An SP takes care of that.
So what you are saying is that I should change the ON CHANGE to NO ACTION as
well?
/Jonah
"ML" <ML@.discussions.microsoft.com> skrev i meddelandet
news:601F4976-E669-4865-9D92-FBAF9A16CA8F@.microsoft.com...
> You say 'A Doc *doesn't necessarily have to be* related to a Folder,
> but *must be* related to a Cust.'
> Keep cascading on FK_DocsInFolders_Docs, but manage the Cust<--Docs
> relationship in procedures. You make it sound as if the importance of the
> Cust<--Docs relationship supercedes the importance of the Folders<--Docs
> relationship.
> Vital relationships should not be cascading.
>
> ML|||If usr_name can be changed then using it as a primary key (and/or referencin
g
it from a foreign key table) is really bad practice. Either disallow usr_nam
e
changes or use a better kandidate key.
I wouldn't allow cascades for this one.
ML|||OK. So if I disallow cascades for usr_name, you would consider the design to
be correct?
/Jonah
"ML" <ML@.discussions.microsoft.com> skrev i meddelandet
news:90F300A3-7104-4924-A594-E28AC05D645D@.microsoft.com...
> If usr_name can be changed then using it as a primary key (and/or
> referencing
> it from a foreign key table) is really bad practice. Either disallow
> usr_name
> changes or use a better kandidate key.
> I wouldn't allow cascades for this one.
>
> ML|||As far as I can see, the design is fine. I would, however, do something abou
t
the Folders and Docs entities. Right now you allow a single document to exis
t
in more than one folder, which can lead to problems. The same goes for
folders - you should focus on preventing circular references.
Oh, and if any given document cannot exist in more than one folder, then the
DocsInFolders table is obsolete. You could simply add a nullable folder_id
foreign key to the Docs table (nullable since you've mentioned that a
document need not exist in any folder).
I hope you started on paper. :) And in case you haven't, maybe you'll do it
next time.
ML|||My design question applied to the cascading paths, not the business rules
themselves. One Doc may actually exist in several Folders.
- One Cust may have zero or more Folders
- One Cust may have zero or more Docs not connected to a Folder
- One Folder may have zero or more Docs related
Thus, my question was only related to if there's a better way of designing
the relations and tables to avoid circular references, which now occurs.
FYI, I didn't start on paper. I use Visio.
Thank you,
Jonah
"ML" <ML@.discussions.microsoft.com> skrev i meddelandet
news:F981A87A-893F-4940-9EBD-566C385CF1CB@.microsoft.com...
> As far as I can see, the design is fine. I would, however, do something
> about
> the Folders and Docs entities. Right now you allow a single document to
> exist
> in more than one folder, which can lead to problems. The same goes for
> folders - you should focus on preventing circular references.
> Oh, and if any given document cannot exist in more than one folder, then
> the
> DocsInFolders table is obsolete. You could simply add a nullable folder_id
> foreign key to the Docs table (nullable since you've mentioned that a
> document need not exist in any folder).
> I hope you started on paper. :) And in case you haven't, maybe you'll do
> it
> next time.
>
> ML|||No pun intended.
You are right - there is a better way to avoid circular references. You
might find more answers studying trees and hierarchies. Consider this model:
ItemInstance : ItemID : BelongsToInstance : ItemType
ItemInstance is unique.
ItemID can be either cust_id, folder_id or doc_id.
BelongsToInstance is a foreign key referencing ItemInstance.
ItemType designates whether ItemID is customer, folder or document.
Valid relationships are:
1) Customer/Folder/Document
2) Customer/Document
3) Folder/Folder (<-- not sure about this one, but seems logical, however:
parent folder_id must should be equal to child folder_id).
A customer can only exist as a root element (BelongsToInstance is null).
A Document can only exist as a leaf element (its ItemInstance is never
referenced in a BelongsToInstance).
The above constraints could be reinforced through the use of indexed views.
ItemInstance and BelongsToInstance prevent circular references while still
allowing all possible relationships between the three entities.
ML|||Thank you for your detailed answer.
I do have a copy of a trees and hierarchies book which I could take a closer
look into (Joe Celko's Trees and Hierarchies in SQL for Smarties). Maybe I
can find some more answers and examples there.
/Jonah
"ML" <ML@.discussions.microsoft.com> skrev i meddelandet
news:D5B9F5D8-4790-4088-BF18-14DE5119891D@.microsoft.com...
> No pun intended.
> You are right - there is a better way to avoid circular references. You
> might find more answers studying trees and hierarchies. Consider this
> model:
> ItemInstance : ItemID : BelongsToInstance : ItemType
> ItemInstance is unique.
> ItemID can be either cust_id, folder_id or doc_id.
> BelongsToInstance is a foreign key referencing ItemInstance.
> ItemType designates whether ItemID is customer, folder or document.
> Valid relationships are:
> 1) Customer/Folder/Document
> 2) Customer/Document
> 3) Folder/Folder (<-- not sure about this one, but seems logical, however:
> parent folder_id must should be equal to child folder_id).
> A customer can only exist as a root element (BelongsToInstance is null).
> A Document can only exist as a leaf element (its ItemInstance is never
> referenced in a BelongsToInstance).
> The above constraints could be reinforced through the use of indexed
> views.
> ItemInstance and BelongsToInstance prevent circular references while still
> allowing all possible relationships between the three entities.
>
> ML|||What about using a nested sets model for the hierarchy? I am not a big
fanof misxed node trees, but this case is pretty easy:
1) If it is the root node, it is a Customer
2) If it is a leaf node, it is a document
3) other it is a folder
CREATE TABLE DocumentHierarchy
(customer_id INTEGER NOT NULL
REFERENCES Customers
ON UPDATE CASCADE
ON DELETE CASCADE,
folder_id INTEGER -- null means no folder
REFERENCES Folders (folder_id)
ON UPDATE CASCADE
ON DELETE CASCADE,
document_id INTEGER -- null means no document
REFERENCES Documents(doc_id)
ON UPDATE CASCADE
ON DELETE CASCADE,
lft INTEGER NOT NULL CHECK (lft >0),
rgt INTEGER NOT NULL CHECK (rgt >lft),
PRIMARY KEY (customer_id, lft, rgt));
untested. You can also get a copy of TREES & HIERARCHIES IN SQL for
more ideas.
