OpenCSVSerDe for processing CSV - Amazon Athena
CSV SerDe (OpenCSVSerDe)
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

OpenCSVSerDe for processing CSV

When you create an Athena table for CSV data, determine the SerDe to use based on the types of values your data contains:

  • If your data contains values enclosed in double quotes ("), you can use the OpenCSV SerDe to deserialize the values in Athena. If your data does not contain values enclosed in double quotes ("), you can omit specifying any SerDe. In this case, Athena uses the default LazySimpleSerDe. For information, see LazySimpleSerDe for CSV, TSV, and custom-delimited files.

  • If your data has UNIX numeric TIMESTAMP values (for example, 1579059880000), use the OpenCSVSerDe. If your data uses the java.sql.Timestamp format, use the LazySimpleSerDe.

CSV SerDe (OpenCSVSerDe)

The OpenCSV SerDe has the following characteristics for string data:

  • Uses double quotes (") as the default quote character, and allows you to specify separator, quote, and escape characters, such as: 

    WITH SERDEPROPERTIES ("separatorChar" = ",", "quoteChar" = "`", "escapeChar" = "\\" )

  • Cannot escape \t or \n directly. To escape them, use "escapeChar" = "\\". See the example in this topic.

  • Does not support embedded line breaks in CSV files.

For data types other than STRING, the OpenCSVSerDe behaves as follows:

  • Recognizes BOOLEAN, BIGINT, INT, and DOUBLE data types.

  • Does not recognize empty or null values in columns defined as a numeric data type, leaving them as string. One workaround is to create the column with the null values as string and then use CAST to convert the field in a query to a numeric data type, supplying a default value of 0 for nulls. For more information, see When I query CSV data in Athena, I get the error HIVE_BAD_DATA: Error parsing field value in the Amazon Knowledge Center.

  • For columns specified with the timestamp data type in your CREATE TABLE statement, recognizes TIMESTAMP data if it is specified in the UNIX numeric format in milliseconds, such as 1579059880000.

    • The OpenCSVSerDe does not support TIMESTAMP in the JDBC-compliant java.sql.Timestamp format, such as "YYYY-MM-DD HH:MM:SS.fffffffff" (9 decimal place precision).

  • For columns specified with the DATE data type in your CREATE TABLE statement, recognizes values as dates if the values represent the number of days that elapsed since January 1, 1970. For example, the value 18276 in a column with the date data type renders as 2020-01-15 when queried. In this UNIX format, each day is considered to have 86,400 seconds.

  • To further convert columns to the desired type in a table, you can create a view over the table and use CAST to convert to the desired type.

Example: Using the TIMESTAMP type and DATE type specified in the UNIX numeric format.

Consider the following three columns of comma-separated data. The values in each column are enclosed in double quotes.

"unixvalue creationdate 18276 creationdatetime 1579059880000","18276","1579059880000"

The following statement creates a table in Athena from the specified Amazon S3 bucket location.

CREATE EXTERNAL TABLE IF NOT EXISTS testtimestamp1(
 `profile_id` string,
 `creationdate` date,
 `creationdatetime` timestamp
 )
 ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.OpenCSVSerde'
 LOCATION 's3://DOC-EXAMPLE-BUCKET'

Next, run the following query: 

SELECT * FROM testtimestamp1

The query returns the following result, showing the date and time data:

profile_id                                                        creationdate     creationdatetime
unixvalue creationdate 18276 creationdatetime 1579146280000       2020-01-15       2020-01-15 03:44:40.000
Example: Escaping \t or \n

Consider the following test data:

" \\t\\t\\n 123 \\t\\t\\n ",abc
" 456 ",xyz

The following statement creates a table in Athena, specifying that "escapeChar" = "\\". 

CREATE EXTERNAL TABLE test1 (
f1 string,
s2 string) 
ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.OpenCSVSerde' 
WITH SERDEPROPERTIES ("separatorChar" = ",", "escapeChar" = "\\") 
LOCATION 's3://DOC-EXAMPLE-BUCKET/dataset/test1/'

Next, run the following query: 

SELECT * FROM test1;

It returns this result, correctly escaping \t or \n:

f1            s2
\t\t\n 123 \t\t\n            abc
456                          xyz

SerDe name

CSV SerDe

Library name

To use this SerDe, specify its fully qualified class name after ROW FORMAT SERDE. Also specify the delimiters inside SERDEPROPERTIES, as follows:

...
ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.OpenCSVSerde'
WITH SERDEPROPERTIES (
  "separatorChar" = ",",
  "quoteChar"     = "`",
  "escapeChar"    = "\\"
)

Ignoring headers

To ignore headers in your data when you define a table, you can use the skip.header.line.count table property, as in the following example.

TBLPROPERTIES ("skip.header.line.count"="1")

For examples, see the CREATE TABLE statements in Querying Amazon VPC flow logs and Querying Amazon CloudFront logs.

Example

This example presumes data in CSV saved in s3://DOC-EXAMPLE-BUCKET/mycsv/ with the following contents:

"a1","a2","a3","a4"
"1","2","abc","def"
"a","a1","abc3","ab4"

Use a CREATE TABLE statement to create an Athena table based on the data. Reference the OpenCSVSerDe class after ROW FORMAT SERDE and specify the character separator, quote character, and escape character in WITH SERDEPROPERTIES, as in the following example.

CREATE EXTERNAL TABLE myopencsvtable (
   col1 string,
   col2 string,
   col3 string,
   col4 string
)
ROW FORMAT SERDE 'org.apache.hadoop.hive.serde2.OpenCSVSerde'
WITH SERDEPROPERTIES (
   'separatorChar' = ',',
   'quoteChar' = '"',
   'escapeChar' = '\\'
   )
STORED AS TEXTFILE
LOCATION 's3://DOC-EXAMPLE-BUCKET/mycsv/';

Query all values in the table:

SELECT * FROM myopencsvtable;

The query returns the following values:

col1     col2    col3    col4
-----------------------------
a1       a2      a3      a4
1        2       abc     def
a        a1      abc3    ab4