Babycino v0.5

This is a minimal compiler for MiniJava, a subset of Java used in the book "Modern Compiler Implementation in Java, Second Edition" by Andrew Appel and Jens Palsberg. See here for more information about MiniJava:

http://www.cambridge.org/resources/052182060X/MCIIJ2e/

The compiler is written in Java, using the parser generator ANTLR 4. It was originally written to support the module CS2CO16: Compilers taught at the University of Reading.

For your convenience, ANTLR 4 and a script to run it are bundled in the archive. There is also a script to run the "grun" grammar testing tool, which produces images of parse trees. Note that you must build the compiler before you can use grun.

Linux Users

Open a command prompt, navigate to the directory containing this README.md, then compile the compiler using the script build.sh:

$ ./build.sh

This runs ANTLR to generate the MiniJava parser, then runs the Java compiler on the Babycino source code and the parser generated by ANTLR.

To compile a program, use the script babycino.sh. For example:

$ ./babycino.sh progs/appel/Factorial.java factorial.c

This runs Babycino, which generates C code as output. To run the C program, you need to compile it with a C compiler. For example:

$ gcc factorial.c -o factorial.o
$ ./factorial.o 
3628800

(The C compiler gcc is installed on most Linux systems, but you can use a different C compiler if you prefer.)

To view a parse tree of a program, use grun. For example:

$ ./grun.sh babycino.MiniJava goal -gui progs/appel/Factorial.java

Windows Users

Open a command prompt, navigate to the directory containing this README.txt, then compile the compiler using the script build.bat:

> build

This runs ANTLR to generate the MiniJava parser, then runs the Java compiler on the Babycino source code and the parser generated by ANTLR.

To compile a program, use the script babycino.bat. For example:

> babycino progs\appel\Factorial.java factorial.c

This runs Babycino, which generates C code as output. To run the C program, you need to compile it with a C compiler. For example:

> tcc factorial.c -o factorial.exe
> factorial.exe
3628800

(The C compiler tcc is included in the distribution, but you can use a different C compiler if you prefer.)

To view a parse tree of a program, use grun. For example:

$ grun babycino.MiniJava goal -gui progs\appel\Factorial.java

Other Scripts

The cleanup script deletes the files generated by ANTLR and .class files created during compilation.

The test script runs Babycino on Factorial.java, compiles the C output to an executable, runs the executable, and checks the result is correct. This shows that some basic features of the compiler are working, but not much else.

The test-my-archive script takes a .zip or .tar.gz archive of the Babycino source code, extracts it to a temporary directory, compiles it and runs the test script. This is useful for checking an archive before releasing it.

Using an IDE

Using an IDE to modify the compiler is complicated, as without careful setup, the IDE will not know to run ANTLR after you change the grammar. It will generally be simpler to use a good text editor. Notepad++ is a popular choice for Windows users.

Bugs

Please report any bugs you find. Examples of bugs are crashes (for example, unhandled null-pointer exceptions) or differences in the behaviour of a MiniJava program when compiled with Babycino, as compared with the official Java compiler.

As Babycino maps Java ints onto C ints, some aspects of integer arithmetic may not match the Java specification exactly. This is intentional.

Licensing

Babycino is distributed under the MIT License.

The sample programs in progs/appel/ are from the aforementioned textbook, "Modern Compiler Implementation in Java, Second Edition". They remain the property of the original authors.

The contents of contrib/antlr/ are from the ANTLR, which is distributed under the BSD License.

The contents of contrib/tcc/ are from TCC, which is distributed under the GNU Lesser General Public License.

The contents of contrib/infozip/ are from Info-ZIP, which is distributed under a BSD-like licence.

Changelog

• Version 0.5: Minor bugfix for 2021-22 academic year.

  • CGenerator: Actually use correct union field for printing ints.

• Version 0.4: More improvements for 2020-21 academic year.

  • Examples: Programs for suggested extra features added in progs/features/.
  • General: Some assertions changed into exceptions, so that they will trigger even when built in release mode.
  • CGenerator: Use correct union field for printing ints with write.
  • CGenerator: Implement read intermediate language instruction.
  • CGenerator: Initialise all local variables to 0 or equivalent.
  • General: Project now hosted on CSGitLab.

• Version 0.3: Various minor improvements for 2019-20 academic year.

  • Grammar: Correct Java precedence observed.
  • TACGenerator: Fixed bug that occurred when generating a label in main().
  • General: Explicit @Override used for classes that override Java visitors/listeners. Stopped using deprecated ANTLR FileStream API.
  • Build: ANTLR libraries bundled to avoid editing classpaths. TCC bundled for Windows users with no C compiler. Build scripts use these. Windows build/compile scripts added. Zip/unzip binaries added.
  • Readme: Updated to reflect the above changes.

• Version 0.2: Peephole optimiser and dead/redundant code removal added.

  • Optimisation: TACBlockOptimiser.java, TACDeadCodeOptimiser.java and TACFlowAnalysis.java added.
  • TAC: NOP (no operation) added to TACOp.java and TACOpType.java.
  • General: Babycino.java tidied up.
  • Readme: Note about ANTLR and -package on Windows added.
  • Build: test-my-archive.sh added to help people check their archives compile OK.

• Version 0.1: Initial release. All of MiniJava supported.