remi-coulom / joedb Goto Github PK

Hello,

Saw a link on cppcast, and on browsing initial documentation this seems very similar to prevayler. I had done quite a bit of work with prevayler a long time ago (10+ years), and was just curious if that was where the original idea came from.

Looks cool, will see if a use case comes up for this.

Hello,

Just some thoughts floating in my mind at present, but had some questions about running joedb as embedded database with a service that provides the following:

• Use Raft to elect leader in cluster
• Write to any node in cluster
  • Lookup leader
  • Write to leader
  • Notify all clients to pull
• Read against any node, should be up to date, eventual consistency is good enough.
• If leader failed, use Raft to elect new leader.
  • Can other clients pull from the new leader without issues?

Do you think there will be issues with that basic workflow?

Thanks
Rakesh

In my .joedbi file i have some table definitions like this where I need a row of 64 doubles:

create_table data_64
add_field data_64 c0 float64
add_field data_64 c1 float64
add_field data_64 c2 float64
...
add_field data_64 c62 float64
add_field data_64 c62 float64
add_field data_64 c63 float64

It would be much easier to access the data if I could do something like:
create_table data_64
add_field data_64 c float64[64]

such that data_of_data_64 looks something like this:
struct data_of_data_64
{
std::array<double, 64> field_value_of_c;
}

Is there a different way to get the same result?

Hi, I've just been reading through joedb a bit (looks great!) and it looks quite interesting for a use case I've been searching for a good solution to:

• log format for a message passing system (currently using capn proto, but this could just as easily be protobuf or some other format) to capture a record of traffic during a 'session' (ie, while the application is running)
• will be write-only during the session, so want to optimize for insertion performance, but data will be queried/replayed/analyzed afterwards so a structured/queryable format is desirable (existing joedb system seems a good match for this)
• store messages (and their schemas) in their native encoded format, so they can be 'replayed' during development (ie messages from a single participant could be retransmitted in a simulation environment for testing)

Ideally, it would be possible to use standard joedb query/access routines to query the natively encoded data. One option that occurred to me was to provide a 'virtual table' interface, such that a 'capnp' table could implement the functionality required to respond to joedb queries against capnp schemas (ie table names, field names etc would map to their definitions in the capnp schema). I believe this approach could be possible to integrate at the joedbc stage.

So, finally a question! 😄 Does this seem feasible/aligned with the goals of joedb? Is there another approach you'd recommend? Thanks for reading.

I see that I can timestamp / comment the log. Is there a method to load a database up to a specified timestamp or comment?

There are other ledger databases (most famously https://aws.amazon.com/qldb/), it would be interesting to see a comparison in README.

(It seems that joedb is a ledger DB, but I'm not sure.)

You can follow this issue if you wish to be informed of the progress of joedb.

It seems github offers no way to follow releases of a project. So I will annouce them here. Please don't comment in this issue. This issue should be strictly reserved to post information about recent new progress of joedb.

How would you handle this?