HLP:SORT.HLP (2 pages)

DEC's SORT-MERGE program (about 2 to 4 times more efficient than CSORT).
This can be used as a stand-alone program or called from Fortran or Cobol.
To be used as a stand-alone basis, first type:
    R SORT
The program will respond with a "*".  It is then ready to receive
a command line.  After a command is entered and completed, a "*"
will again be displayed, then another command may be entered, or
"/EXIT" or <cntl>C or <cntl>Z may be entered to exit from SORT.


Shortest sort command format:
out.fil=in.fil/R:n/K:n:m

The /CORE switch is usually well worth using for moderate to large
files.  Savings of a factor of 2 in cost are not difficult to achieve.
Left to itself SORT will use large amounts of core without much gain
in speed.  Try /CORE:20K and try more or less core to really find the
most efficient values.  Unfortunately the most efficient values not
only depend on the size and nature of the file, but also on the
initial order so no general rules can be given.

General command format for SORT and MERGE:
OUT.FIL/SW,SW,...= IN.FIL/SW,SW,.../FUNCTION

Functions are:
/EXIT   takes no arguments
/HELP   takes no arguments
/MERGE  see below
/RUN (program) file-spec
/SORT   see below (default)

indirect command files may be used:  @filspec

SORT and MERGE Switches are:
/RECORD:<n>     n = maximum record size (does not imply /FIXED)

/KEY:<n>:<m>:<x>  the first two arguments are mandatory
  <n>           position of first KEY character
  <m>           length of the KEY
  <x>           can be ASCENDING, or DESCENDING
Key data type switches:
  /ALPHANUMERIC
  /COMP1
  /COMP3
  /COMPUTATIONAL
  /FORMAT:a<b>.<c> where a<b>.<c> is a FORTRAN format descriptor
     a  = D decimal floating point, double precision
          E decimal floating point, single precision
          F decimal fixed point
          G general
    <b> = KEY length
    <c> = number of decimal places
  /NUMERIC
  /SIGNED       key operational sign is to be used
  /UNSIGNED     key operational sign is to be ignored

/ALIGN          word align all ASCII output records
/ASCII          recording mode
/BINARY         recording mode
/EBCDIC         recording mode
/SIXBIT         recording mode

/BLOCKED:<n>    COBOL blocking factor
/FIXED          fixed length records
/FORTRAN        FORTRAN type data file
/RANDOM         must also use /FORTRAN when sorting Fortran random access files
/SEQUENTIAL     same as /VARIABLE
/VARIABLE       variable length records

/BUFFER-PAGES<n> number of pages for IO buffers
/CHECK          when merging check the sequence of the input record key
/COLLATE:<a>    <a> is one of the following
  ASCII         collating sequence
  EBCDIC        collating sequence
  FILE:file-spec
  LITERAL:/collating-sequence/  first char after colon is delimeter
  ADDRESS:<N> octal address
/CORE:<n>K  or  <n>P   specifies low segment core to be used in K (1024 words,
                       default) or in pages (512 words)
/ERROR-RETURN:<n> octal address
/FATAL-ERROR-CODE:<n> octal address
/LEAVES:<n>     <n> is the number of records to keep in main memory
/PHYSICAL       do not use logical names
/SUPPRESS-ERROR:<a> <a> is one of the following
  ALL
  FATAL
  INFORMATION
  NONE
  WARNING
/TEMPORARY-AREAS:directory-name,...

/ANSI-ASCII     alias STANDARD ASCII for mag-tape
/DENSITY:<a>    <a> is one of the following
  200
  556
  800
  1600
  SYSTEM-DEFAULT
/INDUSTRY       industry compatible (8 bit) mode for mag-tape
/LABEL:<a>      <a> is one of the following
  ANSI
  DEC
  IBM
  NONSTANDARD
  OMITTED
  STANDARD
/PARITY:<a>     <a> is one of the following
  EVEN
  ODD
/REWIND         before use
/UNLOAD         after use

Examples:  (All the switches below may be abbreviated to one character.)

TEST.OUT=TEST.IN/REC:70/KEY:21:10:DESC/CORE:20K
a descending sort (default is ascending)

TEST.OUT=TEST.IN/REC:70/KEY:21:10/FORMAT:F10.4/CORE:20K
sort in numeric order on a field that has a decimal point
(which actually may be anywhere in the field)

TEST.OUT=TEST.IN/REC:70/KEY:21:10/KEY:41:5/CORE:20K
a two key sort, key given first is primary

TEST.OUT=TESTA.IN,TESTB.IN/REC:70/KEY:21:10/CORE:20K/MERG
merge two similar, pre-sorted files

TEST.OUT=TESTA.IN,TESTB.IN/REC:70/KEY:21:10/CORE:20K
merge and sort two similar files.  That's right, without the "/MERG",
the output file will be merged and sorted.

For more information see DEC's SORT/MERGE User's Guide.  This is
a (shock) well written manual that even includes a beginner's section.

Multi-line sets can be sorted if they are fixed length.  When counting
characters, include the <CR> <LF> characters on all lines execpt the
last when giving the record length.  References to positions in the set
must also take into account the <CR> <LF> characters.

Note:  SORT-MERGE is smart enough to look at all the disks available
to you and use them for its temporary sorting files.  But it is not
smart enough to stop using a disk when its quota is being exceeded!
So--If you sort a large file and have a disk with a small quota, you
should DISMOUNT that small-quota-disk first.  The DISMOUNT would be
only a logical opperation--making the system ignore the presence of
that disk as far as your job is concerned.  Example: .DISMOUNT DSKB

[end SORT help file]
las
.KJOB