.SH
Section 7. Hints for Preparing Specifications
.PP
This section contains miscellaneous hints on preparing efficient, easy to change,
and clear specifications.
The individual subsections are, more or less,
independent; the reader seeing Yacc for the first
time may well find that
this entire section could be omitted.
.SH
Input Style
.PP
It is difficult to
input rules with substantial actions
and still have a readable specification file.
The following style hints owe much to Brian Kernighan,
and are officially endorsed by the author.
.IP a.
Use all capital letters for token names, all lower case letters for
nonterminal names.
This rule comes under the heading of ``knowing who to blame when
things go wrong.``
.IP b.
Put grammar rules and actions on separate lines.
This allows either to be changed without
an automatic need to change the other.
.IP c.
Put all rules with the same left hand side together.
Put the left hand side in only once, and let all
following rules begin with a vertical bar.
.IP d.
Indent rule bodies by one tab stop, and action bodies by two tab stops.
.PP
The example in Appendix A is written following this style, as are
the examples in the text of this paper (where space permits).
The user must make up his own mind about these stylistic questions;
the central problem, however, is to make the rules visible through
the morass of action code.
.SH
Common Actions
.PP
When several grammar rules have the same action, the user might well wish to
provide only one code sequence.
A simple, general mechanism is, of course, to use subroutine calls.
It is also possible to put a label on the first statement of an action,
and let other actions be simply a goto to this
label.
Thus, if the user had a routine which built trees,
he might wish to have only one call to it, as follows:
.DS
expr :
	expr \'+\' expr =
	{  binary:
		$$ = btree( $1, $2, $3 );
	} |
	expr \'\-\' expr =
	{
		goto binary;
	} |
	expr \'*\' expr =
	{
		goto binary;
	} ;
.DE
.SH
Left Recursion
.PP
The algorithm used by the Yacc parser encourages so called ``left recursive''
grammar rules: rules of the form
.DS
name : name rest\_of\_rule ;
.DE
These rules frequently arise when
writing specifications of sequences and lists:
.DS
list :
	item |
	list \',\' item ;
.DE
and
.DS
sequence :
	item |
	sequence item ;
.DE
Notice that, in each of these cases, the first rule
will be reduced for the first item only, and the second rule
will be reduced for the second and all succeeding items.
.PP
If the user were to write these rules right recursively, such as
.DS
sequence :
	item |
	item sequence ;
.DE
the parser would be a bit bigger, and the items would be seen, and reduced,
from right to left.
More seriously, an internal stack in the parser
would be in danger of overflowing if a very long sequence were read.
Thus, the user should use left recursion wherever reasonable.
.PP
The user should also consider whether a sequence with zero
elements has any meaning, and if so, consider writing
the sequence specification with an empty rule:
.DS
sequence :
	| /* empty */
	sequence item ;
.DE
Once again, the first rule would always be reduced exactly once, before the
first item was read,
and then the second rule would be reduced once for each item read.
Experience suggests that permitting empty sequences
leads to increased generality, which frequently is not evident at the
time the rule is first written.
There are cases, however, when the Yacc algorithm can fail when
such a change is made.
In effect, conflicts might arise when Yacc is asked to decide
which empty sequence it has seen, when it hasn\'t seen enough to
know!
Nevertheless,
this principle is still worth following wherever possible.
.SH
Lexical Tie-ins
.PP
Frequently, there are lexical decisions which depend on the
presence of various constructions in the specification.
For example, the lexical analyzer might want to
delete blanks normally, but not within quoted strings.
Or names might be entered into a symbol table in declarations,
but not in expressions.
.PP
One way of handling these situations is
to create a global flag which is
examined by the lexical analyzer, and set by actions.
For example, consider a situation where we have a program which
consists of 0 or more declarations, followed by 0 or more statements.
We declare a flag called ``dflag'', which is 1 during declarations, and 0 during
statements.
We may do this as follows:
.DS
%{
	int dflag ;
%}
%%
program :
	decls  stats ;

decls :
	= /* empty */
	{
		dflag = 1;
	} |
	decls declaration ;

stats :
	= /* empty */
	{
		dflag = 0;
	} |
	stats statement ;

	. . .  other rules . . .
.DE
The flag dflag is now set to zero when reading statements, and 1 when reading declarations,
.ul
except for the first token in the first statement.
This token must be seen by the parser before it can tell that
the declaration section has ended and the statements have
begun.
Frequently, however, this single token exception does not
affect the lexical scan required.
.PP
Clearly, this kind of ``backdoor'' approach can be elaborated on
to a noxious degree.
Nevertheless, it represents a way of doing some things
that are difficult, if not impossible, to
do otherwise.
.SH
Bundling
.PP
Bundling is a technique for collecting together various character strings
so that they can be output at some later time.
It is derived from a feature of the same name in the compiler/compiler TMG [6].
.PP
Bundling has two components \- a nice user interface,
and a clever implementation trick.
They will be discussed in that order.
.PP
The user interface consists of two routines, ``bundle'' and ``bprint''.
.DS
bundle( a1, a2, . . ., an )
.DE
accepts a variable number of arguments which are either character strings or bundles, and
returns a bundle,
whose value will be the concatenation of the values of a1, . . ., an.
.DS
bprint( b )
.DE
accepts a bundle as argument and outputs its value.
.PP
For example, suppose that we wish to read arithmetic expressions, and output
function calls to routines called ``add'', ``sub'',
``mul'', ``div'', and ``assign''.
Thus, we wish to translate
.DS
a = b \- c*d
.DE
into
.DS
assign(a,sub(b,mul(c,d)))
.DE
.PP
A Yacc specification file which does this is given in Appendix D; this includes
an implementation of the bundle and bprint
routines.
A rule and action of the form
.DS
expr:
	expr \'+\' expr =
	{
		$$ = bundle( "add(", $1, ",", $3, ")" );
	}
.DE
causes the returned value of
expr to be come a bundle, whose value is the
character string containing the desired function call.
Each NAME token has a value which is a pointer to the
actual name which has been read.
Finally, when the entire input line has been read
and the value has been bundled,
the value is written out
and the bundles and names
are cleared, in preparation for the next input line.
.PP
Bundles are implemented as arrays of pointers, terminated by a zero pointer.
Each pointer either points to a bundle or to a character string.
There is an array, called
.ul
bundle space,
which contains all the bundles.
.PP
The implementation trick is to check the values of the pointers in bundles \-
if the pointer points into bundle space, it is assumed to point to a bundle;
otherwise it is assumed to point to a character string.
.PP
The treatment of functions with a variable number of arguments, like bundle,
is likely to differ from one implementation of C to another.
.PP
In general, one may wish to have a simple storage allocator which
allocates and frees bundles,
in order to handle situations where it is not appropriate to completely
clear all of bundle space at one time.
.SH
Reserved Words
.PP
Some programming languages
permit the user to
use words like ``if'', which are normally reserved,
as label or variable names, provided that such use does not
conflict with the legal use of these names in the programming language.
This is extremely hard to do in the framework of Yacc,
since it is difficult to pass the required information to the lexical analyzer
which tells it ``this instance of if is a keyword, and that instance is a variable''.
The user can make a stab at it, using the
mechanism described in the last subsection,
but it is difficult.
.PP
A number of ways of making this easier are under advisement, and one
will probably be supported eventually.
Until this day comes, I suggest that the keywords be
.ul
reserved;
that is, be forbidden for use as variable names.
There are powerful stylistic reasons for preferring this, anyway
(he said weakly . . . ).
.SH
Non-integer Values
.PP
Frequently, the user wishes to have values which are
bigger than integers;
again, this is an area where Yacc does not make the job as easy as it might,
and some additional support is likely.
Nevertheless, at the cost of writing a storage manager,
the user can return pointers or indices to blocks of storage
big enough to contain the full values desired.
.SH
Previous Work
.PP
There have been many previous applications of Yacc.
The user who is contemplating a big application might well
find that others have developed relevant techniques,
or even portions of grammars.
Yacc specifications appear to be easier to change than
the equivalent computer programs, so that the ``prior art'' is more
relevant here as well.