Skip to content

niksu/motto

 
 

Repository files navigation

Motto is a tool that processes Flick programs:

  • It compiles programs to run on a separate back-end runtime system.

  • Provides a reference runtime for running Flick programs without requiring additional dependencies.

  • Enables querying of Flick programs -- for instance, type-checking Flick expressions from the command-line.

NOTE Motto accepts several switches. At the moment most of them are only described in the code.

Building

To compile Motto:

$ ./build.sh

If compilation fails, check that you have all dependencies listed in the "Dependencies" section below.

Try running ./motto.byte -h to see if you have a binary. See the section on "Testing" for how to run the tool over regression tests.

Using Docker

If you have Docker installed, you can also use a container to run the compiler without installing all the dependencies. The source code can be mounted in /src within the container.

$ docker build -t motto .
$ docker run -v `pwd`/tests:/src motto -q --parser_test_file flick_code/types_process.cp

Getting going

We start with the mathematical cousin of "Hello, world" -- i.e., the factorial program. Look at the factorial function in Flick. To run this, we need an environment that can interpret it -- Motto provides this. This interpreter requires a runtime, which connects it with resources on the machine on which Flick programs are interpreted. For instance, sockets aren't a concept that Flick provides; they are resources that the machine makes available to the runtime, that in turn provides them to running Flick programs as channels.

Motto provides a flexible scriptable-runtime. That is, in addition to running Flick programs, you can programmatically control different aspects of the runtime. This is useful, for instance, if you want to alter the contents of channels, to make tests more repeatable and specific. Or you could simulate phenomena that are beyond the control of Flick programs, such as channels breaking. If you have the time, it's not difficult to turn this into an interactive runtime -- i.e., a "toplevel" or Read-Eval-Print Loop, a.k.a., "REPL".)

Now look at an example of the runtime's use to run the factorial function. Using the scriptable runtime, you can compile and run Flick programs using the OCaml compiler and runtime. To compile the example:

$ ./build.sh factorial.byte

and then run!

$ ./factorial.byte

You should get the following output:

let y = 13 ~> 13
let z = 5 + (y * ?somechan) + y ~> 70
factorial (z - (60 + ?somechan)) ~> 40320

This output reproduces each expression that was evaluated by the runtime, followed by ~>, and followed by the value that the expression evaluated to.

Tracing the evaluation within expressions is not difficult to implement -- essentially by adding and using a hook in the evaluation monad -- but currently this is not implemented.

Take a look at this example for a comprehensive example of using the scriptable runtime.

Compiler

To compile a file named source_file run the following command:

$ ./motto.byte -o output_directory source_file

where output_directory is the name of the directory (that doesn't yet exist) that's intended to hold the compiler's output. The compiler will create output_directory and deposit the output files there, where they can be compiled and linked using a separate toolchain.

Other than the reference runtime system, described above, only a single back-end runtime system is supported. Thus code is generated exclusively for this runtime.

Testing

For testing, I'm currently using the following as source_file:

  • examples/hadoop_wc_type.cp
  • tests/flick_code/simple_function.cp

To run the parser regression tests on both the examples and tests directories:

$ ./motto.byte -q --parser_test_dir examples --parser_test_dir tests/flick_code

Or you might wish to only test a single file:

./motto.byte -q --parser_test_file tests/flick_code/types_process.cp

To run type inference on expressions:

$ ./motto.byte -q --infer_type "(| E |)"

where E is an expression.

To run the type checker on regression tests, see the test_type_checker.sh script.

Background

The Flick language implements the crisp language idea, but simplified as described in the Flick description.

Dependencies

Other than the OCaml compiler (tested with v4.03.0) this project has the following dependencies:

  • ocamlfind. This in turn depends on:
    • m4
  • menhir

License and support

Motto is distributed under the terms of the Apache 2.0 license.

The Motto tool was developed as part of the Network-as-a-Service project (NaaS), which was funded by the UK's Engineering and Physical Sciences Research Council through grants EP/K034723/1, EP/K031724/1, and EP/K032968/1. These grants were held at the following universities respectively:

  • Computer Laboratory, University of Cambridge
  • School of Computer Science, University of Nottingham
  • Department of Computing, Imperial College London

About

Tool for translating and running Flick programs

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • OCaml 94.3%
  • C++ 3.1%
  • Standard ML 2.4%
  • Other 0.2%