[SPARK-47683][PYTHON][BUILD][FOLLOW-UP] Exclude lib/py4j*zip in pyspark-connect package

python/packaging/classic/setup.py

@@ -204,8 +204,13 @@ def run(self):
     copyfile("pyspark/shell.py", "pyspark/python/pyspark/shell.py")
 
     if in_spark:
+        # !!HACK ALTERT!!
+        # `setup.py` has to be located with the same directory with the package.
+        # Therefore, we copy the current file, and place it at `spark/python` directory.
+        # After that, we remove it in the end.
         copyfile("packaging/classic/setup.py", "setup.py")
         copyfile("packaging/classic/setup.cfg", "setup.cfg")
+
         # Construct the symlink farm - this is nein_sparkcessary since we can't refer to
         # the path above the package root and we need to copy the jars and scripts which
         # are up above the python root.

python/packaging/connect/setup.py

@@ -25,7 +25,7 @@
 import sys
 from setuptools import setup
 import os
-from shutil import copyfile
+from shutil import copyfile, move
 import glob
 from pathlib import Path
 
@@ -109,6 +109,13 @@
 
 try:
     if in_spark:
+        # !!HACK ALTERT!!
+        # 1. `setup.py` has to be located with the same directory with the package.
+        #    Therefore, we copy the current file, and place it at `spark/python` directory.
+        #    After that, we remove it in the end.
+        # 2. Here it renames `lib` to `lib.ack` so MANIFEST.in does not pick `py4j` up.
+        #    We rename it back in the end.
+        move("lib", "lib.back")
         copyfile("packaging/connect/setup.py", "setup.py")
         copyfile("packaging/connect/setup.cfg", "setup.cfg")
 
@@ -206,5 +213,6 @@
     )
 finally:
     if in_spark:
+        move("lib.back", "lib")
         os.remove("setup.py")
         os.remove("setup.cfg")