key_buffer
OpenVMS usage:vector_word_unsigned
type: word (unsigned)
access: read only
mechanism: by reference
Array of words describing the keys on which you plan to merge.
The key_buffer argument is the address of an array containing the
key descriptions.
The first word of this array contains the number of keys
described (up to 255). Following the first word, each key is
described (in order of priority) in blocks of four words. The
four words specify the key's data type, order, offset, and
length, respectively.
The first word of the block specifies the key's data type. The
following data types are accepted:
DSC$K_DTYPE_Z Unspecified (uninfluenced by collating
sequence)
DSC$K_DTYPE_B Byte integer (signed)
DSC$K_DTYPE_BU Byte (unsigned)
DSC$K_DTYPE_W Word integer (signed)
DSC$K_DTYPE_WU Word (unsigned)
DSC$K_DTYPE_L Longword integer (signed)
DSC$K_DTYPE_LU Longword (unsigned)
DSC$K_DTYPE_Q Quadword integer (signed)
DSC$K_DTYPE_QU Quadword (unsigned)
DSC$K_DTYPE_O(1) Octaword integer (signed)
DSC$K_DTYPE_OU(1) Octaword (unsigned)
DSC$K_DTYPE_F Single-precision floating
DSC$K_DTYPE_D Double-precision floating
DSC$K_DTYPE_G G-format floating
DSC$K_DTYPE_H(1) H-format floating
DSC$K_DTYPE_FS(2) IEEE single-precision S floating
DSC$K_DTYPE_FT(2) IEEE double-precision T floating
DSC$K_DTYPE_T Text (may be influenced by collating
sequence)
DSC$K_DTYPE_NU Numeric string, unsigned
DSC$K_DTYPE_NL Numeric string, left separate sign
DSC$K_DTYPE_NLO Numeric string, left overpunched sign
DSC$K_DTYPE_NR Numeric string, right separate sign
DSC$K_DTYPE_NRO Numeric string, right overpunched sign
DSC$K_DTYPE_NZ(1) Numeric string, zoned sign
DSC$K_DTYPE_P Packed decimal string
Footnotes:
1. Data type is not currently supported by the high-performance
Sort/Merge utility.
2. Data type is Alpha specific.
The HP OpenVMS Programming Concepts Manual manual describes each
of these data types.
The second word of the block specifies the key order: 0 for
ascending order, 1 for descending order. The third word of the
block specifies the relative offset of the key in the record.
(Note that the first byte in the record is at position 0.) The
fourth word of the block specifies the key length in bytes (in
digits for packed decimal-DSC$K_DTYPE_P).
If you do not specify the key_buffer argument, you must pass
either a key comparison routine or use a specification file to
define the key.
lrl
OpenVMS usage:word_unsigned
type: word (unsigned)
access: read only
mechanism: by reference
Length of the longest record that will be released for merging.
The lrl (longest record length) argument is the address of a
word containing the length. If the input file is on a disk,
this argument is not required. It is required when you use the
record interface. For Vertical Format Control (VFC) records, this
length must include the length of the fixed-length portion of the
record.
options
OpenVMS usage:mask_longword
type: longword (unsigned)
access: read only
mechanism: by reference
Flags that identify merge options. The options argument is the
address of a longword bit mask whose settings determine the merge
options selected.
The following table lists and describes the bit mask values
available:
Flag Description
SOR$M_STABLE Keeps records with equal keys in the same order
as they appeared on input.
SOR$M_EBCDIC Orders ASCII character keys according to EBCDIC
collating sequence. No translation takes place.
SOR$M_MULTI Orders character keys according to the
multinational collating sequence, which collates
the international character set.
SOR$M_ Returns a status code instead of signaling
NOSIGNAL errors.
SOR$M_NODUPS Omits records with duplicate keys. You cannot
use this option if you specify your own equal-key
routine.
SOR$M_SEQ_ Requests an "out of order" error return if
CHECK an input file is not already in sequence. By
default, this check is not done. You must request
sequence checking if you specify an equal-key
routine.
All other bits in the longword are reserved and must be zero.
merge_order
OpenVMS usage:byte_unsigned
type: byte (unsigned)
access: read only
mechanism: by reference
Number of input streams to be merged. The merge_order argument is
the address of a byte containing the number of files (1 through
10) to be merged. (The high-performance Sort/Merge utility allows
you to specify 1 through 12 files.) When you use the record
interface on input, this argument is required.
user_compare
OpenVMS usage:procedure
type: procedure value
access: function call
mechanism: by reference
Routine that compares records to determine their merge order.
(This routine is not currently supported by the high-performance
Sort/Merge utility.) The user_compare argument is the address of
the procedure value for this user-written routine. If you do not
specify the key_buffer argument or if you define key information
in a specification file, this argument is required.
MERGE calls the comparison routine with five reference arguments-
ADRS1, ADRS2, LENG1, LENG2, CNTX-corresponding to the addresses
of the two records to be compared, the lengths of these two
records, and the context longword.
The comparison routine must return a 32-bit integer value:
o -1 if the first record collates before the second
o 0 if the records collate as equal
o 1 if the first record collates after the second
user_equal
OpenVMS usage:procedure
type: procedure value
access: function call
mechanism: by reference
Routine that resolves the merge order when records have duplicate
keys. (This routine is not currently supported by the high-
performance Sort/Merge utility.) The user_equal argument is the
address of the procedure value for this user-written routine. If
you specify SOR$M_STABLE or SOR$M_NODUPS in the options argument,
do not use this argument.
MERGE calls the duplicate key routine with five reference
arguments-ADRS1, ADRS2, LENG1, LENG2, CNTX-corresponding to the
addresses of the two records that compare equally, the lengths of
the two records that compare equally, and the context longword.
The routine must return one of the following 32-bit condition
codes:
Code Description
SOR$_ Delete the first record from the merge.
DELETE1
SOR$_ Delete the second record from the merge.
DELETE2
SOR$_ Delete both records from the merge.
DELBOTH
SS$_NORMAL Keep both records in the merge.
Any other failure value causes the error to be signaled or
returned. Any other success value causes an undefined result.
user_input
OpenVMS usage:procedure
type: procedure value
access: function call
mechanism: by reference
Routine that releases records to the merge operation. The user_
input argument is the address of the procedure value for this
user-written routine. SOR$BEGIN_MERGE and SOR$RETURN_REC call
this routine until all records have been passed.
This input routine must read (or construct) a record, place it in
a record buffer, store its length in an output argument, and then
return control to MERGE.
The input routine must accept the following four arguments:
o A descriptor of the buffer where the routine must place the
record
o A longword, passed by reference, containing the stream number
from which to input a record (the first file is 1, the second
2, and so on)
o A word, passed by reference, where the routine must return the
actual length of the record
o The context longword, passed by reference
The input routine must also return one of the following status
values:
o SS$_NORMAL or any other success status causes the merge
operation to continue.
o SS$_ENDOFFILE indicates that no more records are in the file.
The contents of the buffer are ignored.
o Any other error status terminates the merge operation and
passes the status value back to the caller of SOR$BEGIN_MERGE
or SOR$RETURN_REC.
context
OpenVMS usage:context
type: longword (unsigned)
access: modify
mechanism: by reference
Value that distinguishes between multiple, concurrent SORT/MERGE
operations. The context argument is the address of a longword
containing the context value. When your program makes its first
call to a SORT/MERGE routine for a particular sort or merge
operation, the context longword must equal zero. SORT/MERGE then
stores a value in the longword to identify the operation just
initiated. When you make subsequent routine calls for the same
operation, you must pass the context value that was supplied by
SORT/MERGE.