Data extraction options

The data extraction facility improves performance dramatically for queries with a large result set. This facility currently consists of a set of database options, which are set using the SET OPTION command. Like other database options, the data extraction options can be set either as temporary or permanent. Ordinarily these options are set as temporary. The extract options are set for a connection.

The extract options allow the user to redirect the output of a SELECT statement from the standard interface to go directly to one or more disk files or named pipes. There are two advantages of using the extract options:

• A binary format is supported, which allows loading the output data into the same or a different IQ database.

• A SELECT statement with heavy output will run up to 4 times faster for ASCII output and up to 9 times faster for binary output.

There are 27 options that control the behavior of extract (listed with allowed values for the option, followed by the default value):

*The default values for the Temp_Extract_Sizen options are platform specific:

• AIX and HP-UX: 0 - 64GB

• Sun Solaris: 0 - 512GB

• Windows: 0 - 128GB

• Linux: 0 - 512GB

When large file systems, such as JFS2, support file size larger than the default value, set TEMP_EXTRACT_SIZEn to the value that the file system allows. For example, to support lTB set option:

TEMP_EXTRACT_SIZE1 = 1073741824 KB

The most important of these options is Temp_Extract_Name1. If Temp_Extract_Name1 is set to its default setting (the empty string), extraction is disabled and no output is redirected. To enable extraction, set Temp_Extract_Name1 to a possible pathname. Extract starts extracting into a file with that name. Be sure to choose the pathname to a file that is not otherwise in use. If the file does not already exist, the data extraction facility creates the file.

Temp_Extract_Name1 is also used to specify the name of the output file, when the Temp_Extract_Append option is set ON. Both the directory or folder containing the named file and the named file must have write permission set for the user name used to start IQ (for example, sybase). In append mode, the data extraction facility adds extracted rows to the end of the file and does not overwrite the data that is already in the file. If the file does not exist, the data extraction facility creates the file.

The options Temp_Extract_Name2 through Temp_Extract_Name8 are used in addition to Temp_Extract_Name1 to specify the names of multiple output files. These options must be used sequentially. For example, Temp_Extract_Name3 has no effect unless both the options Temp_Extract_Name1 and Temp_Extract_Name2 are already set.

The options Temp_Extract_Size1 through Temp_Extract_Size8 are used to specify the maximum size of the corresponding output files. Temp_Extract_Size1 specifies the maximum size of the output file specified by Temp_Extract_Name1, Temp_Extract_Size2 specifies the maximum size of the output file specified by Temp_Extract_Name2, and so on.

Note that the default for the data extraction size options is 0. IQ converts this default to the following values:

*Tape devices currently are not supported.

Temp_Extract_Append is not compatible with the Temp_Extract_Sizen options. If you try to restrict the size of the extract append output file, Sybase IQ reports an error.

If you are extracting to a single disk file or a single named pipe, leave the options Temp_Extract_Name2 through Temp_Extract_Name8 and Temp_Extract_Size1 through Temp_Extract_Size8 at their default values.

The Temp_Extract_Directory option controls whether a user is allowed to use the data extraction facility. If the Temp_Extract_Directory option is set to the string FORBIDDEN (case insensitive) for a user, then that user is not allowed to perform data extracts. An attempt by this user to use the data extraction facility results in an error. If Temp_Extract_Directory is set to FORBIDDEN for the PUBLIC group, then no one can run data extraction.

The Temp_Extract_Directory option provides increased security and helps control disk management by restricting the creation of large data extraction files to only the directories for which a user has write access. DBA authority is required to set this option.

There are three types of data extraction:

• binary

• binary/swap

• ASCII

A binary extraction produces a file that can be loaded via a LOAD TABLE statement with an overall "binary" format and with a per column "binary with null byte" format.

The binary/swap extraction is the same as the binary extraction, except it is designed to be loaded on another machine with opposite endianness.

The ASCII extraction produces a text file.

The two options Temp_Extract_Binary and Temp_Extract_Swap determine which of the three types of extraction is done:

The default extraction type is ASCII.

Note that if the data is unloaded using the extraction facility with the TEMP_EXTRACT_BINARY option set ON, then you must use the LOAD TABLE statement BINARY WITH NULL BYTE parameter for each column when you load the binary data.

In the case of an ASCII extraction, the default is to separate column values with commas, and end the row with a newline on UNIX platforms and with a carriage return/newline pair on Windows platforms. The strings are unquoted. If these defaults are not suitable, use the following options to change the delimiters:

• Temp_Extract_Column_Delimiter

• Temp_Extract_Row_Delimiter

• Temp_Extract_Quote

• Temp_Extract_Quotes

• Temp_Extract_Quotes_All

The delimiter must occupy from 1 to a maximum of 4 bytes and must be valid in the collation order you are using, if you are using a multibyte collation order. Be sure to choose delimiters that do not occur in any of the data output strings themselves.

Note that the default for the Temp_Extract_Row_Delimiter option is the empty string. IQ converts the empty string default for this option to the newline on UNIX platforms and to the carriage return/newline pair on Windows platforms.

The option Temp_Extract_Column_Delimiter controls the delimiter between columns. If this option is set to the empty string '' for ASCII extractions, then the extracted data is written in fixed-width ASCII with no column delimiter. Numeric and binary data types are right-justified on a field of n blanks, where n is the maximum number of bytes needed for any value of that type. Character data types are left-justified on a field of n blanks.

During ASCII extraction, the following options control the use of quotes:

The quote string specified in the Temp_Extract_Quote option has the same restrictions as delimiters. The default for this option is the empty string, which IQ converts to the single quote mark.

The Temp_Extract_Null_As_Zero and Temp_Extract_Null_As_Empty options controls the representation of null values for ASCII extractions. When the Temp_Extract_Null_As_Zero option is set to ON, a null value is represented as follows:

• '0' for arithmetic type

• '' (the empty string) for the CHAR and VARCHAR character types

• '' (the empty string) for dates

• '' (the empty string) for times

• '' (the empty string) for timestamps

When the Temp_Extract_Null_As_Empty option is set to ON, a null value is represented as '' (the empty string) for all data types.

Note that the quotes shown above are not present in the extract output file. When the Temp_Extract_Null_As_Zero and Temp_Extract_Null_As_Empty options are set to OFF, the string 'NULL' is used in all cases to represent a NULL value. OFF is the default value.

If Temp_Extract_Null_As_Zero is set to ON, the number of characters that an ASCII extract writes to a file for a CHAR or VARCHAR column equals the number of characters in the column, even if that number is less than four. In previous releases, Sybase IQ always returned at least four characters to accommodate the word NULL.

When the Query_Plan option is ON, a timestamped list of the extracted columns appears in the IQ message log.

The data extraction options must be used with care.

1. Save in a different file any old extract output you need to retain.

2. Remove any previously used extract output files.

3. Set the extraction options you require, making sure to set Temp_Extract_Name1 to the file path that is to receive the output.

   SET [ TEMPORARY ] OPTION option-name = option-value

4. Issue a SELECT statement to extract the data you want.

5. Reset Temp_Extract_Name1 to the empty string, or disconnect if set temporarily, when you are done with extractions.

SET TEMPORARY OPTION Temp_Extract_Name1 = 'daily_report.txt';
SET TEMPORARY OPTION Temp_Extract_Name2 = '';
SELECT ....;
SET TEMPORARY OPTION Temp_Extract_Name1 = '';

Note that Temp_Extract_Name2 is set to the empty string before the SELECT statement is executed, to restrict output to a single file.

Also note that Temp_Extract_Name1 is set to the empty string after the SELECT statement to disable extraction. If extraction is not disabled, then the next SELECT statement executed overwrites the file daily_report.txt.

SET TEMPORARY OPTION Temp_Extract_Append = ON;
SET TEMPORARY OPTION Temp_Extract_Name1 = 'hourly_report.txt';
SET TEMPORARY OPTION Temp_Extract_Name2 = '';
SELECT ....;
SELECT ....;
SELECT ....;
SET TEMPORARY OPTION Temp_Extract_Name1 = '';

All of the output from the three SELECT statements is written to the file hourly_report.txt. Temp_Extract_Name1 is set to the empty string after the last SELECT statement to disable extraction. If extraction is not disabled, then the output from the next SELECT statement executed is also added to the end of the file hourly_report.txt.

First set the filename options:

SET TEMPORARY OPTION Temp_Extract_Name1 = 'file1.out';
SET TEMPORARY OPTION Temp_Extract_Name2 = 'file2.out';
SET TEMPORARY OPTION Temp_Extract_Name3 = 'file3.out';
SET TEMPORARY OPTION Temp_Extract_Name4 = '';

Now limit the size of the files, for example to 1MB each, by setting the corresponding extract size options:

SET TEMPORARY OPTION Temp_Extract_Size1 = '1024';
SET TEMPORARY OPTION Temp_Extract_Size2 = '1024';
SET TEMPORARY OPTION Temp_Extract_Size3 = '1024';

The size options are in KB (1024 bytes).

With these settings, the extraction output is first written to file1.out. When the next row to be written to file1.out would cause the file size to exceed 1MB, the output is redirected to file2.out. When file2.out is full (writing another row to file2.out would cause the file size to exceed 1MB), the output is redirected to file3.out. An error is reported, if the size of file3.out exceeds 1MB before IQ extracts all rows.

The following restrictions and limitations apply to the data extraction facility:

• Extract works only with data stored in the IQ Store.

• Extract does not work on system tables or cross database joins.

• Extract does not work with queries that use user-defined functions or system functions, except for the system functions suser_id() and suser_name().

• A binary LOAD TABLE always trims blanks from VARCHAR data. If you have VARCHAR data with trailing blanks, they are not preserved on insert.

• Trailing zeros are padded onto VARBINARY data during the extract. For example, a field declared as varbinary(6), which contains the data 0x1234, is padded with zeros during extraction and is loaded after extraction as 0x123400.

• Binary format will change in a future release.

• If you need to reproduce floating point data exactly, use the binary option.

• Tape devices are not supported at this time.

• If you run dbisql (Interactive SQL Java) with the -q (quiet mode) option and the data extraction commands are in a command file, you must first set and make permanent the dbisql option “Show multiple result sets.” If this option is not set, the output file is not created.

  To set the “Show multiple result sets” option, click Tools > Options in the dbisql window, then check the box “Show multiple result sets” and click “Make permanent.”

Also note that when Temp_Extract_Name1 is set, you cannot perform these operations:

• LOAD, DELETE, INSERT or INSERT...LOCATION to a table that is the top table in a join

• SYNCHRONIZE JOIN INDEX (issued explicitly or executed as part of CREATE JOIN INDEX)

• INSERT...SELECT

Events do not support execution of statements that return result sets. The server log returns an error similar to the following:

Handler for event 'test_ev' caused SQLSTATE '09W03'
Result set not permitted in 'test_ev'

In order to execute a query through an event, create an event that calls a stored procedure and insert the stored procedure results into a temporary table. If extract is used, then the temporary table is always empty and requires little overhead.

For example:

CREATE PROCEDURE proc1()
BEGIN
	SET TEMPORARY OPTION temp_extract_name1 = 'foo.out';
	SELECT * FROM iq_table;
END;

CREATE EVENT "test_ev" ENABLE HANDLER
BEGIN
	SELECT * INTO #tmp FROM proc1();
END;

TRIGGER EVENT test_ev;

Any UNIX process issuing a read request on a named pipe will wait forever until the process writing data to the pipe either sends data or an EOF (end-of-file). Each time the reading process receives data, it issues another read. If the writing process stops sending data and fails to send an EOF, the read process will wait in the kernel and cannot be interrupted from IQ.

Make sure that any process writing data into a named pipe always finishes with an EOF. If an IQ connection becomes unresponsive while writing out to a named pipe, try dumping the data out of the pipe. For example, issue the following commands form another thread:

cat NamedPipeFile > /dev/null