2017-03-04 18:09:10 +00:00
|
|
|
{ fetchurl, buildPythonPackage, isPy33, lib, six, pythonAtLeast, pythonOlder }:
|
|
|
|
|
|
|
|
buildPythonPackage rec {
|
2017-10-25 19:04:35 +01:00
|
|
|
version = "3.2.5";
|
2017-05-27 10:25:35 +01:00
|
|
|
pname = "nltk";
|
|
|
|
name = "${pname}-${version}";
|
2017-03-04 18:09:10 +00:00
|
|
|
|
|
|
|
src = fetchurl {
|
|
|
|
url = "mirror://pypi/n/nltk/nltk-${version}.tar.gz";
|
2017-10-25 19:04:35 +01:00
|
|
|
sha256 = "2661f9971d983db314bbebd51ba770811a362c6597fd0f303bb1d3beadcb4834";
|
2017-03-04 18:09:10 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
propagatedBuildInputs = [ six ];
|
|
|
|
|
|
|
|
disabled = pythonOlder "2.7" || pythonOlder "3.4" && (pythonAtLeast "3.0");
|
|
|
|
|
|
|
|
# Tests require some data, the downloading of which is impure. It would
|
|
|
|
# probably make sense to make the data another derivation, but then feeding
|
|
|
|
# that into the tests (given that we need nltk itself to download the data,
|
|
|
|
# unless there's an easy way to download it without nltk's downloader) might
|
|
|
|
# be complicated. For now let's just disable the tests and hope for the
|
|
|
|
# best.
|
|
|
|
doCheck = false;
|
|
|
|
|
|
|
|
meta = {
|
|
|
|
description = "Natural Language Processing ToolKit";
|
|
|
|
homepage = http://nltk.org/;
|
|
|
|
license = lib.licenses.asl20;
|
|
|
|
maintainers = with lib.maintainers; [ lheckemann ];
|
|
|
|
};
|
|
|
|
}
|