Use crfsuite (python wrapper over Rust package bindings) to be comp…

…atible with Python 3.10

README.md

@@ -39,7 +39,7 @@ It also does not normalize the address. However, [this library built on top of u
   ```
 
 ## How to use this development code (for the nerds)
-usaddress uses [parserator](https://github.com/datamade/parserator), a library for making and improving probabilistic parsers - specifically, parsers that use [python-crfsuite](https://github.com/tpeng/python-crfsuite)'s implementation of conditional random fields. Parserator allows you to train the usaddress parser's model (a .crfsuite settings file) on labeled training data, and provides tools for adding new labeled training data.
+usaddress uses [parserator](https://github.com/datamade/parserator), a library for making and improving probabilistic parsers - specifically, parsers that use [crfsuite](https://github.com/chokkan/crfsuite)'s implementation of conditional random fields. Parserator allows you to train the usaddress parser's model (a .crf settings file) on labeled training data, and provides tools for adding new labeled training data.
 
 ### Building & testing the code in this repo
 

setup.py

@@ -13,9 +13,9 @@
     description='Parse US addresses using conditional random fields',
     name='usaddress',
     packages=['usaddress'],
-    package_data={'usaddress': ['usaddr.crfsuite']},
+    package_data={'usaddress': ['usaddr.crf']},
     license='The MIT License: http://www.opensource.org/licenses/mit-license.php',
-    install_requires=['python-crfsuite>=0.7',
+    install_requires=['crfsuite>=0.3.1',
                       'future>=0.14',
                       'probableparsing'],
     classifiers=[

training/README.md

@@ -238,11 +238,11 @@ parserator train training/labeled.xml,training/new_addresses.xml usaddress
 After running the command, you should see output that looks something like this:
 
 ```
-renaming old model: usaddress/usaddr.crfsuite -> usaddress/usaddr_2016_12_19_21286.crfsuite
+renaming old model: usaddress/usaddr.crf -> usaddress/usaddr_2016_12_19_21286.crf
 
 training model on 1359 training examples from ['training/labeled.xml', 'trainingnew_addresses.xml']
 
-done training! model file created: usaddress/usaddr.crfsuite
+done training! model file created: usaddress/usaddr.crf
 ```
 
 This output confirms that usaddress has learned from the new training data. Nice!

usaddress/__init__.py

@@ -13,7 +13,7 @@
     from ordereddict import OrderedDict
 import warnings
 
-import pycrfsuite
+import crfsuite
 import probableparsing
 
 # The address components are based upon the `United States Thoroughfare,
@@ -52,7 +52,7 @@
 PARENT_LABEL = 'AddressString'
 GROUP_LABEL = 'AddressCollection'
 
-MODEL_FILE = 'usaddr.crfsuite'
+MODEL_FILE = 'usaddr.crf'
 MODEL_PATH = os.path.split(os.path.abspath(__file__))[0] + '/' + MODEL_FILE
 
 DIRECTIONS = set(['n', 's', 'e', 'w',
@@ -136,12 +136,16 @@
 
 
 try:
-    TAGGER = pycrfsuite.Tagger()
-    TAGGER.open(MODEL_PATH)
+    MODEL = crfsuite.Model(MODEL_PATH)
+    TAGGER = crfsuite.Tagger(MODEL.model)
 except IOError:
     warnings.warn('You must train the model (parserator train --trainfile '
                   'FILES) to create the %s file before you can use the parse '
                   'and tag methods' % MODEL_FILE)
+except Exception:
+    warnings.warn('(Generic `Exception`) You must train the model (parserator '
+                  'train --trainfile FILES) to create the %s file before you '
+                  'can use the parse and tag methods' % MODEL_FILE)
 
 
 def parse(address_string):