Merge pull request #1 from sassoftware/bugfix

Bugfix

Author: jameskochubasas

CHANGELOG.md

@@ -1,9 +1,11 @@
 
 Unreleased
 ----------
-
- - 
-
+**Bugfixes**
+ - Fixed DS2 score code for CAS that was generated when registering a Python model.
+ - `PyMAS.score_code(dest='ESP')` corrected to `dest='EP'`
+ - Fixed an issue where long user-defined properties prevented model registration.
+ 
 
 v1.1.1 (2019-8-6)
 -----------------

src/sasctl/__init__.py

@@ -4,10 +4,10 @@
 # Copyright © 2019, SAS Institute Inc., Cary, NC, USA.  All Rights Reserved.
 # SPDX-License-Identifier: Apache-2.0
 
-__version__ = '1.1.1'
+__version__ = '1.1.2'
 __author__ = 'SAS'
-__credits__ = ['Yi Jian Ching, Lucas De Paula, Peter Tobac, Chris Toth, Jon '
-               'Walker']
+__credits__ = ['Yi Jian Ching, Lucas De Paula, James Kochuba, Peter Tobac, '
+               'Chris Toth, Jon Walker']
 __license__ = 'Apache 2.0'
 __copyright__ = 'Copyright © 2019, SAS Institute Inc., ' \
                 'Cary, NC, USA.  All Rights Reserved.'

src/sasctl/_services/model_management.py

@@ -21,6 +21,19 @@ class ModelManagement(Service):
 
     # TODO:  set ds2MultiType
     def publish_model(self, model, destination, name=None, force=False):
+        """
+
+        Parameters
+        ----------
+        model
+        destination
+        name
+        force
+
+        Returns
+        -------
+
+        """
         from .model_repository import ModelRepository
         from .model_publish import ModelPublish
 

src/sasctl/tasks.py

@@ -25,27 +25,43 @@
 
 
 def _sklearn_to_dict(model):
+    # As of Viya 3.4 model registration fails if character fields are longer
+    # than 1024 characters
+    DESC_MAXLEN = 1024
+
+    # As of Viya 3.4 model registration fails if user-defined properties are
+    # longer than 512 characters.
+    PROP_MAXLEN = 512
+
     # Convert Scikit-learn values to built-in Model Manager values
     mappings = {'LogisticRegression': 'Logistic regression',
                 'LinearRegression': 'Linear regression',
                 'SVC': 'Support vector machine',
                 'GradientBoostingClassifier': 'Gradient boosting',
+                'XGBClassifier': 'Gradient boosting',
+                'XGBRegressor': 'Gradient boosting',
                 'RandomForestClassifier': 'Forest',
                 'DecisionTreeClassifier': 'Decision tree',
                 'DecisionTreeRegressor': 'Decision tree',
                 'classifier': 'Classification',
                 'regressor': 'Prediction'}
 
+    if hasattr(model, '_final_estimator'):
+        estimator = type(model._final_estimator)
+    else:
+        estimator = type(model)
+
     # Can tell if multi-class .multi_class
     result = dict(
-        description=str(model),
-        algorithm=mappings.get(type(model).__name__, type(model).__name__),
+        description=str(model)[:DESC_MAXLEN],
+        algorithm=mappings.get(estimator.__name__, estimator.__name__),
         scoreCodeType='ds2MultiType',
         trainCodeType='Python',
         function=mappings.get(model._estimator_type, model._estimator_type),
         tool='Python %s.%s'
              % (sys.version_info.major, sys.version_info.minor),
-        properties=[{'name': k, 'value': v}
+        properties=[{'name': str(k)[:PROP_MAXLEN],
+                     'value': str(v)[:PROP_MAXLEN]}
                     for k, v in model.get_params().items()]
     )
 
@@ -156,9 +172,6 @@ def register_model(model, name, project, repository=None, input=None,
             files.append({'name': 'dmcas_packagescorecode.sas',
                           'file': mas_module.score_code(),
                           'role': 'Score Code'})
-            files.append({'name': 'dmcas_espscorecode.sas',
-                          'file': mas_module.score_code(dest='ESP'),
-                          'role': 'Score Code'})
             files.append({'name': 'dmcas_epscorecode.sas',
                           'file': mas_module.score_code(dest='CAS'),
                           'role': 'score'})
@@ -251,7 +264,8 @@ def publish_model(model, destination, code=None, max_retries=60,
     def submit_request():
         # Submit a publishing request
         if code is None:
-            publish_req = mm.publish_model(model, destination, force=replace, **kwargs)
+            publish_req = mm.publish_model(model, destination,
+                                           force=replace, **kwargs)
         else:
             publish_req = mp.publish_model(model, destination,
                                            code=code, **kwargs)
@@ -265,7 +279,9 @@ def submit_request():
     # Submit and wait for status
     job = submit_request()
 
-    if job.state.lower() == 'completed' and job.destination.destinationType != 'microAnalyticService':
+    # If model was successfully published and it isn't a MAS module, we're done
+    if job.state.lower() == 'completed' \
+            and job.destination.destinationType != 'microAnalyticService':
             return request_link(job,'self')
 
     # If MAS publish failed and replace=True, attempt to delete the module

src/sasctl/utils/pymas/core.py

@@ -345,7 +345,7 @@ def score_code(self, input_table=None, output_table=None, columns=None, dest='MA
             The name of the table where execution results will be written
         columns : list of str
             Names of the columns from `table` that will be passed to `func`
-        dest : str {'MAS', 'ESP', 'CAS'}
+        dest : str {'MAS', 'EP', 'CAS'}
 
         Returns
         -------
@@ -365,7 +365,7 @@ def score_code(self, input_table=None, output_table=None, columns=None, dest='MA
         # Get package code
         code = tuple(self.package.code().split('\n'))
 
-        if dest == 'ESP':
+        if dest == 'EP':
             code = ('data sasep.out;', ) + code + ('   method run();',
                                                    '      set SASEP.IN;',
                                                    '   end;',

tests/cassettes/tests.integration.test_tasks.TestModels.test_register_sklearn.json

@@ -93,7 +93,7 @@ def test_register_sklearn(self, sklearn_model):
         files = mr.get_model_contents(model)
         filenames = [f.name for f in files]
         assert 'model.pkl' in filenames
-        assert 'dmcas_espscorecode.sas' in filenames
+        assert 'dmcas_epscorecode.sas' in filenames
         assert 'dmcas_packagescorecode.sas' in filenames
 
     def test_publish_sklearn(self):

tests/integration/test_tasks.py

@@ -33,14 +33,11 @@ def test_score_code():
     assert isinstance(p, PyMAS)
 
     mas_code = p.score_code()
-    esp_code = p.score_code(dest='esp')
+    esp_code = p.score_code(dest='ep')
 
     assert mas_code.lower().startswith('package _')
     assert esp_code.lower().startswith('data sasep.out;')
 
-    with pytest.raises(ValueError):
-        cas_code = p.score_code(dest='cas')
-
     cas_code = p.score_code('in_table', 'out_table', ['in1', 'in2'], dest='cas')
 
     pytest.xfail('Not implemented.')