This section contains information on the fie formats and compression schemes that are supported for input to and output of 

NOTE: To work with formats that are proprietary to a desktop application, such as Microsoft Excel, you do not need the supporting application installed on your desktop.

Filenames

NOTE: During import, the identifies file formats based on the extension of the filename. If no extension is provided, the assumes that the submitted file is a text file of some kind. Non-text file formats, such as Avro and Parquet, require filename extensions.


NOTE: Filenames that include special characters can cause problems during import or when publishing to a file-based datastore.

Forbidden characters in import filenames:


The following list of characters present issues in the listed area of the product. If you encounter issues, the following listings may provide some guidance on where the issue occurred.

Tip: You should avoid using any of these characters in your import filenames. This list may not be complete for all available running environments.

Native Input File Formats

 can read and import directly these file formats:

For more information on data is handled initially, see Initial Parsing Steps in the User Guide.

Converted file formats

Files of the following type are not read into the product in their native format. Instead, these file types are converted using the Conversion Service into a file format that is natively supported, stored in the base storage layer, and then ingested for use in the product.

NOTE: Compressed files that require conversion of the underlying file format are not supported for use in the product.

Converted file formats:


Notes on JSON:

There are two methods of ingesting JSON files for use in the product. 

Native Output File Formats

 can write to these file formats:

Compression Algorithms

When a file is imported, the  attempts to infer the compression algorithm in use based on the filename extension. For example, .gz files are assumed to be compressed with GZIP. 

NOTE: Import of a compressed file whose underlying format requires conversion through the Conversion Service is not supported.


NOTE: Importing a compressed file with a high compression ratio can overload the available memory for the application. In such cases, you can decompress the file before uploading. If decompression fails, you should contact your administrator about increasing the Java Heap Size memory.

NOTE: Publication of results in Snappy format may require additional configuration. See below.

NOTE: GZIP files on Hadoop are not split across multiple nodes. As a result, jobs can crash when processing it through a single Hadoop task. This is a known issue with GZIP on Hadoop.

Where possible, limit the size of your GZIP files to 100 MB of less, or use BZIP2 as an alternative compression method. As a workaround, you can try to run the job on the unzipped file. You may also disable profiling for the job. See Run Job Page in the User Guide.

Tip: If preferred, you can configure the to infer the compression scheme based on first few bytes of the file. For more information, see Miscellaneous Configuration.

Read Native File Formats

 GZIPBZIPSnappyNotes
CSV SupportedSupportedSupported
JSON v2Not supportedNot supportedNot supportedA converted file format. See above.
JSON v1SupportedSupportedSupportedNot a converted file format. See above.
Avro  Supported
Hive  Supported


Write Native File Formats

 GZIPBZIPSnappy
CSVSupportedSupportedSupported
JSONSupportedSupportedSupported
Avro  Supported; always on
Hive  Supported; always on


Snappy compression formats

supports the following variants of Snappy compression format:

File extension

Format name

Notes

.sz

Framing2 format

See: https://github.com/google/snappy/blob/master/framing_format.txt 

.snappy

Hadoop-snappy format

See: https://code.google.com/p/hadoop-snappy/

NOTE: Xerial's snappy-java format, which is also written with a .snappy file extension by default, is not supported.


Additional Configuration for File Format Support

Publication of some formats requires execute permissions

When job results are generated and published in the following formats, the includes a JAR, from which is extracted a binary executable into a temporary directory. From this directory, the binary is then executed to generate the results in the proper format. By default, this directory is set to /tmp on the .

In many environments, execute permissions are disabled on /tmp for security reasons. Use the steps below to specify the temporary directory where this binary can be moved and executed.

Steps:

  1. Login to the application as an administrator.
  2. From the menu, select User menu > Admin console > Admin settings.
  3. For each of the following file formats, locate the listed parameter, where the related binary code can be executed:

    File FormatParameterSetting to Add
    Snappy
    "data-service.jvmOptions"
    -Dorg.xerial.snappy.tempdir=<some executable directory>
    HyperSee previous.See previous.
  4. Save your changes and restart the platform.

  5. Run a job configured for direct publication of the modified file format.