[python] Load CSV file with Spark

I'm new to Spark and I'm trying to read CSV data from a file with Spark. Here's what I am doing :

sc.textFile('file.csv')
    .map(lambda line: (line.split(',')[0], line.split(',')[1]))
    .collect()

I would expect this call to give me a list of the two first columns of my file but I'm getting this error :

File "<ipython-input-60-73ea98550983>", line 1, in <lambda>
IndexError: list index out of range

although my CSV file as more than one column.

And yet another option which consist in reading the CSV file using Pandas and then importing the Pandas DataFrame into Spark.

For example:

from pyspark import SparkContext
from pyspark.sql import SQLContext
import pandas as pd

sc = SparkContext('local','example')  # if using locally
sql_sc = SQLContext(sc)

pandas_df = pd.read_csv('file.csv')  # assuming the file contains a header
# pandas_df = pd.read_csv('file.csv', names = ['column 1','column 2']) # if no header
s_df = sql_sc.createDataFrame(pandas_df)

Simply splitting by comma will also split commas that are within fields (e.g. a,b,"1,2,3",c), so it's not recommended. zero323's answer is good if you want to use the DataFrames API, but if you want to stick to base Spark, you can parse csvs in base Python with the csv module:

# works for both python 2 and 3
import csv
rdd = sc.textFile("file.csv")
rdd = rdd.mapPartitions(lambda x: csv.reader(x))

EDIT: As @muon mentioned in the comments, this will treat the header like any other row so you'll need to extract it manually. For example, header = rdd.first(); rdd = rdd.filter(lambda x: x != header) (make sure not to modify header before the filter evaluates). But at this point, you're probably better off using a built-in csv parser.

If your csv data happens to not contain newlines in any of the fields, you can load your data with textFile() and parse it

import csv
import StringIO

def loadRecord(line):
    input = StringIO.StringIO(line)
    reader = csv.DictReader(input, fieldnames=["name1", "name2"])
    return reader.next()

input = sc.textFile(inputFile).map(loadRecord)

from pyspark.sql import SparkSession

spark = SparkSession \
    .builder \
    .appName("Python Spark SQL basic example") \
    .config("spark.some.config.option", "some-value") \
    .getOrCreate()

df = spark.read.csv("/home/stp/test1.csv",header=True,sep="|")

print(df.collect())

This is in-line with what JP Mercier initially suggested about using Pandas, but with a major modification: If you read data into Pandas in chunks, it should be more malleable. Meaning, that you can parse a much larger file than Pandas can actually handle as a single piece and pass it to Spark in smaller sizes. (This also answers the comment about why one would want to use Spark if they can load everything into Pandas anyways.)

from pyspark import SparkContext
from pyspark.sql import SQLContext
import pandas as pd

sc = SparkContext('local','example')  # if using locally
sql_sc = SQLContext(sc)

Spark_Full = sc.emptyRDD()
chunk_100k = pd.read_csv("Your_Data_File.csv", chunksize=100000)
# if you have headers in your csv file:
headers = list(pd.read_csv("Your_Data_File.csv", nrows=0).columns)

for chunky in chunk_100k:
    Spark_Full +=  sc.parallelize(chunky.values.tolist())

YourSparkDataFrame = Spark_Full.toDF(headers)
# if you do not have headers, leave empty instead:
# YourSparkDataFrame = Spark_Full.toDF()
YourSparkDataFrame.show()

This is in PYSPARK

path="Your file path with file name"

df=spark.read.format("csv").option("header","true").option("inferSchema","true").load(path)

Then you can check

df.show(5)
df.count()

If you want to load csv as a dataframe then you can do the following:

from pyspark.sql import SQLContext
sqlContext = SQLContext(sc)

df = sqlContext.read.format('com.databricks.spark.csv') \
    .options(header='true', inferschema='true') \
    .load('sampleFile.csv') # this is your csv file

It worked fine for me.

If you are having any one or more row(s) with less or more number of columns than 2 in the dataset then this error may arise.

I am also new to Pyspark and trying to read CSV file. Following code worked for me:

In this code I am using dataset from kaggle the link is: https://www.kaggle.com/carrie1/ecommerce-data

1. Without mentioning the schema:

from pyspark.sql import SparkSession  
scSpark = SparkSession \
    .builder \
    .appName("Python Spark SQL basic example: Reading CSV file without mentioning schema") \
    .config("spark.some.config.option", "some-value") \
    .getOrCreate()

sdfData = scSpark.read.csv("data.csv", header=True, sep=",")
sdfData.show()

Now check the columns: sdfData.columns

Output will be:

['InvoiceNo', 'StockCode','Description','Quantity', 'InvoiceDate', 'CustomerID', 'Country']

Check the datatype for each column:

sdfData.schema
StructType(List(StructField(InvoiceNo,StringType,true),StructField(StockCode,StringType,true),StructField(Description,StringType,true),StructField(Quantity,StringType,true),StructField(InvoiceDate,StringType,true),StructField(UnitPrice,StringType,true),StructField(CustomerID,StringType,true),StructField(Country,StringType,true)))

This will give the data frame with all the columns with datatype as StringType

2. With schema: If you know the schema or want to change the datatype of any column in the above table then use this (let's say I am having following columns and want them in a particular data type for each of them)

from pyspark.sql import SparkSession  
from pyspark.sql.types import StructType, StructField
from pyspark.sql.types import DoubleType, IntegerType, StringType
    schema = StructType([\
        StructField("InvoiceNo", IntegerType()),\
        StructField("StockCode", StringType()), \
        StructField("Description", StringType()),\
        StructField("Quantity", IntegerType()),\
        StructField("InvoiceDate", StringType()),\
        StructField("CustomerID", DoubleType()),\
        StructField("Country", StringType())\
    ])

scSpark = SparkSession \
    .builder \
    .appName("Python Spark SQL example: Reading CSV file with schema") \
    .config("spark.some.config.option", "some-value") \
    .getOrCreate()

sdfData = scSpark.read.csv("data.csv", header=True, sep=",", schema=schema)

Now check the schema for datatype of each column:

sdfData.schema

StructType(List(StructField(InvoiceNo,IntegerType,true),StructField(StockCode,StringType,true),StructField(Description,StringType,true),StructField(Quantity,IntegerType,true),StructField(InvoiceDate,StringType,true),StructField(CustomerID,DoubleType,true),StructField(Country,StringType,true)))

Edited: We can use the following line of code as well without mentioning schema explicitly:

sdfData = scSpark.read.csv("data.csv", header=True, inferSchema = True)
sdfData.schema

The output is:

StructType(List(StructField(InvoiceNo,StringType,true),StructField(StockCode,StringType,true),StructField(Description,StringType,true),StructField(Quantity,IntegerType,true),StructField(InvoiceDate,StringType,true),StructField(UnitPrice,DoubleType,true),StructField(CustomerID,IntegerType,true),StructField(Country,StringType,true)))

The output will look like this:

sdfData.show()

+---------+---------+--------------------+--------+--------------+----------+-------+
|InvoiceNo|StockCode|         Description|Quantity|   InvoiceDate|CustomerID|Country|
+---------+---------+--------------------+--------+--------------+----------+-------+
|   536365|   85123A|WHITE HANGING HEA...|       6|12/1/2010 8:26|      2.55|  17850|
|   536365|    71053| WHITE METAL LANTERN|       6|12/1/2010 8:26|      3.39|  17850|
|   536365|   84406B|CREAM CUPID HEART...|       8|12/1/2010 8:26|      2.75|  17850|
|   536365|   84029G|KNITTED UNION FLA...|       6|12/1/2010 8:26|      3.39|  17850|
|   536365|   84029E|RED WOOLLY HOTTIE...|       6|12/1/2010 8:26|      3.39|  17850|
|   536365|    22752|SET 7 BABUSHKA NE...|       2|12/1/2010 8:26|      7.65|  17850|
|   536365|    21730|GLASS STAR FROSTE...|       6|12/1/2010 8:26|      4.25|  17850|
|   536366|    22633|HAND WARMER UNION...|       6|12/1/2010 8:28|      1.85|  17850|
|   536366|    22632|HAND WARMER RED P...|       6|12/1/2010 8:28|      1.85|  17850|
|   536367|    84879|ASSORTED COLOUR B...|      32|12/1/2010 8:34|      1.69|  13047|
|   536367|    22745|POPPY'S PLAYHOUSE...|       6|12/1/2010 8:34|       2.1|  13047|
|   536367|    22748|POPPY'S PLAYHOUSE...|       6|12/1/2010 8:34|       2.1|  13047|
|   536367|    22749|FELTCRAFT PRINCES...|       8|12/1/2010 8:34|      3.75|  13047|
|   536367|    22310|IVORY KNITTED MUG...|       6|12/1/2010 8:34|      1.65|  13047|
|   536367|    84969|BOX OF 6 ASSORTED...|       6|12/1/2010 8:34|      4.25|  13047|
|   536367|    22623|BOX OF VINTAGE JI...|       3|12/1/2010 8:34|      4.95|  13047|
|   536367|    22622|BOX OF VINTAGE AL...|       2|12/1/2010 8:34|      9.95|  13047|
|   536367|    21754|HOME BUILDING BLO...|       3|12/1/2010 8:34|      5.95|  13047|
|   536367|    21755|LOVE BUILDING BLO...|       3|12/1/2010 8:34|      5.95|  13047|
|   536367|    21777|RECIPE BOX WITH M...|       4|12/1/2010 8:34|      7.95|  13047|
+---------+---------+--------------------+--------+--------------+----------+-------+
only showing top 20 rows

Spark 2.0.0+

You can use built-in csv data source directly:

spark.read.csv(
    "some_input_file.csv", header=True, mode="DROPMALFORMED", schema=schema
)

or

(spark.read
    .schema(schema)
    .option("header", "true")
    .option("mode", "DROPMALFORMED")
    .csv("some_input_file.csv"))

without including any external dependencies.

Spark < 2.0.0:

Instead of manual parsing, which is far from trivial in a general case, I would recommend spark-csv:

Make sure that Spark CSV is included in the path (--packages, --jars, --driver-class-path)

And load your data as follows:

(df = sqlContext
    .read.format("com.databricks.spark.csv")
    .option("header", "true")
    .option("inferschema", "true")
    .option("mode", "DROPMALFORMED")
    .load("some_input_file.csv"))

It can handle loading, schema inference, dropping malformed lines and doesn't require passing data from Python to the JVM.

Note:

If you know the schema, it is better to avoid schema inference and pass it to DataFrameReader. Assuming you have three columns - integer, double and string:

from pyspark.sql.types import StructType, StructField
from pyspark.sql.types import DoubleType, IntegerType, StringType

schema = StructType([
    StructField("A", IntegerType()),
    StructField("B", DoubleType()),
    StructField("C", StringType())
])

(sqlContext
    .read
    .format("com.databricks.spark.csv")
    .schema(schema)
    .option("header", "true")
    .option("mode", "DROPMALFORMED")
    .load("some_input_file.csv"))

Now, there's also another option for any general csv file: https://github.com/seahboonsiew/pyspark-csv as follows:

Assume we have the following context

sc = SparkContext
sqlCtx = SQLContext or HiveContext

First, distribute pyspark-csv.py to executors using SparkContext

import pyspark_csv as pycsv
sc.addPyFile('pyspark_csv.py')

Read csv data via SparkContext and convert it to DataFrame

plaintext_rdd = sc.textFile('hdfs://x.x.x.x/blah.csv')
dataframe = pycsv.csvToDataFrame(sqlCtx, plaintext_rdd)

When using spark.read.csv, I find that using the options escape='"' and multiLine=True provide the most consistent solution to the CSV standard, and in my experience works the best with CSV files exported from Google Sheets.

That is,

#set inferSchema=False to read everything as string
df = spark.read.csv("myData.csv", escape='"', multiLine=True,
     inferSchema=False, header=True)