Merge pull request #275 from henningpeters/unidecode

remove text-unidecode dependency

requirements.txt

@@ -4,7 +4,6 @@ pathlib
 preshed>=0.46.1,<0.47.0
 thinc>=5.0.0,<5.1.0
 murmurhash>=0.26,<0.27
-text-unidecode
 numpy
 plac
 six

setup.py

@@ -174,7 +174,7 @@ def setup_package():
             license=about['__license__'],
             ext_modules=ext_modules,
             install_requires=['numpy', 'murmurhash>=0.26,<0.27', 'cymem>=1.30,<1.32.0', 'preshed>=0.46.1,<0.47',
-                              'thinc>=5.0.0,<5.1.0', 'text_unidecode', 'plac', 'six',
+                              'thinc>=5.0.0,<5.1.0', 'plac', 'six',
                               'ujson', 'cloudpickle', 'sputnik>=0.9.0,<0.10.0'],
             cmdclass = {
                 'build_ext': build_ext_subclass},

spacy/orth.pxd

@@ -10,4 +10,3 @@ cpdef bint like_email(unicode string)
 cpdef bint like_url(unicode string)
 cpdef bint like_number(unicode string)
 cpdef unicode word_shape(unicode string)
-cpdef bytes asciied(unicode string)

spacy/orth.pyx

@@ -2,21 +2,8 @@
 # cython: infer_types=True
 from __future__ import unicode_literals
 import unicodedata
-
-# If your license is not GPL compatible, use text_unidecode. But if your code
-# is, you should use the unidecode library, because its performance is better.
-# spaCy does not list unidecode as a dependency, in case your license is not
-# GPL compatible.
-try:
-    from unidecode import unidecode
-except ImportError:
-    from text_unidecode import unidecode
-
-
 import re
 
-import math
-
 
 TAGS = 'adj adp adv conj det noun num pdt pos pron prt punct verb'.upper().split()
 
@@ -171,23 +158,6 @@ cpdef unicode word_shape(unicode string):
     return ''.join(shape)
 
 
-cpdef unicode norm1(unicode string, lower_pc=0.0, upper_pc=0.0, title_pc=0.0):
-    """Apply level 1 normalization:
-
-    * Case is canonicalized, using frequency statistics
-    * Unicode mapped to ascii, via unidecode
-    * Regional spelling variations are normalized
-    """
-    pass
-
-
-cpdef bytes asciied(unicode string):
-    stripped = unidecode(string)
-    if not stripped:
-        return b'???'
-    return stripped.encode('ascii')
-
-
 # Exceptions --- do not convert these
 _uk_us_except = set([
     'our',