DIFFERENCES - differences between TF and ROFF.

Summary:

32 character register names
underbar (_) now accepted in identifiers
must be a space after macro commands
tabbing done differently
obsolete ROFF commands removed
.if for conditional processing
.sq squeezes lines
.zt deletes macros
.tt causes break
additional registers
hyphenation default .hy 2
virtually unlimited nesting of macro expansions
.ix changed to .di
.nx changed to .dn
new request ".sl" to control formfeeds
new request ".ln" to control line-numbering
default .tb 1
titles MUST have three fields

Details:

TF is a descendant of the ROFF text formatter with several enchancements and repairs. In particular, it allows almost unlimited nesting of macro expansions (text registers), 32 character register names, and handles the operation of escaped new-line characters properly.

TF accepts the syntax of nearly all ROFF commands with the following exceptions:

TF has no short/long name mode. This means that commands involving parameters after the request (on the same line) must have a delimiter between the request and the parameter. For example, .ce5 will call the macro expander to attempt to expand the name ce5, not centre the next five lines. The center command must be given as ".ce 5".
Because identifiers can be up to 32 characters in length, a break character has been added to the set of characters allowed in names. Name characters allowed are now a-z, 0-9, and underbar (_). Note that upper case alphabetics in names are mapped automatically to lower case. It is recommended that names beginning with _, # and % be avoided. In future, these characters will be used to identify special information registers maintained by TF. For example, "%null_name" has special internal significance and anomalous results may occur if it is used in TF requests.
Tab handling in TF is significantly different. EVERY input line containing a tab character undergoes tab processing; in particular, the presence of a tab character in an input line causes an automatic line break.
Tabs are used to separate fields. Field 0, the field before the first tab character, is output beginning in column 1. The ".ta" request specifies the positions of each of the remaining fields. This request has the following form.
```
.ta <format_item> [<format_item>]*
```
Acceptable format items include numbers, signed numbers, and the letters "L", "C", and "R" (upper or lower case).

An unsigned number in ".ta" indicates an absolute column position. For example, the pair "8 16" refer to column 8 and column 16 on the output line. Signed numbers indicate relative column positions. For example, the pair "8 +4" refer columns 8 and 12.

An "L" indicates left-alignment of a field; the column used for this alignment is given by the closest absolute or relative position to the LEFT of the "L". Similarly, an "R" indicates right-alignment of a field; the column used is given by the closest absolute or relative position to the RIGHT of the "R".

A "C" indicates that a field should be centred between two columns. The left column will be the closest absolute or relative position on the LEFT of the "C" and the right column will be the closest absolute or relative position on the RIGHT of the "C".

In most cases, every field will require a letter indicating whether the field should be left-aligned, right-aligned, or centred. If two column positions are adjacent on the ".ta" request, there is assumed to be an inaccesible gutter between them which will be filled with non-fillable blanks.

If the first format item for ".ta" is a letter, a "1" will be implicitly assumed to precede that letter. If the last format item is a letter, TF will implicitly assume that the letter is followed by an unsigned number equal to the current line length. Thus if the current line length is 60, ".ta C" is equivalent to ".ta 1 C 60".

Examples:
```
.ta 10 20 30
```
Unlike roff, all that happens is that an unfillable gutter is created from columns 11 to 30. No tabs stops are set.
```
.ta 10 L C R 65
```
indicates that the first field is left-aligned in column 10, the second is centred between 10 and 65, and the third is right-aligned in column 65. Note that the "C" does NOT need numbers immediately surrounding it; it uses the numbers CLOSEST to it, even if those numbers are also used by another letter. Note that it is necessary to have the correct number of tab characters to reach each field. That is, each format "letter" defines the existence of a tab stop.

It is possible to set multiple tab stops in a column. If fields overlap because of the way stops are set, they will overstrike each other on output.
Certain obsolete Roff constructs are not recognized by TF, and others will be weeded out as development continues. These commands include the following:
```
.ab     .ad     .bm     .na     .nc
.tm     .un     .va-z   .v%     .v#
.ya-z   .y%     .y#     .n0-2
```
A new directive has been added:
```
.zt R    remove text register (zap text)
```
This request has been added so that macros may be defined, then deleted when no longer needed. This feature can also be useful in removing items from the macro list that are no longer needed for processing the output. Use of the ".zt" request can speed up processing by reducing the length of the macro list (kept as a top down linked list).
The request
```
.if E L
```
can be used for conditional processing. The lines following the ".if" will be processed only if the numeric expression E is non-zero; otherwise, TF will ignore everything until it finds an ".en" with a matching label L.
The request
```
.sq
```
can be used to squeeze lines onto the current page (in order to avoid widows). It is equivalent to
```
.m3 -<ls>
.br
.m3 +<ls>
```
where <ls> is used to represent the current line spacing (e.g. 1 for single spacing).
TF hyphenation is the same as ROFF except that the default is set to ".hy 2".
The ".tt" request now does an automatic line break. Also, if pagination has been turned off, then .tt is ignored except for the line break.

The following text registers are maintained automatically by TF:

(%amon) -- contains the current month name,
           e.g. "January"
(%wday) -- contains the current week day,
           e.g. "Thursday"

In the interests of consistency, index files will be known as "diversions" (since other commands use the term "index file" for something much different). To correspond to this, TF uses the request ".di" in place of ".ix" and ".dn" in place of ".nx".
TF does NOT begin a file with a formfeed character, even if "+FormFeed" is in effect.
The request ".sl 0" tells TF not to output formfeeds at the end of a page (thus it corresponds to -FormFeed on the command line). The request ".sl 1" tells TF to output formfeeds even if -FormFeed was specified.
The request ".n0", ".n1", and ".n2" have been replaced by ".ln 0", ".ln 1", and ".ln 2".
The default setting for ".tb" is 1 (in ROFF it is 0).
Parentheses are not required around register names and labels that act as positional parameters for requests. Thus you can say either
```
 .an (reg) 2
 .an reg 2
```
This only applies in situations where the name is positional in requests. Parentheses are still compulsory for register names used in insertion constructs, e.g. ^(reg).
Titles MUST consist of three fields. This means that they must contain four quote marks. For example, when specifying a footnote separator of "----" you must say
```
 .fs '----'''
```
Copyright © 1996, Thinkage Ltd.