[pspp] / doc / flow-control.texi

@c PSPP - a program for statistical analysis.
@c Copyright (C) 2017, 2020 Free Software Foundation, Inc.
@c Permission is granted to copy, distribute and/or modify this document
@c under the terms of the GNU Free Documentation License, Version 1.3
@c or any later version published by the Free Software Foundation;
@c with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts.
@c A copy of the license is included in the section entitled "GNU
@c Free Documentation License".
@c

@node Conditionals and Looping
@chapter Conditional and Looping Constructs
@cindex conditionals
@cindex loops
@cindex flow of control
@cindex control flow

This chapter documents @pspp{} commands used for conditional execution,
looping, and flow of control.

@menu
* BREAK::                       Exit a loop.
* DEFINE::                      Define a macro.
* DO IF::                       Conditionally execute a block of code.
* DO REPEAT::                   Textually repeat a code block.
* LOOP::                        Repeat a block of code.
@end menu

@node BREAK
@section BREAK
@vindex BREAK

@display
BREAK.
@end display

@cmd{BREAK} terminates execution of the innermost currently executing
@cmd{LOOP} construct.

@cmd{BREAK} is allowed only inside @cmd{LOOP}@dots{}@cmd{END LOOP}.
@xref{LOOP}, for more details.

@node DEFINE
@section DEFINE
@vindex DEFINE
@cindex macro

@node Macro Overview
@subsection Overview

@display
@t{DEFINE} @i{macro_name}@t{(}@r{[}@i{argument}@r{[}@t{/}@i{argument}@r{]@dots{}]}@t{)}
@dots{}@i{body}@dots{}
@t{!ENDDEFINE.}
@end display

@noindent
Each @i{argument} takes the following form:
@display
@r{@{}@i{!arg_name}@t{=} @math{|} @t{!POSITIONAL}@r{@}}
@r{[}@t{!DEFAULT(}@i{default}@t{)}@r{]}
@r{[}@t{!NOEXPAND}@r{]}
@r{@{}@t{!TOKENS(}@i{count}@t{)} @math{|} @t{!CHAREND('}@i{token}@t{')} @math{|} @t{!ENCLOSE('}@i{start}@t{' @math{|} '}@i{end}@t{')} @math{|} @t{!CMDEND}@}
@end display

@noindent
The following directives may be used within @i{body}:
@example
!OFFEXPAND
!ONEXPAND
@end example

@noindent
The following functions may be used within the body:
@display
@t{!BLANKS(}@i{count}@t{)}
@t{!CONCAT(}@i{arg}@dots{}@t{)}
@t{!EVAL(}@i{arg}@t{)}
@t{!HEAD(}@i{arg}@t{)}
@t{!INDEX(}@i{haystack}@t{,} @i{needle}@t{)}
@t{!LENGTH(}@i{arg}@t{)}
@t{!NULL}
@t{!QUOTE(}@i{arg}@t{)}
@t{!SUBSTR(}@i{arg}@t{,} @i{start}[@t{,} @i{count}]@t{)}
@t{!TAIL(}@i{arg}@t{)}
@t{!UNQUOTE(}@i{arg}@t{)}
@t{!UPCASE(}@i{arg}@t{)}
@end display

@noindent
The body may also include the following constructs:
@display
@t{!IF (}@i{condition}@t{) !THEN} @i{true-expansion} @t{!ENDIF}
@t{!IF (}@i{condition}@t{) !THEN} @i{true-expansion} @t{!ELSE} @i{false-expansion} @t{!ENDIF}

@t{!DO} @i{!var} @t{=} @i{start} @t{!TO} @i{end} [@t{!BY} @i{step}]
  @i{body}
@t{!DOEND}
@t{!DO} @i{!var} @t{!IN} @t{(}@i{expression}@t{)}
  @i{body}
@t{!DOEND}

@t{!LET} @i{!var} @t{=} @i{expression}
@end display

@node Macro Introduction
@subsection Introduction

The DEFINE command creates a @dfn{macro}, which is a name for a
fragment of PSPP syntax called the macro's @dfn{body}.  Following the
DEFINE command, syntax may @dfn{call} the macro by name any number of
times.  Each call substitutes, or @dfn{expands}, the macro's body in
place of the call, as if the body had been written in its place.

The following syntax defines a macro named @code{!vars} that expands
to the variable names @code{v1 v2 v3}.  The macro's name begins with
@samp{!}, which is optional for macro names.  The @code{()} following
the macro name are required:

@example
DEFINE !vars()
v1 v2 v3
!ENDDEFINE.
@end example

Here are two ways that @code{!vars} might be called given the
preceding definition:

@example
DESCRIPTIVES !vars.
FREQUENCIES /VARIABLES=!vars.
@end example

With macro expansion, the above calls are equivalent to the following:

@example
DESCRIPTIVES v1 v2 v3.
FREQUENCIES /VARIABLES=v1 v2 v3.
@end example

The @code{!vars} macro expands to a fixed body.  Macros may have more
sophisticated contents:

@itemize @bullet
@item
Macro @dfn{arguments} that are substituted into the body whenever they
are named.  The values of a macro's arguments are specified each time
it is called.  @xref{Macro Arguments}.

@item
Macro @dfn{functions}, expanded when the macro is called.  @xref{Macro
Functions}.

@item
@code{!IF} constructs, for conditional expansion.  @xref{Macro
Conditional Expansion}.

@item
Two forms of @code{!DO} construct, for looping over a numerical range
or a collection of tokens.  @xref{Macro Loops}.

@item
@code{!LET} constructs, for assigning to macro variables.  @xref{Macro
Variable Assignment}.
@end itemize

Many identifiers associated with macros begin with @samp{!}, a
character not normally allowed in identifiers.  These identifiers are
reserved only for use with macros, which helps keep them from being
confused with other kinds of identifiers.

The following sections provide more details on macro syntax and
semantics.

@node Macro Bodies
@subsection Macro Bodies

As previously shown, a macro body may contain a fragment of a PSPP
command (such as a variable name).  A macro body may also contain full
PSPP commands.  In the latter case, the macro body should also contain
the command terminators.

Most PSPP commands may occur within a macro.  The @code{DEFINE}
command itself is one exception, because the inner @code{!ENDDEFINE}
ends the outer macro definition.  For compatibility, @code{BEGIN
DATA}@dots{}@code{END DATA.} should not be used within a macro.

The body of a macro may call another macro.  The following shows one
way that could work:

@example
DEFINE !commands()
DESCRIPTIVES !vars.
FREQUENCIES /VARIABLES=!vars.
!ENDDEFINE.

DEFINE !vars() v1 v2 v3 !ENDDEFINE.
!commands

* We can redefine the variables macro to analyze different variables:
DEFINE !vars() v4 v5 !ENDDEFINE.
!commands
@end example

The @code{!commands} macro would be easier to use if it took the
variables to analyze as an argument rather than through another macro.
The following section shows how to do that.

@node Macro Arguments
@subsection Macro Arguments

This section explains how to use macro arguments.  As an initial
example, the following syntax defines a macro named @code{!analyze}
that takes all the syntax up to the first command terminator as an
argument:

@example
DEFINE !analyze(!POSITIONAL !CMDEND)
DESCRIPTIVES !1.
FREQUENCIES /VARIABLES=!1.
!ENDDEFINE.
@end example

@noindent When @code{!analyze} is called, it expands to a pair of analysis
commands with each @code{!1} in the body replaced by the argument.
That is, these calls:

@example
!analyze v1 v2 v3.
!analyze v4 v5.
@end example

@noindent act like the following:

@example
DESCRIPTIVES v1 v2 v3.
FREQUENCIES /VARIABLES=v1 v2 v3.
DESCRIPTIVES v4 v5.
FREQUENCIES /VARIABLES=v4 v5.
@end example

Macros may take any number of arguments, described within the
parentheses in the DEFINE command.  Arguments come in two varieties
based on how their values are specified when the macro is called:

@itemize @bullet
@item
A @dfn{positional} argument has a required value that follows the
macro's name.  Use the @code{!POSITIONAL} keyword to declare a
positional argument.

When a macro is called, the positional argument values appear in the
same order as their definitions, before any keyword argument values.

References to a positional argument in a macro body are numbered:
@code{!1} is the first positional argument, @code{!2} the second, and
so on.  In addition, @code{!*} expands to all of the positional
arguments' values, separated by spaces.

The following example uses a positional argument:

@example
DEFINE !analyze(!POSITIONAL !CMDEND)
DESCRIPTIVES !1.
FREQUENCIES /VARIABLES=!1.
!ENDDEFINE.

!analyze v1 v2 v3.
!analyze v4 v5.
@end example

@item
A @dfn{keyword} argument has a name.  In the macro call, its value is
specified with the syntax @code{@i{name}=@i{value}}.  The names allow
keyword argument values to take any order in the call.

In declaration and calls, a keyword argument's name may not begin with
@samp{!}, but references to it in the macro body do start with a
leading @samp{!}.

The following example uses a keyword argument that defaults to ALL if
the argument is not assigned a value:

@example
DEFINE !analyze_kw(vars=!DEFAULT(ALL) !CMDEND)
DESCRIPTIVES !vars.
FREQUENCIES /VARIABLES=!vars.
!ENDDEFINE.

!analyze_kw vars=v1 v2 v3.  /* Analyze specified variables.
!analyze_kw.                /* Analyze all variables.
@end example
@end itemize

If a macro has both positional and keyword arguments, then the
positional arguments must come first in the DEFINE command, and their
values also come first in macro calls.  A keyword argument may be
omitted by leaving its keyword out of the call, and a positional
argument may be omitted by putting a command terminator where it would
appear.  (The latter case also omits any following positional
arguments and all keyword arguments, if there are any.)  When an
argument is omitted, a default value is used: either the value
specified in @code{!DEFAULT(@i{value})}, or an empty value otherwise.

Each argument declaration specifies the form of its value:

@table @code
@item !TOKENS(@i{count})
Exactly @var{count} tokens, e.g.@: @code{!TOKENS(1)} for a single
token.  Each identifier, number, quoted string, operator, or
punctuator is a token.  @xref{Tokens}, for a complete definition.

The following variant of @code{!analyze_kw} accepts only a single
variable name (or @code{ALL}) as its argument:

@example
DEFINE !analyze_one_var(!POSITIONAL !TOKENS(1))
DESCRIPTIVES !1.
FREQUENCIES /VARIABLES=!1.
!ENDDEFINE.

!analyze_one_var v1.
@end example

@item !CHAREND('@var{token}')
Any number of tokens up to @var{token}, which should be an operator or
punctuator token such as @samp{/} or @samp{+}.  The @var{token} does
not become part of the value.

With the following variant of @code{!analyze_kw}, the variables must
be following by @samp{/}:

@example
DEFINE !analyze_parens(vars=!CHARNED('/'))
DESCRIPTIVES !vars.
FREQUENCIES /VARIABLES=!vars.
!ENDDEFINE.

!analyze_parens vars=v1 v2 v3/.
@end example

@item !ENCLOSE('@var{start}','@var{end}')
Any number of tokens enclosed between @var{start} and @var{end}, which
should each be operator or punctuator tokens.  For example, use
@code{!ENCLOSE('(',')')} for a value enclosed within parentheses.
(Such a value could never have right parentheses inside it, even
paired with left parentheses.)  The start and end tokens are not part
of the value.

With the following variant of @code{!analyze_kw}, the variables must
be specified within parentheses:

@example
DEFINE !analyze_parens(vars=!ENCLOSE('(',')'))
DESCRIPTIVES !vars.
FREQUENCIES /VARIABLES=!vars.
!ENDDEFINE.

!analyze_parens vars=(v1 v2 v3).
@end example

@item !CMDEND
Any number of tokens up to the end of the command.  This should be
used only for the last positional parameter, since it consumes all of
the tokens in the command calling the macro.

The following variant of @code{!analyze_kw} takes all the variable
names up to the end of the command as its argument:

@example
DEFINE !analyze_kw(vars=!CMDEND)
DESCRIPTIVES !vars.
FREQUENCIES /VARIABLES=!vars.
!ENDDEFINE.

!analyze_kw vars=v1 v2 v3.
@end example
@end table

By default, when an argument's value contains a macro call, the call
is expanded each time the argument appears in the macro's body.  The
@code{!NOEXPAND} keyword in an argument declaration suppresses this
expansion.  @xref{Controlling Macro Expansion}.

@node Controlling Macro Expansion
@subsection Controlling Macro Expansion

Multiple factors control whether macro calls are expanded in different
situations.  At the highest level, @code{SET MEXPAND} controls whether
macro calls are expanded.  By default, it is enabled.  @xref{SET
MEXPAND}, for details.

A macro body may contain macro calls.  By default, these are expanded.
If a macro body contains @code{!OFFEXPAND} or @code{!ONEXPAND}
directives, then @code{!OFFEXPAND} disables expansion of macro calls
until the following @code{!ONEXPAND}.

A macro argument's value may contain a macro call.  These macro calls
are expanded, unless the argument was declared with the
@code{!NOEXPAND} keyword.

The argument to a macro function is a special context that does not
expand macro calls.  For example, if @code{!vars} is the name of a
macro, then @code{!LENGTH(!vars)} expands to 5, as does
@code{!LENGTH(!1)} if positional argument 1 has value @code{!vars}.
To expand macros in these cases, use the @code{!EVAL} macro function,
e.g.@: @code{!LENGTH(!EVAL(!vars))} or @code{!LENGTH(!EVAL(!1))}.
@xref{Macro Functions}, for details.

These rules apply to macro calls, not to uses within a macro body of
macro functions, macro arguments, and macro variables created by
@code{!DO} or @code{!LET}, which are always expanded.

@node Macro Functions
@subsection Macro Functions

Macro bodies may manipulate syntax using macro functions.  Macro
functions accept tokens as arguments and expand to sequences of
characters.

The arguments to macro functions have a restricted form.  They may
only be a single token (such as an identifier or a string), a macro
argument, or a call to a macro function.  Thus, the following are
valid macro arguments:
@example
x    5.0    x    !1    "5 + 6"    !CONCAT(x,y)
@end example
@noindent and the following are not:
@example
x y    5+6
@end example

Macro functions expand to sequences of characters.  When these
character strings are processed further as character strings, e.g.@:
with @code{!LENGTH}, any character string is valid.  When they are
interpreted as PSPP syntax, e.g.@: when the expansion becomes part of
a command, they need to be valid for that purpose.  For example,
@code{!UNQUOTE("It's")} will yield an error if the expansion
@code{It's} becomes part of a PSPP command, because it contains
unbalanced single quotes, but @code{!LENGTH(!UNQUOTE("It's"))} expands
to 4.

The following macro functions are available.  Each function's
documentation includes examples in the form @code{@var{call}
@expansion{} @var{expansion}}.

@deffn {Macro Function} !BLANKS (count)
Expands to @var{count} unquoted spaces, where @var{count} is a
nonnegative integer.  Outside quotes, any positive number of spaces
are equivalent; for a quoted string of spaces, use
@code{!QUOTE(!BLANKS(@var{count}))}.

In the examples below, @samp{_} stands in for a space to make the
results visible.

@c Keep these examples in sync with the test for !BLANKS in
@c tests/language/control/define.at:
@example
!BLANKS(0)                  @expansion{} @r{empty}
!BLANKS(1)                  @expansion{} _
!BLANKS(2)                  @expansion{} __
!QUOTE(!BLANKS(5))          @expansion{} '_____'
@end example
@end deffn

@deffn {Macro Function} !CONCAT (arg@dots{})
Expands to the concatenation of all of the arguments.  Before
concatenation, each quoted string argument is unquoted, as if
@code{!UNQUOTE} were applied.  This allows for ``token pasting'',
combining two (or more) tokens into a single one:

@c Keep these examples in sync with the test for !CONCAT in
@c tests/language/control/define.at:
@example
!CONCAT(x, y)                @expansion{} xy
!CONCAT('x', 'y')            @expansion{} xy
!CONCAT(12, 34)              @expansion{} 1234
!CONCAT(!NULL, 123)          @expansion{} 123
@end example

@code{!CONCAT} is often used for constructing a series of similar
variable names from a prefix followed by a number and perhaps a
suffix.  For example:

@c Keep these examples in sync with the test for !CONCAT in
@c tests/language/control/define.at:
@example
!CONCAT(x, 0)                @expansion{} x0
!CONCAT(x, 0, y)             @expansion{} x0y
@end example

An identifier token must begin with a letter (or @samp{#} or
@samp{@@}), which means that attempting to use a number as the first
part of an identifier will produce a pair of distinct tokens rather
than a single one.  For example:

@c Keep these examples in sync with the test for !CONCAT in
@c tests/language/control/define.at:
@example
!CONCAT(0, x)                @expansion{} 0 x
!CONCAT(0, x, y)             @expansion{} 0 xy
@end example
@end deffn

@deffn {Macro Function} !EVAL (arg)
Expands macro calls in @var{arg}.  This is especially useful if
@var{arg} is the name of a macro or a macro argument that expands to
one, because arguments to macro functions are not expanded by default
(@pxref{Controlling Macro Expansion}).

The following examples assume that @code{!vars} is a macro that
expands to @code{a b c}:

@example
!vars                        @expansion{} a b c
!QUOTE(!vars)                @expansion{} '!vars'
!EVAL(!vars)                 @expansion{} a b c
!QUOTE(!EVAL(!vars))         @expansion{} 'a b c'
@end example

These examples additionally assume that argument @code{!1} has value
@code{!vars}:

@example
!1                           @expansion{} a b c
!QUOTE(!1)                   @expansion{} '!vars'
!EVAL(!1)                    @expansion{} a b c
!QUOTE(!EVAL(!1))            @expansion{} 'a b c'
@end example
@end deffn

@deffn {Macro Function} !HEAD (arg)
@deffnx {Macro Function} !TAIL (arg)
@code{!HEAD} expands to just the first token in an unquoted version of
@var{arg}, and @code{!TAIL} to all the tokens after the first.

@example
!HEAD('a b c')               @expansion{} a
!HEAD('a')                   @expansion{} a
!HEAD(!NULL)                 @expansion{} @r{empty}
!HEAD('')                    @expansion{} @r{empty}

!TAIL('a b c')               @expansion{} b c
!TAIL('a')                   @expansion{} @r{empty}
!TAIL(!NULL)                 @expansion{} @r{empty}
!TAIL('')                    @expansion{} @r{empty}
@end example
@end deffn

@deffn {Macro Function} !INDEX (haystack, needle)
Looks for @var{needle} in @var{haystack}.  If it is present, expands
to the 1-based index of its first occurrence; if not, expands to 0.

@example
!INDEX(banana, an)           @expansion{} 2
!INDEX(banana, nan)          @expansion{} 3
!INDEX(banana, apple)        @expansion{} 0
!INDEX("banana", nan)        @expansion{} 4
!INDEX("banana", "nan")      @expansion{} 0
!INDEX(!UNQUOTE("banana"), !UNQUOTE("nan")) @expansion{} 3
@end example
@end deffn

@deffn {Macro Function} !LENGTH (arg)
Expands to a number token representing the number of characters in
@var{arg}.

@example
!LENGTH(123)                 @expansion{} 3
!LENGTH(123.00)              @expansion{} 6
!LENGTH( 123 )               @expansion{} 3
!LENGTH("123")               @expansion{} 5
!LENGTH(xyzzy)               @expansion{} 5
!LENGTH("xyzzy")             @expansion{} 7
!LENGTH("xy""zzy")           @expansion{} 9
!LENGTH(!UNQUOTE("xyzzy"))   @expansion{} 5
!LENGTH(!UNQUOTE("xy""zzy")) @expansion{} 6
!LENGTH(!1)                  @expansion{} 5 @r{if @t{!1} is @t{a b c}}
!LENGTH(!1)                  @expansion{} 0 @r{if @t{!1} is empty}
!LENGTH(!NULL)               @expansion{} 0
@end example
@end deffn

@deffn {Macro Function} !NULL
Expands to an empty character sequence.

@c Keep these examples in sync with the test for !NULL in
@c tests/language/control/define.at:
@example
!NULL                        @expansion{} @r{empty}
!QUOTE(!NULL)                @expansion{} ''
@end example
@end deffn

@deffn {Macro Function} !QUOTE (arg)
@deffnx {Macro Function} !UNQUOTE (arg)
The @code{!QUOTE} function expands to its argument surrounded by
apostrophes, doubling any apostrophes inside the argument to make sure
that it is valid PSPP syntax for a string.  If the argument was
already a quoted string, @code{!QUOTE} expands to it unchanged.

Given a quoted string argument, the @code{!UNQUOTED} function expands
to the string's contents, with the quotes removed and any doubled
quote marks reduced to singletons.  If the argument was not a quoted
string, @code{!UNQUOTE} expands to the argument unchanged.

@c Keep these examples in sync with the test for !QUOTE and !UNQUOTE in
@c tests/language/control/define.at:
@example
!QUOTE(123.0)                @expansion{} '123.0'
!QUOTE( 123 )                @expansion{} '123'
!QUOTE('a b c')              @expansion{} 'a b c'
!QUOTE("a b c")              @expansion{} "a b c"
!QUOTE(!1)                   @expansion{} 'a ''b'' c' @r{if @t{!1} is @t{a 'b' c}}

!UNQUOTE(123.0)              @expansion{} 123.0
!UNQUOTE( 123 )              @expansion{} 123
!UNQUOTE('a b c')            @expansion{} a b c
!UNQUOTE("a b c")            @expansion{} a b c
!UNQUOTE(!1)                 @expansion{} a 'b' c @r{if @t{!1} is @t{a 'b' c}}

!QUOTE(!UNQUOTE(123.0))      @expansion{} '123.0'
!QUOTE(!UNQUOTE( 123 ))      @expansion{} '123'
!QUOTE(!UNQUOTE('a b c'))    @expansion{} 'a b c'
!QUOTE(!UNQUOTE("a b c"))    @expansion{} 'a b c'
!QUOTE(!UNQUOTE(!1))         @expansion{} 'a ''b'' c' @r{if @t{!1} is @t{a 'b' c}}
@end example
@end deffn

@deffn {Macro Function} !SUBSTR (arg, start[, count])
Expands to a substring of @var{arg} starting from 1-based position
@var{start}.  If @var{count} is given, it limits the number of
characters in the expansion; if it is omitted, then the expansion
extends to the end of @var{arg}.

@example
!SUBSTR(banana, 3)           @expansion{} nana
!SUBSTR(banana, 3, 3)        @expansion{} nan
!SUBSTR("banana", 1, 3)         @expansion{} @r{error (@code{"ba} is not a valid token)}
!SUBSTR(!UNQUOTE("banana"), 3) @expansion{} nana
!SUBSTR("banana", 3, 3)      @expansion{} ana

!SUBSTR(banana, 3, 0)        @expansion{} @r{empty}
!SUBSTR(banana, 3, 10)       @expansion{} nana
!SUBSTR(banana, 10, 3)       @expansion{} @r{empty}
@end example
@end deffn

@deffn {Macro Function} !UPCASE (arg)
Expands to an unquoted version of @var{arg} with all letters converted
to uppercase.

@example
!UPCASE(freckle)             @expansion{} FRECKLE
!UPCASE('freckle')           @expansion{} FRECKLE
!UPCASE('a b c')             @expansion{} A B C
!UPCASE('A B C')             @expansion{} A B C
@end example
@end deffn

@node Macro Expressions
@subsection Macro Expressions

Macro expressions are used in conditional expansion and loops, which
are described in the following sections.  A macro expression may use
the following operators, listed in descending order of operator
precedence:

@table @code
@item ()
Parentheses override the default operator precedence.

@item !EQ !NE !GT !LT !GE !LE = ~= <> > < >= <=
Relational operators compare their operands and yield a Boolean
result, either @samp{0} for false or @samp{1} for true.

These operators always compare their operands as strings.  This can be
surprising when the strings are numbers because, e.g.,@: @code{1 <
1.0} and @code{10 < 2} both evaluate to @samp{1} (true).

Comparisons are case sensitive, so that @code{a = A} evaluates to
@samp{0} (false).

@item !NOT ~
@itemx !AND &
@itemx !OR |
Logical operators interpret their operands as Boolean values, where
quoted or unquoted @samp{0} is false and anything else is true, and
yield a Boolean result, either @samp{0} for false or @samp{1} for
true.
@end table

Macro expressions do not include any arithmetic operators.

An operand in an expression may be a single token (including a macro
argument name) or a macro function invocation.  Either way, the
expression evaluator unquotes the operand, so that @code{1 = '1'} is
true.

@node Macro Conditional Expansion
@subsection Macro Conditional Expansion

The @code{!IF} construct may be used inside a macro body to allow for
conditional expansion.  It takes the following forms:

@example
!IF (@var{expression}) !THEN @var{true-expansion} !IFEND
!IF (@var{expression}) !THEN @var{true-expansion} !ELSE @var{false-expansion} !IFEND
@end example

When @var{expression} evaluates to true, the macro processor expands
@var{true-expansion}; otherwise, it expands @var{false-expansion}, if
it is present.  The macro processor considers quoted or unquoted
@samp{0} to be false, and anything else to be true.

@node Macro Loops
@subsection Macro Loops

The body of a macro may include two forms of loops: loops over
numerical ranges and loops over tokens.  Both forms expand a @dfn{loop
body} multiple times, each time setting a named @dfn{loop variable} to
a different value.  The loop body typically expands the loop variable
at least once.

The MITERATE setting (@pxref{SET MITERATE}) limits the number of
iterations in a loop.  This is a safety measure to ensure that macro
expansion terminates.  PSPP issues a warning when the MITERATE limit
is exceeded.

@subsubheading Loops Over Ranges

@example
!DO @var{!var} = @var{start} !TO @var{end} [!BY @var{step}]
  @var{body}
!DOEND
@end example

A loop over a numerical range has the form shown above.  @var{start},
@var{end}, and @var{step} (if included) must be expressions with
numeric values.  The macro processor accepts both integers and real
numbers.  The macro processor expands @var{body} for each numeric
value from @var{start} to @var{end}, inclusive.

The default value for @var{step} is 1.  If @var{step} is positive and
@math{@var{first} > @var{last}}, or if @var{step} is negative and
@math{@var{first} < @var{last}}, then the macro processor doesn't
expand the body at all.  @var{step} may not be zero.

@subsubheading Loops Over Tokens

@example
!DO @var{!var} !IN (@var{expression})
  @var{body}
!DOEND
@end example

A loop over tokens takes the form shown above.  The macro processor
evaluates @var{expression} and expands @var{body} once per token in
the result, substituting the token for @var{!var} each time it
appears.

@node Macro Variable Assignment
@subsection Macro Variable Assignment

The @code{!LET} construct evaluates an expression and assigns the
result to a macro variable.  It may create a new macro variable or
change the value of one created by a previous @code{!LET} or
@code{!DO}, but it may not change the value of a macro argument.
@code{!LET} has the following form:

@example
!LET @var{!var} = @var{expression}
@end example

If @var{expression} is more than one token, it must be enclosed in
parentheses.

@node Macro Settings
@subsection Macro Settings

Some macro behavior is controlled through the SET command
(@pxref{SET}).  This section describes these settings.

Any SET command that changes these settings within a macro body only
takes effect following the macro.  This is because PSPP expands a
macro's entire body at once, so that the SET command inside the body
only executes afterwards.

The MEXPAND setting (@pxref{SET MEXPAND}) controls whether macros will
be expanded at all.  By default, macro expansion is on.  To avoid
expansion of macros called within a macro body, use @code{!OFFEXPAND}
and @code{!ONEXPAND} (@pxref{Controlling Macro Expansion}).

When MPRINT (@pxref{SET MPRINT}) is turned on, PSPP outputs an
expansion of each macro called.  This feature can be useful for
debugging macro definitions.  For reading the expanded version, note
that macro expansion removes comments and standardizes white space.

MNEST (@pxref{SET MNEST}) limits the depth of expansion of macro
calls, that is, the nesting level of macro expansion.  The default is
50.  This is mainly useful to avoid infinite expansion in the case of
a macro that calls itself.

MITERATE (@pxref{SET MITERATE}) limits the number of iterations in a
@code{!DO} construct.  The default is 1000.

PRESERVE...RESTORE

SET MEXPAND, etc. doesn't work inside macro bodies.

@node Macro Notes
@subsection Additional Notes

@subsubsection Calling Macros from Macros

If the body of macro A includes a call to macro B, the call can use
macro arguments (including @code{!*}) and macro variables as part of
arguments to B.  For @code{!TOKENS} arguments, the argument or
variable name counts as one token regardless of the number that
expands into, for @code{!CHAREND} and @code{!ENCLOSE} arguments the
delimiters come only from the call, not the expansions, and
@code{!CMDEND} ends at the calling command, not any end of command
within an argument or variable.

Macro functions are not supported as part of the arguments in a macro
call.  To get the same effect, use @code{!LET} to define a macro
variable, then pass the macro variable to the macro.

When macro A calls macro B, the order of their @code{DEFINE} commands
doesn't matter, as long as macro B has been defined when A is called.

@subsubsection Command Terminators

Macros and command terminators require care.  Macros honor the syntax
differences between interactive and batch syntax (@pxref{Syntax
Variants}), which means that the interpretation of a macro can vary
depending on the syntax mode in use.  We assume here that interactive
mode is in use, in which @samp{.}@: at the end of a line is the
primary way to end a command.

The @code{DEFINE} command needs to end with @samp{.}@: following the
@code{!ENDDEFINE}.  The macro body may contain @samp{.}@: if it is
intended to expand to whole commands, but using @samp{.}@: within a
macro body that expands to just syntax fragments (such as a list of
variables) will cause syntax errors.

Macro directives such as @code{!IF} and @code{!DO} do not end with
@samp{.}.

@subsubsection Expansion Contexts

Macros do not expand within comments, whether introduced within a line
by @code{/*} or as a separate COMMENT or @samp{*} commands
(@pxref{COMMENT}).  (SPSS does expand macros in COMMENT and @samp{*}.)

Macros do not expand within quoted strings.

Macros are expanded in the @code{TITLE} and @code{SUBTITLE} commands
as long as their arguments are not quoted strings.

@subsubsection PRESERVE and RESTORE

Some macro bodies might use the SET command to change certain
settings.  When this is the case, consider using the PRESERVE and
RESTORE commands to save and then restore these settings.
@xref{PRESERVE and RESTORE}.

@node DO IF
@section DO IF
@vindex DO IF

@display
DO IF condition.
        @dots{}
[ELSE IF condition.
        @dots{}
]@dots{}
[ELSE.
        @dots{}]
END IF.
@end display

@cmd{DO IF} allows one of several sets of transformations to be
executed, depending on user-specified conditions.

If the specified boolean expression evaluates as true, then the block
of code following @cmd{DO IF} is executed.  If it evaluates as
missing, then
none of the code blocks is executed.  If it is false, then
the boolean expression on the first @cmd{ELSE IF}, if present, is tested in
turn, with the same rules applied.  If all expressions evaluate to
false, then the @cmd{ELSE} code block is executed, if it is present.

When @cmd{DO IF} or @cmd{ELSE IF} is specified following @cmd{TEMPORARY}
(@pxref{TEMPORARY}), the @cmd{LAG} function may not be used
(@pxref{LAG}).

@node DO REPEAT
@section DO REPEAT
@vindex DO REPEAT

@display
DO REPEAT dummy_name=expansion@dots{}.
        @dots{}
END REPEAT [PRINT].

expansion takes one of the following forms:
        var_list
        num_or_range@dots{}
        'string'@dots{}
        ALL

num_or_range takes one of the following forms:
        number
        num1 TO num2
@end display

@cmd{DO REPEAT} repeats a block of code, textually substituting
different variables, numbers, or strings into the block with each
repetition.

Specify a dummy variable name followed by an equals sign (@samp{=})
and the list of replacements.  Replacements can be a list of existing
or new variables, numbers, strings, or @code{ALL} to specify all
existing variables.  When numbers are specified, runs of increasing
integers may be indicated as @code{@var{num1} TO @var{num2}}, so that
@samp{1 TO 5} is short for @samp{1 2 3 4 5}.

Multiple dummy variables can be specified.  Each
variable must have the same number of replacements.

The code within @cmd{DO REPEAT} is repeated as many times as there are
replacements for each variable.  The first time, the first value for
each dummy variable is substituted; the second time, the second value
for each dummy variable is substituted; and so on.

Dummy variable substitutions work like macros.  They take place
anywhere in a line that the dummy variable name occurs.  This includes
command and subcommand names, so command and subcommand names that
appear in the code block should not be used as dummy variable
identifiers.  Dummy variable substitutions do not occur inside quoted
strings, comments, unquoted strings (such as the text on the
@cmd{TITLE} or @cmd{DOCUMENT} command), or inside @cmd{BEGIN
DATA}@dots{}@cmd{END DATA}.

Substitution occurs only on whole words, so that, for example, a dummy
variable PRINT would not be substituted into the word PRINTOUT.

New variable names used as replacements are not automatically created
as variables, but only if used in the code block in a context that
would create them, @i{e.g.}@: on a @cmd{NUMERIC} or @cmd{STRING} command
or on the left side of a @cmd{COMPUTE} assignment.

Any command may appear within @subcmd{DO REPEAT}, including nested @subcmd{DO REPEAT}
commands.  If @cmd{INCLUDE} or @cmd{INSERT} appears within @subcmd{DO REPEAT},
the substitutions do not apply to the included file.

If @subcmd{PRINT} is specified on @cmd{END REPEAT}, the commands after
substitutions are made should be printed to the listing file, prefixed
by a plus sign (@samp{+}).  This feature is not yet implemented.

@node LOOP
@section LOOP
@vindex LOOP

@display
LOOP [@var{index_var}=@var{start} TO @var{end} [BY @var{incr}]] [IF @var{condition}].
        @dots{}
END LOOP [IF @var{condition}].
@end display

@cmd{LOOP} iterates a group of commands.  A number of
termination options are offered.

Specify index_var to make that variable count from one value to
another by a particular increment.  @var{index_var} must be a pre-existing
numeric variable.  @var{start}, @var{end}, and @var{incr} are numeric expressions
(@pxref{Expressions}.)

During the first iteration, @var{index_var} is set to the value of @var{start}.
During each successive iteration, @var{index_var} is increased by the value of
@var{incr}.  If @var{end} > @var{start}, then the loop terminates
when @var{index_var} > @var{end};
otherwise it terminates when @var{index_var} < @var{end}.  If @var{incr} is not specified
then it defaults to +1 or -1 as appropriate.

If @var{end} > @var{start} and @var{incr} < 0, or if @var{end} < @var{start} and
 @var{incr} > 0, then the
loop is never executed.  @var{index_var} is nevertheless set to the value of
start.

Modifying @var{index_var} within the loop is allowed, but it has no effect on
the value of @var{index_var} in the next iteration.

Specify a boolean expression for the condition on @cmd{LOOP} to
cause the loop to be executed only if the condition is true.  If the
condition is false or missing before the loop contents are executed the
first time, the loop contents are not executed at all.

If index and condition clauses are both present on @cmd{LOOP}, the
index variable is always set before the condition is evaluated.  Thus,
a condition that makes use of the index variable will always see the
index value to be used in the next execution of the body.

Specify a boolean expression for the condition on @cmd{END LOOP} to cause
the loop to terminate if the condition is true after the enclosed
code block is executed.  The condition is evaluated at the end of the
loop, not at the beginning, so that the body of a loop with only a
condition on @cmd{END LOOP} will always execute at least once.

If the index clause is not present, then the global @code{MXLOOPS}
setting, which defaults to 40, limits the number of iterations
(@pxref{SET MXLOOPS}).

@cmd{BREAK} also terminates @cmd{LOOP} execution (@pxref{BREAK}).

Loop index variables are by default reset to system-missing from one
case to another, not left, unless a scratch variable is used as index.
When loops are nested, this is usually undesired behavior, which can
be corrected with @cmd{LEAVE} (@pxref{LEAVE}) or by using a scratch
variable as the loop index.

When @cmd{LOOP} or @cmd{END LOOP} is specified following @cmd{TEMPORARY}
(@pxref{TEMPORARY}), the @cmd{LAG} function may not be used
(@pxref{LAG}).