A python tool to distributively run GCJ/FB hacker cup testcases on several machines.
How does it work
- First the splitter runs on the input file to split it into seprate test cases.
- Test cases with the actuall code are sent to the server.
- The server distributes the test cases and the actuall code to its registered nodes.
- Each node computes the result for its testcases and sends the result back to the server
- The server collects the results and sort them by their testcase number and sends them back to the calling script.
pip install dTests
- Create new project
dtests_job new project_name language
- The only supported languages now are "cpp" and "java".
- Open the new folder and code your splitter to read from the input.in file.
- Between each test cases print
--split--\n. The split marker can be configured in
config.jsonfile. This will split the input file into single test cases which will be distributed on the machines running.
- Code your program file to read from stdin as if it is reading a single test cases.
- Run the server
- Run one or many nodes on other machines ( or for testing on the same machine )
dtests_node --host host --port port
- Finally run the job
- Check the help of these commands for further customizations.
This is a sample cpp project that takes input and multiply it by itself https://github.com/MohamedBassem/dTests/tree/master/samples/power2
- The config.json file contains the project configuration. The language and the splitter syntax.
- The splitter.cpp file contains the code that reads the input from the
input.infile and splits it to single test cases.
- The program.cpp file that reads a single test cases from stdin and computes the result.
- To Run this project, run the server and at least one node. Then run
dtests_job runin the project's directory.
- Finding a way to pass the testcase number to the program
- Writing Tests.
- Making the system fault tolerant.
- Finding a better way for splitting the input file.
- Supporting more programming languages.
Your contributions ( Ideas, improvments, etc.. ) are more than welcome.
Licensed under the MIT license.