Should I represent two distinct relationships via only one associative table?

Question

I have a database to store the news of my website. The news entity type is associated with two different ones; images (which from the point of view of the app users are grouped in a gallery) and categories.

To represent this situation this, I have some tables like this:

image
id | src

category
id | name

news
id | title | text | author | ...

It is important to mention that each news may have multiple images and belong to multiple categories. So, to create the relationship between images, categories and news, I'll use another table, but my question is:

Should I have (a) two tables (one for image rows and other for category rows) or (b) just one with multiple columns? For example:

(a) Two distinct tables, one called news_relationship_category, with the column headings

id | id_news | id_category

and another one, named news_relationship_gallery set up as follows

 id | id_news | id_gallery

(b) One single table, entitled news_relationships with the headings

id_news | id_category | id_gallery

and INSERT INTO it rows with some columns that accept NULL marks, for instance:

10 | NULL | 3
---+------+------
10 | 27   | NULL
---+------+------

These are simple examples, but imagine if I have an e-commerce table, with products involved in multiple relationships. What is the best way to approach this case?

Answer 1

To delete a news-category pair in schema B, you have to do either a DELETE or an UPDATE depending on whether the relevant row has id_gallery specified or not. How about inserts, are you going to just INSERT or check first for rows with nulls to UPDATE?

So, I strongly recommend option A. However, I would drop the surrogate id columns and use (id_news, id_category) and (id_news, id_gallery) as PK.