[BUG][Spark] DeltaTable.forPath doesn't work with Spark Connect

[stvno]

Bug

Which Delta project/connector is this regarding?

• Spark
• Standalone
• Flink
• Kernel
• Other (fill in here)

Describe the problem

I'm using a Jupyter notebook to connect to a remote Spark cluster, using the Spark Connect functionality in Spark 3.4.1. I can read and write Spark DataFrames in delta format to my S3 bucket. However, when I want to load the data as a deltatable DeltaTable.forPath() it fails because it assumes that the Spark Session has a SparkContext

Steps to reproduce

1. Prepare a spark connect server, see https://spark.apache.org/docs/latest/spark-connect-overview.html eg. ./sbin/start-connect-server.sh --packages org.apache.spark:spark-connect_2.12:3.4.0
2. Connect with python and write some data

from pyspark.sql import SparkSession

spark = SparkSession.builder.remote("sc://localhost").getOrCreate()
columns = ["id","name"]
data = [(1,"Sarah"),(2,"Maria")]
df = spark.createDataFrame(data).toDF(*columns)
df.write.format('delta').mode('overwrite').save('s3a://<bucket>/<name>')

3. Read the data back, this all works fine

data = spark.read.format("delta").load('s3a://<bucket>/<name>')
data.show(n=3, truncate=False)

4. Load the data as a deltatable

from pyspark.sql import SparkSession
from delta.tables import * 

spark = SparkSession.builder.remote("sc://localhost").getOrCreate()

dt = DeltaTable.forPath(spark, "s3a://<bucket>/<name>")

Observed results

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[41], line 8
      4 spark = SparkSession.builder \
      5      .remote("sc://localhost").getOrCreate()
      7 # Lees de delta-tabel in als deltatable
----> 8 dt = DeltaTable.forPath(spark, "s3a://<bucket>/<name>")
     

File /opt/app-root/lib64/python3.8/site-packages/delta/tables.py:384, in DeltaTable.forPath(cls, sparkSession, path, hadoopConf)
    359 """
    360 Instantiate a :class:`DeltaTable` object representing the data at the given path,
    361 If the given path is invalid (i.e. either no table exists or an existing table is
   (...)
    380                    hadoopConf)
    381 """
    382 assert sparkSession is not None
--> 384 jvm: "JVMView" = sparkSession._sc._jvm  # type: ignore[attr-defined]
    385 jsparkSession: "JavaObject" = sparkSession._jsparkSession  # type: ignore[attr-defined]
    387 jdt = jvm.io.delta.tables.DeltaTable.forPath(jsparkSession, path, hadoopConf)

AttributeError: 'SparkSession' object has no attribute '_sc'

Expected results

The data loaded as a DeltaTable object

Further details

I am aware that Spark Connect is very new, not yet feature complete and lacking some documentation on which features are and aren't supported. The best I have found so far are: https://issues.apache.org/jira/browse/SPARK-39375 and databrick's implementation of Spark Connect https://docs.databricks.com/en/dev-tools/databricks-connect-ref.html#limitations where the altter mentions that the SparkContext class and methods are not available.

Environment information

• Delta Lake version: 2.4.0
• Spark version: 3.4.1
• Scala version: 2.12

Willingness to contribute

The Delta Lake Community encourages bug fix contributions. Would you or another member of your organization be willing to contribute a fix for this bug to the Delta Lake code base?

• Yes. I can contribute a fix for this bug independently.
• Yes. I would be willing to contribute a fix for this bug with guidance from the Delta Lake community.
• No. I cannot contribute a bug fix at this time.

[cosmincatalin]

Happens as well for DeltaTable.forName

[allisonport-db]

Cross-linking this with a parent issue for full Delta support using Spark Connect #3240