Mal is a Clojure inspired Lisp interpreter.
Mal is implemented in 40 different languages:
- GNU awk
- Bash shell
- C
- C++
- C#
- Clojure
- CoffeeScript
- Crystal
- Elixir
- Erlang
- ES6 (ECMAScript 6 / ECMAScript 2015)
- F#
- Factor
- Forth
- Go
- Groovy
- GNU Guile
- Haskell
- Java
- JavaScript (Online Demo)
- Julia
- Lua
- GNU Make
- mal itself
- MATLAB
- miniMAL
- Nim
- OCaml
- Perl
- PHP
- Postscript
- Python
- RPython
- R
- Racket
- Ruby
- Rust
- Scala
- Swift
- Visual Basic.NET
Mal is a learning tool. See the make-a-lisp process guide. Each implementation of mal is separated into 11 incremental, self-contained (and testable) steps that demonstrate core concepts of Lisp. The last step is capable of self-hosting (running the mal implementation of mal).
The mal (make a lisp) steps are:
- step0_repl
- step1_read_print
- step2_eval
- step3_env
- step4_if_fn_do
- step5_tco
- step6_file
- step7_quote
- step8_macros
- step9_try
- stepA_mal
Mal was presented publicly for the first time in a lightning talk at Clojure West 2014 (unfortunately there is no video). See mal/clojurewest2014.mal for the presentation that was given at the conference (yes the presentation is a mal program).
If you are interesting in creating a mal implementation (or just interested in using mal for something), please drop by the #mal channel on freenode. In addition to the make-a-lisp process guide there is also a mal/make-a-lisp FAQ where I attempt to answer some common questions.
The GNU awk implementation was created by Miutsuru kariya
The GNU awk implementation of mal has been tested with GNU awk 4.1.1.
cd gawk
gawk -O -f stepX_YYY.awk
cd bash
bash stepX_YYY.sh
The C implementation of mal requires the following libraries (lib and header packages): glib, libffi6 and either the libedit or GNU readline library.
cd c
make
./stepX_YYY
The C++ implementation was created by Stephen Thirlwall (sdt)
The C++ implementation of mal requires g++-4.9 or clang++-3.5 and
a readline compatible library to build. See the cpp/README.md for
more details:
cd cpp
make
# OR
make CXX=clang++-3.5
./stepX_YYY
The C# implementation of mal has been tested on Linux using the Mono C# compiler (mcs) and the Mono runtime (version 2.10.8.1). Both are required to build and run the C# implementation.
cd cs
make
mono ./stepX_YYY.exe
cd clojure
lein with-profile +stepX trampoline run
sudo npm install -g coffee-script
cd coffee
coffee ./stepX_YYY
The Crystal implementation of mal was created by Linda_pp
The Crystal implementation of mal has been tested with Crystal 0.8.0.
cd crystal
crystal run ./stepX_YYY.cr
# OR
make # needed to run tests
./stepX_YYY
The Elixir implementation was created by Martin Ek (ekmartin)
The Elixir implementation of mal has been tested with Elixir 1.0.5.
cd elixir
mix stepX_YYY
# Or with readline/line editing functionality:
iex -S mix stepX_YYY
The Erlang implementation was created by Nathan Fiedler (nlfiedler)
The Erlang implementation of mal requires Erlang/OTP R17 and rebar to build.
cd erlang
make
# OR
MAL_STEP=stepX_YYY rebar compile escriptize # build individual step
./stepX_YYY
The ES6 implementation uses the babel compiler to generate ES5 compatible JavaScript. The generated code has been tested with Node 0.12.4.
cd es6
make
node build/stepX_YYY.js
The F# implementation was created by Peter Stephens (pstephens)
The F# implementation of mal has been tested on Linux using the Mono F# compiler (fsharpc) and the Mono runtime (version 3.12.1). The mono C# compiler (mcs) is also necessary to compile the readline dependency. All are required to build and run the F# implementation.
cd fsharp
make
mono ./stepX_YYY.exe
The Factor implementation was created by Jordan Lewis (jordanlewis)
The Factor implementation of mal has been tested with Factor 0.97 (factorcode.org).
cd factor
FACTOR_ROOTS=src factor -run=stepX_YYY
The Forth implementation was created by Chris Houser (chouser)
cd forth
gforth stepX_YYY.fs
The Go implementation of mal requires that go is installed on on the path. The implementation has been tested with Go 1.3.1.
cd go
make
./stepX_YYY
The Groovy implementation of mal requires Groovy to run and has been tested with Groovy 1.8.6.
cd groovy
make
groovy ./stepX_YYY.groovy
The Guile implementation was created by Mu Lei (NalaGinrut).
cd guile
guile -L ./ stepX_YYY.scm
Install the Haskell compiler (ghc/ghci), the Haskell platform and either the editline package (BSD) or the readline package (GPL). On Ubuntu these packages are: ghc, haskell-platform, libghc-readline-dev/libghc-editline-dev
cd haskell
make
./stepX_YYY
The Java implementation of mal requires maven2 to build.
cd java
mvn compile
mvn -quiet exec:java -Dexec.mainClass=mal.stepX_YYY
# OR
mvn -quiet exec:java -Dexec.mainClass=mal.stepX_YYY -Dexec.args="CMDLINE_ARGS"
cd js
npm update
node stepX_YYY.js
The Julia implementation of mal has been tested with Julia 0.3.7.
cd julia
julia stepX_YYY.jl
Running the Lua implementation of mal requires lua 5.1 or later, luarocks and the lua-rex-pcre library installed.
cd lua
make # to build and link linenoise.so
./stepX_YYY.lua
Running the mal implementation of mal involves running stepA of one of the other implementations and passing the mal step to run as a command line argument.
cd IMPL
IMPL_STEPA_CMD ../mal/stepX_YYY.mal
cd make
make -f stepX_YYY.mk
The Nim implementation was created by Dennis Felsing (def-)
Running the Nim implementation of mal requires Nim 0.11.0 or later.
cd nim
make
# OR
nimble build
./stepX_YYY
The OCaml implementation was created by Chris Houser (chouser)
cd ocaml
make
./stepX_YYY
The MATLAB implementation of mal has been tested with MATLAB version R2014a on Linux. Note that MATLAB is a commercial product. It should be fairly simple to support GNU Octave once it support classdef object syntax.
cd matlab
./stepX_YYY
matlab -nodisplay -nosplash -nodesktop -nojvm -r "stepX_YYY();quit;"
# OR with command line arguments
matlab -nodisplay -nosplash -nodesktop -nojvm -r "stepX_YYY('arg1','arg2');quit;"
miniMAL is small Lisp interpreter implemented in less than 1024 bytes of JavaScript. To run the miniMAL implementation of mal you need to download/install the miniMAL interpreter (which requires Node.js).
cd miniMAL
# Download miniMAL and dependencies
npm install
export PATH=`pwd`/node_modules/minimal-lisp/:$PATH
# Now run mal implementation in miniMAL
miniMAL ./stepX_YYY
For readline line editing support, install Term::ReadLine::Perl or Term::ReadLine::Gnu from CPAN.
cd perl
perl stepX_YYY.pl
The PHP implementation of mal requires the php command line interface to run.
cd php
php stepX_YYY.php
The Postscript implementation of mal requires ghostscript to run. It has been tested with ghostscript 9.10.
cd ps
gs -q -dNODISPLAY -I./ stepX_YYY.ps
cd python
python stepX_YYY.py
You must have rpython on your path (included with pypy).
cd rpython
make # this takes a very long time
./stepX_YYY
The R implementation of mal requires R (r-base-core) to run.
cd r
make libs # to download and build rdyncall
Rscript stepX_YYY.r
The Racket implementation of mal requires the Racket compiler/interpreter to run.
cd racket
./stepX_YYY.rkt
cd ruby
ruby stepX_YYY.rb
The rust implementation of mal requires the rust compiler and build tool (cargo) to build.
cd rust
cargo run --release --bin stepX_YYY
Install scala and sbt (http://www.scala-sbt.org/0.13/tutorial/Installing-sbt-on-Linux.html):
cd scala
sbt 'run-main stepX_YYY'
# OR
sbt compile
scala -classpath target/scala*/classes stepX_YYY
The Swift implementation was created by Keith Rollin
The Swift implementation of mal requires the Swift 2.0 compiler (XCode 7.0) to build. Older versions will not work due to changes in the language and standard library.
cd swift
make
./stepX_YYY
The VB.NET implementation of mal has been tested on Linux using the Mono VB compiler (vbnc) and the Mono runtime (version 2.10.8.1). Both are required to build and run the VB.NET implementation.
cd vb
make
mono ./stepX_YYY.exe
The are nearly 500 generic functional tests (for all implementations)
in the tests/ directory. Each step has a corresponding test file
containing tests specific to that step. The runtest.py test harness
launches a Mal step implementation and then feeds the tests one at
a time to the implementation and compares the output/return value to
the expected output/return value.
To simplify the process of running tests, a top level Makefile is provided with convenient test targets.
- To run all the tests across all implementations (be prepared to wait):
make test
- To run all tests against a single implementation:
make test^IMPL
# e.g.
make test^clojure
make test^js
- To run tests for a single step against all implementations:
make test^stepX
# e.g.
make test^step2
make test^step7
- To run tests for a specific step against a single implementation:
make test^IMPL^stepX
# e.g
make test^ruby^step3
make test^ps^step4
- To run the functional tests in self-hosted mode, you specify
malas the test implementation and use theMAL_IMPLmake variable to change the underlying host language (default is JavaScript):
make MAL_IMPL=IMPL test^mal^step2
# e.g.
make test^mal^step2 # js is default
make MAL_IMPL=ruby test^mal^step2
make MAL_IMPL=python test^mal^step2
Warning: These performance tests are neither statistically valid nor comprehensive; runtime performance is a not a primary goal of mal. If you draw any serious conclusions from these performance tests, then please contact me about some amazing oceanfront property in Kansas that I'm willing to sell you for cheap.
- To run performance tests against a single implementation:
make perf^IMPL
# e.g.
make perf^js
- To run performance tests against all implementations:
make perf
- To report line and byte statistics for a single implementation:
make stats^IMPL
# e.g.
make stats^js
- To report line and bytes statistics for general Lisp code (env, core and stepA):
make stats-lisp^IMPL
# e.g.
make stats-lisp^js
There is a Dockerfile included in the tests/docker directory that
builds a docker image based on Ubuntu Utopic that contains everything
needed to run tests against all the implementations (except for MATLAB
which is proprietary/licensed).
Build the docker image using a provided script. WARNING: this will likely take over an hour to build from scratch and use more 3 GB of disk:
./tests/docker-build.shLaunch a docker container from that image built above. This will
volume mount the mal directory to /mal and then give you a bash
prompt in the container. You can then run individual mal
implementations and tests:
./tests/docker-run.shYou can also specify a command to run within the container. For example, to run step2 tests for every implementation (except MATLAB):
./tests/docker-run.sh make SKIP_IMPLS="matlab" test^step2Notes:
- JVM-based language implementations (Java, Clojure, Scala): you will need to run these implementations once manually first before you can run tests because runtime dependencies need to be downloaded to avoid the tests timing out. These dependencies are download to dot-files in the /mal directory so they will persist between runs.
- Compiled languages: if your host system is different enough from Ubuntu Utopic then you may need to re-compile your compiled languages from within the container to avoid linker version mismatches.
Mal (make-a-lisp) is licensed under the MPL 2.0 (Mozilla Public License 2.0). See LICENSE.txt for more details.