sparql triple store and graph (nodes, mostly)

1 files changed, 243 insertions, 0 deletions

diff --git a/bsfs/graph/nodes.py b/bsfs/graph/nodes.py
new file mode 100644
index 0000000..7d2e9b3
--- /dev/null
+++ b/bsfs/graph/nodes.py
@@ -0,0 +1,243 @@
+"""
+
+Part of the BlackStar filesystem (bsfs) module.
+A copy of the license is provided with the project.
+Author: Matthias Baumgartner, 2022
+"""
+# imports
+import itertools
+import time
+import typing
+
+# bsfs imports
+from bsfs import schema as _schema
+from bsfs.namespace import ns
+from bsfs.triple_store import TripleStoreBase
+from bsfs.utils import errors, URI, typename
+
+# inner-module imports
+from . import ac
+
+# exports
+__all__: typing.Sequence[str] = (
+    'Nodes',
+    )
+
+
+## code ##
+
+class Nodes():
+    """
+    NOTE: guids may or may not exist. This is not verified as nodes are created on demand.
+    """
+
+    # triple store backend.
+    __backend: TripleStoreBase
+
+    # user uri.
+    __user: URI
+
+    # node type.
+    __node_type: _schema.Node
+
+    # guids of nodes. Can be empty.
+    __guids: typing.Set[URI]
+
+    def __init__(
+            self,
+            backend: TripleStoreBase,
+            user: URI,
+            node_type: _schema.Node,
+            guids: typing.Iterable[URI],
+            ):
+        self.__backend = backend
+        self.__user = user
+        self.__node_type = node_type
+        self.__guids = set(guids)
+        self.__ac = ac.NullAC(self.__backend, self.__user)
+
+    def __eq__(self, other: typing.Any) -> bool:
+        return isinstance(other, Nodes) \
+           and self.__backend == other.__backend \
+           and self.__user == other.__user \
+           and self.__node_type == other.__node_type \
+           and self.__guids == other.__guids
+
+    def __hash__(self) -> int:
+        return hash((type(self), self.__backend, self.__user, self.__node_type, tuple(sorted(self.__guids))))
+
+    def __repr__(self) -> str:
+        return f'{typename(self)}({self.__backend}, {self.__user}, {self.__node_type}, {self.__guids})'
+
+    def __str__(self) -> str:
+        return f'{typename(self)}({self.__node_type}, {self.__guids})'
+
+    @property
+    def node_type(self) -> _schema.Node:
+        """Return the node's type."""
+        return self.__node_type
+
+    @property
+    def guids(self) -> typing.Iterator[URI]:
+        """Return all node guids."""
+        return iter(self.__guids)
+
+    def set(
+            self,
+            pred: URI, # FIXME: URI or _schema.Predicate?
+            value: typing.Any,
+            ) -> 'Nodes':
+        """
+        """
+        try:
+            # insert triples
+            self.__set(pred, value)
+            # save changes
+            self.__backend.commit()
+
+        except (
+                errors.PermissionDeniedError, # tried to set a protected predicate (ns.bsm.t_created)
+                errors.ConsistencyError, # node types are not in the schema or don't match the predicate
+                errors.InstanceError, # guids/values don't have the correct type
+                TypeError, # value is supposed to be a Nodes instance
+                ValueError, # multiple values passed to unique predicate
+                ):
+            # revert changes
+            self.__backend.rollback()
+            # notify the client
+            raise
+
+        return self
+
+    def set_from_iterable(
+            self,
+            predicate_values: typing.Iterable[typing.Tuple[URI, typing.Any]], # FIXME: URI or _schema.Predicate?
+            ) -> 'Nodes':
+        """
+        """
+        # TODO: Could group predicate_values by predicate to gain some efficiency
+        # TODO: ignore errors on some predicates; For now this could leave residual
+        #       data (e.g. some nodes were created, some not).
+        try:
+            # insert triples
+            for pred, value in predicate_values:
+                self.__set(pred, value)
+            # save changes
+            self.__backend.commit()
+
+        except (
+                errors.PermissionDeniedError, # tried to set a protected predicate (ns.bsm.t_created)
+                errors.ConsistencyError, # node types are not in the schema or don't match the predicate
+                errors.InstanceError, # guids/values don't have the correct type
+                TypeError, # value is supposed to be a Nodes instance
+                ValueError, # multiple values passed to unique predicate
+                ):
+            # revert changes
+            self.__backend.rollback()
+            # notify the client
+            raise
+
+        return self
+
+    def __set(
+            self,
+            predicate: URI,
+            value: typing.Any,
+            #on_error: str = 'ignore', # ignore, rollback
+            ):
+        """
+        """
+        # get normalized predicate. Raises KeyError if *pred* not in the schema.
+        pred = self.__backend.schema.predicate(predicate)
+
+        # node_type must be a subclass of the predicate's domain
+        node_type = self.node_type
+        if not node_type <= pred.domain:
+            raise errors.ConsistencyError(f'{node_type} must be a subclass of {pred.domain}')
+
+        # check reserved predicates (access controls, metadata, internal structures)
+        # FIXME: Needed? Could be integrated into other AC methods (by passing the predicate!)
+        #        This could allow more fine-grained predicate control (e.g. based on ownership)
+        #        rather than a global approach like this.
+        if self.__ac.is_protected_predicate(pred):
+            raise errors.PermissionDeniedError(pred)
+
+        # set operation affects all nodes (if possible)
+        guids = set(self.guids)
+
+        # ensure subject node existence; create nodes if need be
+        guids = set(self._ensure_nodes(node_type, guids))
+
+        # check value
+        if isinstance(pred.range, _schema.Literal):
+            # check write permissions on existing nodes
+            # As long as the user has write permissions, we don't restrict
+            # the creation or modification of literal values.
+            guids = set(self.__ac.write_literal(node_type, guids))
+
+            # insert literals
+            # TODO: Support passing iterators as values for non-unique predicates
+            self.__backend.set(
+                node_type,
+                guids,
+                pred,
+                [value],
+                )
+
+        elif isinstance(pred.range, _schema.Node):
+            # check value type
+            if not isinstance(value, Nodes):
+                raise TypeError(value)
+            # value's node_type must be a subclass of the predicate's range
+            if not value.node_type <= pred.range:
+                raise errors.ConsistencyError(f'{value.node_type} must be a subclass of {pred.range}')
+
+            # check link permissions on source nodes
+            # Link permissions cover adding and removing links on the source node.
+            # Specifically, link permissions also allow to remove links to other
+            # nodes if needed (e.g. for unique predicates).
+            guids = set(self.__ac.link_from_node(node_type, guids))
+
+            # get link targets
+            targets = set(value.guids)
+            # ensure existence of value nodes; create nodes if need be
+            targets = set(self._ensure_nodes(value.node_type, targets))
+            # check link permissions on target nodes
+            targets = set(self.__ac.link_to_node(value.node_type, targets))
+
+            # insert node links
+            self.__backend.set(
+                node_type,
+                guids,
+                pred,
+                targets,
+                )
+
+        else:
+            raise errors.UnreachableError()
+
+    def _ensure_nodes(
+            self,
+            node_type: _schema.Node,
+            guids: typing.Iterable[URI],
+            ):
+        # check node existence
+        guids = set(guids)
+        existing = set(self.__backend.exists(node_type, guids))
+        # get nodes to be created
+        missing = guids - existing
+        # create nodes if need be
+        if len(missing) > 0:
+            # check which missing nodes can be created
+            missing = set(self.__ac.createable(node_type, missing))
+            # create nodes
+            self.__backend.create(node_type, missing)
+            # add bookkeeping triples
+            self.__backend.set(node_type, missing,
+                self.__backend.schema.predicate(ns.bsm.t_created), [time.time()])
+            # add permission triples
+            self.__ac.create(node_type, missing)
+        # return available nodes
+        return existing | missing
+
+## EOF ##