[pspp-builds.git] / doc / data-io.texi

@node Data Input and Output, System and Portable Files, Expressions, Top
@chapter Data Input and Output
@cindex input
@cindex output
@cindex data
@cindex cases
@cindex observations

Data are the focus of the PSPP language.  
Each datum  belongs to a @dfn{case} (also called an @dfn{observation}).
Each case represents an individual or ``experimental unit''.
For example, in the results of a survey, the names of the respondents,
their sex, age, etc.@: and their responses are all data and the data
pertaining to single respondent is a case.
This chapter examines
the PSPP commands for defining variables and reading and writing data.

@quotation Note
These commands tell PSPP how to read data, but the data will not
actually be read until a procedure is executed.
@end quotation

@menu
* BEGIN DATA::                  Embed data within a syntax file.
* CLEAR TRANSFORMATIONS::       Clear pending transformations.
* CLOSE FILE HANDLE::           Close a file handle.
* DATA LIST::                   Fundamental data reading command.
* END CASE::                    Output the current case.
* END FILE::                    Terminate the current input program.
* FILE HANDLE::                 Support for special file formats.
* INPUT PROGRAM::               Support for complex input programs.
* LIST::                        List cases in the active file.
* MATRIX DATA::                 Read matrices in text format.
* NEW FILE::                    Clear the active file and dictionary.
* PRINT::                       Display values in print formats.
* PRINT EJECT::                 Eject the current page then print.
* PRINT SPACE::                 Print blank lines.
* REREAD::                      Take another look at the previous input line.
* REPEATING DATA::              Multiple cases on a single line.
* WRITE::                       Display values in write formats.
@end menu

@node BEGIN DATA
@section BEGIN DATA
@vindex BEGIN DATA
@vindex END DATA
@cindex Embedding data in syntax files
@cindex Data, embedding in syntax files

@display
BEGIN DATA.
@dots{}
END DATA.
@end display

@cmd{BEGIN DATA} and @cmd{END DATA} can be used to embed raw ASCII
data in a PSPP syntax file.  @cmd{DATA LIST} or another input
procedure must be used before @cmd{BEGIN DATA} (@pxref{DATA LIST}).
@cmd{BEGIN DATA} and @cmd{END DATA} must be used together.  @cmd{END
DATA} must appear by itself on a single line, with no leading
white space and exactly one space between the words @code{END} and
@code{DATA}, like this:

@example
END DATA.
@end example

@node CLEAR TRANSFORMATIONS
@section CLEAR TRANSFORMATIONS
@vindex CLEAR TRANSFORMATIONS

@display
CLEAR TRANSFORMATIONS.
@end display

@cmd{CLEAR TRANSFORMATIONS} clears out all pending
transformations.  It does not cancel the current input program.  It is
valid only when PSPP is interactive, not in syntax files.

@node CLOSE FILE HANDLE
@section CLOSE FILE HANDLE

@display
CLOSE FILE HANDLE handle_name.
@end display

@cmd{CLOSE FILE HANDLE} disassociates the name of a file handle with a
given file.  The only specification is the name of the handle to close.
Afterward
@cmd{FILE HANDLE}.

If the file handle name refers to a scratch file, then the storage
associated with the scratch file in memory or on disk will be freed.
If the scratch file is in use, e.g.@: it has been specified on a
@cmd{GET} command whose execution has not completed, then freeing is
delayed until it is no longer in use.

The file named INLINE, which represents data entered between @cmd{BEGIN
DATA} and @cmd{END DATA}, cannot be closed.  Attempts to close it with
@cmd{CLOSE FILE HANDLE} have no effect.

@cmd{CLOSE FILE HANDLE} is a PSPP extension.

@node DATA LIST
@section DATA LIST
@vindex DATA LIST
@cindex reading data from a file
@cindex data, reading from a file
@cindex data, embedding in syntax files
@cindex embedding data in syntax files

Used to read text or binary data, @cmd{DATA LIST} is the most
fundamental data-reading command.  Even the more sophisticated input
methods use @cmd{DATA LIST} commands as a building block.
Understanding @cmd{DATA LIST} is important to understanding how to use
PSPP to read your data files.

There are two major variants of @cmd{DATA LIST}, which are fixed
format and free format.  In addition, free format has a minor variant,
list format, which is discussed in terms of its differences from vanilla
free format.

Each form of @cmd{DATA LIST} is described in detail below.

@menu
* DATA LIST FIXED::             Fixed columnar locations for data.
* DATA LIST FREE::              Any spacing you like.
* DATA LIST LIST::              Each case must be on a single line.
@end menu

@node DATA LIST FIXED
@subsection DATA LIST FIXED
@vindex DATA LIST FIXED
@cindex reading fixed-format data
@cindex fixed-format data, reading
@cindex data, fixed-format, reading
@cindex embedding fixed-format data

@display
DATA LIST [FIXED]
        @{TABLE,NOTABLE@}
        FILE='filename'
        RECORDS=record_count
        END=end_var
        /[line_no] var_spec@dots{}

where each var_spec takes one of the forms
        var_list start-end [type_spec]
        var_list (fortran_spec)
@end display

@cmd{DATA LIST FIXED} is used to read data files that have values at fixed
positions on each line of single-line or multiline records.  The
keyword FIXED is optional.

The FILE subcommand must be used if input is to be taken from an
external file.  It may be used to specify a filename as a string or a
file handle (@pxref{File Handles}).  If the FILE subcommand is not used,
then input is assumed to be specified within the command file using
@cmd{BEGIN DATA}@dots{}@cmd{END DATA} (@pxref{BEGIN DATA}).

The optional RECORDS subcommand, which takes a single integer as an
argument, is used to specify the number of lines per record.  If RECORDS
is not specified, then the number of lines per record is calculated from
the list of variable specifications later in @cmd{DATA LIST}.

The END subcommand is only useful in conjunction with @cmd{INPUT
PROGRAM}.  @xref{INPUT PROGRAM}, for details.

@cmd{DATA LIST} can optionally output a table describing how the data file
will be read.  The TABLE subcommand enables this output, and NOTABLE
disables it.  The default is to output the table.

The list of variables to be read from the data list must come last.
Each line in the data record is introduced by a slash (@samp{/}).
Optionally, a line number may follow the slash.  Following, any number
of variable specifications may be present.

Each variable specification consists of a list of variable names
followed by a description of their location on the input line.  Sets of
variables may be specified using the @code{DATA LIST} TO convention
(@pxref{Sets of
Variables}).  There are two ways to specify the location of the variable
on the line: columnar style and FORTRAN style.

In columnar style, the starting column and ending column for the field
are specified after the variable name, separated by a dash (@samp{-}).
For instance, the third through fifth columns on a line would be
specified @samp{3-5}.  By default, variables are considered to be in
@samp{F} format (@pxref{Input/Output Formats}).  (This default can be
changed; see @ref{SET} for more information.)

In columnar style, to use a variable format other than the default,
specify the format type in parentheses after the column numbers.  For
instance, for alphanumeric @samp{A} format, use @samp{(A)}.  

In addition, implied decimal places can be specified in parentheses
after the column numbers.  As an example, suppose that a data file has a
field in which the characters @samp{1234} should be interpreted as
having the value 12.34.  Then this field has two implied decimal places,
and the corresponding specification would be @samp{(2)}.  If a field
that has implied decimal places contains a decimal point, then the
implied decimal places are not applied.

Changing the variable format and adding implied decimal places can be
done together; for instance, @samp{(N,5)}.

When using columnar style, the input and output width of each variable is
computed from the field width.  The field width must be evenly divisible
into the number of variables specified.

FORTRAN style is an altogether different approach to specifying field
locations.  With this approach, a list of variable input format
specifications, separated by commas, are placed after the variable names
inside parentheses.  Each format specifier advances as many characters
into the input line as it uses.

Implied decimal places also exist in FORTRAN style.  A format
specification with @var{d} decimal places also has @var{d} implied
decimal places.

In addition to the standard format specifiers (@pxref{Input/Output
Formats}), FORTRAN style defines some extensions:

@table @asis
@item @code{X}
Advance the current column on this line by one character position.

@item @code{T}@var{x}
Set the current column on this line to column @var{x}, with column
numbers considered to begin with 1 at the left margin.

@item @code{NEWREC}@var{x}
Skip forward @var{x} lines in the current record, resetting the active
column to the left margin.

@item Repeat count
Any format specifier may be preceded by a number.  This causes the
action of that format specifier to be repeated the specified number of
times.

@item (@var{spec1}, @dots{}, @var{specN})
Group the given specifiers together.  This is most useful when preceded
by a repeat count.  Groups may be nested arbitrarily.
@end table

FORTRAN and columnar styles may be freely intermixed.  Columnar style
leaves the active column immediately after the ending column
specified.  Record motion using @code{NEWREC} in FORTRAN style also
applies to later FORTRAN and columnar specifiers.
 
@menu
* DATA LIST FIXED Examples::    Examples of DATA LIST FIXED.
@end menu

@node DATA LIST FIXED Examples
@unnumberedsubsubsec Examples

@enumerate
@item
@example
DATA LIST TABLE /NAME 1-10 (A) INFO1 TO INFO3 12-17 (1).

BEGIN DATA.
John Smith 102311
Bob Arnold 122015
Bill Yates  918 6
END DATA.
@end example

Defines the following variables:

@itemize @bullet
@item
@code{NAME}, a 10-character-wide long string variable, in columns 1
through 10.

@item
@code{INFO1}, a numeric variable, in columns 12 through 13.

@item
@code{INFO2}, a numeric variable, in columns 14 through 15.

@item
@code{INFO3}, a numeric variable, in columns 16 through 17.
@end itemize

The @code{BEGIN DATA}/@code{END DATA} commands cause three cases to be
defined:

@example
Case   NAME         INFO1   INFO2   INFO3
   1   John Smith     10      23      11
   2   Bob Arnold     12      20      15
   3   Bill Yates      9      18       6
@end example

The @code{TABLE} keyword causes PSPP to print out a table
describing the four variables defined.

@item
@example
DAT LIS FIL="survey.dat"
        /ID 1-5 NAME 7-36 (A) SURNAME 38-67 (A) MINITIAL 69 (A)
        /Q01 TO Q50 7-56
        /.
@end example

Defines the following variables:

@itemize @bullet
@item
@code{ID}, a numeric variable, in columns 1-5 of the first record.

@item
@code{NAME}, a 30-character long string variable, in columns 7-36 of the
first record.

@item
@code{SURNAME}, a 30-character long string variable, in columns 38-67 of
the first record.

@item
@code{MINITIAL}, a 1-character short string variable, in column 69 of
the first record.

@item
Fifty variables @code{Q01}, @code{Q02}, @code{Q03}, @dots{}, @code{Q49},
@code{Q50}, all numeric, @code{Q01} in column 7, @code{Q02} in column 8,
@dots{}, @code{Q49} in column 55, @code{Q50} in column 56, all in the second
record.
@end itemize

Cases are separated by a blank record.

Data is read from file @file{survey.dat} in the current directory.

This example shows keywords abbreviated to their first 3 letters.

@end enumerate

@node DATA LIST FREE
@subsection DATA LIST FREE
@vindex DATA LIST FREE

@display
DATA LIST FREE
        [(@{TAB,'c'@}, @dots{})]
        [@{NOTABLE,TABLE@}]
        FILE='filename'
        END=end_var
        /var_spec@dots{}

where each var_spec takes one of the forms
        var_list [(type_spec)]
        var_list *
@end display

In free format, the input data is, by default, structured as a series
of fields separated by spaces, tabs, commas, or line breaks.  Each
field's content may be unquoted, or it may be quoted with a pairs of
apostrophes (@samp{'}) or double quotes (@samp{"}).  Unquoted white
space separates fields but is not part of any field.  Any mix of
spaces, tabs, and line breaks is equivalent to a single space for the
purpose of separating fields, but consecutive commas will skip a
field.

Alternatively, delimiters can be specified explicitly, as a
parenthesized, comma-separated list of single-character strings
immediately following FREE.  The word TAB may also be used to specify
a tab character as a delimiter.  When delimiters are specified
explicitly, only the given characters, plus line breaks, separate
fields.  Furthermore, leading spaces at the beginnings of fields are
not trimmed, consecutive delimiters define empty fields, and no form
of quoting is allowed.

The NOTABLE and TABLE subcommands are as in @cmd{DATA LIST FIXED} above.
NOTABLE is the default.

The FILE and END subcommands are as in @cmd{DATA LIST FIXED} above.

The variables to be parsed are given as a single list of variable names.
This list must be introduced by a single slash (@samp{/}).  The set of
variable names may contain format specifications in parentheses
(@pxref{Input/Output Formats}).  Format specifications apply to all
variables back to the previous parenthesized format specification.  

In addition, an asterisk may be used to indicate that all variables
preceding it are to have input/output format @samp{F8.0}.

Specified field widths are ignored on input, although all normal limits
on field width apply, but they are honored on output.

@node DATA LIST LIST
@subsection DATA LIST LIST
@vindex DATA LIST LIST

@display
DATA LIST LIST
        [(@{TAB,'c'@}, @dots{})]
        [@{NOTABLE,TABLE@}]
        FILE='filename'
        END=end_var
        /var_spec@dots{}

where each var_spec takes one of the forms
        var_list [(type_spec)]
        var_list *
@end display

With one exception, @cmd{DATA LIST LIST} is syntactically and
semantically equivalent to @cmd{DATA LIST FREE}.  The exception is
that each input line is expected to correspond to exactly one input
record.  If more or fewer fields are found on an input line than
expected, an appropriate diagnostic is issued.

@node END CASE
@section END CASE
@vindex END CASE

@display
END CASE.
@end display

@cmd{END CASE} is used only within @cmd{INPUT PROGRAM} to output the
current case.  @xref{INPUT PROGRAM}, for details.

@node END FILE
@section END FILE
@vindex END FILE

@display
END FILE.
@end display

@cmd{END FILE} is used only within @cmd{INPUT PROGRAM} to terminate
the current input program.  @xref{INPUT PROGRAM}.

@node FILE HANDLE
@section FILE HANDLE
@vindex FILE HANDLE

@display
For text files:
        FILE HANDLE handle_name
                /NAME='filename'
                [/MODE=CHARACTER]
                /TABWIDTH=tab_width

For binary files with fixed-length records:
        FILE HANDLE handle_name
                /NAME='filename'
                /MODE=IMAGE
                [/LRECL=rec_len]

To explicitly declare a scratch handle:
        FILE HANDLE handle_name
                /MODE=SCRATCH
@end display

Use @cmd{FILE HANDLE} to associate a file handle name with a file and
its attributes, so that later commands can refer to the file by its
handle name.  Names of text files can be specified directly on
commands that access files, so that @cmd{FILE HANDLE} is only needed when a
file is not an ordinary file containing lines of text.  However,
@cmd{FILE HANDLE} may be used even for text files, and it may be
easier to specify a file's name once and later refer to it by an
abstract handle.

Specify the file handle name as the identifier immediately following the
@cmd{FILE HANDLE} command name.  The identifier INLINE is reserved for
representing data embedded in the syntax file (@pxref{BEGIN DATA}) The
file handle name must not already have been used in a previous
invocation of @cmd{FILE HANDLE}, unless it has been closed by an
intervening command (@pxref{CLOSE FILE HANDLE}).

MODE specifies a file mode.  In CHARACTER mode, the default, the data
file is read as a text file, according to the local system's 
conventions, and each text line is read as one record.
In CHARACTER mode, most input programs will expand tabs to spaces
(@cmd{DATA LIST FREE} with explicitly specified delimiters is an
exception).  By default, each tab is 4 characters wide, but an
alternate width may be specified on TABWIDTH.  A tab width of 0
suppresses tab expansion entirely.

In IMAGE mode, the data file is opened in ANSI C binary mode and records
are fixed in length.  In IMAGE mode, LRECL specifies the record length in
bytes, with a default of 1024.  Tab characters are never expanded to
spaces in binary mode.

The NAME subcommand specifies the name of the file associated with the
handle.  It is required in CHARACTER and IMAGE modes.

The SCRATCH mode designates the file handle as a scratch file handle.
Its use is usually unnecessary because file handle names that begin with
@samp{#} are assumed to refer to scratch files.  @pxref{File Handles},
for more information.

@node INPUT PROGRAM
@section INPUT PROGRAM
@vindex INPUT PROGRAM

@display
INPUT PROGRAM.
@dots{} input commands @dots{}
END INPUT PROGRAM.
@end display

@cmd{INPUT PROGRAM}@dots{}@cmd{END INPUT PROGRAM} specifies a
complex input program.  By placing data input commands within @cmd{INPUT
PROGRAM}, PSPP programs can take advantage of more complex file
structures than available with only @cmd{DATA LIST}.

The first sort of extended input program is to simply put multiple @cmd{DATA
LIST} commands within the @cmd{INPUT PROGRAM}.  This will cause all of
the data
files to be read in parallel.  Input will stop when end of file is
reached on any of the data files.

Transformations, such as conditional and looping constructs, can also be
included within @cmd{INPUT PROGRAM}.  These can be used to combine input
from several data files in more complex ways.  However, input will still
stop when end of file is reached on any of the data files.

To prevent @cmd{INPUT PROGRAM} from terminating at the first end of
file, use
the END subcommand on @cmd{DATA LIST}.  This subcommand takes a
variable name,
which should be a numeric scratch variable (@pxref{Scratch Variables}).
(It need not be a scratch variable but otherwise the results can be
surprising.)  The value of this variable is set to 0 when reading the
data file, or 1 when end of file is encountered.

Two additional commands are useful in conjunction with @cmd{INPUT PROGRAM}.
@cmd{END CASE} is the first.  Normally each loop through the
@cmd{INPUT PROGRAM}
structure produces one case.  @cmd{END CASE} controls exactly
when cases are output.  When @cmd{END CASE} is used, looping from the end of
@cmd{INPUT PROGRAM} to the beginning does not cause a case to be output.

@cmd{END FILE} is the second.  When the END subcommand is used on @cmd{DATA
LIST}, there is no way for the @cmd{INPUT PROGRAM} construct to stop
looping,
so an infinite loop results.  @cmd{END FILE}, when executed,
stops the flow of input data and passes out of the @cmd{INPUT PROGRAM}
structure.

All this is very confusing.  A few examples should help to clarify.

@example
INPUT PROGRAM.
        DATA LIST NOTABLE FILE='a.data'/X 1-10.
        DATA LIST NOTABLE FILE='b.data'/Y 1-10.
END INPUT PROGRAM.
LIST.
@end example

The example above reads variable X from file @file{a.data} and variable
Y from file @file{b.data}.  If one file is shorter than the other then
the extra data in the longer file is ignored.

@example
INPUT PROGRAM.
        NUMERIC #A #B.
        
        DO IF NOT #A.
                DATA LIST NOTABLE END=#A FILE='a.data'/X 1-10.
        END IF.
        DO IF NOT #B.
                DATA LIST NOTABLE END=#B FILE='b.data'/Y 1-10.
        END IF.
        DO IF #A AND #B.
                END FILE.
        END IF.
        END CASE.
END INPUT PROGRAM.
LIST.
@end example

The above example reads variable X from @file{a.data} and variable Y from
@file{b.data}.  If one file is shorter than the other then the missing
field is set to the system-missing value alongside the present value for
the remaining length of the longer file.

@example
INPUT PROGRAM.
        NUMERIC #A #B.

        DO IF #A.
                DATA LIST NOTABLE END=#B FILE='b.data'/X 1-10.
                DO IF #B.
                        END FILE.
                ELSE.
                        END CASE.
                END IF.
        ELSE.
                DATA LIST NOTABLE END=#A FILE='a.data'/X 1-10.
                DO IF NOT #A.
                        END CASE.
                END IF.
        END IF.
END INPUT PROGRAM.
LIST.
@end example

The above example reads data from file @file{a.data}, then from
@file{b.data}, and concatenates them into a single active file.

@example
INPUT PROGRAM.
        NUMERIC #EOF.

        LOOP IF NOT #EOF.
                DATA LIST NOTABLE END=#EOF FILE='a.data'/X 1-10.
                DO IF NOT #EOF.
                        END CASE.
                END IF.
        END LOOP.

        COMPUTE #EOF = 0.
        LOOP IF NOT #EOF.
                DATA LIST NOTABLE END=#EOF FILE='b.data'/X 1-10.
                DO IF NOT #EOF.
                        END CASE.
                END IF.
        END LOOP.

        END FILE.
END INPUT PROGRAM.
LIST.
@end example

The above example does the same thing as the previous example, in a
different way.

@example
INPUT PROGRAM.
        LOOP #I=1 TO 50.
                COMPUTE X=UNIFORM(10).
                END CASE.
        END LOOP.
        END FILE.
END INPUT PROGRAM.
LIST/FORMAT=NUMBERED.
@end example

The above example causes an active file to be created consisting of 50
random variates between 0 and 10.

@node LIST
@section LIST
@vindex LIST

@display
LIST
        /VARIABLES=var_list
        /CASES=FROM start_index TO end_index BY incr_index
        /FORMAT=@{UNNUMBERED,NUMBERED@} @{WRAP,SINGLE@} 
                @{NOWEIGHT,WEIGHT@}
@end display

The @cmd{LIST} procedure prints the values of specified variables to the
listing file.

The VARIABLES subcommand specifies the variables whose values are to be
printed.  Keyword VARIABLES is optional.  If VARIABLES subcommand is not
specified then all variables in the active file are printed.

The CASES subcommand can be used to specify a subset of cases to be
printed.  Specify FROM and the case number of the first case to print,
TO and the case number of the last case to print, and BY and the number
of cases to advance between printing cases, or any subset of those
settings.  If CASES is not specified then all cases are printed.

The FORMAT subcommand can be used to change the output format.  NUMBERED
will print case numbers along with each case; UNNUMBERED, the default,
causes the case numbers to be omitted.  The WRAP and SINGLE settings are
currently not used.  WEIGHT will cause case weights to be printed along
with variable values; NOWEIGHT, the default, causes case weights to be
omitted from the output.

Case numbers start from 1.  They are counted after all transformations
have been considered.

@cmd{LIST} attempts to fit all the values on a single line.  If needed
to make them fit, variable names are displayed vertically.  If values
cannot fit on a single line, then a multi-line format will be used.

@cmd{LIST} is a procedure.  It causes the data to be read.

@node MATRIX DATA
@section MATRIX DATA
@vindex MATRIX DATA

@display
MATRIX DATA
        /VARIABLES=var_list
        /FILE='filename'
        /FORMAT=@{LIST,FREE@} @{LOWER,UPPER,FULL@} @{DIAGONAL,NODIAGONAL@}
        /SPLIT=@{new_var,var_list@}
        /FACTORS=var_list
        /CELLS=n_cells
        /N=n
        /CONTENTS=@{N_VECTOR,N_SCALAR,N_MATRIX,MEAN,STDDEV,COUNT,MSE,
                   DFE,MAT,COV,CORR,PROX@}
@end display

@cmd{MATRIX DATA} command reads square matrices in one of several textual
formats.  @cmd{MATRIX DATA} clears the dictionary and replaces it and
reads a
data file.

Use VARIABLES to specify the variables that form the rows and columns of
the matrices.  You may not specify a variable named @code{VARNAME_}.  You
should specify VARIABLES first.

Specify the file to read on FILE, either as a file name string or a file
handle (@pxref{File Handles}).  If FILE is not specified then matrix data
must immediately follow @cmd{MATRIX DATA} with a @cmd{BEGIN
DATA}@dots{}@cmd{END DATA}
construct (@pxref{BEGIN DATA}).

The FORMAT subcommand specifies how the matrices are formatted.  LIST,
the default, indicates that there is one line per row of matrix data;
FREE allows single matrix rows to be broken across multiple lines.  This
is analogous to the difference between @cmd{DATA LIST FREE} and
@cmd{DATA LIST LIST}
(@pxref{DATA LIST}).  LOWER, the default, indicates that the lower
triangle of the matrix is given; UPPER indicates the upper triangle; and
FULL indicates that the entire matrix is given.  DIAGONAL, the default,
indicates that the diagonal is part of the data; NODIAGONAL indicates
that it is omitted.  DIAGONAL/NODIAGONAL have no effect when FULL is
specified.

The SPLIT subcommand is used to specify @cmd{SPLIT FILE} variables for the
input matrices (@pxref{SPLIT FILE}).  Specify either a single variable
not specified on VARIABLES, or one or more variables that are specified
on VARIABLES.  In the former case, the SPLIT values are not present in
the data and ROWTYPE_ may not be specified on VARIABLES.  In the latter
case, the SPLIT values are present in the data.

Specify a list of factor variables on FACTORS.  Factor variables must
also be listed on VARIABLES.  Factor variables are used when there are
some variables where, for each possible combination of their values,
statistics on the matrix variables are included in the data.

If FACTORS is specified and ROWTYPE_ is not specified on VARIABLES, the
CELLS subcommand is required.  Specify the number of factor variable
combinations that are given.  For instance, if factor variable A has 2
values and factor variable B has 3 values, specify 6.

The N subcommand specifies a population number of observations.  When N
is specified, one N record is output for each @cmd{SPLIT FILE}.

Use CONTENTS to specify what sort of information the matrices include.
Each possible option is described in more detail below.  When ROWTYPE_
is specified on VARIABLES, CONTENTS is optional; otherwise, if CONTENTS
is not specified then /CONTENTS=CORR is assumed.

@table @asis
@item N
@item N_VECTOR
Number of observations as a vector, one value for each variable.
@item N_SCALAR
Number of observations as a single value.
@item N_MATRIX
Matrix of counts.
@item MEAN
Vector of means.
@item STDDEV
Vector of standard deviations.
@item COUNT
Vector of counts.
@item MSE
Vector of mean squared errors.
@item DFE
Vector of degrees of freedom.
@item MAT
Generic matrix.
@item COV
Covariance matrix.
@item CORR
Correlation matrix.
@item PROX
Proximities matrix.
@end table

The exact semantics of the matrices read by @cmd{MATRIX DATA} are complex.
Right now @cmd{MATRIX DATA} isn't too useful due to a lack of procedures
accepting or producing related data, so these semantics aren't
documented.  Later, they'll be described here in detail.

@node NEW FILE
@section NEW FILE
@vindex NEW FILE

@display
NEW FILE.
@end display

@cmd{NEW FILE} command clears the current active file.

@node PRINT
@section PRINT
@vindex PRINT

@display
PRINT 
        OUTFILE='filename'
        RECORDS=n_lines
        @{NOTABLE,TABLE@}
        /[line_no] arg@dots{}

arg takes one of the following forms:
        'string' [start-end]
        var_list start-end [type_spec]
        var_list (fortran_spec)
        var_list *
@end display

The @cmd{PRINT} transformation writes variable data to an output file.
@cmd{PRINT} is executed when a procedure causes the data to be read.
Follow @cmd{PRINT} by @cmd{EXECUTE} to print variable data without
invoking a procedure (@pxref{EXECUTE}).

All @cmd{PRINT} subcommands are optional.

The OUTFILE subcommand specifies the file to receive the output.  The
file may be a file name as a string or a file handle (@pxref{File
Handles}).  If OUTFILE is not present then output will be sent to PSPP's
output listing file.

The RECORDS subcommand specifies the number of lines to be output.  The
number of lines may optionally be surrounded by parentheses.

TABLE will cause the PRINT command to output a table to the listing file
that describes what it will print to the output file.  NOTABLE, the
default, suppresses this output table.

Introduce the strings and variables to be printed with a slash
(@samp{/}).  Optionally, the slash may be followed by a number
indicating which output line will be specified.  In the absence of this
line number, the next line number will be specified.  Multiple lines may
be specified using multiple slashes with the intended output for a line
following its respective slash.

Literal strings may be printed.  Specify the string itself.  Optionally
the string may be followed by a column number or range of column
numbers, specifying the location on the line for the string to be
printed.  Otherwise, the string will be printed at the current position
on the line.

Variables to be printed can be specified in the same ways as available
for @cmd{DATA LIST FIXED} (@pxref{DATA LIST FIXED}).  In addition, a
variable
list may be followed by an asterisk (@samp{*}), which indicates that the
variables should be printed in their dictionary print formats, separated
by spaces.  A variable list followed by a slash or the end of command
will be interpreted the same way.

If a FORTRAN type specification is used to move backwards on the current
line, then text is written at that point on the line, the line will be
truncated to that length, although additional text being added will
again extend the line to that length.

@node PRINT EJECT
@section PRINT EJECT
@vindex PRINT EJECT

@display
PRINT EJECT 
        OUTFILE='filename'
        RECORDS=n_lines
        @{NOTABLE,TABLE@}
        /[line_no] arg@dots{}

arg takes one of the following forms:
        'string' [start-end]
        var_list start-end [type_spec]
        var_list (fortran_spec)
        var_list *
@end display

@cmd{PRINT EJECT} writes data to an output file.  Before the data is
written, the current page in the listing file is ejected.

@xref{PRINT}, for more information on syntax and usage.

@node PRINT SPACE
@section PRINT SPACE
@vindex PRINT SPACE

@display
PRINT SPACE OUTFILE='filename' n_lines.
@end display

@cmd{PRINT SPACE} prints one or more blank lines to an output file.

The OUTFILE subcommand is optional.  It may be used to direct output to
a file specified by file name as a string or file handle (@pxref{File
Handles}).  If OUTFILE is not specified then output will be directed to
the listing file.

n_lines is also optional.  If present, it is an expression
(@pxref{Expressions}) specifying the number of blank lines to be
printed.  The expression must evaluate to a nonnegative value.

@node REREAD
@section REREAD
@vindex REREAD

@display
REREAD FILE=handle COLUMN=column.
@end display

The @cmd{REREAD} transformation allows the previous input line in a
data file
already processed by @cmd{DATA LIST} or another input command to be re-read
for further processing.

The FILE subcommand, which is optional, is used to specify the file to
have its line re-read.  The file must be specified as the name of a file
handle (@pxref{File Handles}).  If FILE is not specified then the last
file specified on @cmd{DATA LIST} will be assumed (last file specified
lexically, not in terms of flow-of-control).

By default, the line re-read is re-read in its entirety.  With the
COLUMN subcommand, a prefix of the line can be exempted from
re-reading.  Specify an expression (@pxref{Expressions}) evaluating to
the first column that should be included in the re-read line.  Columns
are numbered from 1 at the left margin.

Issuing @code{REREAD} multiple times will not back up in the data
file.  Instead, it will re-read the same line multiple times.

@node REPEATING DATA
@section REPEATING DATA
@vindex REPEATING DATA

@display
REPEATING DATA
        /STARTS=start-end
        /OCCURS=n_occurs
        /FILE='filename'
        /LENGTH=length
        /CONTINUED[=cont_start-cont_end]
        /ID=id_start-id_end=id_var
        /@{TABLE,NOTABLE@}
        /DATA=var_spec@dots{}

where each var_spec takes one of the forms
        var_list start-end [type_spec]
        var_list (fortran_spec)
@end display

@cmd{REPEATING DATA} parses groups of data repeating in
a uniform format, possibly with several groups on a single line.  Each
group of data corresponds with one case.  @cmd{REPEATING DATA} may only be
used within an @cmd{INPUT PROGRAM} structure (@pxref{INPUT PROGRAM}).
When used with @cmd{DATA LIST}, it
can be used to parse groups of cases that share a subset of variables
but differ in their other data.

The STARTS subcommand is required.  Specify a range of columns, using
literal numbers or numeric variable names.  This range specifies the
columns on the first line that are used to contain groups of data.  The
ending column is optional.  If it is not specified, then the record
width of the input file is used.  For the inline file (@pxref{BEGIN
DATA}) this is 80 columns; for a file with fixed record widths it is the
record width; for other files it is 1024 characters by default.

The OCCURS subcommand is required.  It must be a number or the name of a
numeric variable.  Its value is the number of groups present in the
current record.

The DATA subcommand is required.  It must be the last subcommand
specified.  It is used to specify the data present within each repeating
group.  Column numbers are specified relative to the beginning of a
group at column 1.  Data is specified in the same way as with @cmd{DATA LIST
FIXED} (@pxref{DATA LIST FIXED}).

All other subcommands are optional.

FILE specifies the file to read, either a file name as a string or a
file handle (@pxref{File Handles}).  If FILE is not present then the
default is the last file handle used on @cmd{DATA LIST} (lexically, not in
terms of flow of control).

By default @cmd{REPEATING DATA} will output a table describing how it will
parse the input data.  Specifying NOTABLE will disable this behavior;
specifying TABLE will explicitly enable it.

The LENGTH subcommand specifies the length in characters of each group.
If it is not present then length is inferred from the DATA subcommand.
LENGTH can be a number or a variable name.

Normally all the data groups are expected to be present on a single
line.  Use the CONTINUED command to indicate that data can be continued
onto additional lines.  If data on continuation lines starts at the left
margin and continues through the entire field width, no column
specifications are necessary on CONTINUED.  Otherwise, specify the
possible range of columns in the same way as on STARTS.

When data groups are continued from line to line, it is easy
for cases to get out of sync through careless hand editing.  The
ID subcommand allows a case identifier to be present on each line of
repeating data groups.  @cmd{REPEATING DATA} will check for the same
identifier on each line and report mismatches.  Specify the range of
columns that the identifier will occupy, followed by an equals sign
(@samp{=}) and the identifier variable name.  The variable must already
have been declared with @cmd{NUMERIC} or another command.

@cmd{REPEATING DATA} should be the last command given within an
@cmd{INPUT PROGRAM}.  It should not be enclosed within a @cmd{LOOP}
structure (@pxref{LOOP}).  Use @cmd{DATA LIST} before, not after,
@cmd{REPEATING DATA}.

@node WRITE
@section WRITE
@vindex WRITE

@display
WRITE 
        OUTFILE='filename'
        RECORDS=n_lines
        @{NOTABLE,TABLE@}
        /[line_no] arg@dots{}

arg takes one of the following forms:
        'string' [start-end]
        var_list start-end [type_spec]
        var_list (fortran_spec)
        var_list *
@end display

@code{WRITE} writes text or binary data to an output file.  

@xref{PRINT}, for more information on syntax and usage.  The main
difference between @code{PRINT} and @code{WRITE} is that @cmd{WRITE}
uses write formats by default, where PRINT uses print formats.

The sole additional difference is that if @cmd{WRITE} is used to send output
to a binary file, carriage control characters will not be output.
@xref{FILE HANDLE}, for information on how to declare a file as binary.
@setfilename ignored