- awk
- make (GNU or BSD)
- C compiler (tcc and gcc are used for development)
- libbio,libfmt,libutf
cd src
make
You may need to tweak the CC,LD,CFLAGS and LDFLAGS variables in the Makefile. Also, if your C compiler does not support flexible array members like the example
struct foo { int x; char m[] };
You'll need to tweak the declaration for array and ASIZE in apl.h. The source is developed using gcc and tcc, on both x86 and x86_64 platforms.
This project is my endeavour to create an open-source APL interpreter that is useful for every-day use. I want this interpreter to be used much in the same way as 'awk' is used in the Unix environment; for one-off scripts, to poke and prod at data. To this end, I have the following design goals in mind:
More informative of this interpreter is the list of features it WON'T have: no object system. No foreign function interface. No graphical interface. No networking functionality.
The one feature it will have is the ability to spawn processes and write to their stdin/stdout. With this feature, most other functionality can be implemented using helper programs.
APL caught a lot of flak for its use of non-ascii characters that required a special encoding. Many APL systems used their own private encoding, making it difficult for users of different APL systems to share programs.
These days Unicode is pretty wide-spread, and the encoding problem is no longer a problem anymore. This interpreter uses unicode for everything. Identifiers, primitives, box-drawing characters, everything. So
nước ← 'USA'⊃'日本'⊃'Việt Nam'⊃'Россия'
右 ← ⊢
左 ← ⊣
Are all examples of valid code. There is no special IDE or terminal that is required to input the characters; most modern operating systems provide methods to input unicode characters. Documentation and keymaps will be provided with the distribution for inputting the common APL characters, and a dedicated terminal for each OS may (or may not) be written. I tend to do just fine just using rxvt-unicode or 9term on linux, but for windows a dedicated ide may be in order.
As mentioned before, this interpreter has the ability to spawn os commands and manipulate their stdin/stdout. Using this we can implement many features that are usually baked into the core of the language or loaded as libraries, like:
- Matrix visualization
- Function plotting
- Networking (via
netcat
or similar)
'A dictionary of APL' was written by the creator of APL himself and is a good formal standard to follow. This interpreter will conform to this standard where possible. Notably, bracketed array indexing/assignment(a[n;]) will not be in the language in favor of the 'from' ({) function and 'merge' (}) operator. All functions will have the notion of function rank as well.
I may diverge from this standard in a few areas. For example I'm still undecided on function definition; the Dictionary uses the function ∇ with string arguments for the monadic and dyadic cases. I find this clumsy. I like Dyalog's Dfns, but the '{}' symbols they use for delimiters are already in use as functions/operators.
Work in progress - rudimentary lexer and parser are complete. Unified array data structure is implemented. Basic memory management based on ref-counting implemented. Handy error-reporting system is in place. Function definition syntax using bracket delimiters is recognized. Multi-dimensional arrays are printed correctly, as are arbitrarily nested boxed arrays using unicode box-drawing characters.
Most of the boring stuff is out of the way. I am still planning out the core execution loop and function dispatch/definition. One-by-one standard functions will be added. The project is not currently at a state where outside help would be beneficial, or a fork useful. The lexer, while trivial, may be of interest as it accepts utf-8 input. An earlier version of the lexer also output tokens in a fairly portable streaming format.