Merge pull request #303 from dkpro/feature/249-Set-sofa-string-and-do…

…cument-language-in-Cas-constructor

#249 - Set sofa string and document language in Cas constructor

cassis/cas.py

@@ -21,6 +21,8 @@
     TypeCheckError,
     TypeSystem,
     TypeSystemMode,
+    TYPE_NAME_DOCUMENT_ANNOTATION,
+    FEATURE_BASE_NAME_LANGUAGE,
 )
 
 _validator_optional_string = validators.optional(validators.instance_of(str))
@@ -32,6 +34,7 @@
 def _get_size_in_utf16_bytes(c: str) -> int:
     return len(c.encode("utf-16-le")) // 2
 
+
 class IdGenerator:
     def __init__(self, initial_id: int = 1):
         self._next_id = initial_id
@@ -56,7 +59,6 @@ def __init__(self):
         self._external_to_python: Union[Dict[int, int], None] = None
         self._python_to_external: Union[Dict[int, int], None] = None
 
-
     def create_offset_mapping(self, sofa_string: str) -> None:
         if sofa_string is None:
             return
@@ -202,7 +204,13 @@ def __init__(self, typesystem: TypeSystem):
 class Cas:
     """A CAS object is a container for text (sofa) and annotations"""
 
-    def __init__(self, typesystem: TypeSystem = None, lenient: bool = False):
+    def __init__(
+        self,
+        typesystem: TypeSystem = None,
+        lenient: bool = False,
+        sofa_string: str = None,
+        sofa_mime: str = None,
+    ):
         """Creates a CAS with the specified typesystem. If no typesystem is given, then the default one
         is used which only contains UIMA-predefined types.
 
@@ -226,6 +234,13 @@ def __init__(self, typesystem: TypeSystem = None, lenient: bool = False):
         self._add_view("_InitialView")
         self._current_view: View = self._views["_InitialView"]
 
+        if sofa_string is not None:
+            self.sofa_string = sofa_string
+            if sofa_mime is not None:
+                self.sofa_mime = sofa_mime
+            else:
+                self.sofa_mime = "text/plain"
+
     @property
     def typesystem(self) -> TypeSystem:
         return self._typesystem

cassis/json.py

@@ -37,7 +37,9 @@
 NEGATIVE_INFINITE_VALUE_ABBR = "-Inf"
 
 
-def load_cas_from_json(source: Union[IO, str], typesystem: TypeSystem = None, lenient: bool = False, merge_typesystem: bool =True) -> Cas:
+def load_cas_from_json(
+    source: Union[IO, str], typesystem: TypeSystem = None, lenient: bool = False, merge_typesystem: bool = True
+) -> Cas:
     """Loads a CAS from a JSON source.
 
     Args:
@@ -64,7 +66,13 @@ def __init__(self):
         self._max_sofa_num = 0
         self._post_processors = []
 
-    def deserialize(self, source: Union[IO, str], typesystem: Optional[TypeSystem] = None, lenient: bool = False, merge_typesystem: bool =True) -> Cas:
+    def deserialize(
+        self,
+        source: Union[IO, str],
+        typesystem: Optional[TypeSystem] = None,
+        lenient: bool = False,
+        merge_typesystem: bool = True,
+    ) -> Cas:
         if isinstance(source, str):
             data = json.loads(source)
         else:
@@ -103,6 +111,7 @@ def deserialize(self, source: Union[IO, str], typesystem: Optional[TypeSystem] =
         feature_structures = {}
         json_feature_structures = data.get(FEATURE_STRUCTURES_FIELD)
         if isinstance(json_feature_structures, list):
+
             def parse_and_add(json_fs_):
                 parsed = self._parse_feature_structure(typesystem, json_fs_.get(ID_FIELD), json_fs_, feature_structures)
                 feature_structures[parsed.xmiID] = parsed
@@ -126,6 +135,7 @@ def parse_and_add(json_fs_):
                     parse_and_add(json_fs)
 
         if isinstance(json_feature_structures, dict):
+
             def parse_and_add(fs_id_, json_fs_):
                 parsed = self._parse_feature_structure(typesystem, int(fs_id_), json_fs_, feature_structures)
                 feature_structures[parsed.xmiID] = parsed
@@ -174,7 +184,7 @@ def _parse_features(self, typesystem: TypeSystem, type_name: str, json_type: Dic
 
             range_type = json_feature[RANGE_FIELD]
             element_type = json_feature.get(ELEMENT_TYPE_FIELD)
-            if range_type.endswith('[]'):
+            if range_type.endswith("[]"):
                 element_type = range_type[:-2]
                 range_type = array_type_name_for_type(element_type)
             typesystem.create_feature(
@@ -222,7 +232,7 @@ def _parse_feature_structure(
         self, typesystem: TypeSystem, fs_id: int, json_fs: Dict[str, any], feature_structures: Dict[int, any]
     ):
         type_name = json_fs.get(TYPE_FIELD)
-        if type_name.endswith('[]'):
+        if type_name.endswith("[]"):
             type_name = array_type_name_for_type(type_name)
         AnnotationType = typesystem.get_type(type_name)
 

cassis/typesystem.py

@@ -198,6 +198,7 @@ class TypeSystemMode(Enum):
     MINIMAL = auto()
     NONE = auto()
 
+
 def array_type_name_for_type(type_: Union[str, "Type"]) -> str:
     type_name = type_ if isinstance(type_, str) else type_.name
     if type_name == TYPE_NAME_BYTE:
@@ -238,6 +239,8 @@ def element_type_name_for_array_type(type_: Union[str, "Type"]) -> str:
     if type_name == TYPE_NAME_STRING_ARRAY:
         return TYPE_NAME_STRING
     return TYPE_NAME_TOP
+
+
 def _string_to_valid_classname(name: str):
     return re.sub("[^a-zA-Z0-9_]", "_", name)
 

tests/test_cas.py

@@ -113,6 +113,20 @@ def test_sofa_uri_can_be_set_and_read():
     assert cas.sofa_uri == "https://raw.githubusercontent.com/dkpro/dkpro-cassis/master/README.rst"
 
 
+def test_sofa_string_can_be_set_using_constructor():
+    cas = Cas(sofa_string="I am a test sofa string!")
+
+    assert cas.sofa_string == "I am a test sofa string!"
+    assert cas.sofa_mime == "text/plain"
+
+
+def test_sofa_string_and_mime_type_can_be_set_using_constructor():
+    cas = Cas(sofa_string="I am a <b>test sofa string!</b>", sofa_mime="text/html")
+
+    assert cas.sofa_string == "I am a <b>test sofa string!</b>"
+    assert cas.sofa_mime == "text/html"
+
+
 # Select
 
 

tests/test_json.py

@@ -92,9 +92,10 @@
     (
         os.path.join(ONE_WAY_DIR, "tsv3-testSimpleSlotFeature"),
         [],
-    )
+    ),
 ]
 
+
 @pytest.mark.parametrize("json_path, annotations", ROUND_TRIP_FIXTURES)
 def test_deserialization_serialization(json_path, annotations):
     with open(os.path.join(json_path, "data.json"), "rb") as f:

tests/test_typesystem.py

@@ -16,7 +16,8 @@
     TYPE_NAME_STRING_ARRAY,
     TYPE_NAME_TOP,
     TypeCheckError,
-    is_predefined, TYPE_NAME_DOCUMENT_ANNOTATION,
+    is_predefined,
+    TYPE_NAME_DOCUMENT_ANNOTATION,
 )
 from tests.fixtures import *
 from tests.util import assert_xml_equal