You can use the SYSCS_UTIL.COMPUTE_SPLIT_KEY system procedure to compute the split keys for a table or index prior to calling the   SYSCS_UTIL.SYSCS_SPLIT_TABLE_OR_INDEX_AT_POINTS procedure to pre-split the data that you’re importing into HFiles.

Splice Machine recommends using the SYSCS_UTIL.SYSCS_SPLIT_TABLE_OR_INDEX system procedure instead of this one unless you’re an expert user. The combination of using SYSCS_UTIL.COMPUTE_SPLIT_KEY with SYSCS_UTIL.SYSCS_SPLIT_TABLE_OR_INDEX_AT_POINTS is exactly equivalent to using SYSCS_UTIL.SYSCS_SPLIT_TABLE_OR_INDEX.

For more information about splitting your tables and indexes into HFiles, see the Bulk Importing Flat Files section of our Best Practices Guide.


        columnList | null,
        columnDelimiter | null,
        characterDelimiter | null,
        timestampFormat | null,
        dateFormat | null,
        timeFormat | null,
        badRecordDirectory | null,
        oneLineRecords | null,
        charset | null,


The parameter values that you pass into this procedure should match the values that you use when you subsequently call the  SYSCS_UTIL.BULK_IMPORT_HFILE procedure to perform the import.

This table includes a brief description of each parameter; additional information is available in the Ingestion Parameter Values topic of our Importing Data tutorial.

Parameter Description Example Value
schemaName The name of the schema of the table into which to import. SPLICE
tableName The name of the table into which to import. playerTeams

The names, in single quotes, of the columns to import. If this is null, all columns are imported.

The individual column names in the insertColumnList do not need to be double-quoted, even if they contain special characters. However, if you do double-quote any column name, you must double-quote all of the column names.


Either a single file or a directory. If this is a single file, that file is imported; if this is a directory, all of the files in that directory are imported. You can import compressed or uncompressed files.

On a cluster, the files to be imported MUST be on S3, HDFS (or MapR-FS). If you're using our Database Service product, files can only be imported from S3.

See the Configuring an S3 Bucket for Splice Machine Access topic for information about accessing data on S3.



oneLineRecords A Boolean value that specifies whether (true) each record in the import file is contained in one input line, or (false) if a record can span multiple lines. true
charset The character encoding of the import file. The default value is UTF-8. null
columnDelimiter The character used to separate columns, Specify null if using the comma (,) character as your delimiter. '|', '\t'
characterDelimiter The character used to delimit strings in the imported data. '"', ''''

The format of timestamps stored in the file. You can set this to null if there are no time columns in the file, or if the format of any timestamps in the file match the Java.sql.Timestamp default format, which is: "yyyy-MM-dd HH:mm:ss".

All of the timestamps in the file you are importing must use the same format.

'yyyy-MM-dd HH:mm:ss.SSZ'

dateFormat The format of datestamps stored in the file. You can set this to null if there are no date columns in the file, or if the format of any dates in the file match pattern: "yyyy-MM-dd". yyyy-MM-dd
timeFormat The format of time values stored in the file. You can set this to null if there are no time columns in the file, or if the format of any times in the file match pattern: "HH:mm:ss". HH:mm:ss
badRecordsAllowed The number of rejected (bad) records that are tolerated before the import fails. If this count of rejected records is reached, the import fails, and any successful record imports are rolled back. Specify 0 to indicate that no bad records are tolerated, and specify -1 to indicate that all bad records should be logged and allowed. 25

The directory in which bad record information is logged. Splice Machine logs information to the <import_file_name>.bad file in this directory; for example, bad records in an input file named foo.csv would be logged to a file named badRecordDirectory/foo.csv.bad.

On a cluster, this directory MUST be on S3, HDFS (or MapR-FS). If you're using our Database Service product, files can only be imported from S3.

bulkImportDirectory (outputDirectory)

For SYSCS_UTIL.BULK_IMPORT_HFILE, this is the name of the directory into which the generated HFiles are written prior to being imported into your database.

For the SYSCS_UTIL.COMPUTE_SPLIT_KEY procedure, where it is named outputDirectory, this parameter specifies the directory into which the split keys are written.


The skipSampling parameter is a Boolean value that specifies how you want the split keys used for the bulk HFile import to be computed. Set to false to have SYSCS_UTIL.BULK_IMPORT_HFILE automatically determine splits for you.

This parameter is only used with the SYSCS_UTIL.BULK_IMPORT_HFILE system procedure.



This procedure generates a split keys file in CSV format, which you pass into the SYSCS_UTIL.SYSCS_SPLIT_TABLE_OR_INDEX_AT_POINTS. That procedure pre-splits the data that you then import with   SYSCS_UTIL.BULK_IMPORT_HFILE.

The functionality of the [SYSCS_UTIL.COMPUTE_SPLIT_KEY] and SYSCS_UTIL.SYSCS_SPLIT_TABLE_OR_INDEX_AT_POINTS procedures has been combined into one simplified procedure:   SYSCS_UTIL.SYSCS_SPLIT_TABLE_OR_INDEX. We recommend using the simplified procedure, which performs exactly the same functions.

The Best Practices: Bulk Importing Flat Files section of our Importing Data Tutorial describes the different methods for using our bulk HFile import functionality.

See Also