I have been working on a database design and I'm stuck hitting a wall. I'm ending up with what I'm reading is not a normalized database structure but I'm having issues trying to find a "more correct" design and if this design is acceptable how do I execute it in Access?
TLDR: If a table with a single column set as an auto number is an acceptable design, how do you go about inserting a record in it using Access?
The segment of the database of concern is creating a structure for storing companies. Requirements for this is that any changes need to be approved by another user and all historical changes need to be captured so that it can be easily reverted also a company can have multiple aliases but only one legal name.
There is three tables in my solution but one of them is a single column table. From what I've read 95% of people on stack overflow all think its a very bad idea but I've found one post were people are that there are cases for it. I think this is not normal also because I can't find a way to just create a new record in a table with only an auto number column (In Access I have not tried others yet).
Table Structure
Company Names : ID, Company ID, Is Legal Name, Created By, Created On, Approved On, Approved By, Event ID, Is Active
(A company could have a few different names known to the public: TD vs Toronto Dominion. Each name is inserted here with a reference to the company it belongs to)
Companies : ID (Auto Number)
(A company exists and this is its ID)
Companies History : ID, Company ID, Market ID, Holding Company ID, Created By, Created On, Approved On, Approved By, Event ID, Is Active
(These are the historical changes that have been made to the company and who did them and who approved them)
Column Notes:
Event ID : is a FK reference to a table holding each record of actions that have either created, updated or deleted records. (User Research using method [y], Typo Fix, ...)
Is Active : Since deleting records is not possible (historical records need to be kept) this column is used to track if this record is to be included in queries.
Options I see and their issues:
I could get rid of the companies table and make Companies History : ID be the new company id but I find that in that case each time I want to update a company I would need to update each FK reference to the previous company id (I don't think this would be a very normalized approach)
Another Option I see is that I get rid of Companies table and use Company Names : ID as the company id and I would add a column to Company Names called Alias of Company ID. I find that solution adds a log of complexity to my stored data where an alias has company information that differs from the entry that was aliased.
Another Option is that I could add the columns: Created By, Created On, Approved On, Approved By, Event ID and Is Active but this would be duplicating information found in the first record for this company in the Companies History table and this isn't adding any real description to this record.
Anther Option is that I make the Companies table a mirror of Companies History and that when I update or insert a record in Companies I would also insert a record Companies History. With this solution I find that again I duplicate information, that newest record in "Companies History" would hold the same information found in last Inserted or Updated record in in Companies
Another option but is to replace the Companies : ID auto number with a short text and I just get the hash of the current timestamp + a random int. I can now insert new records into this table using access but I feel that this is overkill since I just need the exact same functionality as the auto number.
Another option is move only the legal name into Companies table but now when the legal name of a company changes I have no way of tracking this. Also if I want a list of all names I need to use a union on Companies and Company Names. I find that using unions can reduce performances of queries and I use them only when explicitly needed.
If I don't want to duplicate any information and I don't want to update all FK it seems that I need a table with a single column. If this is acceptable how do I go about inserting a record into a table with a single column set to auto number in Access.
If Companies can be derived from CompanyNames (select distinct CompanyId from CompanyNames), there is no point storing again that information. Just replace that table by a view if you want it (but it as little added value).
On the other hand, if CreatedOn refers to the Company creation (not the row creation) then it is obviously a property of the Company, and I would rather work with
Companies --> Aliases.
But of course I don't know the ins and outs of the reality you're dealing with.
I have a linked table that contains descriptive information of restaurants such as name, address, city, etc.
tbl_Restaurant
REST_SITE_CD (PK)
Restaurant
Address
City
…
I have created a query based off of this linked table and a different table that hosts auditing information.
tbl_Audit
AuditID (PK)
REST_ID (FK)
Date
…
From this query, a form was created to allow the user to easily choose a restaurant by its name and location, and enter in the necessary auditing information.
However, some audited restaurants are not located inside tbl_Restaurants. I would like to manually enter these locations into the query, but the query does not allow me to enter new addresses that do not match an existing primary key in tbl_Restaurant.
The only solution I can think of is to create a redundant table just for unlisted restaurants and combine it with the query later on via a union query or something like that. However, I’m sure this violations a bunch of normalization principles. Additionally, it would be ideal if the custom locations are only reflected upon the query and not the linked table, if that makes any sense.
Any help would be appreciated! Please let me know if there’s any other information I can provide.
Records need to be stored in a table.
You can use a local table, and use a UNION ALL to add the records from your local table to the query. However, you can't enter them directly in the query, since union queries are not updateable. You need to add them to the table.
I am currently working on a report-generating application (in Visual Basic) for a pre-existing database (in mySQL), in which I need to submit queries across multiple tables to access all of the information needed; however, my experience with this sort of project is limited. I am hoping someone can point me toward an efficient method of achieving this.
In the initial report, I need data from 3 tables.
Table 1) 'invoice' - table structure contains a date, an invoice number, and a customer number (and other non-pertinent columns)
Table 2) 'transaction_data' - table structure contains an invoice number, a billing code, and an item description (and other non-pertinent columns). Each row of the table contains a single line-item from a single invoice (so there can be several rows in this table containing the same invoice number).
Table 3) 'customers' - table structure contains customer number, name, address, phone (and other non-pertinent columns).
I need to be able to search 'invoice' based on dates, to get a list of all invoice numbers within the specified time frame (and their corresponding customer number). I then need to take that list of invoice numbers, and search 'transaction_data' for each row that contains one of the invoice numbers, and check for a specific billing code. If the billing code does not exist, I need to use the customer number (obtained during the invoice search) to put together a list of invoice number, customer name, address, phone number.
This can be accomplished fairly easily by populating an array variable utilizing for/while loops, but will require multiple queries across separate tables... of which 'invoice' and 'customers' have 20k+ entries and 'transaction_data' table has over 100k entries. Surely, this is not the most efficient manner of compiling said data.
Can someone please direct me as to how the query SHOULD be structured efficiently? Thanks in advance for helping a database noobie!
Look up JOINS in the MySQL manual. You would join the three tables together and with a WHERE clause get the specific rows you need that match the criteria you are searching for. No loops needed.
This is a bit embarrassing but it's been a while.
When working with relational databases and linking to separate tables by primary:foreign keys, in 1-∞, such that, in a particular table, the table's foreign relationship returns an integer ID... how do you go about making this relationship intelligible to users?
For example:
You have a company table
CID
CompanyName
CompanyLocation
CompanyBusinessType
etc (not actual column names)
and an employee table
EID
firstName
surName
DOB
email
company
Employee company is related to CID. But some user putting in information about an employee would have to know the ID code for the particular company due to the nature of the relationship. You can run a query to return 'CompanyName's along with their associated 'CID's which can then be searched, but this cipher is hardly an ideal solution. I just cannot remember the theory of how one approaches this, even though I'm certain I've done it before. I need to implement it in both Microsoft Access 2010 and Microsoft SSMS (separate databases fwiw); but the execution should be relatively straightforward as soon as I remember how it's done!
Create a combo box with 2 columns. One column as the company name and the other as the company ID. You can set the width of the company ID to 0 if you don't want the user to see it.
This can be done with the Property Sheet on MS-Access Forms
Set the Row Source of the combo box to:
SELECT CompanyName, CompanyID FROM Company;
Set the Control Source of the combo box to:Employee.CompanyID
Also make sure the Row Source Type is set to Table/Query
If you need a many to many relationship (many employees belong to many companies) then you will need another table CompanyEmployees, something like this: -
If an employee only belongs to one company then this structure will do fine: -
In either case you will need to display the user with some form of drop down list when adding / editing a user to associate the users with one or more companies.
In Meetup.com, when you join a meetup group, you are usually required to complete a profile for that particular group. For example, if you join a movie meetup group, you may need to list the genres of movies you enjoy, etc.
I'm building a similar application, wherein users can join various groups and complete different profile details for each group. Assume the 2 possibilities:
Users can create their own groups and define what details to ask users that join that group (so, something a bit dynamic -- perhaps suggesting that at least an EAV design is required)
The developer decides now which groups to create and specify what details to ask users who join that group (meaning that the profile details will be predefined and "hard coded" into the system)
What's the best way to model such data?
More elaborate example:
The "Movie Goers" group request their members to specify the following:
Name
Birthdate (to be used to compute member's age)
Gender (must select from "male" or "female")
Favorite Genres (must select 1 or more from a list of specified genres)
The "Extreme Sports" group request their member to specify the following:
Name
Description of Activities Enjoyed (narrative form)
Postal Code
The bottom line is that each group may require different details from members joining their group. Ideally, I would like anyone to create a group (ala MeetUp.com). However, I also need the ability to query for members fairly well (e.g. find all women movie goers between the ages of 25 and 30).
For something like this....you'd want maximum normalization, so you wouldn't have duplicate data anywhere. Because your user-defined tables could possibly contain the same type of record, I think that you might have to go above 3NF for this.
My suggestion would be this - explode your tables so that you have something close to 6NF with EAV, so that each question that users must answer will have its own table. Then, your user-created tables will all reference one of your question tables. This avoids the duplication of data issue. (For instance, you don't want an entry in the "MovieGoers" group with the name "John Brown" and one in the "Extreme Sports" group with the name "Johnny B." for the same user; you also don't want his "what is your favorite color" answer to be "Blue" in one group and "Red" in another. Any data that can span across groups, like common questions, would be normalized in this form.)
The main drawback to this is that you'd end up with a lot of tables, and you'd probably want to create views for your statistical queries. However, in terms of pure data integrity, this would work well.
Note that you could probably get away with only factoring out the common fields, if you really wanted to. Examples of common fields would include Name, Location, Gender, and others; you could also do the same for common questions, like "what is your favorite color" or "do you have pets" or something to that extent. Group-specific questions that don't span across groups could be stored in a separate table for that group, un-exploded. I wouldn't advise this because it wouldn't be as flexible as the pure 6NF option and you run the risk of duplication (how do you predetermine which questions won't be common questions?) but if you really wanted to, you could do this.
There's a good question about 6NF here: Would like to Understand 6NF with an Example
I hope that made some sense and I hope it helps. If you have any questions, leave a comment.
Really, this is exactly a problem for which SQL is not a right solution. Forget normalization. This is exactly the job for NoSQL document stores. Every user as a document, having some essential fields like id, name, pwd etc. And every group adds possibility to add some fields. Unique fields can have names group-id-prefixed, shared fields (that grasp some more general concept) can have that field name free.
Except users (and groups) then you will have field descriptions with name, type, possible values, ... which is also very good for a document store.
If you use key-value document store from the beginning, you gain this freeform possibility of structuring your data plus querying them (though not by SQL, but by the means this or that NoSQL database provides).
First i'd like to note that the following structure is just a basis to your DB and you will need to expand/reduce it.
There are the following entities in DB:
user (just user)
group (any group)
template (list of requirement united into template to simplify assignment)
requirement (single requirement. For example: date of birth, gender, favorite sport)
"Modeling":
**User**
user_id
user_name
**Group**
name
group_id
user_group
user_id (FK)
group_id (FK)
**requirement**:
requirement_id
requirement_name
requirement_type (FK) (means the type: combo, free string, date) - should refers to dictionary)
**template**
template_id
template_name
**template_requirement**
r_id (FK)
t_id (FK)
The next step is to model appropriate schema for storing restrictions, i.e. validating rule for any requirement in any template. We have to separate it because for different groups the same restrictions can be different (for example: "age"). You can use the following table:
**restrictions**
group_id
template_id
requirement_id (should be here as template_id because the same requirement can exists in different templates and any group can consists of many templates)
restriction_type (FK) (points to another dict: value, length, regexp, at_least_one_value_choosed and so on)
So, as i said it is the basis. You can feel free to simplify this schema (wipe out tables, multiple templates for group). Or you can make it more general adding opportunity to create and publish temaplate, requirements and so on.
Hope you find this idea useful
You could save such data as JSON or XML (Structure, Data)
User Table
Userid
Username
Password
Groups -> JSON Array of all Groups
GroupStructure Table
Groupid
Groupname
Groupstructure -> JSON Structure (with specified Fields)
GroupData Table
Userid
Groupid
Groupdata -> JSON Data
I think this covers most of your constraints:
users
user_id, user_name, password, birth_date, gender
1, Robert Jones, *****, 2011-11-11, M
group
group_id, group_name
1, Movie Goers
2, Extreme Sports
group_membership
user_id, group_id
1, 1
1, 2
group_data
group_data_id, group_id, group_data_name
1, 1, Favorite Genres
2, 2, Favorite Activities
group_data_value
id, group_data_id, group_data_value
1,1,Comedy
2,1,Sci-Fi
3,1,Documentaries
4,2,Extreme Cage Fighting
5,2,Naked Extreme Bike Riding
user_group_data
user_id, group_id, group_data_id, group_data_value_id
1,1,1,1
1,1,1,2
1,2,2,4
1,2,2,5
I've had similar issues to this. I'm not sure if this would be the best recommendation for your specific situation but consider this.
Provide a means of storing data as XML, or JSON, or some other format that delimits the data, but basically stores it in field that has no specific format.
Provide a way to store the definition of that data
Provide a lookup/index table for the data.
This is a combination of techniques indicated already.
Essentially, you would create some interface to your clients to create a "form" for what they want saved. This form would indicated what pieces of information they want from the user. It would also indicate what pieces of information you want to search on.
Save this information to the definition table.
The definition table is then used to describe the user interface for entering data.
Once user data is entered, save the data (as xml or whatever) to one table with a unique id. At the same time, another table will be populated as an index with
id where the xml data was saved
name of field data is stored in
value of field data stored.
id of data definition.
now when a search commences, there should be no issue in searching for the information in the index table by name, value and definition id and getting back the id of the xml/json (or whatever) data you stored in the table that the data form was stored.
That data should be transformable once it is retrieved.
I was seriously sketchy on the details here, I hope this is enough of an answer to get you started. If you would like any explanation or additional details, let me know and I'll be happy to help.
if you're not stuck to mysql, i suggest you to use postgresql which provides build-in array datatypes.
you can define a define an array of varchar field to store group specific fields, in your groups table. to store values you can do the same in the membership table.
comparing to string parsing based xml types, this array approach will be really fast.
if you dont like array approach you can check out xml datatypes and an optional hstore datatype which is a key-value store.