[NEW] Introduce database logical names

[xbasel]

Allow users to select databases using logical names (strings). This will simplify client code by eliminating the need to map database numbers to names manually. For example, users should be able to run:

select orders
select emails
select monitoring
.
.
select 1
select 2

This feature can be implemented part of #1597

[ranshid]

I agree this is a nice usability feature.

some thoughts that pop to mind:

1. what is the default db? is it still 0?
2. currently when the server starts it logically (and in reality as we observed in other issues) there already exists dbs with numbering names. So just selecting a new name would basically exceed the dbnum right? I think it might be good if we can support both. ie you can assign a name to a DB and can either use the name or a number. It might complicate the API though, since it will be hard to understand if a number provided to swapdb/select is a name or a number.
3. Regarding replication.i think db names is something that should persist. IMO, this means we will probably have to introduce a new command to create/set db name.

[hwware]

I agree this is a nice usability feature.

some thoughts that pop to mind:

1. what is the default db? is it still 0?
2. currently when the server starts it logically (and in reality as we observed in other issues) there already exists dbs with numbering names. So just selecting a new name would basically exceed the dbnum right? I think it might be good if we can support both. ie you can assign a name to a DB and can either use the name or a number. It might complicate the API though, since it will be hard to understand if a number provided to swapdb/select is a name or a number.
3. Regarding replication.i think db names is something that should persist. IMO, this means we will probably have to introduce a new command to create/set db name.

We need consider how to write the database name in the config file, such as: database 0 orders

[madolson]

We need consider how to write the database name in the config file, such as: database 0 orders

I presume we would add it as an auxiliary field, but we have the problem of making sure that across a cluster everyone has the same numeric mapping <-> logical identifier.

[lucasfcnunes]

We need consider how to write the database name in the config file, such as: database 0 orders

I presume we would add it as an auxiliary field, but we have the problem of making sure that across a cluster everyone has the same numeric mapping <-> logical identifier.

Wouldn't a simple standard hash function on the text generate the same unique number in each node?

[xbasel]

We need consider how to write the database name in the config file, such as: database 0 orders

I presume we would add it as an auxiliary field, but we have the problem of making sure that across a cluster everyone has the same numeric mapping <-> logical identifier.

Wouldn't a simple standard hash function on the text generate the same unique number in each node?

Hash functions give consistent values across nodes but don't guarantee uniqueness, so collisions are still possible...

[lucasfcnunes]

Hash functions give consistent values across nodes but don't guarantee uniqueness, so collisions are still possible...

@xbasel
Mathematically yes, practically no. Right?

[xbasel]

Hash functions give consistent values across nodes but don't guarantee uniqueness, so collisions are still possible...

@xbasel Mathematically yes, practically no. Right?

@lucasfcnunes
Mathematically yes, but even a theoretical collision risk is probably unacceptable, better to avoid it entirely.

Even if collision risk is acceptable, if we hash strings to integers for IDs, we lose the ability to use arrays for O(1) random access - hashing makes the ID space sparse, requiring a different structure (likely a hash table), which adds complexity and can reduce performance.

Collisions can be mitigated if limit the length if the name to 13 chars lowercase, in this case we can fit 13 chars in an integer, still we need to think about how to implement random access.