Create Pregel and Neo4j Ingeration tutorials and add Pregel API to user guide - #578

Previously missed (5) — in code that hasn't changed since the last review.

python/graphframes/tutorials/neo4j/loaders.py:35

• get_container_path() uses relpath() without verifying that host_path is under data_dir. If a caller passes a path outside the mounted directory, the resulting file:// URI can include ".." segments and potentially reference paths outside Neo4j’s import mount. Validate host_path is within data_dir before constructing the container URI.

    # Get the path relative to the data_dir
    abs_data_dir = os.path.abspath(data_dir)
    abs_host_path = os.path.abspath(host_path)

    # Calculate relative path from data_dir
    rel_path = os.path.relpath(abs_host_path, abs_data_dir)

python/graphframes/tutorials/neo4j/load.py:10

• The usage example points to python/graphframes/tutorials/neo4j_load.py, but the script lives at python/graphframes/tutorials/neo4j/load.py. This makes the example command fail if copied verbatim.

    spark-submit \\
      --packages org.neo4j:neo4j-connector-apache-spark_2.12:5.3.1_for_spark_3 \\
      --driver-memory 4g \\
      --executor-memory 4g \\
      python/graphframes/tutorials/neo4j_load.py

python/graphframes/tutorials/neo4j/connected_components.py:54

• The Neo4j read query selects properties using capitalized keys (e.g. n.DisplayName, n.Title), but the APOC loaders in this PR write properties using lower camelCase (e.g. displayName, title). With the current query, these fields will always be null when data is loaded via the CLI/APOC path; use coalesce() to support both schemas.

        MATCH (n)
        RETURN
            n.id as id,
            labels(n)[0] as Type,
            n.Id as Id,

python/graphframes/tutorials/neo4j/connected_components.py:166

• Writing back to Neo4j uses .mode("Overwrite"). The tutorial text/examples in this PR use append semantics for updates, and overwrite mode can be interpreted by connectors as destructive depending on configuration. Use Append here to avoid any chance of dropping existing data while running the tutorial.

    component_results.write
    .format("org.neo4j.spark.DataSource")
    .mode("Overwrite")
    .option("query", """
        UNWIND $rows AS row

python/graphframes/tutorials/neo4j/connected_components.py:9

• The usage example points to python/graphframes/tutorials/neo4j_connected_components.py, but the script is python/graphframes/tutorials/neo4j/connected_components.py. This makes the example command fail if copied verbatim.

    spark-submit \\
      --packages org.neo4j:neo4j-connector-apache-spark_2.12:5.3.1_for_spark_3 \\
      --driver-memory 4g \\
      --executor-memory 4g \\
      python/graphframes/tutorials/neo4j_connected_components.py

python/graphframes/console.py:5

• graphframes.console imports the Neo4j CLI unconditionally, but the neo4j Python driver is not declared in python/pyproject.toml (and isn’t present in python/poetry.lock). As a result, importing graphframes.console / running the graphframes CLI will raise ModuleNotFoundError in environments that don’t have the optional Neo4j dependency installed.

import click

from graphframes.tutorials import download
from graphframes.tutorials.neo4j import neo4j