I am converting a spreadsheet of game notes into a database.

I have a table of character Classes for an RPG (fighter, mage, etc)

One of the columns shows the Races that a Class can be in a comma seperated list:

ClassId       ClassName     Race 
------------- ------------- ------------
1             Rogue         Man           
2             Mage          Man,Elf       
3             Warrior       Man,Dwarf,Elf

I'm new to SQL and not sure how to translate this into a database file.

Here is what I'm thinking:

1. I make a 3rd table that holds the relation.

I think this is the most proper solution. The drawback for me are that I would have to create and manage an extra table. Also, to me the comma separated list like seen above is the preferable way for me to display the data and I'm not sure what other ways there would be to display the three possible Races together on one row (I don't think SQL/SQLite has syntax to concatenate strings or build lists?)

2. I make a column for each race and use a boolean 0 or 1 value

This could be pretty clear. The only disadvantage I see is that I would have to create new columns whenever adding new races.

3. I just leave the Race column as a comma seperated value

I could still query the column with statements like

select * from Class where Race like '%Elf%';

But I think this would make it hard to put any kind of constraints on the field (like a FOREIGN KEY constraint to make sure the Race exists in the Race table).

Thoughts or suggestions are appreciated.

Disclaimer: I do not write tons of database stuff.

I would go with option 1. And, in fact, I would have two new tables: a table for Race and a table that maps classes to race, which I would call something like ClassToRace. ClassToRace would just have two columns, a ClassId column and a RaceId column, which are foreign keys into the Class and Race tables, respectively. The Race table would have a RaceId and whatever other information is relevant to races (RaceName, stats, etc).

The Class table now wouldn't have any Race column at all. If you want information about the races a class supports, you would have to query the ClassToRace table to get all the rows that match ClassId, which will give you the RaceId of each race that ClassId supports.

It won't display as nicely as the snippet you posted in the SQL program, but is that really going to matter? It seems more likely to me that all you really want is the information, which your application will use as it sees fit. And you application can display the information however it likes. You shouldn't be making database design decisions based on how SQL displays the information, because that's largely irrelevant to your needs.

EDIT: the ClassToRace table also has the benefit that you can query it by RaceId, and see all the ClassIds that are available to that race.

There is a set of 'best practices' for architecting databases, called Database Normalization. You don't necessarily have to follow all of the guidelines, but at least it will give you some ideas and insight that you can apply if you want to.

http://en.wikipedia.org/wiki/Database_normalization


In your class/race relationship, the normalized way I would do it is:

Classes
(PK)ID      Name
----------------
1           Rogue
2           Mage
3           Warrior

Races
(PK)ID      Name
----------------
1           Man
2           Dwarf
3           Elf

ValidClassRaces
(FK)ClassID  (FK)RaceID
-----------------------
1            1
2            1
2            3
3            1
3            2
3            3

This allows you to ask "what races can class X be?" and "what classes can race Y be?" with the same query complexity since your tables are not biased anymore.

It allows you to rename classes or races without affecting the relationships.

It allows you to delete or change relationships quickly.

It allows you to prevent deletion of a class or race which is still in use by the relationship table through constraints.

I would go with option 1. And, in fact, I would have two new tables: a table for Race and a table that maps classes to race, which I would call something like ClassToRace. ClassToRace would just have two columns, a ClassId column and a RaceId column, which are foreign keys into the Class and Race tables, respectively. The Race table would have a RaceId and whatever other information is relevant to races (RaceName, stats, etc).

The Class table now wouldn't have any Race column at all. If you want information about the races a class supports, you would have to query the ClassToRace table to get all the rows that match ClassId, which will give you the RaceId of each race that ClassId supports.

I already have a Race table so this is why I thought it would be the most correct way to relate the two.

It won't display as nicely as the snippet you posted in the SQL program, but is that really going to matter? It seems more likely to me that all you really want is the information, which your application will use as it sees fit. And you application can display the information however it likes. You shouldn't be making database design decisions based on how SQL displays the information, because that's largely irrelevant to your needs.

At this point I am still in the process of converting from CSV files dumped from spreadsheets. First I am reading the CSV file and then dumping from sqlite3 as an SQL script so that I can edit the definitions and then read them back in as complete tables.

I'm thinking the database is essentially static while my program is running (I have a main-memory array-based database system for in-game entities). But as it is a work in progress I will still be making changes to the database outside of the application.

So, is this a bad thing? I wanted a more regularized representation than a spreadsheet. And it's more natural for me to be able to edit database files from the command-line.

There is a set of 'best practices' for architecting databases, called Database Normalization. You don't necessarily have to follow all of the guidelines, but at least it will give you some ideas and insight that you can apply if you want to.

http://en.wikipedia.org/wiki/Database_normalization

Thanks for this. SQL has been harder for me to learn than programming languages because it seems like resources are less centralized, so more resources on general design topics are great for me right now.

select * from Class where Race like '%Elf%';

This has terrible performance. It is an unindexed query. This means that you are reading your entire table every time you retrieve a line.

If you ever want to make the "who is an elf" query, you should definetly normalize your database.

Is it going to matter if a table has only between 3 and say, 500 rows?

Edit: The database is more for character creation. Player chooses a Race, next screen might list the Classes that are available to that race. But this database is not the in-game entity database.