Redis implementation in golang for educational purposes, using the book Build your own Redis as reference.
There are executables available under the cmd dir. One for a test client and another one for the server.
# On the root dir.
go run ./cmd/server
# action-key-value
# available actions: g|s|d
go run ./cmd/client g key valueAn example flow for the client bin below:
To execute the tests run go test -v ./... to recursively expand on subdirs. Verbosity flag provides information about the tests that have been run.
This repo contains both a server and a client to test the server with. The communication is done thanks to a tiny protocol.
The server is structured around an event loop that polls for events in a main file descriptor. This main file descriptor is the one taking care of new connections.
The server supports setting a key:value pair, retrieving the value of a key or deleting it.
These are the basic conceptual components that are in place here.
The communication protocol is a very simple one.
Reserve the 4 initial bytes to communicate the payload length. Follows the amount of fragments that the message contains, and then the fragments themselves. Each fragment is defined by the ordered tuple of (length, fragment).
| Protocol Header | Number of fragments | Fragment length | Fragment length |
|---|---|---|---|
| 4 byte | 4 byte | 4 byte | variable |
Each request is then built into a struct:
type ProtocolRequest struct {
Action string
Key string
Value *string
}The fragments are serialized and deserialized in the same order that appears in the defined struct; first the action, then the key and, optionally, the value. This is what allows us to GET|SET|DEL a key and its value.
The event loop is implemented synchronously, although the following diagram denotes our possible point of delegation that will articulate the work done by background threads.
