The PXF HDFS Connector supports plain delimited and comma-separated value form text data. This section describes how to use PXF to access HDFS text data, including how to create, query, and insert data into an external table that references files in the HDFS data store.

PXF supports reading or writing text files compressed with the default, bzip2, and gzip codecs.

Prerequisites

Ensure that you have met the PXF Hadoop Prerequisites before you attempt to read data from or write data to HDFS.

Reading Text Data

Use the hdfs:text profile when you read plain text delimited, and hdfs:csv when reading .csv data where each row is a single record. The following syntax creates a Greenplum Database readable external table that references such a text file on HDFS: 

CREATE EXTERNAL TABLE <table_name> 
    ( <column_name> <data_type> [, ...] | LIKE <other_table> )
LOCATION ('pxf://<path-to-hdfs-file>?PROFILE=hdfs:text|csv[&SERVER=<server_name>][&IGNORE_MISSING_PATH=<boolean>][&SKIP_HEADER_COUNT=<numlines>]')
FORMAT '[TEXT|CSV]' (delimiter[=|<space>][E]'<delim_value>');

The specific keywords and values used in the Greenplum Database CREATE EXTERNAL TABLE command are described in the table below.

Keyword Value
<path‑to‑hdfs‑file> The path to the directory or file in the HDFS data store. When the <server_name> configuration includes a pxf.fs.basePath property setting, PXF considers <path‑to‑hdfs‑file> to be relative to the base path specified. Otherwise, PXF considers it to be an absolute path. <path‑to‑hdfs‑file> must not specify a relative path nor include the dollar sign ($) character.
PROFILE Use PROFILE hdfs:text when <path-to-hdfs-file> references plain text delimited data.
Use PROFILE hdfs:csv when <path-to-hdfs-file> references comma-separated value data.
SERVER=<server_name> The named server configuration that PXF uses to access the data. PXF uses the default server if not specified.
IGNORE_MISSING_PATH=<boolean> Specify the action to take when <path-to-hdfs-file> is missing or invalid. The default value is false, PXF returns an error in this situation. When the value is true, PXF ignores missing path errors and returns an empty fragment.
SKIP_HEADER_COUNT=<numlines> Specify the number of header lines that PXF should skip in the first split of each <hdfs-file> before reading the data. The default value is 0, do not skip any lines.
FORMAT Use FORMAT 'TEXT' when <path-to-hdfs-file> references plain text delimited data.
Use FORMAT 'CSV' when <path-to-hdfs-file> references comma-separated value data.
delimiter The delimiter character in the data. For FORMAT 'CSV', the default <delim_value> is a comma (,). Preface the <delim_value> with an E when the value is an escape sequence. Examples: (delimiter=E'\t'), (delimiter ':').

Note: PXF does not support the (HEADER) formatter option in the CREATE EXTERNAL TABLE command. If your text file includes header line(s), use SKIP_HEADER_COUNT to specify the number of lines that PXF should skip at the beginning of the first split of each file.

Example: Reading Text Data on HDFS

Perform the following procedure to create a sample text file, copy the file to HDFS, and use the hdfs:text and hdfs:csv profiles and the default PXF server to create two PXF external tables to query the data:

  1. Create an HDFS directory for PXF example data files. For example:

    $ hdfs dfs -mkdir -p /data/pxf_examples

  2. Create a delimited plain text data file named pxf_hdfs_simple.txt:

    $ echo 'Prague,Jan,101,4875.33
Rome,Mar,87,1557.39
Bangalore,May,317,8936.99
Beijing,Jul,411,11600.67' > /tmp/pxf_hdfs_simple.txt

    Note the use of the comma (,) to separate the four data fields.

  3. Add the data file to HDFS:

    $ hdfs dfs -put /tmp/pxf_hdfs_simple.txt /data/pxf_examples/

  4. Display the contents of the pxf_hdfs_simple.txt file stored in HDFS:

    $ hdfs dfs -cat /data/pxf_examples/pxf_hdfs_simple.txt

  5. Start the psql subsystem:

    $ psql -d postgres

  6. Use the PXF hdfs:text profile to create a Greenplum Database external table that references the pxf_hdfs_simple.txt file that you just created and added to HDFS:

    postgres=# CREATE EXTERNAL TABLE pxf_hdfs_textsimple(location text, month text, num_orders int, total_sales float8)
            LOCATION ('pxf://data/pxf_examples/pxf_hdfs_simple.txt?PROFILE=hdfs:text')
          FORMAT 'TEXT' (delimiter=E',');

  7. Query the external table:

    postgres=# SELECT * FROM pxf_hdfs_textsimple;          

       location    | month | num_orders | total_sales 
---------------+-------+------------+-------------
 Prague        | Jan   |        101 |     4875.33
 Rome          | Mar   |         87 |     1557.39
 Bangalore     | May   |        317 |     8936.99
 Beijing       | Jul   |        411 |    11600.67
(4 rows)

  8. Create a second external table that references pxf_hdfs_simple.txt, this time specifying the hdfs:csv PROFILE and the CSV FORMAT:

    postgres=# CREATE EXTERNAL TABLE pxf_hdfs_textsimple_csv(location text, month text, num_orders int, total_sales float8)
            LOCATION ('pxf://data/pxf_examples/pxf_hdfs_simple.txt?PROFILE=hdfs:csv')
          FORMAT 'CSV';
postgres=# SELECT * FROM pxf_hdfs_textsimple_csv;

    When you specify FORMAT 'CSV' for comma-separated value data, no delimiter formatter option is required because comma is the default delimiter value.

Reading Text Data with Quoted Linefeeds

Use the hdfs:text:multi profile to read plain text data with delimited single- or multi- line records that include embedded (quoted) linefeed characters. The following syntax creates a Greenplum Database readable external table that references such a text file on HDFS:

CREATE EXTERNAL TABLE <table_name>
    ( <column_name> <data_type> [, ...] | LIKE <other_table> )
LOCATION ('pxf://<path-to-hdfs-file>?PROFILE=hdfs:text:multi[&SERVER=<server_name>][&IGNORE_MISSING_PATH=<boolean>][&SKIP_HEADER_COUNT=<numlines>]')
FORMAT '[TEXT|CSV]' (delimiter[=|<space>][E]'<delim_value>');

The specific keywords and values used in the CREATE EXTERNAL TABLE command are described in the table below.

Keyword Value
<path‑to‑hdfs‑file> The path to the directory or file in the HDFS data store. When the <server_name> configuration includes a pxf.fs.basePath property setting, PXF considers <path‑to‑hdfs‑file> to be relative to the base path specified. Otherwise, PXF considers it to be an absolute path. <path‑to‑hdfs‑file> must not specify a relative path nor include the dollar sign ($) character.
PROFILE The PROFILE keyword must specify hdfs:text:multi.
SERVER=<server_name> The named server configuration that PXF uses to access the data. PXF uses the default server if not specified.
IGNORE_MISSING_PATH=<boolean> Specify the action to take when <path-to-hdfs-file> is missing or invalid. The default value is false, PXF returns an error in this situation. When the value is true, PXF ignores missing path errors and returns an empty fragment.
SKIP_HEADER_COUNT=<numlines> Specify the number of header lines that PXF should skip in the first split of each <hdfs-file> before reading the data. The default value is 0, do not skip any lines.
FORMAT Use FORMAT 'TEXT' when <path-to-hdfs-file> references plain text delimited data.
Use FORMAT 'CSV' when <path-to-hdfs-file> references comma-separated value data.
delimiter The delimiter character in the data. For FORMAT 'CSV', the default <delim_value> is a comma (,). Preface the <delim_value> with an E when the value is an escape sequence. Examples: (delimiter=E'\t'), (delimiter ':').

Note: PXF does not support the (HEADER) formatter option in the CREATE EXTERNAL TABLE command. If your text file includes header line(s), use SKIP_HEADER_COUNT to specify the number of lines that PXF should skip at the beginning of the first split of each file.

Example: Reading Multi-Line Text Data on HDFS

Perform the following steps to create a sample text file, copy the file to HDFS, and use the PXF hdfs:text:multi profile and the default PXF server to create a Greenplum Database readable external table to query the data:

  1. Create a second delimited plain text file:

    $ vi /tmp/pxf_hdfs_multi.txt

  2. Copy/paste the following data into pxf_hdfs_multi.txt:

    "4627 Star Rd.
San Francisco, CA  94107":Sept:2017
"113 Moon St.
San Diego, CA  92093":Jan:2018
"51 Belt Ct.
Denver, CO  90123":Dec:2016
"93114 Radial Rd.
Chicago, IL  60605":Jul:2017
"7301 Brookview Ave.
Columbus, OH  43213":Dec:2018

    Notice the use of the colon : to separate the three fields. Also notice the quotes around the first (address) field. This field includes an embedded line feed separating the street address from the city and state.

  3. Copy the text file to HDFS:

    $ hdfs dfs -put /tmp/pxf_hdfs_multi.txt /data/pxf_examples/

  4. Use the hdfs:text:multi profile to create an external table that references the pxf_hdfs_multi.txt HDFS file, making sure to identify the : (colon) as the field separator:

    postgres=# CREATE EXTERNAL TABLE pxf_hdfs_textmulti(address text, month text, year int)
            LOCATION ('pxf://data/pxf_examples/pxf_hdfs_multi.txt?PROFILE=hdfs:text:multi')
          FORMAT 'CSV' (delimiter ':');

    Notice the alternate syntax for specifying the delimiter.

  5. Query the pxf_hdfs_textmulti table:

    postgres=# SELECT * FROM pxf_hdfs_textmulti;

             address          | month | year 
--------------------------+-------+------
 4627 Star Rd.            | Sept  | 2017
 San Francisco, CA  94107           
 113 Moon St.             | Jan   | 2018
 San Diego, CA  92093               
 51 Belt Ct.              | Dec   | 2016
 Denver, CO  90123                  
 93114 Radial Rd.         | Jul   | 2017
 Chicago, IL  60605                 
 7301 Brookview Ave.      | Dec   | 2018
 Columbus, OH  43213                
(5 rows)

Writing Text Data to HDFS

The PXF HDFS connector profiles hdfs:text and hdfs:csv support writing single line plain text data to HDFS. When you create a writable external table with the PXF HDFS connector, you specify the name of a directory on HDFS. When you insert records into a writable external table, the block(s) of data that you insert are written to one or more files in the directory that you specified.

Note: External tables that you create with a writable profile can only be used for INSERT operations. If you want to query the data that you inserted, you must create a separate readable external table that references the HDFS directory.

Use the following syntax to create a Greenplum Database writable external table that references an HDFS directory: 

CREATE WRITABLE EXTERNAL TABLE <table_name> 
    ( <column_name> <data_type> [, ...] | LIKE <other_table> )
LOCATION ('pxf://<path-to-hdfs-dir>
    ?PROFILE=hdfs:text|csv[&SERVER=<server_name>][&<custom-option>=<value>[...]]')
FORMAT '[TEXT|CSV]' (delimiter[=|<space>][E]'<delim_value>');
[DISTRIBUTED BY (<column_name> [, ... ] ) | DISTRIBUTED RANDOMLY];

The specific keywords and values used in the CREATE EXTERNAL TABLE command are described in the table below.

Keyword Value
<path‑to‑hdfs‑dir> The path to the directory in the HDFS data store. When the <server_name> configuration includes a pxf.fs.basePath property setting, PXF considers <path‑to‑hdfs‑dir> to be relative to the base path specified. Otherwise, PXF considers it to be an absolute path. <path‑to‑hdfs‑dir> must not specify a relative path nor include the dollar sign ($) character.
PROFILE Use PROFILE hdfs:text to write plain, delimited text to <path-to-hdfs-file>.
Use PROFILE hdfs:csv to write comma-separated value text to <path-to-hdfs-dir>.
SERVER=<server_name> The named server configuration that PXF uses to access the data. PXF uses the default server if not specified.
<custom‑option> <custom-option>s are described below.
FORMAT Use FORMAT 'TEXT' to write plain, delimited text to <path-to-hdfs-dir>.
Use FORMAT 'CSV' to write comma-separated value text to <path-to-hdfs-dir>.
delimiter The delimiter character in the data. For FORMAT 'CSV', the default <delim_value> is a comma (,). Preface the <delim_value> with an E when the value is an escape sequence. Examples: (delimiter=E'\t'), (delimiter ':').
DISTRIBUTED BY If you want to load data from an existing Greenplum Database table into the writable external table, consider specifying the same distribution policy or <column_name> on both tables. Doing so will avoid extra motion of data between segments on the load operation.

Writable external tables that you create using the hdfs:text or the hdfs:csv profiles can optionally use record or block compression. You specify the compression codec via a custom option in the CREATE EXTERNAL TABLE LOCATION clause. The hdfs:text and hdfs:csv profiles support the following custom write option:

Option Value Description
COMPRESSION_CODEC The compression codec alias. Supported compression codecs for writing text data include: default, bzip2, gzip, and uncompressed. If this option is not provided, Greenplum Database performs no data compression.

Example: Writing Text Data to HDFS

This example utilizes the data schema introduced in Example: Reading Text Data on HDFS.

Column Name Data Type
location text
month text
number_of_orders int
total_sales float8

This example also optionally uses the Greenplum Database external table named pxf_hdfs_textsimple that you created in that exercise.

Procedure

Perform the following procedure to create Greenplum Database writable external tables utilizing the same data schema as described above, one of which will employ compression. You will use the PXF hdfs:text profile and the default PXF server to write data to the underlying HDFS directory. You will also create a separate, readable external table to read the data that you wrote to the HDFS directory.

  1. Create a Greenplum Database writable external table utilizing the data schema described above. Write to the HDFS directory /data/pxf_examples/pxfwritable_hdfs_textsimple1. Create the table specifying a comma (,) as the delimiter:

    postgres=# CREATE WRITABLE EXTERNAL TABLE pxf_hdfs_writabletbl_1(location text, month text, num_orders int, total_sales float8)
            LOCATION ('pxf://data/pxf_examples/pxfwritable_hdfs_textsimple1?PROFILE=hdfs:text')
          FORMAT 'TEXT' (delimiter=',');

    You specify the FORMAT subclause delimiter value as the single ascii comma character ,.

  2. Write a few individual records to the pxfwritable_hdfs_textsimple1 HDFS directory by invoking the SQL INSERT command on pxf_hdfs_writabletbl_1:

    postgres=# INSERT INTO pxf_hdfs_writabletbl_1 VALUES ( 'Frankfurt', 'Mar', 777, 3956.98 );
postgres=# INSERT INTO pxf_hdfs_writabletbl_1 VALUES ( 'Cleveland', 'Oct', 3812, 96645.37 );

  3. (Optional) Insert the data from the pxf_hdfs_textsimple table that you created in [Example: Reading Text Data on HDFS] (#profile_text_query) into pxf_hdfs_writabletbl_1:

    postgres=# INSERT INTO pxf_hdfs_writabletbl_1 SELECT * FROM pxf_hdfs_textsimple;

  4. In another terminal window, display the data that you just added to HDFS:

    $ hdfs dfs -cat /data/pxf_examples/pxfwritable_hdfs_textsimple1/*
Frankfurt,Mar,777,3956.98
Cleveland,Oct,3812,96645.37
Prague,Jan,101,4875.33
Rome,Mar,87,1557.39
Bangalore,May,317,8936.99
Beijing,Jul,411,11600.67

    Because you specified comma (,) as the delimiter when you created the writable external table, this character is the field separator used in each record of the HDFS data.

  5. Greenplum Database does not support directly querying a writable external table. To query the data that you just added to HDFS, you must create a readable external Greenplum Database table that references the HDFS directory:

    postgres=# CREATE EXTERNAL TABLE pxf_hdfs_textsimple_r1(location text, month text, num_orders int, total_sales float8)
            LOCATION ('pxf://data/pxf_examples/pxfwritable_hdfs_textsimple1?PROFILE=hdfs:text')
		    FORMAT 'CSV';

    You specify the 'CSV' FORMAT when you create the readable external table because you created the writable table with a comma (,) as the delimiter character, the default delimiter for 'CSV' FORMAT.

  6. Query the readable external table:

    postgres=# SELECT * FROM pxf_hdfs_textsimple_r1 ORDER BY total_sales;

     location  | month | num_orders | total_sales 
-----------+-------+------------+-------------
 Rome      | Mar   |         87 |     1557.39
 Frankfurt | Mar   |        777 |     3956.98
 Prague    | Jan   |        101 |     4875.33
 Bangalore | May   |        317 |     8936.99
 Beijing   | Jul   |        411 |    11600.67
 Cleveland | Oct   |       3812 |    96645.37
(6 rows)

    The pxf_hdfs_textsimple_r1 table includes the records you individually inserted, as well as the full contents of the pxf_hdfs_textsimple table if you performed the optional step.

  7. Create a second Greenplum Database writable external table, this time using Gzip compression and employing a colon : as the delimiter:

    postgres=# CREATE WRITABLE EXTERNAL TABLE pxf_hdfs_writabletbl_2 (location text, month text, num_orders int, total_sales float8)
            LOCATION ('pxf://data/pxf_examples/pxfwritable_hdfs_textsimple2?PROFILE=hdfs:text&COMPRESSION_CODEC=gzip')
          FORMAT 'TEXT' (delimiter=':');

  8. Write a few records to the pxfwritable_hdfs_textsimple2 HDFS directory by inserting directly into the pxf_hdfs_writabletbl_2 table:

    gpadmin=# INSERT INTO pxf_hdfs_writabletbl_2 VALUES ( 'Frankfurt', 'Mar', 777, 3956.98 );
gpadmin=# INSERT INTO pxf_hdfs_writabletbl_2 VALUES ( 'Cleveland', 'Oct', 3812, 96645.37 );

  9. In another terminal window, display the contents of the data that you added to HDFS; use the -text option to hdfs dfs to view the compressed data as text:

    $ hdfs dfs -text /data/pxf_examples/pxfwritable_hdfs_textsimple2/*
Frankfurt:Mar:777:3956.98
Cleveland:Oct:3812:96645.3

    Notice that the colon : is the field separator in this HDFS data.

    To query data from the newly-created HDFS directory named pxfwritable_hdfs_textsimple2, you can create a readable external Greenplum Database table as described above that references this HDFS directory and specifies FORMAT 'CSV' (delimiter=':').

About Setting the External Table Encoding

When the external file encoding differs from the database encoding, you must set the external table ENCODING to match that of the data file. For example, if the database encoding is UTF8 and the file encoding is LATIN1, create the external table as follows:

CREATE EXTERNAL TABLE pxf_csv_latin1(location text, month text, num_orders int, total_sales float8)
  LOCATION ('pxf://data/pxf_examples/pxf_hdfs_simple.txt?PROFILE=hdfs:csv')
FORMAT 'CSV' ENCODING 'LATIN1';

About Reading Data Containing Multi-Byte or Multi-Character Delimiters

You can use only a *:csv PXF profile to read data that contains a multi-byte delimiter or multiple delimiter characters. The syntax for creating a readable external table for such data follows:

CREATE EXTERNAL TABLE <table_name>
    ( <column_name> <data_type> [, ...] | LIKE <other_table> )
LOCATION ('pxf://<path-to-hdfs-file>?PROFILE=hdfs:csv[&SERVER=<server_name>][&IGNORE_MISSING_PATH=<boolean>][&SKIP_HEADER_COUNT=<numlines>][&NEWLINE=<bytecode>]')
FORMAT 'CUSTOM' (FORMATTER='pxfdelimited_import' <option>[=|<space>][E]'<value>');

Note the FORMAT line in the syntax block. While the syntax is similar to that of reading CSV, PXF requires a custom formatter to read data containing a multi-byte or multi-character delimiter. You must specify the 'CUSTOM' format and the pxfdelimited_import formatter. You must also specify a delimiter in the formatter options.

PXF recognizes the following formatter options when reading data that contains a multi-byte or multi-character delimiter:

Option Name Value Description Default Value
DELIMITER=<delim_string> The single-byte or multi-byte delimiter string that separates columns. The string may be up to 32 bytes in length, and may not contain quote or escape characters. Required None
QUOTE=<char> The single one-byte ASCII quotation character for all columns. None
ESCAPE=<char> The single one-byte ASCII character used to escape special characters (for example, the DELIM, QUOTE, or NEWLINE value, or the ESCAPE value itself). None, or the QUOTE value if that is set
NEWLINE=<bytecode> The end-of-line indicator that designates the end of a row. Valid values are LF (line feed), CR (carriage return), or CRLF (carriage return plus line feed. LF

The following sections provide further information about, and examples for, specifying the delimiter, quote, escape, and new line options.

Specifying the Delimiter

You must directly specify the delimiter or provide its byte representation. For example, given the following sample data that uses a ¤ currency symbol delimiter:

133¤Austin¤USA
321¤Boston¤USA
987¤Paris¤France

Create the external table as follows:

CREATE READABLE EXTERNAL TABLE mbyte_delim (id int, city text, country text)
  LOCATION ('pxf://multibyte_currency?PROFILE=hdfs:csv')
FORMAT 'CUSTOM' (FORMATTER='pxfdelimited_import', DELIMITER='¤');

About Specifying the Byte Representation of the Delimiter

You can directly specify the delimiter or provide its byte representation. If you choose to specify the byte representation of the delimiter:

  • You must specify the byte representation of the delimiter in E'<value>' format.
  • Because some characters have different byte representations in different encodings, you must specify the byte representation of the delimiter in the database encoding.

For example, if the database encoding is UTF8, the file encoding is LATIN1, and the delimiter is the ¤ currency symbol, you must specify the UTF8 byte representation for ¤, which is \xC2\xA4:

CREATE READABLE EXTERNAL TABLE byterep_delim (id int, city text, country text)
  LOCATION ('pxf://multibyte_example?PROFILE=hdfs:csv')
FORMAT 'CUSTOM' (FORMATTER='pxfdelimited_import', DELIMITER=E'\xC2\xA4') ENCODING 'LATIN1';

About Specifying Quote and Escape Characters

When PXF reads data that contains a multi-byte or multi-character delimiter, its behavior depends on the quote and escape character settings:

QUOTE Set? ESCAPE Set? PXF Behaviour
No1 No PXF reads the data as-is.
Yes2 Yes PXF reads the data between quote characters as-is and un-escapes only the quote and escape characters.
Yes2 No (ESCAPE 'OFF') PXF reads the data between quote characters as-is.
No1 Yes PXF reads the data as-is and un-escapes only the delimiter, newline, and escape itself.

1 All data columns must be un-quoted when you do not specify a quote character.

2 All data columns must quoted when you specify a quote character.

Note

PXF expects that there are no extraneous characters between the quote value and the delimiter value, nor between the quote value and the end-of-line value. Additionally, there must be no white space between delimiters and quotes.

About the NEWLINE Options

PXF requires that every line in the file be terminated with the same new line value.

By default, PXF uses the line feed character (LF) for the new line delimiter. When the new line delimiter for the external file is also a line feed, you need not specify the NEWLINE formatter option.

If the NEWLINE formatter option is provided and contains CR or CRLF, you must also specify the same NEWLINE option in the external table LOCATION URI. For example, if the new line delimiter is CRLF, create the external table as follows:

CREATE READABLE EXTERNAL TABLE mbyte_newline_crlf (id int, city text, country text)
  LOCATION ('pxf://multibyte_example_crlf?PROFILE=hdfs:csv&NEWLINE=CRLF')
FORMAT 'CUSTOM' (FORMATTER='pxfdelimited_import', DELIMITER='¤', NEWLINE='CRLF');

Examples

Delimiter with Quoted Data

Given the following sample data that uses the double-quote (") quote character and the delimiter ¤:

"133"¤"Austin"¤"USA"
"321"¤"Boston"¤"USA"
"987"¤"Paris"¤"France"

Create the external table as follows:

CREATE READABLE EXTERNAL TABLE mbyte_delim_quoted (id int, city text, country text)
  LOCATION ('pxf://multibyte_q?PROFILE=hdfs:csv')
FORMAT 'CUSTOM' (FORMATTER='pxfdelimited_import', DELIMITER='¤', QUOTE '"');

Delimiter with Quoted and Escaped Data

Given the following sample data that uses the quote character ", the escape character \, and the delimiter ¤:

"\"hello, my name is jane\" she said. let's escape something \"¤"123"

Create the external table as follows:

CREATE READABLE EXTERNAL TABLE mybte_delim_quoted_escaped (sentence text, num int)
  LOCATION ('pxf://multibyte_qe?PROFILE=hdfs:csv')
FORMAT 'CUSTOM' (FORMATTER='pxfdelimited_import', DELIMITER='¤', QUOTE '"', ESCAPE '\');

With this external table definition, PXF reads the sentence text field as:

SELECT sentence FROM mbyte_delim_quoted_escaped;

                          sentence 
-------------------------------------------------------------
 "hello, my name is jane" she said. let's escape something \
(1 row)
check-circle-line exclamation-circle-line close-line
Scroll to top icon