Now I have a hierarchical structure that looks like this:
.stage
...grade
.....semester
.......subject
.........unit
...........lesson
At first they were all stored inside the same table with an id parent relationship; however, each level of those 6 levels has its own data. For example, a lesson has featured_video_id and a subject has a language_id.
Now I'm thinking of creating 6 tables: one for each level and connect them with foreign keys. I'm thinking of all the possibilities and thought that a professional opinion will help. Will that division be of any good or it will be something that I will regret in the future?
I t makes a lot of sense to have separate tables for stages, grades, semesters, ...
You have already mentioned the best reason for this, you can add individual data for each of these levels. You can name the foreign keys in a sensible way (i.e. stage_id in the table for the grades). And I doubt you will ever need a list of subjects mixed in with lessons or semesters.
I'd do a bit of both:
Have a master table that contains all of the items at all levels with a common primary key. Include only the attributes that are common across all levels of the hierarchy. Link them with a parent key here.
Have separate tables for only the attributes that are unique to a particular level of the hierarchy. Use the primary key in the master table as a foreign key in these "detail tables."
This lets you separate your concerns. The structure of the hierarchy is in the master table. The details of each level are in the detail tables.
Related
I have two tables matches and tournaments with below structure,
MATCH
MATCH_ID
PLAYER_ID_1
PLAYER_ID_2
RESULT
TOURNAMENT_ID
and
TOURNAMENT
TOURNAMENT_ID
NAME
OTHER_DETAILS
with one tournament will have multiple matches
and a match may or may not have tournament id
use Cases:
retrieve all matches
retrieve all matches by tournaments
Is it good to have tournament id in match table? Or should I create a separate joining table for tournament and match mapping? Which will have good performance when the volume increases?
TOURNAMENT_ID has a 1:M relationship to MATCH. It seems to be a straightforward foreign key. The standard way of implementing foreign keys - even optional foreign keys - is a column on the child table with a foreign key constraint. This would support both your use cases.
A separate table would normally be a head scratcher. I say "normally" because there are schools of thought which abominate NULL columns in databases; either for practical reasons - NULLs can do weird things to our code and need wrangling - and academic reasons - NULL is contrary to Relational Algebra. So, if you have a data model which forbids the use of nulls you will need a TOURNAMENT_MATCH table to hold Matches which are part of a Tournament. It also would be likely to perform slightly worse than a foreign key column on MATCH, but unless you have a vast amount of data you won't notice the difference.
There is a use case for join tables (also known as junction or intersection tables) and that is implementing many-to-many relationships. Suppose we add a third table to the mix, PLAYER. A Player can participate in many Tournaments and a Tournament has many Players. Classic M:N relationship. So we can resolve it with a join table REGISTERED_PLAYER. which as a compound key of (TOURNAMENT_ID,PLAYER_ID) and the appropriate foreign keys to TOURNAMENT and PLAYER.
For the sake of completeness I will mention Link tables from Data Vault modelling. This is an interesting modelling technique for Data Warehouses, where - gross simplification alert - tables are defined as Hubs (business and technical keys) and Satellites (immutable attribute records). This approach allows for the capture of data changes over time. Foreign key relationships between Hubs are implemented through Link tables, to support changing relationships over time.
There are several benefits to Data Vault for wrangling large amounts of data in a time-sensitive fashion but an easy-to-understand physical data model isn't one of them. Anyway, find out more.
The simple rule: for one-to-many mapping always prefer a foreign key association to a join table association.
It is hard to control a join table using a standard #OneToMany Hibernate mapping — you can't just delete rows from a join table, or add an additional row. You will need to use list on the Tournament side to do things like that. Another option is to use an additional entity for a join table.
Note: Match can has a tournaments list too, but looks like Tournament is the owner of the association.
A few opinions have been offered in other answers, here is mine.
You do NOT want a separate join table, you would only need that if a Match can be in multiple Tournaments. In your example, just use a foreign key.
The only other suggestion is that if the Match is not part of a Tournament then it is not actually "unknown" which is the meaning of NULL, it is actually something else like "Individual Match". So consider adding a row to your Tournament table, maybe using a known key like 0 or -1, and using that for matches that are not part of a tournament.
I guess this is more of a philosophical question. I have one item with 4 variations. These variations will share about 80% of their columns. The item in question is a portfolio example and the variations are photography/video/web/graphic. They will share column names like client, date, etc.
My question is, should I have 1 table for each variation or 1 base table and an additional table for variation-specific columns. I'm "never" going to need to pull down a photo/web example at the same time or what have you.
I wasn't sure how to Google search for this despite it being fairly basic
You have a one-of relationship. The key question is how other tables will reference this entity.
There are three options:
All references are to the parent entity. For instance, all items have price histories, categories, or comments and the references are only to the parent.
All references are to the children. For instance, photos might have objects in them, videos might have lists of producers, and so on.
References can be to either.
What is the answer? (1) suggests that all the rows should be in a single table with a single id shared for all. (2) suggests that the rows should be in a single table for each variation.
(3) is a little more difficult. I would recommend a parent table with the 80% of the columns. Then children tables with the remaining ones. The primary keys of the children tables would be the same as the primary key of the parent (yes, a column can be both a primary key and a foreign key).
For this situation you need to do a little work to ensure that a given parent entity appears in exactly one child table. That can be handled with foreign keys and triggers (in MySQL).
As long as there are no specific rules for a variant id just create a Media table.
If you decide to keep them seperated i'd expect tables to have a date and client_id.
I have a table that stores information about categories ('cats' let's say) and another table (say 'cars' meaning it stores infos about cars - a lame example, but that doesn't matter) that has a column called 'parent_cat_id' which should reference 'id' col in 'cats'.
Now comes the problem: any car from 'cars' could belong to more then one category.
There's an obvious, but very lame, solution: store in parent_cat_id a delimited list of numbers, but, if i apply that, I couldn't have any foreign key constraint and should jump into parsing these values. Although parsing isn't such a big deal, but I can't believe MySQL doesn't have anything pre-built.
Also it looks like selecting cars by category will involve some substring and other things.
Having no foreign key is a far bigger problem.
The only thing I was able to find is a 'SET' type, but it should be pre-defined what values it can store during database development, also it can have only 64 pre-defined values while there could be 100s of categories.
The normal way of modeling this kind of many-to-many relationship is to introduce a third table (a junction table) that holds information about relationships between items in the two other tables.
In your example you would have three tables:
Category
Cars
CarCategory which has foreign keys to both the primary keys in Category and Cars
This is pretty fundamental relational modeling and you might want to look at a nice tutorial or book to get further knowledge about the concepts.
Just as you recognize storing delimited lists of keys is a bad idea as it breaks the model.
I have the following tables created:
Animes(id,title,date), Comics(id,title,date), TVSeries(id,title,season,episode,date)
Some of them have already foreign keys (for one-to-many or many-to-many relations) of directors, genres, articles and so on.
Now i would like to create two more tables Reviews(id,rating,date) and Posts(id,thumbid,articleid,reviewid).
A review is about one Anime and/or Comic TVSerie and vise-versa but properties of a review may be in more than one table. Its the same about a posts.
Is this a typical example of one-to-one relation in separate table or is it more efficient to add more properties to the existing tables?
So more tables and relations or less tables more columns?
Thank you and i hope my question isnt that stupid but im a bit confused.
In my view, It is better to avoid foreign key relationship for one-to-one relationship. It is best suitable for one - many relationships.
I'm not exactly sure what your requirements are, but the choices are as follows:
Have Reviews have 2 columns, either being a foreign key to the applicable table, can be NULL. This is really for when a single review can be about both.
Have a ReviewsComics and ReviewsAnime table. You'd then have all the fields from Reviews in each table (and no Reviews table).
An alternative (2) is to use them in conjunction with a Reviews table, then those 2 tables only has 2 fields which are foreign keys to Reviews and Comics/Anime respectively (thus no direct link between Reviews and Comics/Anime).
Have a base table to which Anime and Comics are linked to 1-to-1 and have reviews link to that table instead.
(Edit) If all the fields are all going to be the same (or similar) for Anime/Comics, you can merge them into a single table and add a type field, indicating Anime/Comics, then the problem goes away. This is similar to the base table option.
EDIT: The 2 reviews tables will probably give the best performance (unless you want to select all reviews for either, often), but with proper indices the performance shouldn't be an issue with any of the above.
I've read this question: What's the difference between identifying and non-identifying relationships?
But I'm still not too sure...
What I have is three tables.
Users
Objects
Pictures
A user can own many objects and can also post many pictures per individual object.
My gut feeling tells me this is an identifying relationship, because I'll need the userID in the objects table and I'll need the objectID in the pictures tables...
Or am I wrong? The explanations in the other topic limit themselves to the theoretical explanation of the way the database interprets it after it's already been coded, not how the objects are connected in real life. I'm kinda confused as to how to make the decision of identifying versus non-identifying when thinking about how I'm going to build the database.
Both sound like identifying relationships to me. If you have heard the terms one-to-one or one-to-many, and many-to-many, one-to- relationships are identifying relationships, and many-to-many relationships are non-identifying relationships.
If the child identifies its parent, it is an identifying relationship. In the link you have given, if you have a phone number, you know who it belongs to (it only belongs to one).
If the child does not identify its parent, it is a non-identifying relationship. In the link, it mentions states. Think of a state as a row in a table representing mood. "Happy" doesn't identify a particular person, but many people.
Edit: Other real life examples:
A physical address is a non-identifying relationship, because many people may reside at one address. On the other hand, an email address is (usually considered) an identifying relationship.
A Social Security Number is an identifying relationship, because it only belongs to one person
Comments on Youtube videos are identifying relationships, because they only belong to one video.
An original of a painting only has one owner (identifying), while many people may own reprints of the painting (non-identifying).
I think that an easier way to visualize it is to ask yourself if the child record can exist without the parent. For example, an order line item requires an order header to exist. Thus, an order line item must have the order header identifier as part of its key and hence, this is an example of an identifying relationship.
On the other hand, telephone numbers can exist without ownership of a person, although a person may have several phone numbers. In this case, the person who owns the phone number is a non-key or non-identifying relationship since the phone numbers can exist irrespective of the owner person (hence, the phone number owner person can be null whereas in the order line item example, the order header identifier cannot be null.
NickC Said: one-to- relationships are identifying relationships, and many-to-many relationships are non-identifying relationships
The explanation seems totally wrong to me. You can have:
Ono-to-One Non-identifying Relationships
One-to-Many Non-identifying Relationships
One-to-One Identifying Relationships
One-to-Many Identifying Relationships
Many-to-Many Identifying Relationships
Imagine you have the following tables: customer, products and feedback. All of them are based on the customer_id which exists on the cutomer table. So, by NickC definition there shouldn't be exists any kind of Many-to-Many Identifying Relationships, however in my example, you can clearly see that: A Feedback can exists only if the relevant Product exists and has been bought by the Customer, so Customer, Products and Feedback should be Identifying.
You can take a look at MySQL Manual, explaining how to add Foreign Keys on MySQL Workbench as well.
Mahdi, your instincts are correct. This is a duplicate question and this up-voted answer is not correct or complete.
Look at the top two answers here:
difference between identifying non-identifying
Identifying vs non-identifying has nothing to do with identity.
Simply ask yourself can the child record exist without the parent? If the answer is yes, the it is non-identifying.
The core issue whether the primary key of the child includes the foreign key of the parent. In the non-identifying relationship the child's primary key (PK) cannot include the foreign key (FK).
Ask yourself this question
Can the child record exist without the parent record?
If the child can exist without the parent, then the relationship is non-identifying. (Thank you MontrealDevOne for stating it more clearly)
One-to-one identifying relationship
Social security numbers fit nicely in to this category. Let's imagine for example that social security numbers cannot exist with out a person (perhaps they can in reality, but not in our database) The person_id would be the PK for the person table, including columns such as a name and address. (let's keep it simple). The social_security_number table would include the ssn column and the person_id column as a foreign key. Since this FK can be used as the PK for the social_security_number table it is an identifying relationship.
One-to-one non-identifying relationship
At a large office complex you might have an office table that includes the room numbers by floor and building number with a PK, and a separate employee table. The employee table (child) has a FK which is the office_id column from the office table PK. While each employee has only one office and (for this example) every office only has one employee this is a non-identifying relationship since offices can exist without employees, and employees can change offices or work in the field.
One-to-many relationships
One-to-many relationships can be categorized easily by asking the same question.
Many-to-many relationships
Many-to-many relationships are always identifying relationships. This may seem counter intuitive, but bear with me. Take two tables libary and books, each library has many books, and a copy of each book exists in many libraries.
Here's what makes it and identifying relationship:
In order to implement this you need a linking table with two columns which are the primary keys of each table. Call them the library_id column and the ISBN column. This new linking table has no separate primary key, but wait! The foreign keys become a multi-column primary key for the linking table since duplicate records in the linking table would be meaningless. The links cannot exist with out the parents; therefore, this is an identifying relationship. I know, yuck right?
Most of the time the type of relationship does not matter.
All that said, usually you don't have to worry about which you have. Just assign the proper primary and foreign keys to each table and the relationship will discover itself.
EDIT: NicoleC, I read the answer you linked and it does agree with mine. I take his point about SSN, and agree that is a bad example. I'll try to think up another clearer example there. However if we start to use real-world analogies in defining a database relationship the analogies always break down. It matters not, whether an SSN identifies a person, it matters whether you used it as a foreign key.