Provide schema while reading csv file as a dataframe in Scala Spark

I am trying to read a csv file into a dataframe. I know what the schema of my dataframe should be since I know my csv file. Also I am using spark csv package to read the file. I trying to specify the schema like below.

val pagecount = sqlContext.read.format("csv")
.option("delimiter"," ").option("quote","")
.option("schema","project: string ,article: string ,requests: integer ,bytes_served: long")
.load("dbfs:/databricks-datasets/wikipedia-datasets/data-001/pagecounts/sample/pagecounts-20151124-170000")

But when I check the schema of the data frame I created, it seems to have taken its own schema. Am I doing anything wrong ? how to make spark to pick up the schema I mentioned ?

> pagecount.printSchema
root
|-- _c0: string (nullable = true)
|-- _c1: string (nullable = true)
|-- _c2: string (nullable = true)
|-- _c3: string (nullable = true)
205431 次浏览
小开

Try the below code, you need not specify the schema. When you give inferSchema as true it should take it from your csv file.

val pagecount = sqlContext.read.format("csv")
.option("delimiter"," ").option("quote","")
.option("header", "true")
.option("inferSchema", "true")
.load("dbfs:/databricks-datasets/wikipedia-datasets/data-001/pagecounts/sample/pagecounts-20151124-170000")

If you want to manually specify the schema, you can do it as below:

import org.apache.spark.sql.types._


val customSchema = StructType(Array(
StructField("project", StringType, true),
StructField("article", StringType, true),
StructField("requests", IntegerType, true),
StructField("bytes_served", DoubleType, true))
)


val pagecount = sqlContext.read.format("csv")
.option("delimiter"," ").option("quote","")
.option("header", "true")
.schema(customSchema)
.load("dbfs:/databricks-datasets/wikipedia-datasets/data-001/pagecounts/sample/pagecounts-20151124-170000")
小开

Thanks to the answer by @Nulu, it works for pyspark with minimal tweaking

from pyspark.sql.types import LongType, StringType, StructField, StructType, BooleanType, ArrayType, IntegerType


customSchema = StructType(Array(
StructField("project", StringType, true),
StructField("article", StringType, true),
StructField("requests", IntegerType, true),
StructField("bytes_served", DoubleType, true)))


pagecount = sc.read.format("com.databricks.spark.csv")
.option("delimiter"," ")
.option("quote","")
.option("header", "false")
.schema(customSchema)
.load("dbfs:/databricks-datasets/wikipedia-datasets/data-001/pagecounts/sample/pagecounts-20151124-170000")
小开

Here's how you can work with a custom schema, a complete demo:

$> shell code,

echo "
Slingo, iOS
Slingo, Android
" > game.csv

Scala code:

import org.apache.spark.sql.types._


val customSchema = StructType(Array(
StructField("game_id", StringType, true),
StructField("os_id", StringType, true)
))


val csv_df = spark.read.format("csv").schema(customSchema).load("game.csv")
csv_df.show


csv_df.orderBy(asc("game_id"), desc("os_id")).show
csv_df.createOrReplaceTempView("game_view")
val sort_df = sql("select * from game_view order by game_id, os_id desc")
sort_df.show
小开

I'm using the solution provided by Arunakiran Nulu in my analysis (see the code). Despite it is able to assign the correct types to the columns, all the values returned are null. Previously, I've tried to the option .option("inferSchema", "true") and it returns the correct values in the dataframe (although different type).

val customSchema = StructType(Array(
StructField("numicu", StringType, true),
StructField("fecha_solicitud", TimestampType, true),
StructField("codtecnica", StringType, true),
StructField("tecnica", StringType, true),
StructField("finexploracion", TimestampType, true),
StructField("ultimavalidacioninforme", TimestampType, true),
StructField("validador", StringType, true)))


val df_explo = spark.read
.format("csv")
.option("header", "true")
.option("delimiter", "\t")
.option("timestampFormat", "yyyy/MM/dd HH:mm:ss")
.schema(customSchema)
.load(filename)

Result

root




|-- numicu: string (nullable = true)
|-- fecha_solicitud: timestamp (nullable = true)
|-- codtecnica: string (nullable = true)
|-- tecnica: string (nullable = true)
|-- finexploracion: timestamp (nullable = true)
|-- ultimavalidacioninforme: timestamp (nullable = true)
|-- validador: string (nullable = true)

and the table is:

|numicu|fecha_solicitud|codtecnica|tecnica|finexploracion|ultimavalidacioninforme|validador|
+------+---------------+----------+-------+--------------+-----------------------+---------+
|  null|           null|      null|   null|          null|                   null|     null|
|  null|           null|      null|   null|          null|                   null|     null|
|  null|           null|      null|   null|          null|                   null|     null|
|  null|           null|      null|   null|          null|                   null|     null|
小开

For those interested in doing this in Python here is a working version.

customSchema = StructType([
StructField("IDGC", StringType(), True),
StructField("SEARCHNAME", StringType(), True),
StructField("PRICE", DoubleType(), True)
])
productDF = spark.read.load('/home/ForTesting/testProduct.csv', format="csv", header="true", sep='|', schema=customSchema)


testProduct.csv
ID|SEARCHNAME|PRICE
6607|EFKTON75LIN|890.88
6612|EFKTON100HEN|55.66

Hope this helps.

小开

This is one of option where we can pass the column names to the dataframe while loading CSV.

import pandas
names = ['sepal-length', 'sepal-width', 'petal-length', 'petal-width', 'class']
dataset = pandas.read_csv("C:/Users/NS00606317/Downloads/Iris.csv", names=names, header=0)
print(dataset.head(10))

Output 

    sepal-length  sepal-width  petal-length  petal-width        class
1            5.1          3.5           1.4          0.2  Iris-setosa
2            4.9          3.0           1.4          0.2  Iris-setosa
3            4.7          3.2           1.3          0.2  Iris-setosa
4            4.6          3.1           1.5          0.2  Iris-setosa
5            5.0          3.6           1.4          0.2  Iris-setosa
6            5.4          3.9           1.7          0.4  Iris-setosa
7            4.6          3.4           1.4          0.3  Iris-setosa
8            5.0          3.4           1.5          0.2  Iris-setosa
9            4.4          2.9           1.4          0.2  Iris-setosa
10           4.9          3.1           1.5          0.1  Iris-setosa
小开 
// import Library
import java.io.StringReader ;


import au.com.bytecode.opencsv.CSVReader


//filename


var train_csv = "/Path/train.csv";


//read as text file


val train_rdd = sc.textFile(train_csv)


//use string reader to convert in proper format


var full_train_data  = train_rdd.map{line =>  var csvReader = new CSVReader(new StringReader(line)) ; csvReader.readNext();  }


//declares  types


type s = String


// declare case class for schema


case class trainSchema (Loan_ID :s ,Gender :s, Married :s, Dependents :s,Education :s,Self_Employed :s,ApplicantIncome :s,CoapplicantIncome :s,
LoanAmount :s,Loan_Amount_Term :s, Credit_History :s, Property_Area :s,Loan_Status :s)


//create DF RDD with custom schema


var full_train_data_with_schema = full_train_data.mapPartitionsWithIndex{(idx,itr)=> if (idx==0) itr.drop(1);
itr.toList.map(x=> trainSchema(x(0),x(1),x(2),x(3),x(4),x(5),x(6),x(7),x(8),x(9),x(10),x(11),x(12))).iterator }.toDF
小开

here my solution is:

import org.apache.spark.sql.types._
val spark = org.apache.spark.sql.SparkSession.builder.
master("local[*]").
appName("Spark CSV Reader").
getOrCreate()


val movie_rating_schema = StructType(Array(
StructField("UserID", IntegerType, true),
StructField("MovieID", IntegerType, true),
StructField("Rating", DoubleType, true),
StructField("Timestamp", TimestampType, true)))


val df_ratings: DataFrame = spark.read.format("csv").
option("header", "true").
option("mode", "DROPMALFORMED").
option("delimiter", ",").
//option("inferSchema", "true").
option("nullValue", "null").
schema(movie_rating_schema).
load(args(0)) //"file:///home/hadoop/spark-workspace/data/ml-20m/ratings.csv"


val movie_avg_scores = df_ratings.rdd.map(_.toString()).
map(line => {
// drop "[", "]" and then split the str
val fileds = line.substring(1, line.length() - 1).split(",")
//extract (movie id, average rating)
(fileds(1).toInt, fileds(2).toDouble)
}).
groupByKey().
map(data => {
val avg: Double = data._2.sum / data._2.size
(data._1, avg)
})
小开

schema definition as simple string

Just in case if some one is interested in schema definition as simple string with date and time stamp

data file creation from Terminal or shell

echo "
2019-07-02 22:11:11.000999, 01/01/2019, Suresh, abc
2019-01-02 22:11:11.000001, 01/01/2020, Aadi, xyz
" > data.csv

Defining the schema as String

    user_schema = 'timesta TIMESTAMP,date DATE,first_name STRING , last_name STRING'

reading the data

    df = spark.read.csv(path='data.csv', schema = user_schema, sep=',', dateFormat='MM/dd/yyyy',timestampFormat='yyyy-MM-dd HH:mm:ss.SSSSSS')


df.show(10, False)


+-----------------------+----------+----------+---------+
|timesta                |date      |first_name|last_name|
+-----------------------+----------+----------+---------+
|2019-07-02 22:11:11.999|2019-01-01| Suresh   | abc     |
|2019-01-02 22:11:11.001|2020-01-01| Aadi     | xyz     |
+-----------------------+----------+----------+---------+

Please note defining the schema explicitly instead of letting spark infer the schema also improves the spark read performance.

小开

In pyspark 2.4 onwards, you can simply use header parameter to set the correct header:

data = spark.read.csv('data.csv', header=True)

Similarly, if using scala you can use header parameter as well.

小开

You can also do like this by using sparkSession and implicit 

import sparkSession.implicits._
val pagecount:DataFrame = sparkSession.read
.option("delimiter"," ")
.option("quote","")
.option("inferSchema","true")
.csv("dbfs:/databricks-datasets/wikipedia-datasets/data-001/pagecounts/sample/pagecounts-20151124-170000")
.toDF("project","article","requests","bytes_served")
小开

The previous solutions have used the custom StructType.

With spark-sql 2.4.5 (scala version 2.12.10) it is now possible to specify the schema as a string using the schema function

import org.apache.spark.sql.SparkSession;

val sparkSession = SparkSession.builder()
.appName("sample-app")
.master("local[2]")
.getOrCreate();


val pageCount = sparkSession.read
.format("csv")
.option("delimiter","|")
.option("quote","")
.schema("project string ,article string ,requests integer ,bytes_served long")
.load("dbfs:/databricks-datasets/wikipedia-datasets/data-001/pagecounts/sample/pagecounts-20151124-170000")
小开

if your spark version is 3.0.1, you can use following Scala scripts:

val df = spark.read.format("csv").option("delimiter",",").option("header",true).load("file:///LOCAL_CSV_FILE_PATH")

but in this way, all datatypes will be set as String.


提交答案