PSPP

Table of Contents

GNU PSPP Developers Guide

This manual is for GNU PSPP version 0.8.5, software for statistical analysis.

Copyright © 1997, 1998, 2004, 2005, 2007, 2010, 2014 Free Software Foundation, Inc.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the section entitled "GNU Free Documentation License".

1 Introduction

This manual is a guide to PSPP internals. Its intended audience is developers who wish to modify or extend PSPP’s capabilities. The use of PSPP is documented in a separate manual. See Introduction in PSPP Users Guide.

This manual is both a tutorial and a reference manual for PSPP developers. It is ultimately intended to cover everything that developers who wish to implement new PSPP statistical procedures and other commands should know. It is currently incomplete, partly because existing developers have not yet spent enough time on writing, and partly because the interfaces not yet documented are not yet mature enough to making documenting them worthwhile.

PSPP developers should have some familiarity with the basics of PSPP from a user’s perspective. This manual attempts to refer to the PSPP user manual’s descriptions of concepts that PSPP users should find familiar at the time of their first reference. However, it is probably a good idea to at least skim the PSPP manual before reading this one, if you are not already familiar with PSPP.

2 Basic Concepts

This chapter introduces basic data structures and other concepts needed for developing in PSPP.

2.1 Values

The unit of data in PSPP is a value.

Values are classified by type and width. The type of a value is either numeric or string (sometimes called alphanumeric). The width of a string value ranges from 1 to MAX_STRING bytes. The width of a numeric value is artificially defined to be 0; thus, the type of a value can be inferred from its width.

Some support is provided for working with value types and widths, in data/val-type.h:

Macro: int MAX_STRING

Maximum width of a string value, in bytes, currently 32,767.

Function: bool val_type_is_valid (enum val_type val_type)

Returns true if val_type is a valid value type, that is, either VAL_NUMERIC or VAL_STRING. Useful for assertions.

Function: enum val_type val_type_from_width (int width)

Returns VAL_NUMERIC if width is 0 and thus represents the width of a numeric value, otherwise VAL_STRING to indicate that width is the width of a string value.

The following subsections describe how values of each type are represented.

2.1.1 Numeric Values

A value known to be numeric at compile time is represented as a double. PSPP provides three values of double for special purposes, defined in data/val-type.h:

Macro: double SYSMIS

The system-missing value, used to represent a datum whose true value is unknown, such as a survey question that was not answered by the respondent, or undefined, such as the result of division by zero. PSPP propagates the system-missing value through calculations and compensates for missing values in statistical analyses. See Missing Observations in PSPP Users Guide, for a PSPP user’s view of missing values.

PSPP currently defines SYSMIS as -DBL_MAX, that is, the greatest finite negative value of double. It is best not to depend on this definition, because PSPP may transition to using an IEEE NaN (not a number) instead at some point in the future.

Macro: double LOWEST

Macro: double HIGHEST

The greatest finite negative (except for SYSMIS) and positive values of double, respectively. These values do not ordinarily appear in user data files. Instead, they are used to implement endpoints of open-ended ranges that are occasionally permitted in PSPP syntax, e.g. 5 THRU HI as a range of missing values (see MISSING VALUES in PSPP Users Guide).

2.1.2 String Values

A value known at compile time to have string type is represented as an array of char. String values do not necessarily represent readable text strings and may contain arbitrary 8-bit data, including null bytes, control codes, and bytes with the high bit set. Thus, string values are not null-terminated strings, but rather opaque arrays of bytes.

SYSMIS, LOWEST, and HIGHEST have no equivalents as string values. Usually, PSPP fills an unknown or undefined string values with spaces, but PSPP does not treat such a string as a special case when it processes it later.

MAX_STRING, the maximum length of a string value, is defined in data/val-type.h.

2.1.3 Runtime Typed Values

When a value’s type is only known at runtime, it is often represented as a union value, defined in data/value.h. A union value does not identify the type or width of the data it contains. Code that works with union valuess must therefore have external knowledge of its content, often through the type and width of a struct variable (see Variables).

union value has one member that clients are permitted to access directly, a double named ‘f’ that stores the content of a numeric union value. It has other members that store the content of string union value, but client code should use accessor functions instead of referring to these directly.

PSPP provides some functions for working with union values. The most useful are described below. To use these functions, recall that a numeric value has a width of 0.

Function: void value_init (union value *value, int width)

Initializes value as a value of the given width. After initialization, the data in value are indeterminate; the caller is responsible for storing initial data in it.

Function: void value_destroy (union value *value, int width)

Frees auxiliary storage associated with value, which must have the given width.

Function: bool value_needs_init (int width)

For some widths, value_init and value_destroy do not actually do anything, because no additional storage is needed beyond the size of union value. This function returns true if width is such a width, which case there is no actual need to call those functions. This can be a useful optimization if a large number of union values of such a width are to be initialized or destroyed.

This function returns false if value_init and value_destroy are actually required for the given width.

Function: double value_num (const union value *value)

Returns the numeric value in value, which must have been initialized as a numeric value. Equivalent to value->f.

Function: const char * value_str (const union value *value, int width)

Function: char * value_str_rw (union value *value, int width)

Returns the string value in value, which must have been initialized with positive width width. The string returned is not null-terminated. Only width bytes of returned data may be accessed.

The two different functions exist only for const-correctness. Otherwise they are identical.

It is important that width be the correct value that was passed to value_init. Passing a smaller or larger value (e.g. because that number of bytes will be accessed) will not always work and should be avoided.

Function: void value_copy (union value *dst, const union value *src, int width)

Copies the contents of union value src to dst. Both dst and src must have been initialized with the specified width.

Function: void value_set_missing (union value *value, int width)

Sets value to SYSMIS if it is numeric or to all spaces if it is alphanumeric, according to width. value must have been initialized with the specified width.

Function: bool value_is_resizable (const union value *value, int old_width, int new_width)

Determines whether value, which must have been initialized with the specified old_width, may be resized to new_width. Resizing is possible if the following criteria are met. First, old_width and new_width must be both numeric or both string widths. Second, if new_width is a short string width and less than old_width, resizing is allowed only if bytes new_width through old_width in value contain only spaces.

These rules are part of those used by mv_is_resizable and val_labs_can_set_width.

Function: void value_resize (union value *value, int old_width, int new_width)

Resizes value from old_width to new_width, which must be allowed by the rules stated above. value must have been initialized with the specified old_width before calling this function. After resizing, value has width new_width.

If new_width is greater than old_width, value will be padded on the right with spaces to the new width. If new_width is less than old_width, the rightmost bytes of value are truncated.

Function: bool value_equal (const union value *a, const union value *b, int width)

Compares of a and b, which must both have width width. Returns true if their contents are the same, false if they differ.

Function: int value_compare_3way (const union value *a, const union value *b, int width)

Compares of a and b, which must both have width width. Returns -1 if a is less than b, 0 if they are equal, or 1 if a is greater than b.

Numeric values are compared numerically, with SYSMIS comparing less than any real number. String values are compared lexicographically byte-by-byte.

Function: size_t value_hash (const union value *value, int width, unsigned int basis)

Computes and returns a hash of value, which must have the specified width. The value in basis is folded into the hash.

2.2 Input and Output Formats

Input and output formats specify how to convert data fields to and from data values (see Input and Output Formats in PSPP Users Guide). PSPP uses struct fmt_spec to represent input and output formats.

Function prototypes and other declarations related to formats are in the <data/format.h> header.

Structure: struct fmt_spec

An input or output format, with the following members:

enum fmt_type type

The format type (see below).

int w

Field width, in bytes. The width of numeric fields is always between 1 and 40 bytes, and the width of string fields is always between 1 and 65534 bytes. However, many individual types of formats place stricter limits on field width (see fmt_max_input_width, fmt_max_output_width).

int d

Number of decimal places, in character positions. For format types that do not allow decimal places to be specified, this value must be 0. Format types that do allow decimal places have type-specific and often width-specific restrictions on d (see fmt_max_input_decimals, fmt_max_output_decimals).

Enumeration: enum fmt_type

An enumerated type representing an input or output format type. Each PSPP input and output format has a corresponding enumeration constant prefixed by ‘FMT’: FMT_F, FMT_COMMA, FMT_DOT, and so on.

The following sections describe functions for manipulating formats and the data in fields represented by formats.

2.2.1 Constructing and Verifying Formats

These functions construct struct fmt_specs and verify that they are valid.

Function: struct fmt_spec fmt_for_input (enum fmt_type type, int w, int d)

Function: struct fmt_spec fmt_for_output (enum fmt_type type, int w, int d)

Constructs a struct fmt_spec with the given type, w, and d, asserts that the result is a valid input (or output) format, and returns it.

Function: struct fmt_spec fmt_for_output_from_input (const struct fmt_spec *input)

Given input, which must be a valid input format, returns the equivalent output format. See Input and Output Formats in PSPP Users Guide, for the rules for converting input formats into output formats.

Function: struct fmt_spec fmt_default_for_width (int width)

Returns the default output format for a variable of the given width. For a numeric variable, this is F8.2 format; for a string variable, it is the A format of the given width.

The following functions check whether a struct fmt_spec is valid for various uses and return true if so, false otherwise. When any of them returns false, it also outputs an explanatory error message using msg. To suppress error output, enclose a call to one of these functions by a msg_disable/msg_enable pair.

Function: bool fmt_check (const struct fmt_spec *format, bool for_input)

Function: bool fmt_check_input (const struct fmt_spec *format)

Function: bool fmt_check_output (const struct fmt_spec *format)

Checks whether format is a valid input format (for fmt_check_input, or fmt_check if for_input) or output format (for fmt_check_output, or fmt_check if not for_input).

Function: bool fmt_check_type_compat (const struct fmt_spec *format, enum val_type type)

Checks whether format matches the value type type, that is, if type is VAL_NUMERIC and format is a numeric format or type is VAL_STRING and format is a string format.

Function: bool fmt_check_width_compat (const struct fmt_spec *format, int width)

Checks whether format may be used as an output format for a value of the given width.

fmt_var_width, described in the following section, can be also be used to determine the value width needed by a format.

2.2.2 Format Utility Functions

These functions work with struct fmt_specs.

Function: int fmt_var_width (const struct fmt_spec *format)

Returns the width for values associated with format. If format is a numeric format, the width is 0; if format is an A format, then the width format->w; otherwise, format is an AHEX format and its width is format->w / 2.

Function: char *fmt_to_string (const struct fmt_spec *format, char s[FMT_STRING_LEN_MAX + 1])

Converts format to a human-readable format specifier in s and returns s. format need not be a valid input or output format specifier, e.g. it is allowed to have an excess width or decimal places. In particular, if format has decimals, they are included in the output string, even if format’s type does not allow decimals, to allow accurately presenting incorrect formats to the user.

Function: bool fmt_equal (const struct fmt_spec *a, const struct fmt_spec *b)

Compares a and b memberwise and returns true if they are identical, false otherwise. format need not be a valid input or output format specifier.

Function: void fmt_resize (struct fmt_spec *fmt, int width)

Sets the width of fmt to a valid format for a union value of size width.

2.2.3 Obtaining Properties of Format Types

These functions work with enum fmt_types instead of the higher-level struct fmt_specs. Their primary purpose is to report properties of each possible format type, which in turn allows clients to abstract away many of the details of the very heterogeneous requirements of each format type.

The first group of functions works with format type names.

Function: const char *fmt_name (enum fmt_type type)

Returns the name for the given type, e.g. "COMMA" for FMT_COMMA.

Function: bool fmt_from_name (const char *name, enum fmt_type *type)

Tries to find the enum fmt_type associated with name. If successful, sets *type to the type and returns true; otherwise, returns false without modifying *type.

The functions below query basic limits on width and decimal places for each kind of format.

Function: bool fmt_takes_decimals (enum fmt_type type)

Returns true if a format of the given type is allowed to have a nonzero number of decimal places (the d member of struct fmt_spec), false if not.

Function: int fmt_min_input_width (enum fmt_type type)

Function: int fmt_max_input_width (enum fmt_type type)

Function: int fmt_min_output_width (enum fmt_type type)

Function: int fmt_max_output_width (enum fmt_type type)

Returns the minimum or maximum width (the w member of struct fmt_spec) allowed for an input or output format of the specified type.

Function: int fmt_max_input_decimals (enum fmt_type type, int width)

Function: int fmt_max_output_decimals (enum fmt_type type, int width)

Returns the maximum number of decimal places allowed for an input or output format, respectively, of the given type and width. Returns 0 if the specified type does not allow any decimal places or if width is too narrow to allow decimal places.

Function: int fmt_step_width (enum fmt_type type)

Returns the “width step” for a struct fmt_spec of the given type. A struct fmt_spec’s width must be a multiple of its type’s width step. Most format types have a width step of 1, so that their formats’ widths may be any integer within the valid range, but hexadecimal numeric formats and AHEX string formats have a width step of 2.

These functions allow clients to broadly determine how each kind of input or output format behaves.

Function: bool fmt_is_string (enum fmt_type type)

Function: bool fmt_is_numeric (enum fmt_type type)

Returns true if type is a format for numeric or string values, respectively, false otherwise.

Function: enum fmt_category fmt_get_category (enum fmt_type type)

Returns the category within which type falls.

Enumeration: enum fmt_category

A group of format types. Format type categories correspond to the input and output categories described in the PSPP user documentation (see Input and Output Formats in PSPP Users Guide).

Each format is in exactly one category. The categories have bitwise disjoint values to make it easy to test whether a format type is in one of multiple categories, e.g.

if (fmt_get_category (type) & (FMT_CAT_DATE | FMT_CAT_TIME))
  {
    /* …type is a date or time format… */
  }

The format categories are:

FMT_CAT_BASIC

Basic numeric formats.

FMT_CAT_CUSTOM

Custom currency formats.

FMT_CAT_LEGACY

Legacy numeric formats.

FMT_CAT_BINARY

Binary formats.

FMT_CAT_HEXADECIMAL

Hexadecimal formats.

FMT_CAT_DATE

Date formats.

FMT_CAT_TIME

Time formats.

FMT_CAT_DATE_COMPONENT

Date component formats.

FMT_CAT_STRING

String formats.

The PSPP input and output routines use the following pair of functions to convert enum fmt_types to and from the separate set of codes used in system and portable files:

Function: int fmt_to_io (enum fmt_type type)

Returns the format code used in system and portable files that corresponds to type.

Function: bool fmt_from_io (int io, enum fmt_type *type)

Converts io, a format code used in system and portable files, into a enum fmt_type in *type. Returns true if successful, false if io is not valid.

These functions reflect the relationship between input and output formats.

Function: enum fmt_type fmt_input_to_output (enum fmt_type type)

Returns the output format type that is used by default by DATA LIST and other input procedures when type is specified as an input format. The conversion from input format to output format is more complicated than simply changing the format. See fmt_for_output_from_input, for a function that performs the entire conversion.

Function: bool fmt_usable_for_input (enum fmt_type type)

Returns true if type may be used as an input format type, false otherwise. The custom currency formats, in particular, may be used for output but not for input.

All format types are valid for output.

The final group of format type property functions obtain human-readable templates that illustrate the formats graphically.

Function: const char *fmt_date_template (enum fmt_type type)

Returns a formatting template for type, which must be a date or time format type. These formats are used by data_in and data_out to guide parsing and formatting date and time data.

Function: char *fmt_dollar_template (const struct fmt_spec *format)

Returns a string of the form $#,###.## according to format, which must be of type FMT_DOLLAR. The caller must free the string with free.

2.2.4 Numeric Formatting Styles

Each of the basic numeric formats (F, E, COMMA, DOT, DOLLAR, PCT) and custom currency formats (CCA, CCB, CCC, CCD, CCE) has an associated numeric formatting style, represented by struct fmt_number_style. Input and output conversion of formats that have numeric styles is determined mainly by the style, although the formatting rules have special cases that are not represented within the style.

Structure: struct fmt_number_style

A structure type with the following members:

struct substring neg_prefix

struct substring prefix

struct substring suffix

struct substring neg_suffix

A set of strings used a prefix to negative numbers, a prefix to every number, a suffix to every number, and a suffix to negative numbers, respectively. Each of these strings is no more than FMT_STYLE_AFFIX_MAX bytes (currently 16) bytes in length. These strings must be freed with ss_dealloc when no longer needed.

decimal

The character used as a decimal point. It must be either ‘.’ or ‘,’.

grouping

The character used for grouping digits to the left of the decimal point. It may be ‘.’ or ‘,’, in which case it must not be equal to decimal, or it may be set to 0 to disable grouping.

The following functions are provided for working with numeric formatting styles.

Function: void fmt_number_style_init (struct fmt_number_style *style)

Initialises a struct fmt_number_style with all of the prefixes and suffixes set to the empty string, ‘.’ as the decimal point character, and grouping disables.

Function: void fmt_number_style_destroy (struct fmt_number_style *style)

Destroys style, freeing its storage.

Function: struct fmt_number_style *fmt_create (void)

A function which creates an array of all the styles used by pspp, and calls fmt_number_style_init on each of them.

Function: void fmt_done (struct fmt_number_style *styles)

A wrapper function which takes an array of struct fmt_number_style, calls fmt_number_style_destroy on each of them, and then frees the array.

Function: int fmt_affix_width (const struct fmt_number_style *style)

Returns the total length of style’s prefix and suffix.

Function: int fmt_neg_affix_width (const struct fmt_number_style *style)

Returns the total length of style’s neg_prefix and neg_suffix.

PSPP maintains a global set of number styles for each of the basic numeric formats and custom currency formats. The following functions work with these global styles:

Function: const struct fmt_number_style * fmt_get_style (enum fmt_type type)

Returns the numeric style for the given format type.

Function: const char * fmt_name (enum fmt_type type)

Returns the name of the given format type.

2.2.5 Formatted Data Input and Output

These functions provide the ability to convert data fields into union values and vice versa.

Function: bool data_in (struct substring input, const char *encoding, enum fmt_type type, int implied_decimals, int first_column, const struct dictionary *dict, union value *output, int width)

Parses input as a field containing data in the given format type. The resulting value is stored in output, which the caller must have initialized with the given width. For consistency, width must be 0 if type is a numeric format type and greater than 0 if type is a string format type. encoding should be set to indicate the character encoding of input. dict must be a pointer to the dictionary with which output is associated.

If input is the empty string (with length 0), output is set to the value set on SET BLANKS (see SET BLANKS in PSPP Users Guide) for a numeric value, or to all spaces for a string value. This applies regardless of the usual parsing requirements for type.

If implied_decimals is greater than zero, then the numeric result is shifted right by implied_decimals decimal places if input does not contain a decimal point character or an exponent. Only certain numeric format types support implied decimal places; for string formats and other numeric formats, implied_decimals has no effect. DATA LIST FIXED is the primary user of this feature (see DATA LIST FIXED in PSPP Users Guide). Other callers should generally specify 0 for implied_decimals, to disable this feature.

When input contains invalid input data, data_in outputs a message using msg. If first_column is nonzero, it is included in any such error message as the 1-based column number of the start of the field. The last column in the field is calculated as first_column + input - 1. To suppress error output, enclose the call to data_in by calls to msg_disable and msg_enable.

This function returns true on success, false if a message was output (even if suppressed). Overflow and underflow provoke warnings but are not propagated to the caller as errors.

This function is declared in data/data-in.h.

Function: char * data_out (const union value *input, const struct fmt_spec *format)

Function: char * data_out_legacy (const union value *input, const char *encoding, const struct fmt_spec *format)

Converts the data pointed to by input into a string value, which will be encoded in UTF-8, according to output format specifier format. Format must be a valid output format. The width of input is inferred from format using an algorithm equivalent to fmt_var_width.

When input contains data that cannot be represented in the given format, data_out may output a message using msg, although the current implementation does not consistently do so. To suppress error output, enclose the call to data_out by calls to msg_disable and msg_enable.

This function is declared in data/data-out.h.

2.3 User-Missing Values

In addition to the system-missing value for numeric values, each variable has a set of user-missing values (see MISSING VALUES in PSPP Users Guide). A set of user-missing values is represented by struct missing_values.

It is rarely necessary to interact directly with a struct missing_values object. Instead, the most common operation, querying whether a particular value is a missing value for a given variable, is most conveniently executed through functions on struct variable. See Variable Missing Values, for details.

A struct missing_values is essentially a set of union values that have a common value width (see Values). For a set of missing values associated with a variable (the common case), the set’s width is the same as the variable’s width.

Function prototypes and other declarations related to missing values are declared in data/missing-values.h.

Structure: struct missing_values

Opaque type that represents a set of missing values.

The contents of a set of missing values is subject to some restrictions. Regardless of width, a set of missing values is allowed to be empty. A set of numeric missing values may contain up to three discrete numeric values, or a range of numeric values (which includes both ends of the range), or a range plus one discrete numeric value. A set of string missing values may contain up to three discrete string values (with the same width as the set), but ranges are not supported.

In addition, values in string missing values wider than MV_MAX_STRING bytes may contain non-space characters only in their first MV_MAX_STRING bytes; all the bytes after the first MV_MAX_STRING must be spaces. See mv_is_acceptable, for a function that tests a value against these constraints.

Macro: int MV_MAX_STRING

Number of bytes in a string missing value that are not required to be spaces. The current value is 8, a value which is fixed by the system file format. In PSPP we could easily eliminate this restriction, but doing so would also require us to extend the system file format in an incompatible way, which we consider a bad tradeoff.

The most often useful functions for missing values are those for testing whether a given value is missing, described in the following section. Several other functions for creating, inspecting, and modifying struct missing_values objects are described afterward, but these functions are much more rarely useful.

2.3.1 Testing for Missing Values

The most often useful functions for missing values are those for testing whether a given value is missing, described here. However, using one of the corresponding missing value testing functions for variables can be even easier (see Variable Missing Values).

Function: bool mv_is_value_missing (const struct missing_values *mv, const union value *value, enum mv_class class)

Function: bool mv_is_num_missing (const struct missing_values *mv, double value, enum mv_class class)

Function: bool mv_is_str_missing (const struct missing_values *mv, const char value[], enum mv_class class)

Tests whether value is in one of the categories of missing values given by class. Returns true if so, false otherwise.

mv determines the width of value and provides the set of user-missing values to test.

The only difference among these functions in the form in which value is provided, so you may use whichever function is most convenient.

The class argument determines the exact kinds of missing values that the functions test for:

Enumeration: enum mv_class

MV_USER

Returns true if value is in the set of user-missing values given by mv.

MV_SYSTEM

Returns true if value is system-missing. (If mv represents a set of string values, then value is never system-missing.)

MV_ANY

MV_USER | MV_SYSTEM

Returns true if value is user-missing or system-missing.

MV_NONE

Always returns false, that is, value is never considered missing.

2.3.2 Creation and Destruction

These functions create and destroy struct missing_values objects.

Function: void mv_init (struct missing_values *mv, int width)

Initializes mv as a set of user-missing values. The set is initially empty. Any values added to it must have the specified width.

Function: void mv_destroy (struct missing_values *mv)

Destroys mv, which must not be referred to again.

Function: void mv_copy (struct missing_values *mv, const struct missing_values *old)

Initializes mv as a copy of the existing set of user-missing values old.

Function: void mv_clear (struct missing_values *mv)

Empties the user-missing value set mv, retaining its existing width.

2.3.3 Changing User-Missing Value Set Width

A few PSPP language constructs copy sets of user-missing values from one variable to another. When the source and target variables have the same width, this is simple. But when the target variable’s width might be different from the source variable’s, it takes a little more work. The functions described here can help.

In fact, it is usually unnecessary to call these functions directly. Most of the time var_set_missing_values, which uses mv_resize internally to resize the new set of missing values to the required width, may be used instead. See var_set_missing_values, for more information.

Function: bool mv_is_resizable (const struct missing_values *mv, int new_width)

Tests whether mv’s width may be changed to new_width using mv_resize. Returns true if it is allowed, false otherwise.

If mv contains any missing values, then it may be resized only if each missing value may be resized, as determined by value_is_resizable (see value_is_resizable).

Function: void mv_resize (struct missing_values *mv, int width)

Changes mv’s width to width. mv and width must satisfy the constraints explained above.

When a string missing value set’s width is increased, each user-missing value is padded on the right with spaces to the new width.

2.3.4 Inspecting User-Missing Value Sets

These functions inspect the properties and contents of struct missing_values objects.

The first set of functions inspects the discrete values that sets of user-missing values may contain:

Function: bool mv_is_empty (const struct missing_values *mv)

Returns true if mv contains no user-missing values, false if it contains at least one user-missing value (either a discrete value or a numeric range).

Function: int mv_get_width (const struct missing_values *mv)

Returns the width of the user-missing values that mv represents.

Function: int mv_n_values (const struct missing_values *mv)

Returns the number of discrete user-missing values included in mv. The return value will be between 0 and 3. For sets of numeric user-missing values that include a range, the return value will be 0 or 1.

Function: bool mv_has_value (const struct missing_values *mv)

Returns true if mv has at least one discrete user-missing values, that is, if mv_n_values would return nonzero for mv.

Function: const union value * mv_get_value (const struct missing_values *mv, int index)

Returns the discrete user-missing value in mv with the given index. The caller must not modify or free the returned value or refer to it after modifying or freeing mv. The index must be less than the number of discrete user-missing values in mv, as reported by mv_n_values.

The second set of functions inspects the single range of values that numeric sets of user-missing values may contain:

Function: bool mv_has_range (const struct missing_values *mv)

Returns true if mv includes a range, false otherwise.

Function: void mv_get_range (const struct missing_values *mv, double *low, double *high)

Stores the low endpoint of mv’s range in *low and the high endpoint of the range in *high. mv must include a range.

2.3.5 Modifying User-Missing Value Sets

These functions modify the contents of struct missing_values objects.

The next set of functions applies to all sets of user-missing values:

Function: bool mv_add_value (struct missing_values *mv, const union value *value)

Function: bool mv_add_str (struct missing_values *mv, const char value[])

Function: bool mv_add_num (struct missing_values *mv, double value)

Attempts to add the given discrete value to set of user-missing values mv. value must have the same width as mv. Returns true if value was successfully added, false if the set could not accept any more discrete values or if value is not an acceptable user-missing value (see mv_is_acceptable below).

These functions are equivalent, except for the form in which value is provided, so you may use whichever function is most convenient.

Function: void mv_pop_value (struct missing_values *mv, union value *value)

Removes a discrete value from mv (which must contain at least one discrete value) and stores it in value.

Function: bool mv_replace_value (struct missing_values *mv, const union value *value, int index)

Attempts to replace the discrete value with the given index in mv (which must contain at least index + 1 discrete values) by value. Returns true if successful, false if value is not an acceptable user-missing value (see mv_is_acceptable below).

Function: bool mv_is_acceptable (const union value *value, int width)

Returns true if value, which must have the specified width, may be added to a missing value set of the same width, false if it cannot. As described above, all numeric values and string values of width MV_MAX_STRING or less may be added, but string value of greater width may be added only if bytes beyond the first MV_MAX_STRING are all spaces.

The second set of functions applies only to numeric sets of user-missing values:

Function: bool mv_add_range (struct missing_values *mv, double low, double high)

Attempts to add a numeric range covering low…high (inclusive on both ends) to mv, which must be a numeric set of user-missing values. Returns true if the range is successful added, false on failure. Fails if mv already contains a range, or if mv contains more than one discrete value, or if low > high.

Function: void mv_pop_range (struct missing_values *mv, double *low, double *high)

Given mv, which must be a numeric set of user-missing values that contains a range, removes that range from mv and stores its low endpoint in *low and its high endpoint in *high.

2.4 Value Labels

Each variable has a set of value labels (see VALUE LABELS in PSPP Users Guide), represented as struct val_labs. A struct val_labs is essentially a map from union values to strings. All of the values in a set of value labels have the same width, which for a set of value labels owned by a variable (the common case) is the same as its variable.

Sets of value labels may contain any number of entries.

It is rarely necessary to interact directly with a struct val_labs object. Instead, the most common operation, looking up the label for a value of a given variable, can be conveniently executed through functions on struct variable. See Variable Value Labels, for details.

Function prototypes and other declarations related to missing values are declared in data/value-labels.h.

Structure: struct val_labs

Opaque type that represents a set of value labels.

The most often useful function for value labels is val_labs_find, for looking up the label associated with a value.

Function: char * val_labs_find (const struct val_labs *val_labs, union value value)

Looks in val_labs for a label for the given value. Returns the label, if one is found, or a null pointer otherwise.

Several other functions for working with value labels are described in the following section, but these are more rarely useful.

2.4.1 Creation and Destruction

These functions create and destroy struct val_labs objects.

Function: struct val_labs * val_labs_create (int width)

Creates and returns an initially empty set of value labels with the given width.

Function: struct val_labs * val_labs_clone (const struct val_labs *val_labs)

Creates and returns a set of value labels whose width and contents are the same as those of var_labs.

Function: void val_labs_clear (struct val_labs *var_labs)

Deletes all value labels from var_labs.

Function: void val_labs_destroy (struct val_labs *var_labs)

Destroys var_labs, which must not be referenced again.

2.4.2 Value Labels Properties

These functions inspect and manipulate basic properties of struct val_labs objects.

Function: size_t val_labs_count (const struct val_labs *val_labs)

Returns the number of value labels in val_labs.

Function: bool val_labs_can_set_width (const struct val_labs *val_labs, int new_width)

Tests whether val_labs’s width may be changed to new_width using val_labs_set_width. Returns true if it is allowed, false otherwise.

A set of value labels may be resized to a given width only if each value in it may be resized to that width, as determined by value_is_resizable (see value_is_resizable).

Function: void val_labs_set_width (struct val_labs *val_labs, int new_width)

Changes the width of val_labs’s values to new_width, which must be a valid new width as determined by val_labs_can_set_width.

2.4.3 Adding and Removing Labels

These functions add and remove value labels from a struct val_labs object.

Function: bool val_labs_add (struct val_labs *val_labs, union value value, const char *label)

Adds label to in var_labs as a label for value, which must have the same width as the set of value labels. Returns true if successful, false if value already has a label.

Function: void val_labs_replace (struct val_labs *val_labs, union value value, const char *label)

Adds label to in var_labs as a label for value, which must have the same width as the set of value labels. If value already has a label in var_labs, it is replaced.

Function: bool val_labs_remove (struct val_labs *val_labs, union value value)

Removes from val_labs any label for value, which must have the same width as the set of value labels. Returns true if a label was removed, false otherwise.

2.4.4 Iterating through Value Labels

These functions allow iteration through the set of value labels represented by a struct val_labs object. They may be used in the context of a for loop:

struct val_labs val_labs;
const struct val_lab *vl;

…

for (vl = val_labs_first (val_labs); vl != NULL;
     vl = val_labs_next (val_labs, vl))
  {
    …do something with vl…
  }

Value labels should not be added or deleted from a struct val_labs as it is undergoing iteration.

Function: const struct val_lab * val_labs_first (const struct val_labs *val_labs)

Returns the first value label in var_labs, if it contains at least one value label, or a null pointer if it does not contain any value labels.

Function: const struct val_lab * val_labs_next (const struct val_labs *val_labs, const struct val_labs_iterator **vl)

Returns the value label in var_labs following vl, if vl is not the last value label in val_labs, or a null pointer if there are no value labels following vl.

Function: const struct val_lab ** val_labs_sorted (const struct val_labs *val_labs)

Allocates and returns an array of pointers to value labels, which are sorted in increasing order by value. The array has val_labs_count (val_labs) elements. The caller is responsible for freeing the array with free (but must not free any of the struct val_lab elements that the array points to).

The iteration functions above work with pointers to struct val_lab which is an opaque data structure that users of struct val_labs must not modify or free directly. The following functions work with objects of this type:

Function: const union value * val_lab_get_value (const struct val_lab *vl)

Returns the value of value label vl. The caller must not modify or free the returned value. (To achieve a similar result, remove the value label with val_labs_remove, then add the new value with val_labs_add.)

The width of the returned value cannot be determined directly from vl. It may be obtained by calling val_labs_get_width on the struct val_labs that vl is in.

Function: const char * val_lab_get_label (const struct val_lab *vl)

Returns the label in vl as a null-terminated string. The caller must not modify or free the returned string. (Use val_labs_replace to change a value label.)

2.5 Variables

A PSPP variable is represented by struct variable, an opaque type declared in data/variable.h along with related declarations. See Variables in PSPP Users Guide, for a description of PSPP variables from a user perspective.

PSPP is unusual among computer languages in that, by itself, a PSPP variable does not have a value. Instead, a variable in PSPP takes on a value only in the context of a case, which supplies one value for each variable in a set of variables (see Cases). The set of variables in a case, in turn, are ordinarily part of a dictionary (see Dictionaries).

Every variable has several attributes, most of which correspond directly to one of the variable attributes visible to PSPP users (see Attributes in PSPP Users Guide).

The following sections describe variable-related functions and macros.

2.5.1 Variable Name

A variable name is a string between 1 and ID_MAX_LEN bytes long that satisfies the rules for PSPP identifiers (see Tokens in PSPP Users Guide). Variable names are mixed-case and treated case-insensitively.

Macro: int ID_MAX_LEN

Maximum length of a variable name, in bytes, currently 64.

Only one commonly useful function relates to variable names:

Function: const char * var_get_name (const struct variable *var)

Returns var’s variable name as a C string.

A few other functions are much more rarely used. Some of these functions are used internally by the dictionary implementation:

Function: void var_set_name (struct variable *var, const char *new_name)

Changes the name of var to new_name, which must be a “plausible” name as defined below.

This function cannot be applied to a variable that is part of a dictionary. Use dict_rename_var instead (see Dictionary Renaming Variables).

Function: enum dict_class var_get_dict_class (const struct variable *var)

Returns the dictionary class of var’s name (see Dictionary Class).

2.5.2 Variable Type and Width

A variable’s type and width are the type and width of its values (see Values).

Function: enum val_type var_get_type (const struct variable *var)

Returns the type of variable var.

Function: int var_get_width (const struct variable *var)

Returns the width of variable var.

Function: void var_set_width (struct variable *var, int width)

Sets the width of variable var to width. The width of a variable should not normally be changed after the variable is created, so this function is rarely used. This function cannot be applied to a variable that is part of a dictionary.

Function: bool var_is_numeric (const struct variable *var)

Returns true if var is a numeric variable, false otherwise.

Function: bool var_is_alpha (const struct variable *var)

Returns true if var is an alphanumeric (string) variable, false otherwise.

2.5.3 Variable Missing Values

A numeric or short string variable may have a set of user-missing values (see MISSING VALUES in PSPP Users Guide), represented as a struct missing_values (see User-Missing Values).

The most frequent operation on a variable’s missing values is to query whether a value is user- or system-missing:

Function: bool var_is_value_missing (const struct variable *var, const union value *value, enum mv_class class)

Function: bool var_is_num_missing (const struct variable *var, double value, enum mv_class class)

Function: bool var_is_str_missing (const struct variable *var, const char value[], enum mv_class class)

Tests whether value is a missing value of the given class for variable var and returns true if so, false otherwise. var_is_num_missing may only be applied to numeric variables; var_is_str_missing may only be applied to string variables. value must have been initialized with the same width as var.

var_is_type_missing (var, value, class) is equivalent to mv_is_type_missing (var_get_missing_values (var), value, class).

In addition, a few functions are provided to work more directly with a variable’s struct missing_values:

Function: const struct missing_values * var_get_missing_values (const struct variable *var)

Returns the struct missing_values associated with var. The caller must not modify the returned structure. The return value is always non-null.

Function: void var_set_missing_values (struct variable *var, const struct missing_values *miss)

Changes var’s missing values to a copy of miss, or if miss is a null pointer, clears var’s missing values. If miss is non-null, it must have the same width as var or be resizable to var’s width (see mv_resize). The caller retains ownership of miss.

Function: void var_clear_missing_values (struct variable *var)

Clears var’s missing values. Equivalent to var_set_missing_values (var, NULL).

Function: bool var_has_missing_values (const struct variable *var)

Returns true if var has any missing values, false if it has none. Equivalent to mv_is_empty (var_get_missing_values (var)).

2.5.4 Variable Value Labels

A numeric or short string variable may have a set of value labels (see VALUE LABELS in PSPP Users Guide), represented as a struct val_labs (see Value Labels). The most commonly useful functions for value labels return the value label associated with a value:

Function: const char * var_lookup_value_label (const struct variable *var, const union value *value)

Looks for a label for value in var’s set of value labels. value must have the same width as var. Returns the label if one exists, otherwise a null pointer.

Function: void var_append_value_name (const struct variable *var, const union value *value, struct string *str)

Looks for a label for value in var’s set of value labels. value must have the same width as var. If a label exists, it will be appended to the string pointed to by str. Otherwise, it formats value using var’s print format (see Input and Output Formats) and appends the formatted string.

The underlying struct val_labs structure may also be accessed directly using the functions described below.

Function: bool var_has_value_labels (const struct variable *var)

Returns true if var has at least one value label, false otherwise.

Function: const struct val_labs * var_get_value_labels (const struct variable *var)

Returns the struct val_labs associated with var. If var has no value labels, then the return value may or may not be a null pointer.

The variable retains ownership of the returned struct val_labs, which the caller must not attempt to modify.

Function: void var_set_value_labels (struct variable *var, const struct val_labs *val_labs)

Replaces var’s value labels by a copy of val_labs. The caller retains ownership of val_labs. If val_labs is a null pointer, then var’s value labels, if any, are deleted.

Function: void var_clear_value_labels (struct variable *var)

Deletes var’s value labels. Equivalent to var_set_value_labels (var, NULL).

A final group of functions offers shorthands for operations that would otherwise require getting the value labels from a variable, copying them, modifying them, and then setting the modified value labels into the variable (making a second copy):

Function: bool var_add_value_label (struct variable *var, const union value *value, const char *label)

Attempts to add a copy of label as a label for value for the given var. value must have the same width as var. If value already has a label, then the old label is retained. Returns true if a label is added, false if there was an existing label for value. Either way, the caller retains ownership of value and label.

Function: void var_replace_value_label (struct variable *var, const union value *value, const char *label)

Attempts to add a copy of label as a label for value for the given var. value must have the same width as var. If value already has a label, then label replaces the old label. Either way, the caller retains ownership of value and label.

2.5.5 Variable Print and Write Formats

Each variable has an associated pair of output formats, called its print format and write format. See Input and Output Formats in PSPP Users Guide, for an introduction to formats. See Input and Output Formats, for a developer’s description of format representation.

The print format is used to convert a variable’s data values to strings for human-readable output. The write format is used similarly for machine-readable output, primarily by the WRITE transformation (see WRITE in PSPP Users Guide). Most often a variable’s print and write formats are the same.

A newly created variable by default has format F8.2 if it is numeric or an A format with the same width as the variable if it is string. Many creators of variables override these defaults.

Both the print format and write format are output formats. Input formats are not part of struct variable. Instead, input programs and transformations keep track of variable input formats themselves.

The following functions work with variable print and write formats.

Function: const struct fmt_spec * var_get_print_format (const struct variable *var)

Function: const struct fmt_spec * var_get_write_format (const struct variable *var)

Returns var’s print or write format, respectively.

Function: void var_set_print_format (struct variable *var, const struct fmt_spec *format)

Function: void var_set_write_format (struct variable *var, const struct fmt_spec *format)

Function: void var_set_both_formats (struct variable *var, const struct fmt_spec *format)

Sets var’s print format, write format, or both formats, respectively, to a copy of format.

2.5.6 Variable Labels

A variable label is a string that describes a variable. Variable labels may contain spaces and punctuation not allowed in variable names. See VARIABLE LABELS in PSPP Users Guide, for a user-level description of variable labels.

The most commonly useful functions for variable labels are those to retrieve a variable’s label:

Function: const char * var_to_string (const struct variable *var)

Returns var’s variable label, if it has one, otherwise var’s name. In either case the caller must not attempt to modify or free the returned string.

This function is useful for user output.

Function: const char * var_get_label (const struct variable *var)

Returns var’s variable label, if it has one, or a null pointer otherwise.

A few other variable label functions are also provided:

Function: void var_set_label (struct variable *var, const char *label)

Sets var’s variable label to a copy of label, or removes any label from var if label is a null pointer or contains only spaces. Leading and trailing spaces are removed from the variable label and its remaining content is truncated at 255 bytes.

Function: void var_clear_label (struct variable *var)

Removes any variable label from var.

Function: bool var_has_label (const struct variable *var)

Returns true if var has a variable label, false otherwise.

2.5.7 GUI Attributes

These functions and types access and set attributes that are mainly used by graphical user interfaces. Their values are also stored in and retrieved from system files (but not portable files).

The first group of functions relate to the measurement level of numeric data. New variables are assigned a nominal level of measurement by default.

Enumeration: enum measure

Measurement level. Available values are:

MEASURE_NOMINAL

Numeric data values are arbitrary. Arithmetic operations and numerical comparisons of such data are not meaningful.

MEASURE_ORDINAL

Numeric data values indicate progression along a rank order. Arbitrary arithmetic operations such as addition are not meaningful on such data, but inequality comparisons (less, greater, etc.) have straightforward interpretations.

MEASURE_SCALE

Ratios, sums, etc. of numeric data values have meaningful interpretations.

PSPP does not have a separate category for interval data, which would naturally fall between the ordinal and scale measurement levels.

Function: bool measure_is_valid (enum measure measure)

Returns true if measure is a valid level of measurement, that is, if it is one of the enum measure constants listed above, and false otherwise.

Function: enum measure var_get_measure (const struct variable *var)

Function: void var_set_measure (struct variable *var, enum measure measure)

Gets or sets var’s measurement level.

The following set of functions relates to the width of on-screen columns used for displaying variable data in a graphical user interface environment. The unit of measurement is the width of a character. For proportionally spaced fonts, this is based on the average width of a character.

Function: int var_get_display_width (const struct variable *var)

Function: void var_set_display_width (struct variable *var, int display_width)

Gets or sets var’s display width.

Function: int var_default_display_width (int width)

Returns the default display width for a variable with the given width. The default width of a numeric variable is 8. The default width of a string variable is width or 32, whichever is less.

The final group of functions work with the justification of data when it is displayed in on-screen columns. New variables are by default right-justified.

Enumeration: enum alignment

Text justification. Possible values are ALIGN_LEFT, ALIGN_RIGHT, and ALIGN_CENTRE.

Function: bool alignment_is_valid (enum alignment alignment)

Returns true if alignment is a valid alignment, that is, if it is one of the enum alignment constants listed above, and false otherwise.

Function: enum alignment var_get_alignment (const struct variable *var)

Function: void var_set_alignment (struct variable *var, enum alignment alignment)

Gets or sets var’s alignment.

2.5.8 Variable Leave Status

Commonly, most or all data in a case come from an input file, read with a command such as DATA LIST or GET, but data can also be generated with transformations such as COMPUTE. In the latter case the question of a datum’s “initial value” can arise. For example, the value of a piece of generated data can recursively depend on its own value:

COMPUTE X = X + 1.

Another situation where the initial value of a variable arises is when its value is not set at all for some cases, e.g. below, Y is set only for the first 10 cases:

DO IF #CASENUM <= 10.
+ COMPUTE Y = 1.
END IF.

By default, the initial value of a datum in either of these situations is the system-missing value for numeric values and spaces for string values. This means that, above, X would be system-missing and that Y would be 1 for the first 10 cases and system-missing for the remainder.

PSPP also supports retaining the value of a variable from one case to another, using the LEAVE command (see LEAVE in PSPP Users Guide). The initial value of such a variable is 0 if it is numeric and spaces if it is a string. If the command ‘LEAVE X Y’ is appended to the above example, then X would have value 1 in the first case and increase by 1 in every succeeding case, and Y would have value 1 for the first 10 cases and 0 for later cases.

The LEAVE command has no effect on data that comes from an input file or whose values do not depend on a variable’s initial value.

The value of scratch variables (see Scratch Variables in PSPP Users Guide) are always left from one case to another.

The following functions work with a variable’s leave status.

Function: bool var_get_leave (const struct variable *var)

Returns true if var’s value is to be retained from case to case, false if it is reinitialized to system-missing or spaces.

Function: void var_set_leave (struct variable *var, bool leave)

If leave is true, marks var to be left from case to case; if leave is false, marks var to be reinitialized for each case.

If var is a scratch variable, leave must be true.

Function: bool var_must_leave (const struct variable *var)

Returns true if var must be left from case to case, that is, if var is a scratch variable.

2.5.9 Dictionary Class

Occasionally it is useful to classify variables into dictionary classes based on their names. Dictionary classes are represented by enum dict_class. This type and other declarations for dictionary classes are in the <data/dict-class.h> header.

Enumeration: enum dict_class

The dictionary classes are:

DC_ORDINARY

An ordinary variable, one whose name does not begin with ‘$’ or ‘#’.

DC_SYSTEM

A system variable, one whose name begins with ‘$’. See System Variables in PSPP Users Guide.

DC_SCRATCH

A scratch variable, one whose name begins with ‘#’. See Scratch Variables in PSPP Users Guide.

The values for dictionary classes are bitwise disjoint, which allows them to be used in bit-masks. An extra enumeration constant DC_ALL, whose value is the bitwise-or of all of the above constants, is provided to aid in this purpose.

One example use of dictionary classes arises in connection with PSPP syntax that uses a TO b to name the variables in a dictionary from a to b (see Sets of Variables in PSPP Users Guide). This syntax requires a and b to be in the same dictionary class. It limits the variables that it includes to those in that dictionary class.

The following functions relate to dictionary classes.

Function: enum dict_class dict_class_from_id (const char *name)

Returns the “dictionary class” for the given variable name, by looking at its first letter.

Function: const char * dict_class_to_name (enum dict_class dict_class)

Returns a name for the given dict_class as an adjective, e.g. "scratch".

This function should probably not be used in new code as it can lead to difficulties for internationalization.

2.5.10 Variable Creation and Destruction

Only rarely should PSPP code create or destroy variables directly. Ordinarily, variables are created within a dictionary and destroying by individual deletion from the dictionary or by destroying the entire dictionary at once. The functions here enable the exceptional case, of creation and destruction of variables that are not associated with any dictionary. These functions are used internally in the dictionary implementation.

Function: struct variable * var_create (const char *name, int width)

Creates and returns a new variable with the given name and width. The new variable is not part of any dictionary. Use dict_create_var, instead, to create a variable in a dictionary (see Dictionary Creating Variables).

name should be a valid variable name and must be a “plausible” variable name (see Variable Name). width must be between 0 and MAX_STRING, inclusive (see Values).

The new variable has no user-missing values, value labels, or variable label. Numeric variables initially have F8.2 print and write formats, right-justified display alignment, and scale level of measurement. String variables are created with A print and write formats, left-justified display alignment, and nominal level of measurement. The initial display width is determined by var_default_display_width (see var_default_display_width).

The new variable initially has no short name (see Variable Short Names) and no auxiliary data (see Variable Auxiliary Data).

Function: struct variable * var_clone (const struct variable *old_var)

Creates and returns a new variable with the same attributes as old_var, with a few exceptions. First, the new variable is not part of any dictionary, regardless of whether old_var was in a dictionary. Use dict_clone_var, instead, to add a clone of a variable to a dictionary.

Second, the new variable is not given any short name, even if old_var had a short name. This is because the new variable is likely to be immediately renamed, in which case the short name would be incorrect (see Variable Short Names).

Finally, old_var’s auxiliary data, if any, is not copied to the new variable (see Variable Auxiliary Data).

Function: void var_destroy (struct variable *var)

Destroys var and frees all associated storage, including its auxiliary data, if any. var must not be part of a dictionary. To delete a variable from a dictionary and destroy it, use dict_delete_var (see Dictionary Deleting Variables).

2.5.11 Variable Short Names

PSPP variable names may be up to 64 (ID_MAX_LEN) bytes long. The system and portable file formats, however, were designed when variable names were limited to 8 bytes in length. Since then, the system file format has been augmented with an extension record that explains how the 8-byte short names map to full-length names (see Long Variable Names Record), but the short names are still present. Thus, the continued presence of the short names is more or less invisible to PSPP users, but every variable in a system file still has a short name that must be unique.

PSPP can generate unique short names for variables based on their full names at the time it creates the data file. If all variables’ full names are unique in their first 8 bytes, then the short names are simply prefixes of the full names; otherwise, PSPP changes them so that they are unique.

By itself this algorithm interoperates well with other software that can read system files, as long as that software understands the extension record that maps short names to long names. When the other software does not understand the extension record, it can produce surprising results. Consider a situation where PSPP reads a system file that contains two variables named RANKINGSCORE, then the user adds a new variable named RANKINGSTATUS, then saves the modified data as a new system file. A program that does not understand long names would then see one of these variables under the name RANKINGS—either one, depending on the algorithm’s details—and the other under a different name. The effect could be very confusing: by adding a new and apparently unrelated variable in PSPP, the user effectively renamed the existing variable.

To counteract this potential problem, every struct variable may have a short name. A variable created by the system or portable file reader receives the short name from that data file. When a variable with a short name is written to a system or portable file, that variable receives priority over other long names whose names begin with the same 8 bytes but which were not read from a data file under that short name.

Variables not created by the system or portable file reader have no short name by default.

A variable with a full name of 8 bytes or less in length has absolute priority for that name when the variable is written to a system file, even over a second variable with that assigned short name.

PSPP does not enforce uniqueness of short names, although the short names read from any given data file will always be unique. If two variables with the same short name are written to a single data file, neither one receives priority.

The following macros and functions relate to short names.

Macro: SHORT_NAME_LEN

Maximum length of a short name, in bytes. Its value is 8.

Function: const char * var_get_short_name (const struct variable *var)

Returns var’s short name, or a null pointer if var has not been assigned a short name.

Function: void var_set_short_name (struct variable *var, const char *short_name)

Sets var’s short name to short_name, or removes var’s short name if short_name is a null pointer. If it is non-null, then short_name must be a plausible name for a variable. The name will be truncated to 8 bytes in length and converted to all-uppercase.

Function: void var_clear_short_name (struct variable *var)

Removes var’s short name.

2.5.12 Variable Relationships

Variables have close relationships with dictionaries (see Dictionaries) and cases (see Cases). A variable is usually a member of some dictionary, and a case is often used to store data for the set of variables in a dictionary.

These functions report on these relationships. They may be applied only to variables that are in a dictionary.

Function: size_t var_get_dict_index (const struct variable *var)

Returns var’s index within its dictionary. The first variable in a dictionary has index 0, the next variable index 1, and so on.

The dictionary index can be influenced using dictionary functions such as dict_reorder_var (see dict_reorder_var).

Function: size_t var_get_case_index (const struct variable *var)

Returns var’s index within a case. The case index is an index into an array of union value large enough to contain all the data in the dictionary.

The returned case index can be used to access the value of var within a case for its dictionary, as in e.g. case_data_idx (case, var_get_case_index (var)), but ordinarily it is more convenient to use the data access functions that do variable-to-index translation internally, as in e.g. case_data (case, var).

2.5.13 Variable Auxiliary Data

Each struct variable can have a single pointer to auxiliary data of type void *. These functions manipulate a variable’s auxiliary data.

Use of auxiliary data is discouraged because of its lack of flexibility. Only one client can make use of auxiliary data on a given variable at any time, even though many clients could usefully associate data with a variable.

To prevent multiple clients from attempting to use a variable’s single auxiliary data field at the same time, we adopt the convention that use of auxiliary data in the active dataset dictionary is restricted to the currently executing command. In particular, transformations must not attach auxiliary data to a variable in the active dataset in the expectation that it can be used later when the active dataset is read and the transformation is executed. To help enforce this restriction, auxiliary data is deleted from all variables in the active dataset dictionary after the execution of each PSPP command.

This convention for safe use of auxiliary data applies only to the active dataset dictionary. Rules for other dictionaries may be established separately.

Auxiliary data should be replaced by a more flexible mechanism at some point, but no replacement mechanism has been designed or implemented so far.

The following functions work with variable auxiliary data.

Function: void * var_get_aux (const struct variable *var)

Returns var’s auxiliary data, or a null pointer if none has been assigned.

Function: void * var_attach_aux (const struct variable *var, void *aux, void (*aux_dtor) (struct variable *))

Sets var’s auxiliary data to aux, which must not be null. var must not already have auxiliary data.

Before var’s auxiliary data is cleared by var_clear_aux, aux_dtor, if non-null, will be called with var as its argument. It should free any storage associated with aux, if necessary. var_dtor_free may be appropriate for use as aux_dtor:

Function: void var_dtor_free (struct variable *var)

Frees var’s auxiliary data by calling free.

Function: void var_clear_aux (struct variable *var)

Removes auxiliary data, if any, from var, first calling the destructor passed to var_attach_aux, if one was provided.

Use dict_clear_aux to remove auxiliary data from every variable in a dictionary.

Function: void * var_detach_aux (struct variable *var)

Removes auxiliary data, if any, from var, and returns it. Returns a null pointer if var had no auxiliary data.

Any destructor passed to var_attach_aux is not called, so the caller is responsible for freeing storage associated with the returned auxiliary data.

2.5.14 Variable Categorical Values

Some statistical procedures require a list of all the values that a categorical variable takes on. Arranging such a list requires making a pass through the data, so PSPP caches categorical values in struct variable.

When variable auxiliary data is revamped to support multiple clients as described in the previous section, categorical values are an obvious candidate. The form in which they are currently supported is inelegant.

Categorical values are not robust against changes in the data. That is, there is currently no way to detect that a transformation has changed data values, meaning that categorical values lists for the changed variables must be recomputed. PSPP is in fact in need of a general-purpose caching and cache-invalidation mechanism, but none has yet been designed and built.

The following functions work with cached categorical values.

Function: struct cat_vals * var_get_obs_vals (const struct variable *var)

Returns var’s set of categorical values. Yields undefined behavior if var does not have any categorical values.

Function: void var_set_obs_vals (const struct variable *var, struct cat_vals *cat_vals)

Destroys var’s categorical values, if any, and replaces them by cat_vals, ownership of which is transferred to var. If cat_vals is a null pointer, then var’s categorical values are cleared.

Function: bool var_has_obs_vals (const struct variable *var)

Returns true if var has a set of categorical values, false otherwise.

2.6 Dictionaries

Each data file in memory or on disk has an associated dictionary, whose primary purpose is to describe the data in the file. See Variables in PSPP Users Guide, for a PSPP user’s view of a dictionary.

A data file stored in a PSPP format, either as a system or portable file, has a representation of its dictionary embedded in it. Other kinds of data files are usually not self-describing enough to construct a dictionary unassisted, so the dictionaries for these files must be specified explicitly with PSPP commands such as DATA LIST.

The most important content of a dictionary is an array of variables, which must have unique names. A dictionary also conceptually contains a mapping from each of its variables to a location within a case (see Cases), although in fact these mappings are stored within individual variables.

System variables are not members of any dictionary (see System Variables in PSPP Users Guide).

Dictionaries are represented by struct dictionary. Declarations related to dictionaries are in the <data/dictionary.h> header.

The following sections describe functions for use with dictionaries.

2.6.1 Accessing Variables

The most common operations on a dictionary simply retrieve a struct variable * of an individual variable based on its name or position.

Function: struct variable * dict_lookup_var (const struct dictionary *dict, const char *name)

Function: struct variable * dict_lookup_var_assert (const struct dictionary *dict, const char *name)

Looks up and returns the variable with the given name within dict. Name lookup is not case-sensitive.

dict_lookup_var returns a null pointer if dict does not contain a variable named name. dict_lookup_var_assert asserts that such a variable exists.

Function: struct variable * dict_get_var (const struct dictionary *dict, size_t position)

Returns the variable at the given position in dict. position must be less than the number of variables in dict (see below).

Function: size_t dict_get_var_cnt (const struct dictionary *dict)

Returns the number of variables in dict.

Another pair of functions allows retrieving a number of variables at once. These functions are more rarely useful.

Function: void dict_get_vars (const struct dictionary *dict, const struct variable ***vars, size_t *cnt, enum dict_class exclude)

Function: void dict_get_vars_mutable (const struct dictionary *dict, struct variable ***vars, size_t *cnt, enum dict_class exclude)

Retrieves all of the variables in dict, in their original order, except that any variables in the dictionary classes specified exclude, if any, are excluded (see Dictionary Class). Pointers to the variables are stored in an array allocated with malloc, and a pointer to the first element of this array is stored in *vars. The caller is responsible for freeing this memory when it is no longer needed. The number of variables retrieved is stored in *cnt.

The presence or absence of DC_SYSTEM in exclude has no effect, because dictionaries never include system variables.

One additional function is available. This function is most often used in assertions, but it is not restricted to such use.

Function: bool dict_contains_var (const struct dictionary *dict, const struct variable *var)

Tests whether var is one of the variables in dict. Returns true if so, false otherwise.

2.6.2 Creating Variables

These functions create a new variable and insert it into a dictionary in a single step.

There is no provision for inserting an already created variable into a dictionary. There is no reason that such a function could not be written, but so far there has been no need for one.

The names provided to one of these functions should be valid variable names and must be plausible variable names.

If a variable with the same name already exists in the dictionary, the non-assert variants of these functions return a null pointer, without modifying the dictionary. The assert variants, on the other hand, assert that no duplicate name exists.

A variable may be in only one dictionary at any given time.

Function: struct variable * dict_create_var (struct dictionary *dict, const char *name, int width)

Function: struct variable * dict_create_var_assert (struct dictionary *dict, const char *name, int width)

Creates a new variable with the given name and width, as if through a call to var_create with those arguments (see var_create), appends the new variable to dict’s array of variables, and returns the new variable.

Function: struct variable * dict_clone_var (struct dictionary *dict, const struct variable *old_var)

Function: struct variable * dict_clone_var_assert (struct dictionary *dict, const struct variable *old_var)

Creates a new variable as a clone of var, inserts the new variable into dict, and returns the new variable. Other properties of the new variable are copied from old_var, except for those not copied by var_clone (see var_clone).

var does not need to be a member of any dictionary.

Function: struct variable * dict_clone_var_as (struct dictionary *dict, const struct variable *old_var, const char *name)

Function: struct variable * dict_clone_var_as_assert (struct dictionary *dict, const struct variable *old_var, const char *name)

These functions are similar to dict_clone_var and dict_clone_var_assert, respectively, except that the new variable is named name instead of keeping old_var’s name.

2.6.3 Deleting Variables

These functions remove variables from a dictionary’s array of variables. They also destroy the removed variables and free their associated storage.

Deleting a variable to which there might be external pointers is a bad idea. In particular, deleting variables from the active dataset dictionary is a risky proposition, because transformations can retain references to arbitrary variables. Therefore, no variable should be deleted from the active dataset dictionary when any transformations are active, because those transformations might reference the variable to be deleted. The safest time to delete a variable is just after a procedure has been executed, as done by DELETE VARIABLES.

Deleting a variable automatically removes references to that variable from elsewhere in the dictionary as a weighting variable, filter variable, SPLIT FILE variable, or member of a vector.

No functions are provided for removing a variable from a dictionary without destroying that variable. As with insertion of an existing variable, there is no reason that this could not be implemented, but so far there has been no need.

Function: void dict_delete_var (struct dictionary *dict, struct variable *var)

Deletes var from dict, of which it must be a member.

Function: void dict_delete_vars (struct dictionary *dict, struct variable *const *vars, size_t count)

Deletes the count variables in array vars from dict. All of the variables in vars must be members of dict. No variable may be included in vars more than once.

Function: void dict_delete_consecutive_vars (struct dictionary *dict, size_t idx, size_t count)

Deletes the variables in sequential positions idx…idx + count (exclusive) from dict, which must contain at least idx + count variables.

Function: void dict_delete_scratch_vars (struct dictionary *dict)

Deletes all scratch variables from dict.

2.6.4 Changing Variable Order

The variables in a dictionary are stored in an array. These functions change the order of a dictionary’s array of variables without changing which variables are in the dictionary.

Function: void dict_reorder_var (struct dictionary *dict, struct variable *var, size_t new_index)

Moves var, which must be in dict, so that it is at position new_index in dict’s array of variables. Other variables in dict, if any, retain their relative positions. new_index must be less than the number of variables in dict.

Function: void dict_reorder_vars (struct dictionary *dict, struct variable *const *new_order, size_t count)

Moves the count variables in new_order to the beginning of dict’s array of variables in the specified order. Other variables in dict, if any, retain their relative positions.

All of the variables in new_order must be in dict. No duplicates are allowed within new_order, which means that count must be no greater than the number of variables in dict.

2.6.5 Renaming Variables

These functions change the names of variables within a dictionary. The var_set_name function (see var_set_name) cannot be applied directly to a variable that is in a dictionary, because struct dictionary contains an index by name that var_set_name would not update. The following functions take care to update the index as well. They also ensure that variable renaming does not cause a dictionary to contain a duplicate variable name.

Function: void dict_rename_var (struct dictionary *dict, struct variable *var, const char *new_name)

Changes the name of var, which must be in dict, to new_name. A variable named new_name must not already be in dict, unless new_name is the same as var’s current name.

Function: bool dict_rename_vars (struct dictionary *dicT, struct variable **vars, char **new_names, size_t count, char **err_name)

Renames each of the count variables in vars to the name in the corresponding position of new_names. If the renaming would result in a duplicate variable name, returns false and stores one of the names that would be be duplicated into *err_name, if err_name is non-null. Otherwise, the renaming is successful, and true is returned.

2.6.6 Weight Variable

A data set’s cases may optionally be weighted by the value of a numeric variable. See WEIGHT in PSPP Users Guide, for a user view of weight variables.

The weight variable is written to and read from system and portable files.

The most commonly useful function related to weighting is a convenience function to retrieve a weighting value from a case.

Function: double dict_get_case_weight (const struct dictionary *dict, const struct ccase *case, bool *warn_on_invalid)

Retrieves and returns the value of the weighting variable specified by dict from case. Returns 1.0 if dict has no weighting variable.

Returns 0.0 if c’s weight value is user- or system-missing, zero, or negative. In such a case, if warn_on_invalid is non-null and *warn_on_invalid is true, dict_get_case_weight also issues an error message and sets *warn_on_invalid to false. To disable error reporting, pass a null pointer or a pointer to false as warn_on_invalid or use a msg_disable/msg_enable pair.

The dictionary also has a pair of functions for getting and setting the weight variable.

Function: struct variable * dict_get_weight (const struct dictionary *dict)

Returns dict’s current weighting variable, or a null pointer if the dictionary does not have a weighting variable.

Function: void dict_set_weight (struct dictionary *dict, struct variable *var)

Sets dict’s weighting variable to var. If var is non-null, it must be a numeric variable in dict. If var is null, then dict’s weighting variable, if any, is cleared.

2.6.7 Filter Variable

When the active dataset is read by a procedure, cases can be excluded from analysis based on the values of a filter variable. See FILTER in PSPP Users Guide, for a user view of filtering.

These functions store and retrieve the filter variable. They are rarely useful, because the data analysis framework automatically excludes from analysis the cases that should be filtered.

Function: struct variable * dict_get_filter (const struct dictionary *dict)

Returns dict’s current filter variable, or a null pointer if the dictionary does not have a filter variable.

Function: void dict_set_filter (struct dictionary *dict, struct variable *var)

Sets dict’s filter variable to var. If var is non-null, it must be a numeric variable in dict. If var is null, then dict’s filter variable, if any, is cleared.

2.6.8 Case Limit

The limit on cases analyzed by a procedure, set by the N OF CASES command (see N OF CASES in PSPP Users Guide), is stored as part of the dictionary. The dictionary does not, on the other hand, play any role in enforcing the case limit (a job done by data analysis framework code).

A case limit of 0 means that the number of cases is not limited.

These functions are rarely useful, because the data analysis framework automatically excludes from analysis any cases beyond the limit.

Function: casenumber dict_get_case_limit (const struct dictionary *dict)

Returns the current case limit for dict.

Function: void dict_set_case_limit (struct dictionary *dict, casenumber limit)

Sets dict’s case limit to limit.

2.6.9 Split Variables

The user may use the SPLIT FILE command (see SPLIT FILE in PSPP Users Guide) to select a set of variables on which to split the active dataset into groups of cases to be analyzed independently in each statistical procedure. The set of split variables is stored as part of the dictionary, although the effect on data analysis is implemented by each individual statistical procedure.

Split variables may be numeric or short or long string variables.

The most useful functions for split variables are those to retrieve them. Even these functions are rarely useful directly: for the purpose of breaking cases into groups based on the values of the split variables, it is usually easier to use casegrouper_create_splits.

Function: const struct variable *const * dict_get_split_vars (const struct dictionary *dict)

Returns a pointer to an array of pointers to split variables. If and only if there are no split variables, returns a null pointer. The caller must not modify or free the returned array.

Function: size_t dict_get_split_cnt (const struct dictionary *dict)

Returns the number of split variables.

The following functions are also available for working with split variables.

Function: void dict_set_split_vars (struct dictionary *dict, struct variable *const *vars, size_t cnt)

Sets dict’s split variables to the cnt variables in vars. If cnt is 0, then dict will not have any split variables. The caller retains ownership of vars.

Function: void dict_unset_split_var (struct dictionary *dict, struct variable *var)

Removes var, which must be a variable in dict, from dict’s split of split variables.

2.6.10 File Label

A dictionary may optionally have an associated string that describes its contents, called its file label. The user may set the file label with the FILE LABEL command (see FILE LABEL in PSPP Users Guide).

These functions set and retrieve the file label.

Function: const char * dict_get_label (const struct dictionary *dict)

Returns dict’s file label. If dict does not have a label, returns a null pointer.

Function: void dict_set_label (struct dictionary *dict, const char *label)

Sets dict’s label to label. If label is non-null, then its content, truncated to at most 60 bytes, becomes the new file label. If label is null, then dict’s label is removed.

The caller retains ownership of label.

2.6.11 Documents

A dictionary may include an arbitrary number of lines of explanatory text, called the dictionary’s documents. For compatibility, document lines have a fixed width, and lines that are not exactly this width are truncated or padded with spaces as necessary to bring them to the correct width.

PSPP users can use the DOCUMENT (see DOCUMENT in PSPP Users Guide), ADD DOCUMENT (see ADD DOCUMENT in PSPP Users Guide), and DROP DOCUMENTS (see DROP DOCUMENTS in PSPP Users Guide) commands to manipulate documents.

Macro: int DOC_LINE_LENGTH

The fixed length of a document line, in bytes, defined to 80.

The following functions work with whole sets of documents. They accept or return sets of documents formatted as null-terminated strings that are an exact multiple of DOC_LINE_LENGTH bytes in length.

Function: const char * dict_get_documents (const struct dictionary *dict)

Returns the documents in dict, or a null pointer if dict has no documents.

Function: void dict_set_documents (struct dictionary *dict, const char *new_documents)

Sets dict’s documents to new_documents. If new_documents is a null pointer or an empty string, then dict’s documents are cleared. The caller retains ownership of new_documents.

Function: void dict_clear_documents (struct dictionary *dict)

Clears the documents from dict.

The following functions work with individual lines in a dictionary’s set of documents.

Function: void dict_add_document_line (struct dictionary *dict, const char *content)

Appends content to the documents in dict. The text in content will be truncated or padded with spaces as necessary to make it exactly DOC_LINE_LENGTH bytes long. The caller retains ownership of content.

If content is over DOC_LINE_LENGTH, this function also issues a warning using msg. To suppress the warning, enclose a call to one of this function in a msg_disable/msg_enable pair.

Function: size_t dict_get_document_line_cnt (const struct dictionary *dict)

Returns the number of line of documents in dict. If the dictionary contains no documents, returns 0.

Function: void dict_get_document_line (const struct dictionary *dict, size_t idx, struct string *content)

Replaces the text in content (which must already have been initialized by the caller) by the document line in dict numbered idx, which must be less than the number of lines of documents in dict. Any trailing white space in the document line is trimmed, so that content will have a length between 0 and DOC_LINE_LENGTH.

2.7 Coding Conventions

Every .c file should have ‘#include <config.h>’ as its first non-comment line. No .h file should include config.h.

This section needs to be finished.

2.8 Cases

This section needs to be written.

2.9 Data Sets

This section needs to be written.

2.10 Pools

This section needs to be written.

3 Parsing Command Syntax

4 Processing Data

Developer’s Guide

Proposed outline:

* Introduction
* Basic concepts
** Data sets
** Variables
** Dictionaries
** Coding conventions
** Pools
* Syntax parsing
* Data processing
** Reading data
*** Casereaders generalities
*** Casereaders from data files
*** Casereaders from the active dataset
*** Other casereaders
** Writing data
*** Casewriters generally
*** Casewriters to data files
*** Modifying the active dataset
**** Modifying cases obtained from active dataset casereaders has no real effect
**** Transformations; procedures that transform
** Transforming data
*** Sorting and merging
*** Filtering
*** Grouping
**** Ordering and interaction of filtering and grouping
*** Multiple passes over data
*** Counting cases and case weights
** Best practices
*** Multiple passes with filters versus single pass with loops
*** Sequential versus random access
*** Managing memory
*** Passing cases around
*** Renaming casereaders
*** Avoiding excessive buffering
*** Propagating errors
*** Avoid static/global data
*** Don't worry about null filters, groups, etc.
*** Be aware of reference counting semantics for cases

5 Presenting Output

6 Internationalisation

Internationalisation in pspp is complicated. The most annoying aspect is that of character-encoding. This chapter attempts to describe the problems and current ways in which they are addressed.

6.1 The working locales

Pspp has three “working” locales:

• The locale of the user interface.
• The locale of the output.
• The locale of the data. Only the character encoding is relevant.

Each of these locales may, at different times take separate (or identical) values. So for example, a French statistician can use pspp to prepare a report in the English language, using a datafile which has been created by a Japanese researcher hence uses a Japanese character set.

It’s rarely, if ever, necessary to interrogate the system to find out the values of the 3 locales. However it’s important to be aware of the source (destination) locale when reading (writing) string data. When transfering data between a source and a destination, the appropriate recoding must be performed.

6.1.1 The user interface locale

This is the locale which is visible to the person using pspp. Error messages and confidence indications are written in this locale. For example “Cannot open file” will be written in the user interface locale.

This locale is set from the environment of the user who starts pspp{ire} or from the system locale if not set.

6.1.2 The output locale

This locale is the one that should be visible to the person reading a report generated by pspp. Non-data related strings (Eg: “Page number”, “Standard Deviation” etc.) will appear in this locale.

6.1.3 The data locale

This locale is the one associated with the data being analysed with pspp. The only important aspect of this locale is the character encoding. ¹ The dictionary pertaining to the data contains a field denoting the encoding. Any string data stored in a union value will be encoded in the dictionary’s character set.

6.2 System files

*.sav files contain a field which is supposed to identify the encoding of the data they contain (see Machine Integer Info Record). However, many files produced by early versions of spss set this to “2” (ASCII) regardless of the encoding of the data. Later versions contain an additional record (see Character Encoding Record) describing the encoding. When a system file is read, the dictionary’s encoding is set using information gleened from the system file. If the encoding cannot be determined or would be unreliable, then it remains unset.

6.3 GUI

The psppire graphic user interface is written using the Gtk+ api, for which all strings must be encoded in UTF8. All strings passed to the GTK+/GLib library functions (except for filenames) must be UTF-8 encoded otherwise errors will occur. Thus, for the purposes of the programming psppire, the user interface locale should be assumed to be UTF8, even if setlocale and/or nl_langinfo indicates otherwise.

6.3.1 Filenames

The GLib API has some special functions for dealing with filenames. Strings returned from functions like gtk_file_chooser_dialog_get_name are not, in general, encoded in UTF8, but in “filename” encoding. If that filename is passed to another GLib function which expects a filename, no conversion is necessary. If it’s passed to a function for the purposes of displaying it (eg. in a window’s title-bar) it must be converted to UTF8 — there is a special function for this: g_filename_display_name or g_filename_basename. If however, a filename needs to be passed outside of GTK+/GLib (for example to fopen) it must be converted to the local system encoding.

6.4 Existing locale handling functions

The major aspect of locale handling which the programmer has to consider is that of character encoding.

The following function is used to recode strings:

Function: char * recode_string (const char *to, const char *from, const char *text, int len);

Converts the string text, which is encoded in from to a new string encoded in to encoding. If len is not -1, then it must be the number of bytes in text. It is the caller’s responsibility to free the returned string when no longer required.

In order to minimise the number of conversions required, and to simplify design, PSPP attempts to store all internal strings in UTF8 encoding. Thus, when reading system and portable files (or any other data source), the following items are immediately converted to UTF8 encoding:

• Variable names
• Variable labels
• Value labels

Conversely, when writing system files, these are converted back to the encoding of that system file.

String data stored in union values are left in their original encoding. These will be converted by the data_in/data_out functions.

6.5 Quirks

For historical reasons, not all locale handling follows posix conventions. This makes it difficult (impossible?) to elegantly handle the issues. For example, it would make sense for the gui’s datasheet to display numbers formatted according to the LC_NUMERIC category of the data locale. Instead however there is the data_out function (see Obtaining Properties of Format Types) which uses the settings_get_decimal_char function instead of the decimal separator of the locale. Similarly, formatting of monetary values is displayed in a pspp/spss specific fashion instead of using the LC_MONETARY category.

7 Function Index

8 Concept Index

Appendix A Portable File Format

These days, most computers use the same internal data formats for integer and floating-point data, if one ignores little differences like big- versus little-endian byte ordering. However, occasionally it is necessary to exchange data between systems with incompatible data formats. This is what portable files are designed to do.

Please note: This information is gleaned from examination of ASCII-formatted portable files only, so some of it may be incorrect for portable files formatted in EBCDIC or other character sets.

A.1 Portable File Characters

Portable files are arranged as a series of lines of 80 characters each. Each line is terminated by a carriage-return, line-feed sequence (“new-lines”). New-lines are only used to avoid line length limits imposed by some OSes; they are not meaningful.

Most lines in portable files are exactly 80 characters long. The only exception is a line that ends in one or more spaces, in which the spaces may optionally be omitted. Thus, a portable file reader must act as though a line shorter than 80 characters is padded to that length with spaces.

The file must be terminated with a ‘Z’ character. In addition, if the final line in the file does not have exactly 80 characters, then it is padded on the right with ‘Z’ characters. (The file contents may be in any character set; the file contains a description of its own character set, as explained in the next section. Therefore, the ‘Z’ character is not necessarily an ASCII ‘Z’.)

For the rest of the description of the portable file format, new-lines and the trailing ‘Z’s will be ignored, as if they did not exist, because they are not an important part of understanding the file contents.

A.2 Portable File Structure

Every portable file consists of the following records, in sequence:

• File header.
• Version and date info.
• Product identification.
• Author identification (optional).
• Subproduct identification (optional).
• Variable count.
• Case weight variable (optional).
• Variables. Each variable record may optionally be followed by a missing value record and a variable label record.
• Value labels (optional).
• Documents (optional).
• Data.

Most records are identified by a single-character tag code. The file header and version info record do not have a tag.

Other than these single-character codes, there are three types of fields in a portable file: floating-point, integer, and string. Floating-point fields have the following format:

• Zero or more leading spaces.
• Optional asterisk (‘*’), which indicates a missing value. The asterisk must be followed by a single character, generally a period (‘.’), but it appears that other characters may also be possible. This completes the specification of a missing value.
• Optional minus sign (‘-’) to indicate a negative number.
• A whole number, consisting of one or more base-30 digits: ‘0’ through ‘9’ plus capital letters ‘A’ through ‘T’.
• Optional fraction, consisting of a radix point (‘.’) followed by one or more base-30 digits.
• Optional exponent, consisting of a plus or minus sign (‘+’ or ‘-’) followed by one or more base-30 digits.
• A forward slash (‘/’).

Integer fields take a form identical to floating-point fields, but they may not contain a fraction.

String fields take the form of a integer field having value n, followed by exactly n characters, which are the string content.

A.3 Portable File Header

Every portable file begins with a 464-byte header, consisting of a 200-byte collection of vanity splash strings, followed by a 256-byte character set translation table, followed by an 8-byte tag string.

The 200-byte segment is divided into five 40-byte sections, each of which represents the string charset SPSS PORT FILE in a different character set encoding, where charset is the name of the character set used in the file, e.g. ASCII or EBCDIC. Each string is padded on the right with spaces in its respective character set.

It appears that these strings exist only to inform those who might view the file on a screen, and that they are not parsed by SPSS products. Thus, they can be safely ignored. For those interested, the strings are supposed to be in the following character sets, in the specified order: EBCDIC, 7-bit ASCII, CDC 6-bit ASCII, 6-bit ASCII, Honeywell 6-bit ASCII.

The 256-byte segment describes a mapping from the character set used in the portable file to an arbitrary character set having characters at the following positions:

0–60

Control characters. Not important enough to describe in full here.

61–63

Reserved.

64–73

Digits ‘0’ through ‘9’.

74–99

Capital letters ‘A’ through ‘Z’.

100–125

Lowercase letters ‘a’ through ‘z’.

126

Space.

127–130

Symbols .<(+

131

Solid vertical pipe.

132–142

Symbols &[]!$*);^-/

143

Broken vertical pipe.

144–150

Symbols ,%_>?`:

151

British pound symbol.

152–155

Symbols @'=".

156

Less than or equal symbol.

157

Empty box.

158

Plus or minus.

159

Filled box.

160

Degree symbol.

161

Dagger.

162

Symbol ‘~’.

163

En dash.

164

Lower left corner box draw.

165

Upper left corner box draw.

166

Greater than or equal symbol.

167–176

Superscript ‘0’ through ‘9’.

177

Lower right corner box draw.

178

Upper right corner box draw.

179

Not equal symbol.

180

Em dash.

181

Superscript ‘(’.

182

Superscript ‘)’.

183

Horizontal dagger (?).

184–186

Symbols ‘{}\’.

187

Cents symbol.

188

Centered dot, or bullet.

189–255

Reserved.

Symbols that are not defined in a particular character set are set to the same value as symbol 64; i.e., to ‘0’.

The 8-byte tag string consists of the exact characters SPSSPORT in the portable file’s character set, which can be used to verify that the file is indeed a portable file.

A.4 Version and Date Info Record

This record does not have a tag code. It has the following structure:

• A single character identifying the file format version. The letter A represents version 0, and so on.
• An 8-character string field giving the file creation date in the format YYYYMMDD.
• A 6-character string field giving the file creation time in the format HHMMSS.

A.5 Identification Records

The product identification record has tag code ‘1’. It consists of a single string field giving the name of the product that wrote the portable file.

The author identification record has tag code ‘2’. It is optional. If present, it consists of a single string field giving the name of the person who caused the portable file to be written.

The subproduct identification record has tag code ‘3’. It is optional. If present, it consists of a single string field giving additional information on the product that wrote the portable file.

A.6 Variable Count Record

The variable count record has tag code ‘4’. It consists of a single integer field giving the number of variables in the file dictionary.

A.7 Precision Record

The precision record has tag code ‘5’. It consists of a single integer field specifying the maximum number of base-30 digits used in data in the file.

A.8 Case Weight Variable Record

The case weight variable record is optional. If it is present, it indicates the variable used for weighting cases; if it is absent, cases are unweighted. It has tag code ‘6’. It consists of a single string field that names the weighting variable.

A.9 Variable Records

Each variable record represents a single variable. Variable records have tag code ‘7’. They have the following structure:

• Width (integer). This is 0 for a numeric variable, and a number between 1 and 255 for a string variable.
• Name (string). 1–8 characters long. Must be in all capitals.

  A few portable files that contain duplicate variable names have been spotted in the wild. PSPP handles these by renaming the duplicates with numeric extensions: var_1, var_2, and so on.

• Print format. This is a set of three integer fields:
  • - Format type (see Variable Record).
  • - Format width. 1–40.
  • - Number of decimal places. 1–40.

  A few portable files with invalid format types or formats that are not of the appropriate width for their variables have been spotted in the wild. PSPP assigns a default F or A format to a variable with an invalid format.

• Write format. Same structure as the print format described above.

Each variable record can optionally be followed by a missing value record, which has tag code ‘8’. A missing value record has one field, the missing value itself (a floating-point or string, as appropriate). Up to three of these missing value records can be used.

There is also a record for missing value ranges, which has tag code ‘B’. It is followed by two fields representing the range, which are floating-point or string as appropriate. If a missing value range is present, it may be followed by a single missing value record.

Tag codes ‘9’ and ‘A’ represent LO THRU x and x THRU HI ranges, respectively. Each is followed by a single field representing x. If one of the ranges is present, it may be followed by a single missing value record.

In addition, each variable record can optionally be followed by a variable label record, which has tag code ‘C’. A variable label record has one field, the variable label itself (string).

A.10 Value Label Records

Value label records have tag code ‘D’. They have the following format:

• Variable count (integer).
• List of variables (strings). The variable count specifies the number in the list. Variables are specified by their names. All variables must be of the same type (numeric or string), but string variables do not necessarily have the same width.
• Label count (integer).
• List of (value, label) tuples. The label count specifies the number of tuples. Each tuple consists of a value, which is numeric or string as appropriate to the variables, followed by a label (string).

A few portable files that specify duplicate value labels, that is, two different labels for a single value of a single variable, have been spotted in the wild. PSPP uses the last value label specified in these cases.

A.11 Document Record

One document record may optionally follow the value label record. The document record consists of tag code ‘E’, following by the number of document lines as an integer, followed by that number of strings, each of which represents one document line. Document lines must be 80 bytes long or shorter.

A.12 Portable File Data

The data record has tag code ‘F’. There is only one tag for all the data; thus, all the data must follow the dictionary. The data is terminated by the end-of-file marker ‘Z’, which is not valid as the beginning of a data element.

Data elements are output in the same order as the variable records describing them. String variables are output as string fields, and numeric variables are output as floating-point fields.

Appendix B System File Format

A system file encapsulates a set of cases and dictionary information that describes how they may be interpreted. This chapter describes the format of a system file.

System files use four data types: 8-bit characters, 32-bit integers, 64-bit integers, and 64-bit floating points, called here char, int32, int64, and flt64, respectively. Data is not necessarily aligned on a word or double-word boundary: the long variable name record (see Long Variable Names Record) and very long string records (see Very Long String Record) have arbitrary byte length and can therefore cause all data coming after them in the file to be misaligned.

Integer data in system files may be big-endian or little-endian. A reader may detect the endianness of a system file by examining layout_code in the file header record (see layout_code).

Floating-point data in system files may nominally be in IEEE 754, IBM, or VAX formats. A reader may detect the floating-point format in use by examining bias in the file header record (see bias).

PSPP detects big-endian and little-endian integer formats in system files and translates as necessary. PSPP also detects the floating-point format in use, as well as the endianness of IEEE 754 floating-point numbers, and translates as needed. However, only IEEE 754 numbers with the same endianness as integer data in the same file have actually been observed in system files, and it is likely that other formats are obsolete or were never used.

System files use a few floating point values for special purposes:

SYSMIS

The system-missing value is represented by the largest possible negative number in the floating point format (-DBL_MAX).

HIGHEST

HIGHEST is used as the high end of a missing value range with an unbounded maximum. It is represented by the largest possible positive number (DBL_MAX).

LOWEST

LOWEST is used as the low end of a missing value range with an unbounded minimum. It was originally represented by the second-largest negative number (in IEEE 754 format, 0xffeffffffffffffe). System files written by SPSS 21 and later instead use the largest negative number (-DBL_MAX), the same value as SYSMIS. This does not lead to ambiguity because LOWEST appears in system files only in missing value ranges, which never contain SYSMIS.

System files may use most character encodings based on an 8-bit unit. UTF-16 and UTF-32, based on wider units, appear to be unacceptable. rec_type in the file header record is sufficient to distinguish between ASCII and EBCDIC based encodings. The best way to determine the specific encoding in use is to consult the character encoding record (see Character Encoding Record), if present, and failing that the character_code in the machine integer info record (see Machine Integer Info Record). The same encoding should be used for the dictionary and the data in the file, although it is possible to artificially synthesize files that use different encodings (see Character Encoding Record).

B.1 System File Record Structure

System files are divided into records with the following format:

int32               type;
char                data[];

This header does not identify the length of the data or any information about what it contains, so the system file reader must understand the format of data based on type. However, records with type 7, called extension records, have a stricter format:

int32               type;
int32               subtype;
int32               size;
int32               count;
char                data[size * count];

int32 rec_type;

Record type. Always set to 7.

int32 subtype;

Record subtype. This value identifies a particular kind of extension record.

int32 size;

The size of each piece of data that follows the header, in bytes. Known extension records use 1, 4, or 8, for char, int32, and flt64 format data, respectively.

int32 count;

The number of pieces of data that follow the header.

char data[size * count];

Data, whose format and interpretation depend on the subtype.

An extension record contains exactly size * count bytes of data, which allows a reader that does not understand an extension record to skip it. Extension records provide only nonessential information, so this allows for files written by newer software to preserve backward compatibility with older or less capable readers.

Records in a system file must appear in the following order:

• File header record.
• Variable records.
• All pairs of value labels records and value label variables records, if present.
• Document record, if present.
• Extension (type 7) records, in ascending numerical order of their subtypes.
• Dictionary termination record.
• Data record.

We advise authors of programs that read system files to tolerate format variations. Various kinds of misformatting and corruption have been observed in system files written by SPSS and other software alike. In particular, because extension records provide nonessential information, it is generally better to ignore an extension record entirely than to refuse to read a system file.

The following sections describe the known kinds of records.

B.2 File Header Record

A system file begins with the file header, with the following format:

char                rec_type[4];
char                prod_name[60];
int32               layout_code;
int32               nominal_case_size;
int32               compression;
int32               weight_index;
int32               ncases;
flt64               bias;
char                creation_date[9];
char                creation_time[8];
char                file_label[64];
char                padding[3];

char rec_type[4];

Record type code, either ‘$FL2’ for system files with uncompressed data or data compressed with simple bytecode compression, or ‘$FL3’ for system files with ZLIB compressed data.

This is truly a character field that uses the character encoding as other strings. Thus, in a file with an ASCII-based character encoding this field contains 24 46 4c 32 or 24 46 4c 33, and in a file with an EBCDIC-based encoding this field contains 5b c6 d3 f2. (No EBCDIC-based ZLIB-compressed files have been observed.)

char prod_name[60];

Product identification string. This always begins with the characters ‘@(#) SPSS DATA FILE’. PSPP uses the remaining characters to give its version and the operating system name; for example, ‘GNU pspp 0.1.4 - sparc-sun-solaris2.5.2’. The string is truncated if it would be longer than 60 characters; otherwise it is padded on the right with spaces.

int32 layout_code;

Normally set to 2, although a few system files have been spotted in the wild with a value of 3 here. PSPP use this value to determine the file’s integer endianness (see System File Format).

int32 nominal_case_size;

Number of data elements per case. This is the number of variables, except that long string variables add extra data elements (one for every 8 characters after the first 8). However, string variables do not contribute to this value beyond the first 255 bytes. Further, system files written by some systems set this value to -1. In general, it is unsafe for systems reading system files to rely upon this value.

int32 compression;

Set to 0 if the data in the file is not compressed, 1 if the data is compressed with simple bytecode compression, 2 if the data is ZLIB compressed. This field has value 2 if and only if rec_type is ‘$FL3’.

int32 weight_index;

If one of the variables in the data set is used as a weighting variable, set to the dictionary index of that variable, plus 1 (see Dictionary Index). Otherwise, set to 0.

int32 ncases;

Set to the number of cases in the file if it is known, or -1 otherwise.

In the general case it is not possible to determine the number of cases that will be output to a system file at the time that the header is written. The way that this is dealt with is by writing the entire system file, including the header, then seeking back to the beginning of the file and writing just the ncases field. For files in which this is not valid, the seek operation fails. In this case, ncases remains -1.

flt64 bias;

Compression bias, ordinarily set to 100. Only integers between 1 - bias and 251 - bias can be compressed.

By assuming that its value is 100, PSPP uses bias to determine the file’s floating-point format and endianness (see System File Format). If the compression bias is not 100, PSPP cannot auto-detect the floating-point format and assumes that it is IEEE 754 format with the same endianness as the system file’s integers, which is correct for all known system files.

char creation_date[9];

Date of creation of the system file, in ‘dd mmm yy’ format, with the month as standard English abbreviations, using an initial capital letter and following with lowercase. If the date is not available then this field is arbitrarily set to ‘01 Jan 70’.

char creation_time[8];

Time of creation of the system file, in ‘hh:mm:ss’ format and using 24-hour time. If the time is not available then this field is arbitrarily set to ‘00:00:00’.

char file_label[64];

File label declared by the user, if any (see FILE LABEL in PSPP Users Guide). Padded on the right with spaces.

A product that identifies itself as VOXCO INTERVIEWER 4.3 uses CR-only line ends in this field, rather than the more usual LF-only or CR LF line ends.

char padding[3];

Ignored padding bytes to make the structure a multiple of 32 bits in length. Set to zeros.

B.3 Variable Record

There must be one variable record for each numeric variable and each string variable with width 8 bytes or less. String variables wider than 8 bytes have one variable record for each 8 bytes, rounding up. The first variable record for a long string specifies the variable’s correct dictionary information. Subsequent variable records for a long string are filled with dummy information: a type of -1, no variable label or missing values, print and write formats that are ignored, and an empty string as name. A few system files have been encountered that include a variable label on dummy variable records, so readers should take care to parse dummy variable records in the same way as other variable records.

The dictionary index of a variable is its offset in the set of variable records, including dummy variable records for long string variables. The first variable record has a dictionary index of 0, the second has a dictionary index of 1, and so on.

The system file format does not directly support string variables wider than 255 bytes. Such very long string variables are represented by a number of narrower string variables. See Very Long String Record, for details.

A system file should contain at least one variable and thus at least one variable record, but system files have been observed in the wild without any variables (thus, no data either).

int32               rec_type;
int32               type;
int32               has_var_label;
int32               n_missing_values;
int32               print;
int32               write;
char                name[8];

/* Present only if has_var_label is 1. */
int32               label_len;
char                label[];

/* Present only if n_missing_values is nonzero. */
flt64               missing_values[];

int32 rec_type;

Record type code. Always set to 2.

int32 type;

Variable type code. Set to 0 for a numeric variable. For a short string variable or the first part of a long string variable, this is set to the width of the string. For the second and subsequent parts of a long string variable, set to -1, and the remaining fields in the structure are ignored.

int32 has_var_label;

If this variable has a variable label, set to 1; otherwise, set to 0.

int32 n_missing_values;

If the variable has no missing values, set to 0. If the variable has one, two, or three discrete missing values, set to 1, 2, or 3, respectively. If the variable has a range for missing variables, set to -2; if the variable has a range for missing variables plus a single discrete value, set to -3.

A long string variable always has the value 0 here. A separate record indicates missing values for long string variables (see Long String Missing Values Record).

int32 print;

Print format for this variable. See below.

int32 write;

Write format for this variable. See below.

char name[8];

Variable name. The variable name must begin with a capital letter or the at-sign (‘@’). Subsequent characters may also be digits, octothorpes (‘#’), dollar signs (‘$’), underscores (‘_’), or full stops (‘.’). The variable name is padded on the right with spaces.

The ‘name’ fields should be unique within a system file. System files written by SPSS that contain very long string variables with similar names sometimes contain duplicate names that are later eliminated by resolving the very long string names (see Very Long String Record). PSPP handles duplicates by assigning them new, unique names.

int32 label_len;

This field is present only if has_var_label is set to 1. It is set to the length, in characters, of the variable label. The documented maximum length varies from 120 to 255 based on SPSS version, but some files have been seen with longer labels. PSPP accepts labels of any length.

char label[];

This field is present only if has_var_label is set to 1. It has length label_len, rounded up to the nearest multiple of 32 bits. The first label_len characters are the variable’s variable label.

flt64 missing_values[];

This field is present only if n_missing_values is nonzero. It has the same number of 8-byte elements as the absolute value of n_missing_values. Each element is interpreted as a number for numeric variables (with HIGHEST and LOWEST indicated as described in the chapter introduction). For string variables of width less than 8 bytes, elements are right-padded with spaces; for string variables wider than 8 bytes, only the first 8 bytes of each missing value are specified, with the remainder implicitly all spaces.

For discrete missing values, each element represents one missing value. When a range is present, the first element denotes the minimum value in the range, and the second element denotes the maximum value in the range. When a range plus a value are present, the third element denotes the additional discrete missing value.

The print and write members of sysfile_variable are output formats coded into int32 types. The least-significant byte of the int32 represents the number of decimal places, and the next two bytes in order of increasing significance represent field width and format type, respectively. The most-significant byte is not used and should be set to zero.

Format types are defined as follows:

Value	Meaning
0	Not used.
1	A
2	AHEX
3	COMMA
4	DOLLAR
5	F
6	IB
7	PIBHEX
8	P
9	PIB
10	PK
11	RB
12	RBHEX
13	Not used.
14	Not used.
15	Z
16	N
17	E
18	Not used.
19	Not used.
20	DATE
21	TIME
22	DATETIME
23	ADATE
24	JDATE
25	DTIME
26	WKDAY
27	MONTH
28	MOYR
29	QYR
30	WKYR
31	PCT
32	DOT
33	CCA
34	CCB
35	CCC
36	CCD
37	CCE
38	EDATE
39	SDATE

A few system files have been observed in the wild with invalid write fields, in particular with value 0. Readers should probably treat invalid print or write fields as some default format.

B.4 Value Labels Records

The value label records documented in this section are used for numeric and short string variables only. Long string variables may have value labels, but their value labels are recorded using a different record type (see Long String Value Labels Record).

The value label record has the following format:

int32               rec_type;
int32               label_count;

/* Repeated label_cnt times. */
char                value[8];
char                label_len;
char                label[];

int32 rec_type;

Record type. Always set to 3.

int32 label_count;

Number of value labels present in this record.

The remaining fields are repeated count times. Each repetition specifies one value label.

char value[8];

A numeric value or a short string value padded as necessary to 8 bytes in length. Its type and width cannot be determined until the following value label variables record (see below) is read.

char label_len;

The label’s length, in bytes. The documented maximum length varies from 60 to 120 based on SPSS version. PSPP supports value labels up to 255 bytes long.

char label[];

label_len bytes of the actual label, followed by up to 7 bytes of padding to bring label and label_len together to a multiple of 8 bytes in length.

The value label record is always immediately followed by a value label variables record with the following format:

int32               rec_type;
int32               var_count;
int32               vars[];

int32 rec_type;

Record type. Always set to 4.

int32 var_count;

Number of variables that the associated value labels from the value label record are to be applied.

int32 vars[];

A list of dictionary indexes of variables to which to apply the value labels (see Dictionary Index). There are var_count elements.

String variables wider than 8 bytes may not be specified in this list.

B.5 Document Record

The document record, if present, has the following format:

int32               rec_type;
int32               n_lines;
char                lines[][80];

int32 rec_type;

Record type. Always set to 6.

int32 n_lines;

Number of lines of documents present.

char lines[][80];

Document lines. The number of elements is defined by n_lines. Lines shorter than 80 characters are padded on the right with spaces.

B.6 Machine Integer Info Record

The integer info record, if present, has the following format:

/* Header. */
int32               rec_type;
int32               subtype;
int32               size;
int32               count;

/* Data. */
int32               version_major;
int32               version_minor;
int32               version_revision;
int32               machine_code;
int32               floating_point_rep;
int32               compression_code;
int32               endianness;
int32               character_code;

int32 rec_type;

Record type. Always set to 7.

int32 subtype;

Record subtype. Always set to 3.

int32 size;

Size of each piece of data in the data part, in bytes. Always set to 4.

int32 count;

Number of pieces of data in the data part. Always set to 8.

int32 version_major;

PSPP major version number. In version x.y.z, this is x.

int32 version_minor;

PSPP minor version number. In version x.y.z, this is y.

int32 version_revision;

PSPP version revision number. In version x.y.z, this is z.

int32 machine_code;

Machine code. PSPP always set this field to value to -1, but other values may appear.

int32 floating_point_rep;

Floating point representation code. For IEEE 754 systems this is 1. IBM 370 sets this to 2, and DEC VAX E to 3.

int32 compression_code;

Compression code. Always set to 1, regardless of whether or how the file is compressed.

int32 endianness;

Machine endianness. 1 indicates big-endian, 2 indicates little-endian.

int32 character_code;

Character code. The following values have been actually observed in system files:

1

EBCDIC.

2

7-bit ASCII.

1250

The windows-1250 code page for Central European and Eastern European languages.

1252

The windows-1252 code page for Western European languages.

28591

ISO 8859-1.

65001

UTF-8.

The following additional values are known to be defined:

3

8-bit “ASCII”.

4

DEC Kanji.

Other Windows code page numbers are known to be generally valid.

Old versions of SPSS for Unix and Windows always wrote value 2 in this field, regardless of the encoding in use. Newer versions also write the character encoding as a string (see Character Encoding Record).

B.7 Machine Floating-Point Info Record

The floating-point info record, if present, has the following format:

/* Header. */
int32               rec_type;
int32               subtype;
int32               size;
int32               count;

/* Data. */
flt64               sysmis;
flt64               highest;
flt64               lowest;

int32 rec_type;

Record type. Always set to 7.

int32 subtype;

Record subtype. Always set to 4.

int32 size;

Size of each piece of data in the data part, in bytes. Always set to 8.

int32 count;

Number of pieces of data in the data part. Always set to 3.

flt64 sysmis;

flt64 highest;

flt64 lowest;

The system missing value, the value used for HIGHEST in missing values, and the value used for LOWEST in missing values, respectively. See System File Format, for more information.

The SPSSWriter library in PHP, which identifies itself as FOM SPSS 1.0.0 in the file header record prod_name field, writes unexpected values to these fields, but it uses the same values consistently throughout the rest of the file.

B.8 Multiple Response Sets Records

The system file format has two different types of records that represent multiple response sets (see MRSETS in PSPP Users Guide). The first type of record describes multiple response sets that can be understood by SPSS before version 14. The second type of record, with a closely related format, is used for multiple dichotomy sets that use the CATEGORYLABELS=COUNTEDVALUES feature added in version 14.

/* Header. */
int32               rec_type;
int32               subtype;
int32               size;
int32               count;

/* Exactly count bytes of data. */
char                mrsets[];

int32 rec_type;

Record type. Always set to 7.

int32 subtype;

Record subtype. Set to 7 for records that describe multiple response sets understood by SPSS before version 14, or to 19 for records that describe dichotomy sets that use the CATEGORYLABELS=COUNTEDVALUES feature added in version 14.

int32 size;

The size of each element in the mrsets member. Always set to 1.

int32 count;

The total number of bytes in mrsets.

char mrsets[];

Zero or more line feeds (byte 0x0a), followed by a series of multiple response sets, each of which consists of the following:

• The set’s name (an identifier that begins with ‘$’), in mixed upper and lower case.
• An equals sign (‘=’).
• ‘C’ for a multiple category set, ‘D’ for a multiple dichotomy set with CATEGORYLABELS=VARLABELS, or ‘E’ for a multiple dichotomy set with CATEGORYLABELS=COUNTEDVALUES.
• For a multiple dichotomy set with CATEGORYLABELS=COUNTEDVALUES, a space, followed by a number expressed as decimal digits, followed by a space. If LABELSOURCE=VARLABEL was specified on MRSETS, then the number is 11; otherwise it is 1.²
• For either kind of multiple dichotomy set, the counted value, as a positive integer count specified as decimal digits, followed by a space, followed by as many string bytes as specified in the count. If the set contains numeric variables, the string consists of the counted integer value expressed as decimal digits. If the set contains string variables, the string contains the counted string value. Either way, the string may be padded on the right with spaces (older versions of SPSS seem to always pad to a width of 8 bytes; newer versions don’t).
• A space.
• The multiple response set’s label, using the same format as for the counted value for multiple dichotomy sets. A string of length 0 means that the set does not have a label. A string of length 0 is also written if LABELSOURCE=VARLABEL was specified.
• A space.
• The short names of the variables in the set, converted to lowercase, each separated from the previous by a single space.

  Even though a multiple response set must have at least two variables, some system files contain multiple response sets with no variables or one variable. The source and meaning of these multiple response sets is unknown. (Perhaps they arise from creating a multiple response set then deleting all the variables that it contains?)

• One line feed (byte 0x0a). Sometimes multiple, even hundreds, of line feeds are present.

Example: Given appropriate variable definitions, consider the following MRSETS command:

MRSETS /MCGROUP NAME=$a LABEL='my mcgroup' VARIABLES=a b c
       /MDGROUP NAME=$b VARIABLES=g e f d VALUE=55
       /MDGROUP NAME=$c LABEL='mdgroup #2' VARIABLES=h i j VALUE='Yes'
       /MDGROUP NAME=$d LABEL='third mdgroup' CATEGORYLABELS=COUNTEDVALUES
        VARIABLES=k l m VALUE=34
       /MDGROUP NAME=$e CATEGORYLABELS=COUNTEDVALUES LABELSOURCE=VARLABEL
        VARIABLES=n o p VALUE='choice'.

The above would generate the following multiple response set record of subtype 7:

$a=C 10 my mcgroup a b c
$b=D2 55 0  g e f d
$c=D3 Yes 10 mdgroup #2 h i j

It would also generate the following multiple response set record with subtype 19:

$d=E 1 2 34 13 third mdgroup k l m
$e=E 11 6 choice 0  n o p

B.9 Extra Product Info Record

This optional record appears to contain a text string that describes the program that wrote the file and the source of the data. (This is redundant with the file label and product info found in the file header record.)

/* Header. */
int32               rec_type;
int32               subtype;
int32               size;
int32               count;

/* Exactly count bytes of data. */
char                info[];

int32 rec_type;

Record type. Always set to 7.

int32 subtype;

Record subtype. Always set to 10.

int32 size;

The size of each element in the info member. Always set to 1.

int32 count;

The total number of bytes in info.

char info[];

A text string. A product that identifies itself as VOXCO INTERVIEWER 4.3 uses CR-only line ends in this field, rather than the more usual LF-only or CR LF line ends.

B.10 Variable Display Parameter Record

The variable display parameter record, if present, has the following format:

/* Header. */
int32               rec_type;
int32               subtype;
int32               size;
int32               count;

/* Repeated count times. */
int32               measure;
int32               width;           /* Not always present. */
int32               alignment;

int32 rec_type;

Record type. Always set to 7.

int32 subtype;

Record subtype. Always set to 11.

int32 size;

The size of int32. Always set to 4.

int32 count;

The number of sets of variable display parameters (ordinarily the number of variables in the dictionary), times 2 or 3.

The remaining members are repeated count times, in the same order as the variable records. No element corresponds to variable records that continue long string variables. The meanings of these members are as follows:

int32 measure;

The measurement type of the variable:

1

Nominal Scale

2

Ordinal Scale

3

Continuous Scale

SPSS sometimes writes a measure of 0. PSPP interprets this as nominal scale.

int32 width;

The width of the display column for the variable in characters.

This field is present if count is 3 times the number of variables in the dictionary. It is omitted if count is 2 times the number of variables.

int32 alignment;

The alignment of the variable for display purposes:

0

Left aligned

1

Right aligned

2

Centre aligned