SORT Procedure
Concepts: SORT Procedure
Multi-threaded Sorting
The
SAS system option THREADS permits multi-threaded sorting, which is
new with SAS System 9. Multi-threaded sorting achieves a degree of
parallelism in the sorting operations. This parallelism is intended
to reduce the real time to completion for a given operation and therefore
limit the cost of additional CPU resources. For more information,
see Support for Parallel Processing in SAS Language Reference: Concepts.
The value of the SAS
system option CPUCOUNT= affects the performance of the multi-threaded
sort. CPUCOUNT= suggests how many system CPUs are available for use
by the multi-threaded procedures.
For more information,
see the THREADS System Option in SAS System Options: Reference and the CPUCOUNT= System Option in SAS System Options: Reference.
Sorting Orders for Character Variables
Default Collating Sequence
The order in which alphanumeric characters are sorted
is known as the collating sequence. This sort order is determined
by the session encoding.
By default, PROC SORT
uses either the EBCDIC or the ASCII collating sequence when it compares
character values, depending on the environment under which the procedure
is running.
For more information
about the various collating sequences and when they are used, seeCollating Sequence in SAS National Language Support (NLS): Reference Guide.
EBCDIC Order
ASCII Order
Specifying Sorting Orders for Character Variables
The options EBCDIC,
ASCII, NATIONAL, DANISH, SWEDISH, and REVERSE specify collating sequences
that are stored in the HOST catalog.
If you want to provide
your own collating sequences or change a collating sequence provided
for you, then use the TRANTAB procedure to create or modify translation
tables. When you create your own translation tables, they are stored
in your PROFILE catalog, and they override any translation tables
that have the same name in the HOST catalog. For complete details,
see TRANTAB Procedure in SAS National Language Support (NLS): Reference Guide.
Linguistic Collation
sorts data according to rules of language. For detailed information
about Linguistic Collation, see Collating Sequence in SAS National Language Support (NLS): Reference Guide.
Stored Sort Information
PROC
SORT records the BY variables, collating sequence, and character set
that it uses to sort the data set. This information is stored with
the data set to help avoid unnecessary sorts.
Before PROC SORT sorts
a data set, it checks the stored sort information. If you try to
sort a data set how it is currently sorted, then PROC SORT does not
perform the sort and writes a message to the log to that effect. To
override this behavior, use the FORCE option. If you try to sort a
data set how it is currently sorted and you specify an OUT= data set,
then PROC SORT simply makes a copy of the DATA= data set.
To override the sort
information that PROC SORT stores, use the _NULL_ value with the SORTEDBY=
data set option. Refer to the SORTEDBY= Data Set Option in SAS Data Set Options: Reference.
If you want to change
the sort information for an existing data set, then use the SORTEDBY=
data set option in the MODIFY statement in the DATASETS procedure.
For more information, see MODIFY Statement.
To access the sort information
that is stored with a data set, use the CONTENTS statement in PROC
DATASETS. For more information, see CONTENTS Statement.
The number of variables
by which you can sort a data set with PROC SORT is limited only by
available memory. The number of columns by which you can order the
rows of a result set using PROC SQL, is also limited only by available
memory. The sort indicator, whether stored in the metadata of a
Base data set or represented in memory, is limited to 127 variables.
For this reason, up to 127 variables can be stored in the sort indicator
or listed on the SORTEDBY= data set option. If you are sorting by
more than 127 variables, then only the first 127 are recorded in the
sort indicator. If you sort the data set again by the entire list
of BY variables, then the data set will not be recognized as being
sorted, because the additional variables (beyond 127) are not found
within the sort indicator. For a detailed explanation, refer toWhat Is a Sort Indicator? in SAS Language Reference: Concepts.
Presorted Input Data Sets
Specifying the PRESORTED option prevents SAS from sorting an already sorted data
set. Before sorting, SAS checks the sequence of observations within
the input data set to determine whether the observations are in order.
Use the PRESORTED option when you know or strongly suspect that a
data set is already in order according to the key variables specified
in the BY statement. The sequence of observations within the data
set is checked by reading the data set and comparing the BY variables
of each observation read to the BY variables of the preceding observation.
This process continues until either the entire data set has been read
or an out-of-sequence observation is detected.
If the entire data set
has been read and no out-of-sequence observations have been found,
then one of two actions is taken. If no output data set has been specified,
the sort order metadata of the input data set is updated to indicate
that the sequence has been verified. This verification notes that
the data set is validly sorted according to the specified BY variables.
Otherwise, if the observation sequence has been verified and an output
data set is specified, the observations from the input data set are
copied to the output data set, and the metadata for the output data
set indicates that the data is validly sorted according to the BY
variables.
If the NODUPKEY option has been specified, then the sequence checking determines
whether observations with duplicate keys are present in the data set.
Otherwise, the
input data set is deemed not to be sorted if the NODUPKEY option is
specified and observations with duplicate keys are detected.
If the metadata of the
input data set indicates that the data is already sorted according
to the key variables listed in the BY statement and the input data
set has been validated, then neither sequence checking nor sorting
will be performed.
See Sorted Data Sets in SAS Language Reference: Concepts and interactions with
the SORTVALIDATE System Option in SAS System Options: Reference.