How to link PyCharm with PySpark?

Question

I'm new with apache spark and apparently I installed apache-spark with homebrew in my macbook:

Last login: Fri Jan  8 12:52:04 on console
user@MacBook-Pro-de-User-2:~$ pyspark
Python 2.7.10 (default, Jul 13 2015, 12:05:58)
[GCC 4.2.1 Compatible Apple LLVM 6.1.0 (clang-602.0.53)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
Using Spark's default log4j profile: org/apache/spark/log4j-defaults.properties
16/01/08 14:46:44 INFO SparkContext: Running Spark version 1.5.1
16/01/08 14:46:46 WARN NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
16/01/08 14:46:47 INFO SecurityManager: Changing view acls to: user
16/01/08 14:46:47 INFO SecurityManager: Changing modify acls to: user
16/01/08 14:46:47 INFO SecurityManager: SecurityManager: authentication disabled; ui acls disabled; users with view permissions: Set(user); users with modify permissions: Set(user)
16/01/08 14:46:50 INFO Slf4jLogger: Slf4jLogger started
16/01/08 14:46:50 INFO Remoting: Starting remoting
16/01/08 14:46:51 INFO Remoting: Remoting started; listening on addresses :[akka.tcp://[email protected]:50199]
16/01/08 14:46:51 INFO Utils: Successfully started service 'sparkDriver' on port 50199.
16/01/08 14:46:51 INFO SparkEnv: Registering MapOutputTracker
16/01/08 14:46:51 INFO SparkEnv: Registering BlockManagerMaster
16/01/08 14:46:51 INFO DiskBlockManager: Created local directory at /private/var/folders/5x/k7n54drn1csc7w0j7vchjnmc0000gn/T/blockmgr-769e6f91-f0e7-49f9-b45d-1b6382637c95
16/01/08 14:46:51 INFO MemoryStore: MemoryStore started with capacity 530.0 MB
16/01/08 14:46:52 INFO HttpFileServer: HTTP File server directory is /private/var/folders/5x/k7n54drn1csc7w0j7vchjnmc0000gn/T/spark-8e4749ea-9ae7-4137-a0e1-52e410a8e4c5/httpd-1adcd424-c8e9-4e54-a45a-a735ade00393
16/01/08 14:46:52 INFO HttpServer: Starting HTTP Server
16/01/08 14:46:52 INFO Utils: Successfully started service 'HTTP file server' on port 50200.
16/01/08 14:46:52 INFO SparkEnv: Registering OutputCommitCoordinator
16/01/08 14:46:52 INFO Utils: Successfully started service 'SparkUI' on port 4040.
16/01/08 14:46:52 INFO SparkUI: Started SparkUI at http://192.168.1.64:4040
16/01/08 14:46:53 WARN MetricsSystem: Using default name DAGScheduler for source because spark.app.id is not set.
16/01/08 14:46:53 INFO Executor: Starting executor ID driver on host localhost
16/01/08 14:46:53 INFO Utils: Successfully started service 'org.apache.spark.network.netty.NettyBlockTransferService' on port 50201.
16/01/08 14:46:53 INFO NettyBlockTransferService: Server created on 50201
16/01/08 14:46:53 INFO BlockManagerMaster: Trying to register BlockManager
16/01/08 14:46:53 INFO BlockManagerMasterEndpoint: Registering block manager localhost:50201 with 530.0 MB RAM, BlockManagerId(driver, localhost, 50201)
16/01/08 14:46:53 INFO BlockManagerMaster: Registered BlockManager
Welcome to
      ____              __
     / __/__  ___ _____/ /__
    _\ \/ _ \/ _ `/ __/  '_/
   /__ / .__/\_,_/_/ /_/\_\   version 1.5.1
      /_/

Using Python version 2.7.10 (default, Jul 13 2015 12:05:58)
SparkContext available as sc, HiveContext available as sqlContext.
>>>

I would like start playing in order to learn more about MLlib. However, I use Pycharm to write scripts in python. The problem is: when I go to Pycharm and try to call pyspark, Pycharm can not found the module. I tried adding the path to Pycharm as follows:

Then from a blog I tried this:

import os
import sys

# Path for spark source folder
os.environ['SPARK_HOME']="/Users/user/Apps/spark-1.5.2-bin-hadoop2.4"

# Append pyspark  to Python Path
sys.path.append("/Users/user/Apps/spark-1.5.2-bin-hadoop2.4/python/pyspark")

try:
    from pyspark import SparkContext
    from pyspark import SparkConf
    print ("Successfully imported Spark Modules")

except ImportError as e:
    print ("Can not import Spark Modules", e)
    sys.exit(1)

And still can not start using PySpark with Pycharm, any idea of how to "link" PyCharm with apache-pyspark?.

Update:

Then I search for apache-spark and python path in order to set the environment variables of Pycharm:

apache-spark path:

user@MacBook-Pro-User-2:~$ brew info apache-spark
apache-spark: stable 1.6.0, HEAD
Engine for large-scale data processing
https://spark.apache.org/
/usr/local/Cellar/apache-spark/1.5.1 (649 files, 302.9M) *
  Poured from bottle
From: https://github.com/Homebrew/homebrew/blob/master/Library/Formula/apache-spark.rb

python path:

user@MacBook-Pro-User-2:~$ brew info python
python: stable 2.7.11 (bottled), HEAD
Interpreted, interactive, object-oriented programming language
https://www.python.org
/usr/local/Cellar/python/2.7.10_2 (4,965 files, 66.9M) *

Then with the above information I tried to set the environment variables as follows:

Any idea of how to correctly link Pycharm with pyspark?

Then when I run a python script with the above configuration I have this exception:

/usr/local/Cellar/python/2.7.10_2/Frameworks/Python.framework/Versions/2.7/bin/python2.7 /Users/user/PycharmProjects/spark_examples/test_1.py
Traceback (most recent call last):
  File "/Users/user/PycharmProjects/spark_examples/test_1.py", line 1, in <module>
    from pyspark import SparkContext
ImportError: No module named pyspark

UPDATE: Then I tried this configurations proposed by @zero323

Configuration 1:

/usr/local/Cellar/apache-spark/1.5.1/ 

out:

 user@MacBook-Pro-de-User-2:/usr/local/Cellar/apache-spark/1.5.1$ ls
CHANGES.txt           NOTICE                libexec/
INSTALL_RECEIPT.json  README.md
LICENSE               bin/

Configuration 2:

/usr/local/Cellar/apache-spark/1.5.1/libexec 

out:

user@MacBook-Pro-de-User-2:/usr/local/Cellar/apache-spark/1.5.1/libexec$ ls
R/        bin/      data/     examples/ python/
RELEASE   conf/     ec2/      lib/      sbin/

Solution 1

With PySpark package (Spark 2.2.0 and later)

With SPARK-1267 being merged you should be able to simplify the process by pip installing Spark in the environment you use for PyCharm development.

1. Go to File -> Settings -> Project Interpreter

2. Click on install button and search for PySpark

   

3. Click on install package button.

Manually with user provided Spark installation

Create Run configuration:

1. Go to Run -> Edit configurations
2. Add new Python configuration
3. Set Script path so it points to the script you want to execute

4. Edit Environment variables field so it contains at least:

   • SPARK_HOME - it should point to the directory with Spark installation. It should contain directories such as bin (with spark-submit, spark-shell, etc.) and conf (with spark-defaults.conf, spark-env.sh, etc.)

   • PYTHONPATH - it should contain $SPARK_HOME/python and optionally $SPARK_HOME/python/lib/py4j-some-version.src.zip if not available otherwise. some-version should match Py4J version used by a given Spark installation (0.8.2.1 - 1.5, 0.9 - 1.6, 0.10.3 - 2.0, 0.10.4 - 2.1, 0.10.4 - 2.2, 0.10.6 - 2.3, 0.10.7 - 2.4)

     

5. Apply the settings

Add PySpark library to the interpreter path (required for code completion):

1. Go to File -> Settings -> Project Interpreter
2. Open settings for an interpreter you want to use with Spark
3. Edit interpreter paths so it contains path to $SPARK_HOME/python (an Py4J if required)
4. Save the settings

Optionally

1. Install or add to path type annotations matching installed Spark version to get better completion and static error detection (Disclaimer - I am an author of the project).

Finally

Use newly created configuration to run your script.

Solution 2

Here's how I solved this on mac osx.

1. brew install apache-spark

2. Add this to ~/.bash_profile

   export SPARK_VERSION=`ls /usr/local/Cellar/apache-spark/ | sort | tail -1`
   export SPARK_HOME="/usr/local/Cellar/apache-spark/$SPARK_VERSION/libexec"
   export PYTHONPATH=$SPARK_HOME/python/:$PYTHONPATH
   export PYTHONPATH=$SPARK_HOME/python/lib/py4j-0.9-src.zip:$PYTHONPATH
   

3. Add pyspark and py4j to content root (use the correct Spark version):

   /usr/local/Cellar/apache-spark/1.6.1/libexec/python/lib/py4j-0.9-src.zip
   /usr/local/Cellar/apache-spark/1.6.1/libexec/python/lib/pyspark.zip
   

Solution 3

Here is the setup that works for me (Win7 64bit, PyCharm2017.3CE)

Set up Intellisense:

1. Click File -> Settings -> Project: -> Project Interpreter

2. Click the gear icon to the right of the Project Interpreter dropdown

3. Click More... from the context menu

4. Choose the interpreter, then click the "Show Paths" icon (bottom right)

5. Click the + icon two add the following paths:

   \python\lib\py4j-0.9-src.zip

   \bin\python\lib\pyspark.zip

6. Click OK, OK, OK

Go ahead and test your new intellisense capabilities.

Solution 4

Configure pyspark in pycharm (windows)

File menu - settings - project interpreter - (gearshape) - more - (treebelowfunnel) - (+) - [add python folder form spark installation and then py4j-*.zip] - click ok

Ensure SPARK_HOME set in windows environment, pycharm will take from there. To confirm :

Run menu - edit configurations - environment variables - [...] - show

Optionally set SPARK_CONF_DIR in environment variables.

Solution 5

I used the following page as a reference and was able to get pyspark/Spark 1.6.1 (installed via homebrew) imported in PyCharm 5.

http://renien.com/blog/accessing-pyspark-pycharm/

import os
import sys

# Path for spark source folder
os.environ['SPARK_HOME']="/usr/local/Cellar/apache-spark/1.6.1"

# Append pyspark  to Python Path
sys.path.append("/usr/local/Cellar/apache-spark/1.6.1/libexec/python")

try:
    from pyspark import SparkContext
    from pyspark import SparkConf
    print ("Successfully imported Spark Modules")
except ImportError as e:
    print ("Can not import Spark Modules", e)
sys.exit(1)

With the above, pyspark loads, but I get a gateway error when I try to create a SparkContext. There's some issue with Spark from homebrew, so I just grabbed Spark from the Spark website (download the Pre-built for Hadoop 2.6 and later) and point to the spark and py4j directories under that. Here's the code in pycharm that works!

import os
import sys

# Path for spark source folder
os.environ['SPARK_HOME']="/Users/myUser/Downloads/spark-1.6.1-bin-hadoop2.6"

# Need to Explicitly point to python3 if you are using Python 3.x
os.environ['PYSPARK_PYTHON']="/usr/local/Cellar/python3/3.5.1/bin/python3"

#You might need to enter your local IP
#os.environ['SPARK_LOCAL_IP']="192.168.2.138"

#Path for pyspark and py4j
sys.path.append("/Users/myUser/Downloads/spark-1.6.1-bin-hadoop2.6/python")
sys.path.append("/Users/myUser/Downloads/spark-1.6.1-bin-hadoop2.6/python/lib/py4j-0.9-src.zip")

try:
    from pyspark import SparkContext
    from pyspark import SparkConf
    print ("Successfully imported Spark Modules")
except ImportError as e:
    print ("Can not import Spark Modules", e)
    sys.exit(1)

sc = SparkContext('local')
words = sc.parallelize(["scala","java","hadoop","spark","akka"])
print(words.count())

I had a lot of help from these instructions, which helped me troubleshoot in PyDev and then get it working PyCharm - https://enahwe.wordpress.com/2015/11/25/how-to-configure-eclipse-for-developing-with-python-and-spark-on-hadoop/

I'm sure somebody has spent a few hours bashing their head against their monitor trying to get this working, so hopefully this helps save their sanity!

View more solutions

Comments

[Jason Wolosonovich]

@ml_student I'll also mention that if you follow the video method (which would be my recommendation for its speed and ease) you'll need to instantiate a SparkContext object at the beginning of your script as well. I note this because using the interactive pyspark console via the command line automatically creates the context for you, whereas in PyCharm, you need to take care of that yourself; syntax would be: sc = SparkContext()

[ravindrab]

which version of pycharm is this? I am on 2016.1 community edition and I don't see this window.

[sthomps]

2016.1 Im on osx but it should be similar. Go under 'Preferences'. Click on your project on the left.

[OneCricketeer]

Thanks. This helped me on IntelliJ IDEA, which doesn't have the Project Interpreter setting.

[OneCricketeer]

you should be able to get it working in PyCharm by setting the project's interpreter to spark-submit -- Tried it. "The selected file is not a valid home for Python SDK". Same for bin/pyspark

[OneCricketeer]

Could you explain what adding to the content root does? I didn't need to do that... I just put the $SPARK_HOME/python in the interpreter classpath and added the Environment variables and it works as expected.

[Ajeet Shah]

@cricket_007 The 3rd point: Add pyspark and py4j to content root (use the correct Spark version) helped me in code completion. How did you get it done by changing Project Interpreter?

[OneCricketeer]

@Orions I just used a virtualenv interpreter. I don't know if I tested auto-completion. I was saying that I could execute PySpark code as expected.

[AlonL]

BTW, this is how you're editing the interpreter paths, at least in PyCharm 2016: jetbrains.com/help/pycharm/2016.1/… Select the "Show paths for the selected interpreter" button

[Vishwas Shashidhar]

Perfect solution! Thank you.

[Dichen]

Thanks! You saved me from 2 hours of desperately trying to make this work

[H S Rathore]

This way pycharm will think its another lib.

[Random Certainty]

On Mac version of PyCharm (v-2017.2), the Project Interpreter is under Preferences... instead of File/Settings

[lfk]

With option 1, how do you add Spark JARs/packages? e.g., I need com.databricks:spark-redshift_2.10:3.0.0-preview1

[10465355]

@lfk Either through configuration files (spark-defaults.conf) or through submit args - same as with Jupyter notebook. Submit args can defined in PyCharm's Environment variables, instead of code, if you prefer this option.

[Rohit Nimmala]

Much needed answer :)