an experimental frankenstein
what if guido was a type theory fan? what if we went to an alternate dimension, where python starts like a typed functional language, instead of a class-based oop one?
a python-like language with hindley-milner-like type system, which is compiled to c.
code by Robert Smallshire (see below)
We use the new syntax for annotations in Python3
You can add optional annotations to args and returns:
@template(y, z)
def f_map(f: y >> z, s: [y]) -> [z]:
out = []
for i in s:
out = append(out, f(i))
return out
def nope(a: Integer) -> Integer:
return a + 4
print(f_map(nope, [2, 4]))
def sum(n: Integer) -> Integer:
'''sum of the numbers from 0 to n inclusively'''
result = 0
for i in range(0, n + 1):
result += i
return result
print(sum(2000))
owever we just reuse Python3's syntax. We try to preserve the spirit and semantics in many cases, but look at it like a different language.
you can install it and use it like that
git clone https://github.com/alehander42/airtight.git
cd airtight
bin/airtight # and
The core library contains some predefined functions, accessible from airtight and
some template files which implement generic airtight functions in c
(for example core/list.c
has placeholders like %{list_type}
and %{elem_type}
and
different copies of the functions are generated for each %{elem_type}
in a program})
options for the compiler:
bin/airtight filename.py # compiles it to filename.py.c
bin/airtight filename.py --to-binary # compiles it to filename.py.c and then invokes
# c99 and generates a binary filename
bin/airtight filename.py --hm-ast # show the hindley milner ast
bin/airtight filename.py --typed-hm-ast # show the typed hindley milner ast
bin/airtight filename.py --typed-c-ast # show the lower level typed ast
the resulting c code is quite amusing:
a_print_AList_AString_AString(
a_f_map_intREFAString_AList_int_AList_AString(
&a_wtf_int_AString,
AList_intOf(2, 2, 4)));
yeah, you need a shower now, doncha
syntactic sugar : [Integer]
for list types, y >> z
for function types,
Integer | Float
for union types. we have some extensions to the hindley milner
inference algorithm, so the current type system is probably unsound and weird, but
you have haskell/idris for that
airtight is just a little kid now and its quite buggy and a proof of concept however if you're interested, you have some ideas/questions, or you just wanna hack on it, feel free to use the issues tab here
Alexander Ivanov