A simple protocol to query databases based on JSON and WebSocket. It's implemented as a Java client library and a proxy server that can be used to easily access SQL databases in a standardized manner. The proxy server is currently able to communicate with Transbase and PostgreSQL databases.
The project started as a master thesis by Stefan Burnicki in 2015. Therefore it might not be stable or complete, yet.
The project uses gradle as build system. Consequently, the project's file structure is a typical gradle structure:
Java code can be found in src/main/java
, while test code can be found in src/test/java
.
Resources are in src/main/resources
and test resources in src/test/resources
.
The folder libs/ contains non-maven dependencies.
This project consists of several main modules, namely:
- The core (io.sqp.core), which contains definitions for the SQP data types and messages in Java
- The SQP proxy server (io.sqp.proxy) that is able to understand the SQP, talk to a database and answer the client
- The backend module (io.sqp.backend) contains interfaces and some utilities to write a database backend for the server
- The PostgreSQL SQP proxy server backend (org.postgres.sqp) which enables the proxy to talk to PostgreSQL databases
- The Transbase SQP proxy server backend (io.sqp.transbase) which enables the proxy to talk to Transbase databases
- The SchemaMatcher (io.sqp.schemamatcher) which is used by the proxy to match custom JSON schemas to existing ones
- The client library (io.sqp.client) can be used to communicate with an SQP server, as the proxy server.
To understand this project, you probably need to get an overview of some technologies involved in this project first.
In general:
git
as the version control systemgradle
as the build system in useJava 8
as main programming language- unit testing
For unit testing with mocks:
TestNG
as unit test frameworkMockito
as framework for mock objectsHamcrest
for matchers used in test assertions
For the protocol/core:
WebSocket
as the underlying protocolJSON
as the standard serialization/message formatMsgPack
as the binary equivalent for serialization/messagesJackson
as the serialization framework for both JSON and MsgPack used in the Java implementationJSON Schema
for data type descriptions
For the proxy:
- Event-loop/callback based, asynchronous (non-blocking) programming
Vert.x 3.0
as the event-loop and netty based server platformJSON Schema Validator
as a validator input against a custom JSON Schema
For the client:
JSR 356
specification for Java WebSocket supportTyrus
as a glassfish-based reference implementation of JSR 356- Java 8's
CompletableFuture
s for non-blocking, but not event-loop/callback based programming
If you run the gradle build for the first time you will notice that quite a bunch of dependency packages are downloaded. Some packages are dependencies of others that get downloaded, like Netty as the basis of Vert.x. Other package are needed for tests only, like hamcrest, mockito, and TestNG. The "JSON Schema Validator" project pulls pretty much dependencies. So it's probably a good idea to replace this package in the future if the amount of dependencies packages need to be reduced.
Another dependency is the "tbjdbc.jar" that is not available in a maven repository and thus does not get downloaded automatically. However, this package is needed for the Transbase SQP proxy backend.
Make sure you have Java 8 installed and a working internet connection so dependencies can get downloaded. Then just execute the gradle wrapper with the build task:
./gradlew build
The Transbase backend is only built optionally. You can either install the eligible Transbase JDBC
driver to libs/tbjdbc.jar
or define its location on the cmdline:
./gradlew -Ptbjdbc=/path/to/tbjdbc.jar
The gradle plugins "eclipse" and "idea" are used to generate project files for the Eclipse or IntelliJ IDEA IDEs. Just run
./gradlew idea
or
./gradlew eclipse
Check out the plugin documentation for more details on this.
In order for the proxy to work, you need to provide a configuration. See the config.json file for an example. The proxy server itself needs to be configured in a number of ways:
- The
port
to listen on for WebSocket connections. It's optional and the default is8080
. - The
path
to match for incoming connections. It's optional and the default is/
. - The
connectionPoolSize
, i.e. the maximum number of concurrently open connections. It's optional and the default is30
. - The
backends
array which contains backend configurations. Currently only the first is used and the rest is ignored. This is mandatory as there are no defaults.
A backend configuration needs to include two fields:
- The
type
is the full class name of the Backend implementation. E.g.io.sqp.postgresql.PostgreSQLBackend
. - The
config
needs to be an object whose values depend on the concrete backend. Common fields are server credentials of the DBMS.
You can run it directly in your IDE by creating a run configuration that uses the main class
io.vertx.core.Starter
and passes in the arguments -config <conffile>.json run io.sqp.proxy.ServerVerticle
. ".json" should be the path to the proxy's configuration
file, which is mandatory to connect to a database. For details on the configuration file, see the
section "Proxy Configuration" below.
To use a specific Transbase JDBC driver, you might define it just like stated for the building process.
The build.gradle uses the Gradle shadowJar plugin to assemble the application and all it's dependencies into a single "fat" jar. To build the "fat jar", run
./gradlew shadowJar
To run the fat jar, run
java -jar build/libs/sqp-1.0-fat.jar -conf config.json
(You can take that jar and run it anywhere there is a Java 8+ JDK. It contains all the dependencies it needs so you don't need to install Vert.x or any other libraries on the target machine).
The server is implemented as a Vert.x verticle, which allows easy scaling of the server. E.g. let's say you have 8 cores on your server and you want to utilise them all, you can deploy 8 instances as follows:
java -jar build/libs/sqp-1.0-fat.jar -conf config.json -instances 8
You can also enable clustering and ha at the command line, e.g.
java -jar build/libs/sqp-1.0-fat.jar -conf config.json -cluster
java -jar build/libs/sqp-1.0-fat.jar -conf config.json -ha
Please see the Vert.x docs for a full list of Vert.x command line options.
To run the tests you need to care about some prerequisites:
- Install and start a PostgreSQL server.
- Set up a PostgreSQL user called
proxyuser
with passwordproxypw
. If you want to use another user, you need to modify the test configuration and the run configuration accordingly. - Create a
proxytest
PostgreSQL database. - Install and start a Transbase server.
- Create a
proxytest
Transbase database. - For both Transbase and PostgreSQL: Create tables with statements that can be found in
src/test/resources/*Table.sql
. - Run the tests by executing
./gradlew test
The test results can be seen on the console or as a HTML summary at
build/reports/tests/index.html
.
If you want to exclude a backend from the tests, you need to comment out the backend related sections in the common test configuration and exclude the backend-falvored tests from being run in the build.gradle file. In the console output you might see backtraces and logs. This does not mean that tests failed, as some of them tests correct behavior on failure.
Transbase has a free edition that can be used with SQP. To do so, install it at a specific location
and copy the eligible Transbase JDBC driver to that location.
You need to then build and run the server by explicitly defining the location of the tbjdbc.jar
in the
installation location of Transbase Free.
You can also try to define the location of the Transbase Free installation by setting the Java property
LINKED_IN_PATH
. Then you don't need to explicitly copy and define the location of tbjdbc.jar
.
The backend configuration must then set "pipe": true
instead of a host and port.
The integration tests currently do not work with Transbase Free.
Due to some limitations of Transbase Free, the integration tests currently fail with it.
The ideas behind the project are explained in the master thesis.
You can build and read the Javadoc documentation by executing the javadoc
gradle task:
./gradlew javadoc
The documentation can then be found in build/docs
.
SQP uses "dual licensing". Under this model, developers can freely choose to use SQP under the free software/Open Source GNU Affero General Public License Version 3 (commonly known as the "AGPLv3") or under a commercial license.
A copy of AGPLv3 can be found in the license file. For a commercial license, please contact Rothmeyer Consulting (see below).
Copyright Holder: Rothmeyer Consulting
Author: Stefan Burnicki [email protected]