Reading CSV files with options in Apache Spark - Time & Space Complexity

Time Complexity: Reading CSV files with options

O(n)

Understanding Time Complexity

When reading CSV files with options in Apache Spark, it is important to understand how the time taken grows as the file size increases.

We want to know how the reading process scales when the input data gets bigger.

Scenario Under Consideration

Analyze the time complexity of the following code snippet.

val df = spark.read
  .option("header", "true")
  .option("inferSchema", "true")
  .csv("/path/to/large_file.csv")

// Show first 5 rows
df.show(5)
    

This code reads a CSV file with a header and tries to guess the data types automatically.

Identify Repeating Operations

Identify the loops, recursion, array traversals that repeat.

• Primary operation: Reading each line of the CSV file and parsing it.
• How many times: Once for every row in the file (n times).

How Execution Grows With Input

As the number of rows increases, the time to read and parse grows roughly in direct proportion.

Input Size (n)	Approx. Operations
10	About 10 lines read and parsed
100	About 100 lines read and parsed
1000	About 1000 lines read and parsed

Pattern observation: The work grows linearly as the file gets bigger.

Final Time Complexity

Time Complexity: O(n)

This means the time to read the CSV grows directly with the number of rows in the file.

Common Mistake

[X] Wrong: "Reading a CSV file with options is constant time regardless of file size."

[OK] Correct: Each row must be read and parsed, so more rows mean more work and more time.

Interview Connect

Understanding how data reading scales helps you explain performance in real projects and shows you think about efficiency.

Self-Check

"What if we disable schema inference and provide a schema manually? How would the time complexity change?"

Practice

1. What does the header option do when reading a CSV file in Apache Spark?

easy

A. It tells Spark to treat the first line as column names.

B. It specifies the character used to separate columns.

C. It infers the data types of each column automatically.

D. It sets the file reading mode to permissive.

Solution

1. Step 1: Understand the header option

   The header option tells Spark whether the first line of the CSV file contains column names or data.

2. Step 2: Identify its effect

   If header is set to true, Spark uses the first line as column names; otherwise, it treats all lines as data.

3. Final Answer:

   It tells Spark to treat the first line as column names. -> Option A

4. Quick Check:

   header = first line as columns [OK]

Hint: Remember: header=true means first line is column names [OK]

Common Mistakes:

• Confusing header with separator option
• Thinking header infers data types
• Assuming header controls error handling

2. Which of the following is the correct way to read a CSV file with a semicolon separator and header in Spark?

easy

A. spark.read.csv(options={'sep': ';', 'header': 'true'}, 'file.csv')

B. spark.read.csv(sep=';', 'file.csv', header=True)

C. spark.read.option('sep', ';').option('header', 'true').csv('file.csv')

D. spark.read.csv(header=True, sep=';', 'file.csv')

Solution

1. Step 1: Recall Spark CSV reading syntax

   Spark uses option() method to set options before calling csv().

2. Step 2: Check each option usage

   spark.read.option('sep', ';').option('header', 'true').csv('file.csv') correctly chains option('sep', ';') and option('header', 'true') before reading the file.

3. Final Answer:

   spark.read.option('sep', ';').option('header', 'true').csv('file.csv') -> Option C

4. Quick Check:

   Use option() chaining for CSV options [OK]

Hint: Use .option() chaining to set CSV read options [OK]

Common Mistakes:

• Passing options directly inside csv() as kwargs
• Using Python booleans instead of strings for options
• Incorrect method chaining order

3. Given the code below, what will be the schema of the DataFrame?

df = spark.read.option('header', 'true').option('inferSchema', 'true').csv('data.csv')

Assuming data.csv has columns: id (integer), name (string), age (integer).

medium

A. All columns will be read as strings.

B. id and age as integers, name as string.

C. All columns will be inferred as integers.

D. Schema will be empty because no schema is provided.

Solution

1. Step 1: Understand inferSchema option

   When inferSchema is true, Spark tries to detect the data type of each column automatically.

2. Step 2: Apply to given columns

   Since id and age contain integers, they will be inferred as integer types; name remains string.

3. Final Answer:

   id and age as integers, name as string. -> Option B

4. Quick Check:

   inferSchema=true means detect types [OK]

Hint: inferSchema=true detects column data types automatically [OK]

Common Mistakes:

• Assuming all columns are strings by default
• Thinking inferSchema converts all to integers
• Believing schema is empty without explicit definition

4. What is wrong with the following code snippet for reading a CSV file with a header and comma separator?

df = spark.read.csv(header=True, sep=',', 'file.csv')

medium

A. The header option must be set to 'false' to read headers.

B. The separator should be a semicolon, not a comma.

C. The file path must be an absolute path.

D. The options should be set using the option() method, not as parameters in csv().

Solution

1. Step 1: Check how options are passed in Spark CSV reading

   Spark requires options like header and sep to be set using option() method before calling csv().

2. Step 2: Identify the error in the code

   The code incorrectly passes options as parameters inside csv(), which is not supported and causes an error.

3. Final Answer:

   The options should be set using the option() method, not as parameters in csv(). -> Option D

4. Quick Check:

   Use .option() for CSV options, not csv() params [OK]

Hint: Set CSV options with .option(), not inside csv() call [OK]

Common Mistakes:

• Passing options directly inside csv()
• Using Python booleans instead of strings for options
• Confusing separator character

5. You want to read a CSV file that may have some corrupted rows. Which option should you use to skip corrupted rows and still read the rest of the data?

hard

A. mode='dropmalformed'

B. mode='failfast'

C. mode='permissive'

D. mode='ignore'

Solution

1. Step 1: Understand CSV reading modes in Spark

   Spark supports modes like 'permissive', 'failfast', and 'dropmalformed' to handle corrupted rows.

2. Step 2: Identify mode to skip corrupted rows

   The 'dropmalformed' mode skips corrupted rows and reads the rest without error.

3. Final Answer:

   mode='dropmalformed' -> Option A

4. Quick Check:

   Use dropmalformed to skip bad rows [OK]

Hint: Use mode='dropmalformed' to skip bad rows safely [OK]

Common Mistakes:

• Using failfast which stops on error
• Assuming permissive skips corrupted rows
• Using ignore which is not a valid mode