.help lcalc Apr84 "List Calculator"
.nh
Introduction

    The list calculator performs general arithmetic, string, and
conditional operations upon lists.  Examples of such operations include
.ls 4
.ls o
General filtering of lists, i.e., conditionally pass or stop records
within a list.
.le
.ls o
Rearranging the fields of a list.
.le
.ls o
Extracting fields from a list.
.le
.ls o
Merging lists.
.le
.ls o
Arithmetic or string operations upon lists.
.le
.le

A \fBlist\fR is a text file consisting of a sequence of \fBrecords\fR.
Whitespace normally delimits the fields of a record; newline normally
delimits a record.  There may be an arbitrary number of fields within a
record, and records within a list.  A record may span several lines of
text if a special delimiter is used.  Blank lines and comment lines are
passed on to the output but are otherwise ignored.

The input language of the list calculator is optimized to provide a
concise notation for expressing the types of operations most commonly
performed upon lists.  Operations which cannot be specified by this
somewhat specialized language can easily be programmed as CL scripts.

.nh
Lists

    Input may consist of one or more logical input lists.  Multiple input
lists are by default read simultaneously as discreet lists; they may 
optionally be concatenated into a single logical list.
A single list is always generated as output.  Multiple input lists
are referred to in expressions as "lN", i.e., "l1", "l2", etc.
If multiple discreet input lists are used, the output list will be
truncated to the length of the shortest input list.

.nh
Records

    The list calculator breaks each list into a sequence of records.
Each record has the following attributes:

.nf
	record number
	record length
	number of fields
.fi

If the record delimiter is other than newline, newline is treated as
whitespace.  Records are numbered starting at 1.  Record number BOL may
be used to conditionally execute code before processing a list.  Record
number EOL is matched only at the end of the list.

.nh
Fields

    The list calculator breaks each record into a sequence of fields.
Each field has the following attributes:

.nf
	field number
	field width (nchars in input record)
	datatype
	number of significant digits, if a number
.fi

The "current record" may be thought of as an array F of fields.
The individual fields may be referenced in expressions in either of two
ways.  If a single field is to be referenced, the field may be referred
to as the variable fN, e.g., "f1", "f2", and so on.  If multiple
input lists are in use, the notation "li.fj" must be used instead.
Sets of fields may be referenced as sections of the array F;
a \fBsection\fR is an operand of type string, and may not be used
in arithmetic expressions.  The entire record is most simply referred
to as the string R.  The section notation may also be used to refer
to substrings of string type operands, and is discussed in a later section.

.nh
Calling Sequence

    The principal operands to \fBlcalc\fR are the list template and
the program to be executed.  An additional set of positional arguments
may be given; these are passed on to the program and are not used by
\fBlcalc\fR itself.  A number of hidden parameters are available to
modify the behavior of the program.

	lcalc (list, program [, p1,...,p9])

.nh
Parameters
.ls 4
.ls lists
A filename template specifying the list files to be processed.  Not used
if the standard input is redirected.
.le
.ls program
Either the explicit \fBlcalc\fR program to be run, or if the first character
is an @, the name of a file containing the program.
.le
.ls p1 - p9
Optional program parameters, referred to in the program as "$1", "$2", etc.
.le
.ls concatenate = no
Treat multiple input lists independently (concatenate = no) or concatenate
all input lists to form a single list.
.le
.ls field_delim = " \t"
A set of characters, any \fIone\fR of which will delimit the fields of an input
record.  The default field delimiters are blank and tab (denoted \t).
.le
.ls record_delim = "$"
The \fIpattern\fR which delimits input records.  All of the usual pattern
matching meta-characters are recognized.  The default is "$", meaning that
end of line delimits input records (each record is a line of the input list).
Special record delimiters may be used to process multiline records.
.le
.ls ofield_delim = "in_field_delim"
The output field delimiter, inserted by the list calculator in output records
to delimit fields.  The string "in_field_delim" signals that the first input
field delimiter character is to be used.
.le
.ls orecord_delim = "in_record_delim"
The output record delimiter, used to delimit output records.  The default
action is to use the string (not pattern) which delimited the input record.
.le
.ls comments = yes
Pass comment lines and blank lines on to the output list.
.le
.le

.nh
Statements

    A list transformation is specified by a program consisting of a
sequence of statements.  The program may be input either directly on
the command line as a string, or in a file.  Statements may be delimited
by either semicolon or newline.

.nf
	prog	:	stmt
		|	prog stmt
		;

	stmt	:	assign eost
		|	print eost
		;

	eost	:	'\n'
		|	';'
		;
.fi

There are two types of statements, assignment statements and print
statements.  The \fBassignment\fR statements may be used to set or modify
the contents of an internal register.  Registers are created and initialized
to zero or to the null string when first referenced in an assignment statement.
The usual modify and replace assignment statements "+=", "//=", etc. are
recognized.

.nf
	assign	:	IDENT '=' expr
		|	IDENT OPEQ expr
		;

	print	:	expr
		|	expr '?'
		|	expr '?' expr
		;
.fi

The \fBprint\fR statement is used to generate output.  In its simplest
form the print statement is an expression (e.g., "f5"); the value of the
expression is computed and output as the next field of the output record.

As a simple example, suppose we wish to swap fields 2 and 3 of a three field
list, dropping field 1 (any additional fields will be discarded):

	lcalc list, "f3,f2"

.nh
Expressions

    The power of the list calculator derives principally from its ability
to evaluate complex expressions.  The operands of an expression may be
fields, registers, command line arguments, functions, constants, or
expressions.  The datatype of an expression may be boolean, integer, real,
or string.  Conditional expressions may be used to select or reject entire
records, or to conditionally format the fields of output records.

The standard set of operators are supported.  Expressions are evaluated
using the standard operator precedences and associativities.  The following
operators are recognized:


.nf
binary operators:

	+   -    *   /   **			arithmetic
	<   <=   >   >=  ==  !=	  @		comparison
	&&  ||					boolean and, or
	//					string concatenation


unary operators:

	!   -    @
.fi

The only unconventional operator is the string matching operator @,
used to determine if a string matches a pattern.  Usage is as follows:

.nf
	string @ pattern
or
	@ pattern
.fi
	
The expression evaluates to true if the pattern can be matched.  If no
string is given, the entire record is searched.

The expression syntax recognized by the list calculator is defined in
the figure below.


.nf
	expr	:	primary
		|	expr ',' opnl expr
		|	'(' expr ')'
		;

	primary	:	IDENT
		|	CONSTANT
		|	primary BINOP opnl primary
		|	UNOP primary
		|	primary '?' primary ':' primary
		|	IDENT '(' arglist ')'
		|	section
		|	'(' primary ')'
		;

	section	:	IDENT '[' flist ']'
		;

	flist	:	# Empty
		|	fields
		|	flist ',' fields
		;

	fields	:	primary
		|	primary ':'
		|	primary ':' ':' primary
		|	primary ':' primary
		|	primary ':' primary ':' primary
		|	'*'
		|	'-' '*'
		|	'-' '*' ':' primary
		;	
.fi

.nh
Sections

    The section notation is used to extract substrings from string type
variables.  If the variable referenced is F, an array of strings, then
the section indexes the fields of the record; the specified fields are
concatenated to produce the output string.  If the variable is a string
variable, an array of characters, then the section specifies a set of
substrings to be concatenated.

For example, to extract the first three characters of the string
variable "s":

	s[1:3]

To extract characters 4 through the end of the string:

	s[4:]

To reverse the order of the characters in the string:

	s[-*]

To extract the first field of the record:

	f[1]

To extract all fields, any of the following would do (F is the array of
field strings, whereas R is the entire record as a string):

	r, r[*], r[1:], f, f[*], f[1:]

To refer to all fields in reverse order:

	f[-*]

To specify fields 3 through 5:

	f[3:5]

To specify fields 3, 5, and 7 through 10:

	f[3,5,7:10]

If multiple input lists are in use, the notation "r1", "r2", etc. should
be used instead.

.nh
Intrinsic Functions

    All of the standard intrinsic functions are recognized, plus a few
special intrinsic functions.

.nf
      abs     exp     log     min     real    str
      atan2   int     log10   mod     sin     tan
      cos     len     max     nint    sqrt    type
.fi

Certain of these functions deserve further mention:
.ls 4
.ls len (expr)
Returns the length of an array or string, e.g., "len(f)" is the number
of fields in the current record, and "len(f2)" is the number of chars in
field 2.
.le
.ls str (expr)
Converts the argument into an operand of type string.
.le
.ls type (expr)
Returns the datatype of the argument.  Legal values are "b", "i", "r",
and "s".  The string type is the catchall.
.le
.le

.nh
Variables

    The types of variables implemented in the list calculator are
the field variables ("f1", "f2", etc.), user variables, parameters,
and various builtin variables.  All such variables may be used equivalently
within expressions.  Only the builtin variables are writable by the
program.
.nh 2
User Variables

    User variables are named by the user program; only the first 8 characters
of the variable name are significant.  A user variable is created when it
is first used in an assignment statement.  If the first reference is in a
modify and replace assignment, integer and real variables are initialized
to 0 and string variables are initialized to the null string, before the
modify operation takes place.  It is an error if a user variable is first
referenced in an expression.
.nh 2
Program Parameters

    The parameter variables are optional positional arguments to the
\fBlcalc\fR procedure.  There are nine such parameters, "$1" through "$9".
The special parameter "$nargs" specifies the number of these parameters
set on the \fBlcalc\fR command line (the \fBlist\fR and \fBprogram\fR
arguments to \fBlcalc\fR are not counted).  Parameters are read only.
.nh 2
Builtin Variables

    The list calculator manages a number of variables internally.  These
variables describe the list and record currently being processed, and are
read only to the user program (with one exception).  The internal variables
are used primarily for conditional statements to be executed only at certain
times.

.ls
.ls nfiles
The number of files in the input list (file template).
.le
.ls rnum
The record number within the list \fBfile\fR currently being processed.
The first record is number 1.
.le
.ls arnum
The absolute record number.
.le
.ls nfields
The number of fields in the current record.
.le
.ls nchars
The number of characters in the current record.
.le
.ls nlines
The number of input lines of text in the current record.
.le
.ls fname
The name of the input file currently being processed.
.le
.ls outfile
The name of the output file.  This parameter is writable by the program.
If modified by the program, the current output file is closed and the
new output file is created for writing.
.le
.ls atbol
Set when processing the first record of the first file.
.le
.ls ateol
Set once the last record of the last file has been processed.
.le
.ls atbof
Set when processing the first record of a file.
.le
.ls ateof
Set once the last record of a file has been processed, before reading
the next file.
.le
.le

.nh
Examples

    The list calculator is a very powerful tool and it is difficult to 
present enough examples to illustrate all of the possible applications.
Furthermore, in many applications the list calculator is often used in
combination with other tools such as \fBsort\fR and \fBgraph\fR; we
will not use such programs in our examples here.

.nh 2
Simple bandpass filter

Given a three column list {x,y,string}, pass only those lines for which
y is greater than 100 and less than or equal to 140:

	lcalc list, "f2 > 100 && f2 <= 140 ?"

Pass only records in list1 wherein the third field has the same value
as the fifth field of list2:

	lcalc "list1,list2", "l1.f3 == l2.f5 ? r1"

.nh 2
Simple bandstop filter

Pass only those records which do \fInot\fR contain the substrings
"obj1, "obj2", or "obj3":

	lcalc list, "!(@obj1 || @obj2 || @obj3) ?"

.nh 2
Rearranging and extracting fields

Move fields 8 and 9 to the beginning of the list:

	lcalc list, "f8,f9,f[1:7],f[10:]"

Reverse the order of all fields in each record:

	lcalc list, "f[-*]"

Reverse the text of each record:

	lcalc list, "r[-*]"

Extract fields 2 and 3 from a list:

	lcalc list, "f2,f3"

.nh 2
Merging lists

Simple merge of two lists:

	lcalc "list1,list2", "r1 r2"

.nh 2
Arithmetic operations upon lists

Scale the third field of each record of a list by the factor 5e3
(if there is a third field):

	lcalc list, "f1, f2, nfields >= 3 ? f3 * 1e3, f[4:]"

Print the log of the sum of the first three fields of each record:

	lcalc list, "log(f1+f2+f3)"

Print the sum of the second field of \fIall\fR records in the list:

	lcalc list, "s+=f2; ateof ? 'sum is ' s"

.nh 2
Miscellaneous examples

Execute the program in file "prog" on a list, passing two arguments:

	lcalc list, "@prog", 33, "circle"

Process a list according to a complex program entered via the standard input:

	lcalc list, "@STDIN"

Print record 55 of a list:

	lcalc list, "rnum == 55 ?"

Edit a list, replacing the third field in each record with the string
"circle" if the input value of the field is "box":

	lcalc list, "f1,f2, f3 == box ? circle : f3, f[4:]"

.nh 2
Possible looping examples

    A concise notation for repeating the same operation on a set of fields
would be useful.  Use of a very general construct such as the \fBfor\fR
loop is probably not justified, because such operations can already be
performed in a CL script.  For the list calculator it would be preferable
to have a very concise notation, even if it is not as flexible.  One possible
notation is shown in the example below.  The loop variable "fi" is automatically
assigned, and represents field "i" of the record, where i takes on the values
specified by the section.  If more than a single statement is to be looped
on, braces must be used to group statements.  If loops are nested, the
outer (leftmost) loop is assigned the loop variable fi, the next fj, and so on.

Print the sum of all fields in each record:

	lcalc list, "s=0; [*] s += fi; s"

Reverse the characters in each field, without changing the ordering
of the fields:

	lcalc list, "[*] fi[-*]"
.endhelp