One example could be to add a value to the relationship, like a rating or a ranking.

For example a Movie can be seen by many Users and a User can see many Movies. A user can rate the movie they’ve seen between 0 and 10. So, the join table would have 3 collumns:

• he FK for User
• the FK for Movie
• the numerical value of the rating by that particular User for that particualr Movie.

There is no concrete difference between the two options. But in general they will be similar. I think you are talking about these options:

struct Person;
struct Skill;

struct PersonSkills {
    person: PersonId,
    skill: SkillId,
}

vs

struct Person {
    skills: SkillId[],
}

struct Skill;

The main difference that I see is that there is a natural place to put data about this relationship with the “join table”.

struct PersonSkills {
    person: PersonId,
    skill: SkillId,
    acquired: Timestamp,
    experience: Duration,
}

You can still do this at in the second one, but you notice that you are basically heading towards an interleaved join table.

struct PersonSkills {
    skill: SkillId,
    acquired: Timestamp,
    experience: Duration,
}

struct Person {
    skills: PersonSkills[],
}

There are other less abstract concerns. Such as performance (are you always loading the list of skills, what if it is long) or correctness (if you delete a Person do you want to delete these relationships, it comes “for free” if they are stored on the Person) But which is better will depend on your use case.

Conceptually, the benefit of a join table is to allow many-to-many relations. That’s it.

If I understand you correctly, your relations are one-to-many, so a join table would just be needless complexity.

Joins and tables are abstract concepts, they don’t dictate how you store data on memory or disk or how you read it.

If you want a specialized data storage, go with whatever format is easier for you to use. But also, the format that is easier to store is not necessarily the easiest one to work on memory.

It depends entirely on how you want to work with the data.

Have you considered sqlite? The database is just a single file, which gives you all the advantages of a text file (easy backup, sharing, easy editing via sqlite browser) while also providing the benefits of SQL when operating on the data (join, etc).

The real primary benefit of storing your relationships in a separate place is that it becomes a point of entry for scans or alterations instead of scanning all entries of one of the larger entity types. For example, “how many users have favorited movie X” is a query on one smaller table (and likely much better optimized on modern processor architectures) vs across all favorites of all users. And “movie x2 is deleted so let’s remove all references to it” is again a single table to alter.

Another benefit regardless of language is normalization. You can keep your entities distinct, and can operate on only one of either. This matters a lot more the more relationships you have between instances of both entities. You could get away with your json array containing IDs of movies rather than storing the joins separately, but that still loses for efficiency when compared to a third relationship table.

The biggest win for design is normalization. Store entities separately and updates or scans will require significantly less rewriting. And there are degrees of it, each with benefits and trade-offs.

I’m wondering if you would benefit from using MongoDB?

Not sure I’m getting the issue here (what does “join table” mean in the scope of JSON/XML?), but… doesn’t how you lay out your data in JSON/XML file have zero impact in your application’s queries? You won’t be querying the JSON - you’ll be loading data from it into memory and query the memory.

SQL uses join tables to do what traditionally would have been done with dictionaries embedded in the objects. I.e. purpose made data structure. That approach is more efficient but less convenient than using a 3-way join.