1 @node Data Manipulation
2 @chapter Data transformations
3 @cindex transformations
5 The @pspp{} procedures examined in this chapter manipulate data and
6 prepare the active dataset for later analyses. They do not produce output,
10 * AGGREGATE:: Summarize multiple cases into a single case.
11 * AUTORECODE:: Automatic recoding of variables.
12 * COMPUTE:: Assigning a variable a calculated value.
13 * COUNT:: Counting variables with particular values.
14 * FLIP:: Exchange variables with cases.
15 * IF:: Conditionally assigning a calculated value.
16 * RECODE:: Mapping values from one set to another.
17 * SORT CASES:: Sort the active dataset.
26 OUTFILE=@{*,'@var{file_name}',@var{file_handle}@} [MODE=@{REPLACE, ADDVARIABLES@}]
31 /@var{dest_var}['@var{label}']@dots{}=@var{agr_func}(@var{src_vars}, @var{args}@dots{})@dots{}
34 @cmd{AGGREGATE} summarizes groups of cases into single cases.
35 Cases are divided into groups that have the same values for one or more
36 variables called @dfn{break variables}. Several functions are available
37 for summarizing case contents.
39 The @subcmd{OUTFILE} subcommand is required and must appear first. Specify a
40 system file or portable file by file name or file
41 handle (@pxref{File Handles}), or a dataset by its name
43 The aggregated cases are written to this file. If @samp{*} is
44 specified, then the aggregated cases replace the active dataset's data.
45 Use of @subcmd{OUTFILE} to write a portable file is a @pspp{} extension.
47 If @subcmd{OUTFILE=*} is given, then the subcommand @subcmd{MODE} may also be
49 The mode subcommand has two possible values: @subcmd{ADDVARIABLES} or @subcmd{REPLACE}.
50 In @subcmd{REPLACE} mode, the entire active dataset is replaced by a new dataset
51 which contains just the break variables and the destination varibles.
52 In this mode, the new file will contain as many cases as there are
53 unique combinations of the break variables.
54 In @subcmd{ADDVARIABLES} mode, the destination variables will be appended to
55 the existing active dataset.
56 Cases which have identical combinations of values in their break
57 variables, will receive identical values for the destination variables.
58 The number of cases in the active dataset will remain unchanged.
59 Note that if @subcmd{ADDVARIABLES} is specified, then the data @emph{must} be
60 sorted on the break variables.
62 By default, the active dataset will be sorted based on the break variables
63 before aggregation takes place. If the active dataset is already sorted
64 or otherwise grouped in terms of the break variables, specify
65 @subcmd{PRESORTED} to save time.
66 @subcmd{PRESORTED} is assumed if @subcmd{MODE=ADDVARIABLES} is used.
68 Specify @subcmd{DOCUMENT} to copy the documents from the active dataset into the
69 aggregate file (@pxref{DOCUMENT}). Otherwise, the aggregate file will
70 not contain any documents, even if the aggregate file replaces the
73 Normally, only a single case (for @subcmd{SD} and @subcmd{SD}., two cases) need be
74 non-missing in each group for the aggregate variable to be
75 non-missing. Specifying @subcmd{/MISSING=COLUMNWISE} inverts this behavior, so
76 that the aggregate variable becomes missing if any aggregated value is
79 If @subcmd{PRESORTED}, @subcmd{DOCUMENT}, or @subcmd{MISSING} are specified, they must appear
80 between @subcmd{OUTFILE} and @subcmd{BREAK}.
82 At least one break variable must be specified on @subcmd{BREAK}, a
83 required subcommand. The values of these variables are used to divide
84 the active dataset into groups to be summarized. In addition, at least
85 one @var{dest_var} must be specified.
87 One or more sets of aggregation variables must be specified. Each set
88 comprises a list of aggregation variables, an equals sign (@samp{=}),
89 the name of an aggregation function (see the list below), and a list
90 of source variables in parentheses. Some aggregation functions expect
91 additional arguments following the source variable names.
93 Aggregation variables typically are created with no variable label,
94 value labels, or missing values. Their default print and write
95 formats depend on the aggregation function used, with details given in
96 the table below. A variable label for an aggregation variable may be
97 specified just after the variable's name in the aggregation variable
100 Each set must have exactly as many source variables as aggregation
101 variables. Each aggregation variable receives the results of applying
102 the specified aggregation function to the corresponding source
103 variable. The @subcmd{MEAN}, @subcmd{MEDIAN}, @subcmd{SD}, and @subcmd{SUM}
104 aggregation functions may only be
105 applied to numeric variables. All the rest may be applied to numeric
106 and string variables.
108 The available aggregation functions are as follows:
111 @item @subcmd{FGT(@var{var_name}, @var{value})}
112 Fraction of values greater than the specified constant. The default
115 @item @subcmd{FIN(@var{var_name}, @var{low}, @var{high})}
116 Fraction of values within the specified inclusive range of constants.
117 The default format is F5.3.
119 @item @subcmd{FLT(@var{var_name}, @var{value})}
120 Fraction of values less than the specified constant. The default
123 @item @subcmd{FIRST(@var{var_name})}
124 First non-missing value in break group. The aggregation variable
125 receives the complete dictionary information from the source variable.
126 The sort performed by @cmd{AGGREGATE} (and by @cmd{SORT CASES}) is stable, so that
127 the first case with particular values for the break variables before
128 sorting will also be the first case in that break group after sorting.
130 @item @subcmd{FOUT(@var{var_name}, @var{low}, @var{high})}
131 Fraction of values strictly outside the specified range of constants.
132 The default format is F5.3.
134 @item @subcmd{LAST(@var{var_name})}
135 Last non-missing value in break group. The aggregation variable
136 receives the complete dictionary information from the source variable.
137 The sort performed by @cmd{AGGREGATE} (and by @cmd{SORT CASES}) is stable, so that
138 the last case with particular values for the break variables before
139 sorting will also be the last case in that break group after sorting.
141 @item @subcmd{MAX(@var{var_name})}
142 Maximum value. The aggregation variable receives the complete
143 dictionary information from the source variable.
145 @item @subcmd{MEAN(@var{var_name})}
146 Arithmetic mean. Limited to numeric values. The default format is
149 @item @subcmd{MEDIAN(@var{var_name})}
150 The median value. Limited to numeric values. The default format is F8.2.
152 @item @subcmd{MIN(@var{var_name})}
153 Minimum value. The aggregation variable receives the complete
154 dictionary information from the source variable.
156 @item @subcmd{N(@var{var_name})}
157 Number of non-missing values. The default format is F7.0 if weighting
158 is not enabled, F8.2 if it is (@pxref{WEIGHT}).
161 Number of cases aggregated to form this group. The default format is
162 F7.0 if weighting is not enabled, F8.2 if it is (@pxref{WEIGHT}).
164 @item @subcmd{NMISS(@var{var_name})}
165 Number of missing values. The default format is F7.0 if weighting is
166 not enabled, F8.2 if it is (@pxref{WEIGHT}).
168 @item @subcmd{NU(@var{var_name})}
169 Number of non-missing values. Each case is considered to have a weight
170 of 1, regardless of the current weighting variable (@pxref{WEIGHT}).
171 The default format is F7.0.
174 Number of cases aggregated to form this group. Each case is considered
175 to have a weight of 1, regardless of the current weighting variable.
176 The default format is F7.0.
178 @item @subcmd{NUMISS(@var{var_name})}
179 Number of missing values. Each case is considered to have a weight of
180 1, regardless of the current weighting variable. The default format is F7.0.
182 @item @subcmd{PGT(@var{var_name}, @var{value})}
183 Percentage between 0 and 100 of values greater than the specified
184 constant. The default format is F5.1.
186 @item @subcmd{PIN(@var{var_name}, @var{low}, @var{high})}
187 Percentage of values within the specified inclusive range of
188 constants. The default format is F5.1.
190 @item @subcmd{PLT(@var{var_name}, @var{value})}
191 Percentage of values less than the specified constant. The default
194 @item @subcmd{POUT(@var{var_name}, @var{low}, @var{high})}
195 Percentage of values strictly outside the specified range of
196 constants. The default format is F5.1.
198 @item @subcmd{SD(@var{var_name})}
199 Standard deviation of the mean. Limited to numeric values. The
200 default format is F8.2.
202 @item @subcmd{SUM(@var{var_name})}
203 Sum. Limited to numeric values. The default format is F8.2.
206 Aggregation functions compare string values in terms of internal
208 On most modern computers, this is @acronym{ASCII} or a superset thereof.
210 The aggregation functions listed above exclude all user-missing values
211 from calculations. To include user-missing values, insert a period
212 (@samp{.}) at the end of the function name. (e.g.@: @samp{SUM.}).
213 (Be aware that specifying such a function as the last token on a line
214 will cause the period to be interpreted as the end of the command.)
216 @cmd{AGGREGATE} both ignores and cancels the current @cmd{SPLIT FILE}
217 settings (@pxref{SPLIT FILE}).
224 AUTORECODE VARIABLES=@var{src_vars} INTO @var{dest_vars}
228 [ /BLANK = @{VALID, MISSING@} ]
231 The @cmd{AUTORECODE} procedure considers the @var{n} values that a variable
232 takes on and maps them onto values 1@dots{}@var{n} on a new numeric
235 Subcommand @subcmd{VARIABLES} is the only required subcommand and must come
236 first. Specify @subcmd{VARIABLES}, an equals sign (@samp{=}), a list of source
237 variables, @subcmd{INTO}, and a list of target variables. There must the same
238 number of source and target variables. The target variables must not
241 By default, increasing values of a source variable (for a string, this
242 is based on character code comparisons) are recoded to increasing values
243 of its target variable. To cause increasing values of a source variable
244 to be recoded to decreasing values of its target variable (@var{n} down
245 to 1), specify @subcmd{DESCENDING}.
247 @subcmd{PRINT} is currently ignored.
249 The @subcmd{GROUP} subcommand is relevant only if more than one variable is to be
250 recoded. It causes a single mapping between source and target values to
251 be used, instead of one map per variable.
253 If @subcmd{/BLANK=MISSING} is given, then string variables which contain only
254 whitespace are recoded as SYSMIS. If @subcmd{/BLANK=VALID} is given then they
255 will be allocated a value like any other. @subcmd{/BLANK} is not relevant
256 to numeric values. @subcmd{/BLANK=VALID} is the default.
258 @cmd{AUTORECODE} is a procedure. It causes the data to be read.
265 COMPUTE @var{variable} = @var{expression}.
269 COMPUTE vector(@var{index}) = @var{expression}.
272 @cmd{COMPUTE} assigns the value of an expression to a target
273 variable. For each case, the expression is evaluated and its value
274 assigned to the target variable. Numeric and string
275 variables may be assigned. When a string expression's width differs
276 from the target variable's width, the string result of the expression
277 is truncated or padded with spaces on the right as necessary. The
278 expression and variable types must match.
280 For numeric variables only, the target variable need not already
281 exist. Numeric variables created by @cmd{COMPUTE} are assigned an
282 @code{F8.2} output format. String variables must be declared before
283 they can be used as targets for @cmd{COMPUTE}.
285 The target variable may be specified as an element of a vector
286 (@pxref{VECTOR}). In this case, an expression @var{index} must be
287 specified in parentheses following the vector name. The expression @var{index}
288 must evaluate to a numeric value that, after rounding down
289 to the nearest integer, is a valid index for the named vector.
291 Using @cmd{COMPUTE} to assign to a variable specified on @cmd{LEAVE}
292 (@pxref{LEAVE}) resets the variable's left state. Therefore,
293 @code{LEAVE} should be specified following @cmd{COMPUTE}, not before.
295 @cmd{COMPUTE} is a transformation. It does not cause the active dataset to be
298 When @cmd{COMPUTE} is specified following @cmd{TEMPORARY}
299 (@pxref{TEMPORARY}), the @cmd{LAG} function may not be used
307 COUNT @var{var_name} = @var{var}@dots{} (@var{value}@dots{}).
309 Each @var{value} takes one of the following forms:
312 @var{num1} THRU @var{num2}
315 where @var{num1} is a numeric expression or the words @subcmd{LO} or @subcmd{LOWEST}
316 and @var{num2} is a numeric expression or @subcmd{HI} or @subcmd{HIGHEST}.
319 @cmd{COUNT} creates or replaces a numeric @dfn{target} variable that
320 counts the occurrence of a @dfn{criterion} value or set of values over
321 one or more @dfn{test} variables for each case.
323 The target variable values are always nonnegative integers. They are
324 never missing. The target variable is assigned an F8.2 output format.
325 @xref{Input and Output Formats}. Any variables, including
326 string variables, may be test variables.
328 User-missing values of test variables are treated just like any other
329 values. They are @strong{not} treated as system-missing values.
330 User-missing values that are criterion values or inside ranges of
331 criterion values are counted as any other values. However (for numeric
332 variables), keyword @subcmd{MISSING} may be used to refer to all system-
333 and user-missing values.
335 @cmd{COUNT} target variables are assigned values in the order
336 specified. In the command @subcmd{COUNT @var{A}=@var{A} @var{B}(1) /@var{B}=@var{A} @var{B}(2).}, the
337 following actions occur:
341 The number of occurrences of 1 between @var{A} and @var{B} is counted.
344 @var{A} is assigned this value.
347 The number of occurrences of 1 between @var{B} and the @strong{new}
348 value of @var{A} is counted.
351 @var{B} is assigned this value.
354 Despite this ordering, all @cmd{COUNT} criterion variables must exist
355 before the procedure is executed---they may not be created as target
356 variables earlier in the command! Break such a command into two
359 The examples below may help to clarify.
363 Assuming @code{Q0}, @code{Q2}, @dots{}, @code{Q9} are numeric variables,
364 the following commands:
368 Count the number of times the value 1 occurs through these variables
369 for each case and assigns the count to variable @code{QCOUNT}.
372 Print out the total number of times the value 1 occurs throughout
373 @emph{all} cases using @cmd{DESCRIPTIVES}. @xref{DESCRIPTIVES}, for
378 COUNT QCOUNT=Q0 TO Q9(1).
379 DESCRIPTIVES QCOUNT /STATISTICS=SUM.
383 Given these same variables, the following commands:
387 Count the number of valid values of these variables for each case and
388 assigns the count to variable @code{QVALID}.
391 Multiplies each value of @code{QVALID} by 10 to obtain a percentage of
392 valid values, using @cmd{COMPUTE}. @xref{COMPUTE}, for details.
395 Print out the percentage of valid values across all cases, using
396 @cmd{DESCRIPTIVES}. @xref{DESCRIPTIVES}, for details.
400 COUNT QVALID=Q0 TO Q9 (LO THRU HI).
401 COMPUTE QVALID=QVALID*10.
402 DESCRIPTIVES QVALID /STATISTICS=MEAN.
411 FLIP /VARIABLES=@var{var_list} /NEWNAMES=@var{var_name}.
414 @cmd{FLIP} transposes rows and columns in the active dataset. It
415 causes cases to be swapped with variables, and vice versa.
417 All variables in the transposed active dataset are numeric. String
418 variables take on the system-missing value in the transposed file.
420 @subcmd{N} subcommands are required. If specified, the @subcmd{VARIABLES} subcommand
421 selects variables to be transformed into cases, and variables not
422 specified are discarded. If the @subcmd{VARIABLES} subcommand is omitted, all
423 variables are selected for transposition.
425 The variables specified by @subcmd{NEWNAMES}, which must be a
427 used to give names to the variables created by @cmd{FLIP}. Only the
428 first 8 characters of the variable are used. If
429 @subcmd{NEWNAMES} is not
430 specified then the default is a variable named CASE_LBL, if it exists.
431 If it does not then the variables created by @cmd{FLIP} are named VAR000
432 through VAR999, then VAR1000, VAR1001, and so on.
434 When a @subcmd{NEWNAMES} variable is available, the names must be canonicalized
435 before becoming variable names. Invalid characters are replaced by
436 letter @samp{V} in the first position, or by @samp{_} in subsequent
437 positions. If the name thus generated is not unique, then numeric
438 extensions are added, starting with 1, until a unique name is found or
439 there are no remaining possibilities. If the latter occurs then the
440 @cmd{FLIP} operation aborts.
442 The resultant dictionary contains a CASE_LBL variable, a string
443 variable of width 8, which stores the names of the variables in the
444 dictionary before the transposition. Variables names longer than 8
445 characters are truncated. If the active dataset is subsequently
446 transposed using @cmd{FLIP}, this variable can be used to recreate the
447 original variable names.
449 @cmd{FLIP} honors @cmd{N OF CASES} (@pxref{N OF CASES}). It ignores
450 @cmd{TEMPORARY} (@pxref{TEMPORARY}), so that ``temporary''
451 transformations become permanent.
458 IF @var{condition} @var{variable}=@var{expression}.
462 IF @var{condition} vector(@var{index})=@var{expression}.
465 The @cmd{IF} transformation conditionally assigns the value of a target
466 expression to a target variable, based on the truth of a test
469 Specify a boolean-valued expression (@pxref{Expressions}) to be tested
470 following the @cmd{IF} keyword. This expression is evaluated for each case.
471 If the value is true, then the value of the expression is computed and
472 assigned to the specified variable. If the value is false or missing,
473 nothing is done. Numeric and string variables may be
474 assigned. When a string expression's width differs from the target
475 variable's width, the string result of the expression is truncated or
476 padded with spaces on the right as necessary. The expression and
477 variable types must match.
479 The target variable may be specified as an element of a vector
480 (@pxref{VECTOR}). In this case, a vector index expression must be
481 specified in parentheses following the vector name. The index
482 expression must evaluate to a numeric value that, after rounding down
483 to the nearest integer, is a valid index for the named vector.
485 Using @cmd{IF} to assign to a variable specified on @cmd{LEAVE}
486 (@pxref{LEAVE}) resets the variable's left state. Therefore,
487 @code{LEAVE} should be specified following @cmd{IF}, not before.
489 When @cmd{IF} is specified following @cmd{TEMPORARY}
490 (@pxref{TEMPORARY}), the @cmd{LAG} function may not be used
497 The @cmd{RECODE} command is used to transform existing values into other,
498 user specified values.
502 RECODE @var{src_vars}
503 (@var{src_value} @var{src_value} @dots{} = @var{dest_value})
504 (@var{src_value} @var{src_value} @dots{} = @var{dest_value})
505 (@var{src_value} @var{src_value} @dots{} = @var{dest_value}) @dots{}
506 [INTO @var{dest_vars}].
509 Following the @cmd{RECODE} keyword itself comes @var{src_vars} which is a list
510 of variables whose values are to be transformed.
511 These variables may be string variables or they may be numeric.
512 However the list must be homogeneous; you may not mix string variables and
513 numeric variables in the same recoding.
515 After the list of source variables, there should be one or more @dfn{mappings}.
516 Each mapping is enclosed in parentheses, and contains the source values and
517 a destination value separated by a single @samp{=}.
518 The source values are used to specify the values in the dataset which
519 need to change, and the destination value specifies the new value
520 to which they should be changed.
521 Each @var{src_value} may take one of the following forms:
524 If the source variables are numeric then @var{src_value} may be a literal
527 If the source variables are string variables then @var{src_value} may be a
528 literal string (like all strings, enclosed in single or double quotes).
529 @item @var{num1} THRU @var{num2}
530 This form is valid only when the source variables are numeric.
531 It specifies all values in the range between @var{num1} and @var{num2},
532 including both endpoints of the range. By convention, @var{num1}
533 should be less than @var{num2}.
534 Open-ended ranges may be specified using @samp{LO} or @samp{LOWEST}
536 or @samp{HI} or @samp{HIGHEST} for @var{num2}.
538 The literal keyword @samp{MISSING} matches both system missing and user
540 It is valid for both numeric and string variables.
542 The literal keyword @samp{SYSMIS} matches system missing
544 It is valid for both numeric variables only.
546 The @samp{ELSE} keyword may be used to match any values which are
547 not matched by any other @var{src_value} appearing in the command.
548 If this keyword appears, it should be used in the last mapping of the
552 After the source variables comes an @samp{=} and then the @var{dest_value}.
553 The @var{dest_value} may take any of the following forms:
556 A literal numeric value to which the source values should be changed.
557 This implies the destination variable must be numeric.
559 A literal string value (enclosed in quotation marks) to which the source
560 values should be changed.
561 This implies the destination variable must be a string variable.
563 The keyword @samp{SYSMIS} changes the value to the system missing value.
564 This implies the destination variable must be numeric.
566 The special keyword @samp{COPY} means that the source value should not be
568 copied directly to the destination value.
569 This is meaningful only if @samp{INTO @var{dest_vars}} is specified.
572 Mappings are considered from left to right.
573 Therefore, if a value is matched by a @var{src_value} from more than
574 one mapping, the first (leftmost) mapping which matches will be considered.
575 Any subsequent matches will be ignored.
577 The clause @samp{INTO @var{dest_vars}} is optional.
578 The behaviour of the command is slightly different depending on whether it
581 If @samp{INTO @var{dest_vars}} does not appear, then values will be recoded
583 This means that the recoded values are written back to the
584 source variables from whence the original values came.
585 In this case, the @var{dest_value} for every mapping must imply a value which
586 has the same type as the @var{src_value}.
587 For example, if the source value is a string value, it is not permissible for
588 @var{dest_value} to be @samp{SYSMIS} or another forms which implies a numeric
590 It is also not permissible for @var{dest_value} to be longer than the width
591 of the source variable.
593 The following example two numeric variables @var{x} and @var{y} are recoded
595 Zero is recoded to 99, the values 1 to 10 inclusive are unchanged,
596 values 1000 and higher are recoded to the system-missing value and all other
597 values are changed to 999:
599 recode @var{x} @var{y}
602 (1000 THRU HIGHEST = SYSMIS)
606 If @samp{INTO @var{dest_vars}} is given, then recoded values are written
607 into the variables specified in @var{dest_vars}, which must therefore
608 contain a list of valid variable names.
609 The number of variables in @var{dest_vars} must be the same as the number
610 of variables in @var{src_vars}
611 and the respective order of the variables in @var{dest_vars} corresponds to
612 the order of @var{src_vars}.
613 That is to say, recoded values whose
614 original value came from the @var{n}th variable in @var{src_vars} will be
615 placed into the @var{n}th variable in @var{dest_vars}.
616 The source variables will be unchanged.
617 If any mapping implies a string as its destination value, then the respective
618 destination variable must already exist, or
619 have been declared using @cmd{STRING} or another transformation.
620 Numeric variables however will be automatically created if they don't already
622 The following example deals with two source variables, @var{a} and @var{b}
623 which contain string values. Hence there are two destination variables
624 @var{v1} and @var{v2}.
625 Any cases where @var{a} or @var{b} contain the values @samp{apple},
626 @samp{pear} or @samp{pomegranate} will result in @var{v1} or @var{v2} being
627 filled with the string @samp{fruit} whilst cases with
628 @samp{tomato}, @samp{lettuce} or @samp{carrot} will result in @samp{vegetable}.
629 Any other values will produce the result @samp{unknown}:
631 string @var{v1} (a20).
632 string @var{v2} (a20).
634 recode @var{a} @var{b}
635 ("apple" "pear" "pomegranate" = "fruit")
636 ("tomato" "lettuce" "carrot" = "vegetable")
638 into @var{v1} @var{v2}.
641 There is one very special mapping, not mentioned above.
642 If the source variable is a string variable
643 then a mapping may be specified as @samp{(CONVERT)}.
644 This mapping, if it appears must be the last mapping given and
645 the @samp{INTO @var{dest_vars}} clause must also be given and
646 must not refer to a string variable.
647 @samp{CONVERT} causes a number specified as a string to
648 be converted to a numeric value.
649 For example it will convert the string @samp{"3"} into the numeric
650 value 3 (note that it will not convert @samp{three} into 3).
651 If the string cannot be parsed as a number, then the system-missing value
653 In the following example, cases where the value of @var{x} (a string variable)
654 is the empty string, are recoded to 999 and all others are converted to the
655 numeric equivalent of the input value. The results are placed into the
656 numeric variable @var{y}:
664 It is possible to specify multiple recodings on a single command.
665 Introduce additional recodings with a slash (@samp{/}) to
666 separate them from the previous recodings:
669 @var{a} (2 = 22) (else = 99)
670 /@var{b} (1 = 3) into @var{z}
673 @noindent Here we have two recodings. The first affects the source variable
674 @var{a} and recodes in-place the value 2 into 22 and all other values to 99.
675 The second recoding copies the values of @var{b} into the variable @var{z},
676 changing any instances of 1 into 3.
683 SORT CASES BY @var{var_list}[(@{D|A@}] [ @var{var_list}[(@{D|A@}] ] ...
686 @cmd{SORT CASES} sorts the active dataset by the values of one or more
689 Specify @subcmd{BY} and a list of variables to sort by. By default, variables
690 are sorted in ascending order. To override sort order, specify @subcmd{(D)} or
691 @subcmd{(DOWN)} after a list of variables to get descending order, or @subcmd{(A)} or @subcmd{(UP)}
692 for ascending order. These apply to all the listed variables
693 up until the preceding @subcmd{(A)}, @subcmd{(D)}, @subcmd{(UP)} or @subcmd{(DOWN)}.
695 The sort algorithms used by @cmd{SORT CASES} are stable. That is,
696 records that have equal values of the sort variables will have the
697 same relative order before and after sorting. As a special case,
698 re-sorting an already sorted file will not affect the ordering of
701 @cmd{SORT CASES} is a procedure. It causes the data to be read.
703 @cmd{SORT CASES} attempts to sort the entire active dataset in main memory.
704 If workspace is exhausted, it falls back to a merge sort algorithm that
705 involves creates numerous temporary files.
707 @cmd{SORT CASES} may not be specified following @cmd{TEMPORARY}.