An esoteric visual language that takes image files as input based on a multi-tape turing machine, designed for compatibility with C.
An esoteric visual language that takes image files as input based on a multi-tape turing machine, designed for compatibility with C.
Here is an implementation of memcpy
:
Yes, you literally pass an image file to the compiler.
The “parser” is based on computer vision and the backend produces C code.
Here’s how the parser understands the program (produced if you pass --debug-parser
when compiling):
Here’s a C program which calls into the function:
#include <stdio.h>
// Supplied by linking with the vizh object file
void memcopy(uint8_t*,uint8_t*,uint8_t*);
int main() {
uint8_t str[] = "Hello!";
uint8_t size = sizeof(str);
uint8_t to[sizeof(str)];
memcopy(&size, str, to);
puts(to);
}
We can compile this into an executable in a single command:
$ vizh memcopy.png main_memcopy.c -o memcopy
Then run it:
$ ./memcopy
Hello!
But what if you make an error, like this?
Well you get the best compiler errors you’ll ever see:
The provided implementation is called vizh
and compiles to C and can link executables.
You can install vizh
directly from PyPi:
$ pip install vizh
vizh
depends on OpenCV, cffi, and Tesseract OCR.
You can install OpenCV and cffi with pip
:
$ pip install opencv-python cffi
You’ll have to install Tesseract OCR separately. See their documentation for instructions.
Usage: vizh [OPTIONS] [INPUTS]...
Options:
--version Show the version and exit.
-c, --compile-only Only compile, don't link.
-o, --output-file PATH Output file for executables or vizh object files.
-q, --quiet Suppress output.
--debug-parser Display how the parser understands your source file.
--help Show this message and exit.
The compiler can take any combination of image files, C sources files, and object files.
You may need to set the TESSDATA_PREFIX
environment variable to the folder containing Tesseract data. If you’re on Linux this is likely /usr/share/tesseract-ocr/<version>/tessdata
.
The vizh abstract machine consists of:
The initial state of the abstract machine is:
See instructions for the valid operations on the abstract machine.
A vizh program consists of a number of functions, each in its own image file. (What image types are allowed? Ideally at least png and jpg)
The entry point to a vizh program is a function called main
. (Note that the main
function gets mangled as vizh_main
. For all other functions the symbol name is the same as the vizh name).
A vizh function is an image file containing:
Function names are alphanumeric: [a-zA-Z][a-zA-Z0-9]*
.
The tapes available to a vizh function consist of its tape arguments. On entry to the function the r/w head is initialised to the start of the first tape argument, if any.
A function returns when control flow reaches the end of its instructions.
Any tapes allocated by a function are automatically deallocated when the function exits.
When you call a function subsequent pointer arguments are taken from the currently active tape onwards.
For example, given the following state of the abstract machine where ^
is the last position of the r/w head on that tape and $
is the active tape:
t1 01234
^
$t2 99999
^
t3 00000
^
Then a call to a function that takes two tapes would supply the arguments t2, t3
.
The valid instructions in vizh and their encodings are:
1
1
0
When you move the r/w head up or down, the position it was last at for the previous tape is saved. E.g. given this state of the abstract where ^
is the last position of the r/w head on that tape and $
is the active tape:
$t0 01234
^
t1 01234
^
The sequence of instructions “right right right down” would result in this state:
t0 01234
^
$t1 01234
^
Comments in vizh are anything enclosed in a rectangle. Stick what you want in there.
The vizh standard library is called libv
. Much of it is implemented in vizh itself and it is built when you install vizh
. It provides the following functions:
readin
: read an ASCII character from stdin and write its integral representation into the cell pointed to by the r/w headprint
: print the value of the cell pointed to by the r/w head to stout, interpreted as an ASCII characterputstr
: write the null-terminated ASCII string starting at the position pointed to by the r/w head to stdout.geta
: puts the character a
at the current position of the r/w head.getA
: puts the character A
at the current position of the r/w head.add
given tape cells a,b
from the r/w head, results in a+b,0
.mul
given tape cells a,b,c
from the r/w head, results in a*b,0,0
.zero
given tape cell a
from the r/w head, results in 0
newtape
: allocate a new secondary tape underneath the last one currently allocated for this function (or the primary tape if there are no secondary tapes)freetape
: deallocate the bottom-most secondary tape for this function (no-op if there are not any)