Make Result.serialize work more like Graph.serialize

This patch makes the following changes to `Result.serialize`.

* Return str by default instead of bytes.
* Use "txt" as the default tabular serialization format.
* Use "turtle" as the default graph serialization format.
* Support both typing.IO[bytes] and typing.TextIO destinations.

Corresponding changes are made to the specific serializers also.

This patch also changes how text is written to typing.IO[bytes] in
serializers to ensure that the buffer is flushed and
detatched from the TextIOWrapper once the serialization function
completes so it can be used normally afterwards.

This patch further includes a bunch of additional type hints.

Author: aucampia

rdflib/graph.py

@@ -351,15 +351,15 @@ def __init__(
         self.default_union = False
 
     @property
-    def store(self):
+    def store(self) -> Store:  # read-only attr
         return self.__store
 
     @property
-    def identifier(self):
+    def identifier(self) -> Node:  # read-only attr
         return self.__identifier
 
     @property
-    def namespace_manager(self):
+    def namespace_manager(self) -> NamespaceManager:
         """
         this graph's namespace-manager
         """
@@ -368,8 +368,9 @@ def namespace_manager(self):
         return self.__namespace_manager
 
     @namespace_manager.setter
-    def namespace_manager(self, nm):
-        self.__namespace_manager = nm
+    def namespace_manager(self, value: NamespaceManager):
+        """this graph's namespace-manager"""
+        self.__namespace_manager = value
 
     def __repr__(self):
         return "<Graph identifier=%s (%s)>" % (self.identifier, type(self))
@@ -1096,18 +1097,37 @@ def serialize(
         encoding: Optional[str] = None,
         **args: Any,
     ) -> Union[bytes, str, "Graph"]:
-        """Serialize the Graph to destination
-
-        If destination is None serialize method returns the serialization as
-        bytes or string.
-
-        If encoding is None and destination is None, returns a string
-        If encoding is set, and Destination is None, returns bytes
-
-        Format defaults to turtle.
-
-        Format support can be extended with plugins,
-        but "xml", "n3", "turtle", "nt", "pretty-xml", "trix", "trig" and "nquads" are built in.
+        """
+        Serialize the graph.
+
+        :param destination:
+           The destination to serialize the graph to. This can be a path as a
+           :class:`str` or :class:`~pathlib.PurePath` object, or it can be a
+           :class:`~typing.IO[bytes]` like object. If this parameter is not
+           supplied the serialized graph will be returned.
+        :type destination: Optional[Union[str, typing.IO[bytes], pathlib.PurePath]]
+        :param format:
+           The format that the output should be written in. This value
+           references a :class:`~rdflib.serializer.Serializer` plugin. Format
+           support can be extended with plugins, but `"xml"`, `"n3"`,
+           `"turtle"`, `"nt"`, `"pretty-xml"`, `"trix"`, `"trig"`, `"nquads"`
+           and `"json-ld"` are built in. Defaults to `"turtle"`.
+        :type format: str
+        :param base:
+           The base IRI for formats that support it. For the turtle format this
+           will be used as the `@base` directive.
+        :type base: Optional[str]
+        :param encoding: Encoding of output.
+        :type encoding: Optional[str]
+        :param **args:
+           Additional arguments to pass to the
+           :class:`~rdflib.serializer.Serializer` that will be used.
+        :type **args: Any
+        :return: The serialized graph if `destination` is `None`.
+        :rtype: :class:`bytes` if `destination` is `None` and `encoding` is not `None`.
+        :rtype: :class:`bytes` if `destination` is `None` and `encoding` is `None`.
+        :return: `self` (i.e. the :class:`~rdflib.graph.Graph` instance) if `destination` is not None.
+        :rtype: :class:`~rdflib.graph.Graph` if `destination` is not None.
         """
 
         # if base is not given as attribute use the base set for the graph
@@ -1298,7 +1318,7 @@ def query(
         if none are given, the namespaces from the graph's namespace manager
         are used.
 
-        :returntype: rdflib.query.Result
+        :returntype: :class:`~rdflib.query.Result`
 
         """
 

rdflib/plugins/serializers/n3.py

@@ -109,7 +109,7 @@ def p_clause(self, node, position):
             self.write("{")
             self.depth += 1
             serializer = N3Serializer(node, parent=self)
-            serializer.serialize(self.stream)
+            serializer.serialize(self.stream.buffer)
             self.depth -= 1
             self.write(self.indent() + "}")
             return True

rdflib/plugins/serializers/nt.py

@@ -12,6 +12,8 @@
 import warnings
 import codecs
 
+from rdflib.util import as_textio
+
 __all__ = ["NTSerializer"]
 
 
@@ -38,9 +40,15 @@ def serialize(
                 f"Given encoding was: {encoding}"
             )
 
-        for triple in self.store:
-            stream.write(_nt_row(triple).encode())
-        stream.write("\n".encode())
+        with as_textio(
+            stream,
+            encoding=encoding,  # TODO: CHECK: self.encoding set removed, why?
+            errors="_rdflib_nt_escape",
+            write_through=True,
+        ) as text_stream:
+            for triple in self.store:
+                text_stream.write(_nt_row(triple))
+            text_stream.write("\n")
 
 
 class NT11Serializer(NTSerializer):

rdflib/plugins/serializers/rdfxml.py

@@ -1,4 +1,4 @@
-from typing import IO, Dict, Optional, Set
+from typing import IO, Dict, Optional, Set, cast
 from rdflib.plugins.serializers.xmlwriter import XMLWriter
 
 from rdflib.namespace import Namespace, RDF, RDFS  # , split_uri
@@ -173,6 +173,8 @@ def serialize(
         encoding: Optional[str] = None,
         **args,
     ):
+        # TODO FIXME: this should be Optional, but it's not because nothing
+        # treats it as such.
         self.__serialized: Dict[Identifier, int] = {}
         store = self.store
         # if base is given here, use that, if not and a base is set for the graph use that
@@ -241,6 +243,7 @@ def subject(self, subject: IdentifiedNode, depth: int = 1):
         writer = self.writer
 
         if subject in self.forceRDFAbout:
+            subject = cast(URIRef, subject)
             writer.push(RDFVOC.Description)
             writer.attribute(RDFVOC.about, self.relativize(subject))
             writer.pop(RDFVOC.Description)
@@ -282,6 +285,7 @@ def subj_as_obj_more_than(ceil):
 
         elif subject in self.forceRDFAbout:
             # TODO FIXME?: this looks like a duplicate of first condition
+            subject = cast(URIRef, subject)
             writer.push(RDFVOC.Description)
             writer.attribute(RDFVOC.about, self.relativize(subject))
             writer.pop(RDFVOC.Description)

rdflib/plugins/serializers/trig.py

@@ -62,53 +62,45 @@ def serialize(
         spacious: Optional[bool] = None,
         **args,
     ):
-        self.reset()
-        self.stream = stream
-        # if base is given here, use that, if not and a base is set for the graph use that
-        if base is not None:
-            self.base = base
-        elif self.store.base is not None:
-            self.base = self.store.base
-
-        if spacious is not None:
-            self._spacious = spacious
-
-        self.preprocess()
-
-        self.startDocument()
-
-        firstTime = True
-        for store, (ordered_subjects, subjects, ref) in self._contexts.items():
-            if not ordered_subjects:
-                continue
-
-            self._references = ref
-            self._serialized = {}
-            self.store = store
-            self._subjects = subjects
-
-            if self.default_context and store.identifier == self.default_context:
-                self.write(self.indent() + "\n{")
-            else:
-                iri: Optional[str]
-                if isinstance(store.identifier, BNode):
-                    iri = store.identifier.n3()
-                else:
-                    iri = self.getQName(store.identifier)
-                    if iri is None:
-                        iri = store.identifier.n3()
-                self.write(self.indent() + "\n%s {" % iri)
+        self._serialize_init(stream, base, encoding, spacious)
+        try:
+            self.preprocess()
 
-            self.depth += 1
-            for subject in ordered_subjects:
-                if self.isDone(subject):
+            self.startDocument()
+
+            firstTime = True
+            for store, (ordered_subjects, subjects, ref) in self._contexts.items():
+                if not ordered_subjects:
                     continue
-                if firstTime:
-                    firstTime = False
-                if self.statement(subject) and not firstTime:
-                    self.write("\n")
-            self.depth -= 1
-            self.write("}\n")
-
-        self.endDocument()
-        stream.write("\n".encode("latin-1"))
+
+                self._references = ref
+                self._serialized = {}
+                self.store = store
+                self._subjects = subjects
+
+                if self.default_context and store.identifier == self.default_context:
+                    self.write(self.indent() + "\n{")
+                else:
+                    if isinstance(store.identifier, BNode):
+                        iri = store.identifier.n3()
+                    else:
+                        iri = self.getQName(store.identifier)
+                        if iri is None:
+                            iri = store.identifier.n3()
+                    self.write(self.indent() + "\n%s {" % iri)
+
+                self.depth += 1
+                for subject in ordered_subjects:
+                    if self.isDone(subject):
+                        continue
+                    if firstTime:
+                        firstTime = False
+                    if self.statement(subject) and not firstTime:
+                        self.write("\n")
+                self.depth -= 1
+                self.write("}\n")
+
+            self.endDocument()
+            self.write("\n")
+        finally:
+            self._serialize_end()

rdflib/plugins/serializers/turtle.py

@@ -6,10 +6,13 @@
 from collections import defaultdict
 from functools import cmp_to_key
 
+from rdflib.graph import Graph
 from rdflib.term import BNode, Literal, URIRef
 from rdflib.exceptions import Error
 from rdflib.serializer import Serializer
 from rdflib.namespace import RDF, RDFS
+from io import TextIOWrapper
+from typing import IO, Dict, Optional
 
 __all__ = ["RecursiveSerializer", "TurtleSerializer"]
 
@@ -44,10 +47,13 @@ class RecursiveSerializer(Serializer):
     indentString = "  "
     roundtrip_prefixes = ()
 
-    def __init__(self, store):
+    def __init__(self, store: Graph):
 
         super(RecursiveSerializer, self).__init__(store)
-        self.stream = None
+        # TODO FIXME: Ideally stream should be optional, but nothing treats it
+        # as such, so least weird solution is to just type it as not optional
+        # even thoug it can sometimes be null.
+        self.stream: IO[str] = None  # type: ignore[assignment]
         self.reset()
 
     def addNamespace(self, prefix, uri):
@@ -166,9 +172,9 @@ def indent(self, modifier=0):
         """Returns indent string multiplied by the depth"""
         return (self.depth + modifier) * self.indentString
 
-    def write(self, text):
-        """Write text in given encoding."""
-        self.stream.write(text.encode(self.encoding, "replace"))
+    def write(self, text: str):
+        """Write text"""
+        self.stream.write(text)
 
 
 SUBJECT = 0
@@ -184,15 +190,15 @@ class TurtleSerializer(RecursiveSerializer):
     short_name = "turtle"
     indentString = "    "
 
-    def __init__(self, store):
-        self._ns_rewrite = {}
+    def __init__(self, store: Graph):
+        self._ns_rewrite: Dict[str, str] = {}
         super(TurtleSerializer, self).__init__(store)
         self.keywords = {RDF.type: "a"}
         self.reset()
-        self.stream = None
+        self.stream: TextIOWrapper = None  # type: ignore[assignment]
         self._spacious = _SPACIOUS_OUTPUT
 
-    def addNamespace(self, prefix, namespace):
+    def addNamespace(self, prefix: str, namespace: str):
         # Turtle does not support prefix that start with _
         # if they occur in the graph, rewrite to p_blah
         # this is more complicated since we need to make sure p_blah
@@ -223,36 +229,60 @@ def reset(self):
         self._started = False
         self._ns_rewrite = {}
 
-    def serialize(self, stream, base=None, encoding=None, spacious=None, **args):
+    def _serialize_init(
+        self,
+        stream: IO[bytes],
+        base: Optional[str],
+        encoding: Optional[str],
+        spacious: Optional[bool],
+    ) -> None:
         self.reset()
-        self.stream = stream
+        if encoding is not None:
+            self.encoding = encoding
+        self.stream = TextIOWrapper(
+            stream, self.encoding, errors="replace", write_through=True
+        )
         # if base is given here, use that, if not and a base is set for the graph use that
         if base is not None:
             self.base = base
         elif self.store.base is not None:
             self.base = self.store.base
-
         if spacious is not None:
             self._spacious = spacious
 
-        self.preprocess()
-        subjects_list = self.orderSubjects()
-
-        self.startDocument()
-
-        firstTime = True
-        for subject in subjects_list:
-            if self.isDone(subject):
-                continue
-            if firstTime:
-                firstTime = False
-            if self.statement(subject) and not firstTime:
-                self.write("\n")
-
-        self.endDocument()
-        stream.write("\n".encode("latin-1"))
-
-        self.base = None
+    def _serialize_end(self) -> None:
+        self.stream.flush()
+        self.stream.detach()
+        self.stream = None  # type: ignore[assignment]
+
+    def serialize(
+        self,
+        stream: IO[bytes],
+        base: Optional[str] = None,
+        encoding: Optional[str] = None,
+        spacious: Optional[bool] = None,
+        **args,
+    ):
+        self._serialize_init(stream, base, encoding, spacious)
+        try:
+            self.preprocess()
+            subjects_list = self.orderSubjects()
+
+            self.startDocument()
+
+            firstTime = True
+            for subject in subjects_list:
+                if self.isDone(subject):
+                    continue
+                if firstTime:
+                    firstTime = False
+                if self.statement(subject) and not firstTime:
+                    self.write("\n")
+
+            self.endDocument()
+            self.stream.write("\n")
+        finally:
+            self._serialize_end()
 
     def preprocessTriple(self, triple):
         super(TurtleSerializer, self).preprocessTriple(triple)