Calling Haskell functions from Python
Question:
I want to use some Haskell libraries (e.g. Darcs, Pandoc) from Python, but it seems there’s no direct foreign function interface to Haskell in Python. Is there any way to do that?
Answers:
Provided you can get your Python code to call C, you can call Haskell functions that have been exported via the FFI
Another approach would be to write a standard IPC interface, in the case of darcs and pandoc just calling them as vanilla executables and parsing their output might be the way to go.
As to automating the generation of boring, repetitive, FFI and marshalling code on the Haskell side, I’d recommend c2hs, which allows you to auto-generate a lot based on an existing C interface. There’s probably similar things for python.
SWIG, alas, has, to the best of my knowledge, never been implemented for Haskell, presumably because it caters to less strictly-typed languages.
One additional idea: Something less efficient than a direct C binding, but more efficient than shelling out to Haskell is an rpc system such as Apache Thrift: http://incubator.apache.org/thrift/
I’ve found thrift easy to use, well supported, and reasonably performant. Once you have your Haskell server running, the cost of local communication is fairly cheap, although you pay a bit more in marshalling/unmarshalling than using c types directly.
There are also at least two packages for calling Python from Haskell, missingpy (http://hackage.haskell.org/package/MissingPy) and cpython (http://hackage.haskell.org/package/cpython). The latter claims that support in the other direction is planned — although you’d have to ask the author if this is still the case, and if so when.
For pandoc, at least, you can use these C bindings:
https://github.com/toyvo/libpandoc
Noob here.
But I did manage to call user defined Haskell functions from python using Haskell’s FFI.
Basically I compiled the Haskell function to a dll and imported the dll using ctypes in python.
So the function became available in python.
I wrote the procedure here: https://justa0xc0de.wordpress.com/2015/01/08/using_haskell_function_in_python/
Hope this helps.
There is a wrapper that allows one to call Haskell functions from Python here:
https://github.com/sakana/HaPy
From a cursory inspection, it seems to require that the Haskell functions have relatively simple type signatures (basically, all the types involved had better be things like Int and Float which c knows about, or lists of things of this form, or lists of lists, or so on).
An example is provided where one has this Haskell code:
module ExampleModule where
import Data.Char
foo :: Double -> Double -> Double
foo = (*)
bar :: Int -> Int
bar i = sum [1..i]
baz :: Int -> Bool
baz = (> 5)
arr_arg :: [Int] -> Int
arr_arg = sum
arr_ret :: Int -> [Int]
arr_ret i = [1..i]
arr_complex :: [[Int]] -> [[Int]]
arr_complex = map (map (* 2))
string_fun :: String -> String
string_fun str = str ++ reverse str
char_test :: Char -> Int
char_test = ord
and one accesses it like this:
from HaPy import ExampleModule
print "3 * 7 is", ExampleModule.foo(3,7)
print "sum from 1 to 10 is", ExampleModule.bar(10)
print "3 > 5 is", ExampleModule.baz(3)
print "sum from 1 to 100 is", ExampleModule.arr_arg(range(101))
print "numbers from 1 to 10 are", ExampleModule.arr_ret(10)
print "complex array passing:", ExampleModule.arr_complex([range(3), [], range(100)])
print "string fun:", ExampleModule.string_fun("This isn't really a palindrome.")
s = ExampleModule.string_fun("abc 00def")
print "string fun with nulls:", s,
for c in s:
print ord(c),
print
print "char test:", ExampleModule.char_test("t")
Unfortunately, you do need to do some export plumbing on the Haskell side.
Another option is hyphen, which can be found here. Basic usage looks something like:
>>> import hyphen, hs.Prelude
>>> hs.Prelude.sum([1,2,3]) # list converted to Haskell list
6
>>> hs.Prelude.drop(5, "Hello, world")
", world"
>>> hs.Prelude.drop(1, [1,2,3])
<hs.GHC.Types.[] object of Haskell type [GHC.Integer.Integer], containing '[2,3]'>
>>> list(hs.Prelude.drop(1, [1,2,3])) # Convert back to Python list
[2, 3]
This seems to be a less lightweight solution than some of the other options in other answers.
In return for the extra weight, you seem to get a full bridge from Haskell to Python. Whereas HaPy and github.com/nh2/call-haskell-from-anything only allow you to use a Haskell function from Python if that Haskell function has all its arguments from fairly basic types and returns a fairly basic type, hyphen seems to let you use arbitrary functions. It can do this because it introduces into python a type representing an arbitrary object on the Haskell heap.
These ‘haskell objects viewed from python’ behave fairly nicely as python objects. For example Haskell Maps behave a bit like dictionaries:
>>> import hs.Data.Map
>>> my_map = hs.Data.Map.fromList([(1, 'Hello'), (2, 'World')])
>>> my_map[1]
'Hello'
>>> print(sorted([key for key in my_map]))
[1, 2]
See the readme for many more examples!
It also seems to handle various fancy things like converting exceptions between Haskell and Python.
I want to use some Haskell libraries (e.g. Darcs, Pandoc) from Python, but it seems there’s no direct foreign function interface to Haskell in Python. Is there any way to do that?
Provided you can get your Python code to call C, you can call Haskell functions that have been exported via the FFI
Another approach would be to write a standard IPC interface, in the case of darcs and pandoc just calling them as vanilla executables and parsing their output might be the way to go.
As to automating the generation of boring, repetitive, FFI and marshalling code on the Haskell side, I’d recommend c2hs, which allows you to auto-generate a lot based on an existing C interface. There’s probably similar things for python.
SWIG, alas, has, to the best of my knowledge, never been implemented for Haskell, presumably because it caters to less strictly-typed languages.
One additional idea: Something less efficient than a direct C binding, but more efficient than shelling out to Haskell is an rpc system such as Apache Thrift: http://incubator.apache.org/thrift/
I’ve found thrift easy to use, well supported, and reasonably performant. Once you have your Haskell server running, the cost of local communication is fairly cheap, although you pay a bit more in marshalling/unmarshalling than using c types directly.
There are also at least two packages for calling Python from Haskell, missingpy (http://hackage.haskell.org/package/MissingPy) and cpython (http://hackage.haskell.org/package/cpython). The latter claims that support in the other direction is planned — although you’d have to ask the author if this is still the case, and if so when.
For pandoc, at least, you can use these C bindings:
https://github.com/toyvo/libpandoc
Noob here.
But I did manage to call user defined Haskell functions from python using Haskell’s FFI.
Basically I compiled the Haskell function to a dll and imported the dll using ctypes in python.
So the function became available in python.
I wrote the procedure here: https://justa0xc0de.wordpress.com/2015/01/08/using_haskell_function_in_python/
Hope this helps.
There is a wrapper that allows one to call Haskell functions from Python here:
https://github.com/sakana/HaPy
From a cursory inspection, it seems to require that the Haskell functions have relatively simple type signatures (basically, all the types involved had better be things like Int and Float which c knows about, or lists of things of this form, or lists of lists, or so on).
An example is provided where one has this Haskell code:
module ExampleModule where
import Data.Char
foo :: Double -> Double -> Double
foo = (*)
bar :: Int -> Int
bar i = sum [1..i]
baz :: Int -> Bool
baz = (> 5)
arr_arg :: [Int] -> Int
arr_arg = sum
arr_ret :: Int -> [Int]
arr_ret i = [1..i]
arr_complex :: [[Int]] -> [[Int]]
arr_complex = map (map (* 2))
string_fun :: String -> String
string_fun str = str ++ reverse str
char_test :: Char -> Int
char_test = ord
and one accesses it like this:
from HaPy import ExampleModule
print "3 * 7 is", ExampleModule.foo(3,7)
print "sum from 1 to 10 is", ExampleModule.bar(10)
print "3 > 5 is", ExampleModule.baz(3)
print "sum from 1 to 100 is", ExampleModule.arr_arg(range(101))
print "numbers from 1 to 10 are", ExampleModule.arr_ret(10)
print "complex array passing:", ExampleModule.arr_complex([range(3), [], range(100)])
print "string fun:", ExampleModule.string_fun("This isn't really a palindrome.")
s = ExampleModule.string_fun("abc 00def")
print "string fun with nulls:", s,
for c in s:
print ord(c),
print
print "char test:", ExampleModule.char_test("t")
Unfortunately, you do need to do some export plumbing on the Haskell side.
Another option is hyphen, which can be found here. Basic usage looks something like:
>>> import hyphen, hs.Prelude
>>> hs.Prelude.sum([1,2,3]) # list converted to Haskell list
6
>>> hs.Prelude.drop(5, "Hello, world")
", world"
>>> hs.Prelude.drop(1, [1,2,3])
<hs.GHC.Types.[] object of Haskell type [GHC.Integer.Integer], containing '[2,3]'>
>>> list(hs.Prelude.drop(1, [1,2,3])) # Convert back to Python list
[2, 3]
This seems to be a less lightweight solution than some of the other options in other answers.
In return for the extra weight, you seem to get a full bridge from Haskell to Python. Whereas HaPy and github.com/nh2/call-haskell-from-anything only allow you to use a Haskell function from Python if that Haskell function has all its arguments from fairly basic types and returns a fairly basic type, hyphen seems to let you use arbitrary functions. It can do this because it introduces into python a type representing an arbitrary object on the Haskell heap.
These ‘haskell objects viewed from python’ behave fairly nicely as python objects. For example Haskell Maps behave a bit like dictionaries:
>>> import hs.Data.Map
>>> my_map = hs.Data.Map.fromList([(1, 'Hello'), (2, 'World')])
>>> my_map[1]
'Hello'
>>> print(sorted([key for key in my_map]))
[1, 2]
See the readme for many more examples!
It also seems to handle various fancy things like converting exceptions between Haskell and Python.