Catalog Arrow Endpoints¶

class graphdatascience.procedure_surface.arrow.catalog.catalog_arrow_endpoints.CatalogArrowEndpoints¶

__new__(**kwargs)¶

construct(graph_name: str, nodes: DataFrame | List[DataFrame], relationships: DataFrame | List[DataFrame] | None = None, concurrency: int | None = None, undirected_relationship_types: List[str] | None = None, inverse_indexed_relationship_types: List[str] | None = None, batch_size: int = 100000, overwrite: bool = False) → Graph¶

Construct a graph from a list of node and relationship dataframes.

Parameters:

graph_name (str) – Name of the graph to be created
nodes (DataFrame | List[DataFrame]) –
Node dataframes. A dataframe should follow the schema:
- nodeId to identify uniquely the node overall dataframes
- labels to specify the labels of the node as a list of strings (optional)
- other columns are treated as node properties
relationships (DataFrame | List[DataFrame] | None) –
Relationship dataframes. A dataframe should follow the schema:
- sourceNodeId to identify the start node of the relationship
- targetNodeId to identify the end node of the relationship
- relationshipType to specify the type of the relationship (optional)
- other columns are treated as relationship properties
concurrency (int | None) – Number of concurrent threads to use.
undirected_relationship_types (List[str] | None) – List of relationship types to treat as undirected.
inverse_indexed_relationship_types (List[str] | None) – List of relationship types for which to create an inverse index.
batch_size (int) – Batch size to use when sending data to GDS.
overwrite (bool) – If True, drop an existing graph with the same name before constructing the new one. Defaults to False.

Returns:

Constructed graph object.

Return type:

Graph

property datasets: DatasetEndpoints¶: Endpoints for loading predefined datasets into the graph catalog.

drop(G: Graph | str | List[Graph | str], fail_if_missing: bool = True) → List[GraphInfo]¶

Drop graphs from the graph catalog.

Parameters:

G (Graph | str | List[Graph | str]) – Graphs to drop by name or object.
fail_if_missing (bool) – Whether to fail if the graph is missing.

Returns:

Metadata of the dropped graphs.

Return type:

List[GraphInfo]

exists(graph_name: str) → bool¶

Check if a graph exists in the catalog.

Parameters:: graph_name (str) – Name of the graph
Returns:: True if the graph exists, False otherwise.
Return type:: bool

property export: GraphExportEndpoints¶: Endpoints for exporting graphs to a new database or CSV files.

filter(G: Graph, graph_name: str, node_filter: str, relationship_filter: str, parameters: dict[str, Any] | None = None, concurrency: int | None = None, job_id: str | None = None, sudo: bool = False, log_progress: bool = True, username: str | None = None, overwrite: bool = False) → GraphWithFilterResult¶

Create a subgraph of a graph based on a filter expression.

Parameters:

G (Graph) – Graph object to use
graph_name (str) – Name of the graph to be created
node_filter (str) – Filter expression for nodes
relationship_filter (str) – Filter expression for relationships
parameters (dict[str, Any] | None) – A map of user-defined query parameters that are passed into the node and relationship filters.
concurrency (int | None) – Number of concurrent threads to use.
job_id (str | None) – Identifier for the computation.
sudo (bool) – Disable the memory guard.
log_progress (bool) – Display progress logging.
username (str | None) – As an administrator, impersonate a different user for accessing their graphs.
overwrite (bool) – If True, drop an existing graph with the same name before creating the filtered subgraph. Defaults to False.

Returns:

tuple of the filtered graph object and the information like graph name, node count, relationship count, etc.

Return type:

GraphWithFilterResult

filter_async(G: Graph, graph_name: str, node_filter: str, relationship_filter: str, parameters: dict[str, Any] | None = None, concurrency: int | None = None, job_id: str | None = None, sudo: bool = False, log_progress: bool = True, username: str | None = None, overwrite: bool = False) → ProjectionJobHandle¶

Kick off a graph filter operation and return a ProjectionJobHandle.

Unlike filter(), this method does not block on completion.

Parameters:

G (Graph)
graph_name (str)
node_filter (str)
relationship_filter (str)
parameters (dict[str, Any] | None)
concurrency (int | None)
job_id (str | None)
sudo (bool)
log_progress (bool)
username (str | None)
overwrite (bool)

Return type:

ProjectionJobHandle

generate(graph_name: str, node_count: int, average_degree: float, *, relationship_distribution: str | None = 'UNIFORM', relationship_seed: int | None = None, relationship_property: RelationshipPropertySpec | None = None, orientation: str | None = 'NATURAL', aggregation: str | None = 'NONE', allow_self_loops: bool | None = False, read_concurrency: int | None = None, job_id: str | None = None, sudo: bool = False, log_progress: bool = True, username: str | None = None, overwrite: bool = False) → GraphWithGenerationStats¶

Generates a random graph and store it in the graph catalog.

Parameters:

graph_name (str) – Name of the graph to be created
node_count (int) – The number of nodes in the generated graph
average_degree (float) – The average out-degree of the generated nodes
relationship_distribution (str | None) – Determines the relationship distribution strategy.
relationship_seed (int | None) – Seed value for generating deterministic relationships.
relationship_property (RelationshipPropertySpec | None) – Configure generated relationship properties.
orientation (str | None) – Specifies the orientation of the generated relationships.
aggregation (str | None) – The relationship aggregation method of Relationship Projection.
allow_self_loops (bool | None) – Whether nodes in the graph can have relationships where start and end nodes are the same.
read_concurrency (int | None) – Number of concurrent threads/processes to use during graph generation.
job_id (str | None) – Identifier for the computation.
sudo (bool) – Disable the memory guard.
log_progress (bool) – Display progress logging.
username (str | None) – As an administrator, impersonate a different user for accessing their graphs.
overwrite (bool) – If True, drop an existing graph with the same name before generating the new one. Defaults to False.

Returns:

tuple of the generated graph object and the result object containing stats about the generation.

Return type:

GraphWithGenerationStats

generate_async(graph_name: str, node_count: int, average_degree: float, *, relationship_distribution: str | None = 'UNIFORM', relationship_seed: int | None = None, relationship_property: RelationshipPropertySpec | None = None, orientation: str | None = 'NATURAL', aggregation: str | None = 'NONE', allow_self_loops: bool | None = False, read_concurrency: int | None = None, job_id: str | None = None, sudo: bool = False, log_progress: bool = True, username: str | None = None, overwrite: bool = False) → ProjectionJobHandle¶

Kick off a graph generation and return a ProjectionJobHandle.

Unlike generate(), this method does not block on completion.

Parameters:

graph_name (str)
node_count (int)
average_degree (float)
relationship_distribution (str | None)
relationship_seed (int | None)
relationship_property (RelationshipPropertySpec | None)
orientation (str | None)
aggregation (str | None)
allow_self_loops (bool | None)
read_concurrency (int | None)
job_id (str | None)
sudo (bool)
log_progress (bool)
username (str | None)
overwrite (bool)

Return type:

ProjectionJobHandle

get(graph_name: str) → Graph¶

Retrieve a handle to a graph from the graph catalog.

Parameters:: graph_name (str) – Name of the graph
Returns:: A handle to the graph.
Return type:: Graph

list(G: Graph | str | None = None) → list[GraphInfoWithDegrees]¶

List graphs in the graph catalog.

Parameters:: G (Graph | str | None) – GraphV2 object or name to filter results. If None, list all graphs.
Returns:: List of graph metadata objects containing information like node count.
Return type:: list[GraphInfoWithDegrees]

property node_labels: NodeLabelEndpoints¶: Endpoints for node label operations.

property node_properties: NodePropertiesEndpoints¶: Endpoints for node property operations.

property node_property: NodePropertyEndpoints¶: Endpoints for streaming a single node property.

property project: ProjectArrowEndpoints¶: Provides access to the project endpoints.

property relationship_properties: RelationshipPropertiesEndpoints¶: Endpoints for streaming several relationship properties.

property relationship_property: RelationshipPropertyEndpoints¶: Endpoints for streaming a single relationship property.

property relationships: RelationshipsEndpoints¶: Endpoints for relationship operations.

property sample: GraphSamplingEndpoints¶: Endpoints for graph sampling.

class graphdatascience.procedure_surface.arrow.catalog.GraphWithProjectResult¶

Result object for graph projection jobs, containing the projected graph and the projection result. Can be used as a context manager to ensure the projected graph is dropped after use.

static __new__(_cls, graph: Graph, result: ProjectionResult | StoreProjectionResult)¶

Create new instance of GraphWithProjectResult(graph, result)

Parameters:

graph (Graph)
result (ProjectionResult | StoreProjectionResult)

graph: Graph¶: Alias for field number 0

result: ProjectionResult | StoreProjectionResult¶: Alias for field number 1

class graphdatascience.procedure_surface.arrow.catalog.ProjectArrowEndpoints¶

Endpoints for projecting graphs via the Arrow pipeline.

cypher(graph_name: str, query: str, *, query_parameters: dict[str, Any] | None = None, job_id: str | None = None, concurrency: int | None = None, undirected_relationship_types: List[str] | None = None, inverse_indexed_relationship_types: List[str] | None = None, batch_size: int | None = None, logging: bool = True, overwrite: bool = False) → GraphWithProjectResult¶

Projects a graph from the Neo4j database into the GDS graph catalog using Cypher projection.

Parameters:

graph_name (str) – Name of the graph to be created
query (str) – Cypher query to select nodes and relationships for the graph projection. Must contain gds.graph.project.remote. Example: MATCH (n)–>(m) RETURN gds.graph.project.remote(n, m)
query_parameters (dict[str, Any] | None) – Parameters that will be passed to the Cypher query.
job_id (str | None) – Identifier for the computation.
concurrency (int | None) – Number of concurrent threads to use.
undirected_relationship_types (List[str] | None) – List of relationship types to treat as undirected.
inverse_indexed_relationship_types (List[str] | None) – List of relationship types to index in both directions.
batch_size (int | None) – Number of rows to process in each batch when projecting the graph.
logging (bool) – Whether to log progress during graph projection.
overwrite (bool) – If True, drop an existing graph with the same name before projecting the new one. Defaults to False.

Returns:

A result object containing information about the projected graph.

Return type:

GraphWithProjectResult

cypher_async(graph_name: str, query: str, *, query_parameters: dict[str, Any] | None = None, job_id: str | None = None, concurrency: int | None = None, undirected_relationship_types: List[str] | None = None, inverse_indexed_relationship_types: List[str] | None = None, batch_size: int | None = None, overwrite: bool = False) → ProjectionJobHandle¶

Kick off a cypher graph projection and return a ProjectionJobHandle.

Unlike cypher(), this method does not block on completion. Use the returned handle to query status or retrieve the projected graph and result.

Parameters:

graph_name (str)
query (str)
query_parameters (dict[str, Any] | None)
job_id (str | None)
concurrency (int | None)
undirected_relationship_types (List[str] | None)
inverse_indexed_relationship_types (List[str] | None)
batch_size (int | None)
overwrite (bool)

Return type:

ProjectionJobHandle

native(graph_name: str, node_label_filter: List[str], relationship_type_filter: List[str], *, node_properties: List[str] | None = None, relationship_properties: List[str] | None = None, job_id: str | None = None, concurrency: int | None = None, undirected_relationship_types: List[str] | None = None, inverse_indexed_relationship_types: List[str] | None = None, batch_size: int | None = None, logging: bool = True, overwrite: bool = False) → GraphWithProjectResult¶

Projects a graph from the Neo4j database into the GDS graph catalog.

Parameters:

graph_name (str) – Name of the graph to be created
node_label_filter (List[str]) – List of node labels to include in the graph projection.
relationship_type_filter (List[str]) – List of relationship types to include in the graph projection.
node_properties (List[str] | None) – List of node properties to include in the graph projection.
relationship_properties (List[str] | None) – List of relationship properties to include in the graph projection.
job_id (str | None) – Identifier for the computation.
concurrency (int | None) – Number of concurrent threads to use.
undirected_relationship_types (List[str] | None) – List of relationship types to treat as undirected.
inverse_indexed_relationship_types (List[str] | None) – List of relationship types to index in both directions.
batch_size (int | None) – Number of rows to process in each batch when projecting the graph.
logging (bool) – Whether to log progress during graph projection.
overwrite (bool) – If True, drop an existing graph with the same name before projecting the new one. Defaults to False.

Returns:

A result object containing information about the projected graph.

Return type:

GraphWithProjectResult

native_async(graph_name: str, node_label_filter: List[str], relationship_type_filter: List[str], *, node_properties: List[str] | None = None, relationship_properties: List[str] | None = None, job_id: str | None = None, concurrency: int | None = None, undirected_relationship_types: List[str] | None = None, inverse_indexed_relationship_types: List[str] | None = None, batch_size: int | None = None, overwrite: bool = False) → ProjectionJobHandle¶

Kick off a native graph projection and return a ProjectionJobHandle.

Unlike native(), this method does not block on completion. The returned handle can be used to await completion and retrieve the projected graph and result.

Parameters:

graph_name (str)
node_label_filter (List[str])
relationship_type_filter (List[str])
node_properties (List[str] | None)
relationship_properties (List[str] | None)
job_id (str | None)
concurrency (int | None)
undirected_relationship_types (List[str] | None)
inverse_indexed_relationship_types (List[str] | None)
batch_size (int | None)
overwrite (bool)

Return type:

ProjectionJobHandle

pydantic model graphdatascience.procedure_surface.arrow.catalog.ProjectionResult¶

Result object for graph projection jobs.

field configuration: dict[str, Any]¶

field graph_name: str¶

field node_count: int¶

field project_millis: int¶

field query: str¶

field relationship_count: int¶

pydantic model graphdatascience.procedure_surface.arrow.catalog.StoreProjectionResult¶

Result object for native graph projection jobs.

field graph_name: str¶

field node_count: int¶

field project_millis: int¶

field relationship_count: int¶