lookahead.html | searchcode

/projects/javacc-5.0/www/doc/lookahead.html

https://gitlab.com/essere.lab.public/qualitas.class-corpus
HTML | 986 lines | 894 code | 61 blank | 31 comment | 0 complexity | d595b141bd0f05c48ffd8f11760c2d4a MD5 | raw file

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<!--

Copyright (c) 2006, Sun Microsystems, Inc.
All rights reserved.

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:

    * Redistributions of source code must retain the above copyright notice,
      this list of conditions and the following disclaimer.
    * Redistributions in binary form must reproduce the above copyright
      notice, this list of conditions and the following disclaimer in the
      documentation and/or other materials provided with the distribution.
    * Neither the name of the Sun Microsystems, Inc. nor the names of its
      contributors may be used to endorse or promote products derived from
      this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
THE POSSIBILITY OF SUCH DAMAGE.

-->
<head>
 <title>JavaCC: LOOKAHEAD MiniTutorial</title>
<!-- Changed by: Michael Van De Vanter, 14-Jan-2003 -->
</head>
<body bgcolor="#FFFFFF" >

<h1>JavaCC [tm]: LOOKAHEAD MiniTutorial</h1>


<p>
<strong>This tutorial refers
to examples that are available in the Lookahead directory under the
examples directory of the release.  Currently, this page is a copy of
the contents of the README file within that directory.
</strong>
</p>


<h1>Lookahead tutorial</h1>

<p>
We assume that you have already taken a look at some of the simple
examples provided in the release before you read this section.
</p>

<h2>WHAT IS LOOKAHEAD?</h2>

<p>
The job of a parser is to read an input stream and determine whether
or not the input stream conforms to the grammar.
</p>
<p>
This determination in its most general form can be quite time
consuming.  Consider the following example (file Example1.jj):
</p>
<pre><tt>

	void Input() :
	{}
	{
	  "a" BC() "c"
	}

	void BC() :
	{}
	{
	  "b" [ "c" ]
	}
</tt></pre>
<p>
In this simple example, it is quite clear that there are exactly two
strings that match the above grammar, namely:
</p>
<pre><tt>
	abc
	abcc
</tt></pre>
<p>
The general way to perform this match is to walk through the grammar
based on the string as follows.  Here, we use "abc" as the input
string:
</p>
<ul>
<li><b>Step 1.</b>
There is only one choice here - the first input character
must be 'a' - and since that is indeed the case, we are OK.
</li>
<li><b>Step 2.</b>
We now proceed on to non-terminal BC.  Here again, there is
only one choice for the next input character - it must be 'b'.  The
input matches this one too, so we are still OK.
</li>
<li><b>Step 3.</b>
We now come to a "choice point" in the grammar.
We can either
go inside the [...] and match it, or ignore it altogether.  We decide
to go inside.  So the next input character must be a 'c'.  We are
again OK.
</li>
<li><b>Step 4.</b>
Now we have completed with non-terminal BC and go back to
non-terminal Input.  Now the grammar says the next character must be
yet another 'c'.  But there are no more input characters.  So we have
a problem.
</li>
<li><b>Step 5.</b>
When we have such a problem in the general case, we conclude
that we may have made a bad choice somewhere.  In this case, we made
the bad choice in Step 3.  So we retrace our steps back to step 3 and
make another choice and try that.  This process is called
"backtracking".
</li>
<li><b>Step 6.</b>
 We have now backtracked and made the other choice we could
have made at Step 3 - namely, ignore the [...].  Now we have completed
with non-terminal BC and go back to non-terminal Input.  Now the
grammar says the next character must be yet another 'c'.  The next
input character is a 'c', so we are OK now.
</li>
<li><b>Step 7.</b>
 We realize we have reached the end of the grammar (end of
non-terminal Input) successfully.  This means we have successfully
matched the string "abc" to the grammar.
</li>
</ul>
<p>
As the above example indicates, the general problem of matching an
input with a grammar may result in large amounts of backtracking and
making new choices and this can consume a lot of time.  The amount of
time taken can also be a function of how the grammar is written.  Note
that many grammars can be written to cover the same set of inputs - or
the same language (i.e., there can be multiple equivalent grammars for
the same input language).
</p>
<hr />
<p>
For example, the following grammar would speed up the parsing of the
same language as compared to the previous grammar:
</p>
<pre><tt>
	void Input() :
	{}
	{
	  "a" "b" "c" [ "c" ]
	}
</tt></pre>
<p>
while the following grammar slows it down even more since the parser
has to backtrack all the way to the beginning:
</p>
<pre><tt>
	void Input() :
	{}
	{
	  "a" "b" "c" "c"
	|
	  "a" "b" "c"
	}
</tt></pre>
<p>
One can even have a grammar that looks like the following:
</p>
<pre><tt>

	void Input() :
	{}
	{
	  "a" ( BC1() | BC2() )
	}

	void BC1() :
	{}
	{
	  "b" "c" "c"
	}

	void BC2() :
	{}
	{
	  "b" "c" [ "c" ]
	}
</tt></pre>
<p>
This grammar can match "abcc" in two ways, and is therefore considered
"ambiguous".
</p>
<hr />
<p>
The performance hit from such backtracking is unacceptable for most
systems that include a parser.  Hence most parsers do not backtrack in
this general manner (or do not backtrack at all), rather they make
decisions at choice points based on limited information and then
commit to it.
</p>
<p>
Parsers generated by Java Compiler Compiler make decisions at choice
points based on some exploration of tokens further ahead in the input
stream, and once they make such a decision, they commit to it.  i.e.,
No backtracking is performed once a decision is made.
</p>
<p>

The process of exploring tokens further in the input stream is termed
"looking ahead" into the input stream - hence our use of the term
"LOOKAHEAD".
</p>
<p>
Since some of these decisions may be made with less than perfect
information (JavaCC [tm] will warn you in these situations, so you don't
have to worry), you need to know something about LOOKAHEAD to make
your grammar work correctly.
</p>
<p>
The two ways in which you make the choice decisions work properly are:
</p>
<ul>
<li>
 Modify the grammar to make it simpler.
</li>
<li>
 Insert hints at the more complicated choice points to help the
  parser make the right choices.
</li>
</ul>

<h2>CHOICE POINTS IN JAVACC GRAMMARS</h2>

<p>
There are 4 different kinds of choice points in JavaCC:
</p>
<ol>
<li>
  An expansion of the form: ( exp1 | exp2 | ... ).  In this case, the
  generated parser has to somehow determine which of exp1, exp2, etc.
  to select to continue parsing.
</li>
<li>
  An expansion of the form: ( exp )?.  In this case, the generated parser
  must somehow determine whether to choose exp or to continue beyond
  the ( exp )? without choosing exp.  Note: ( exp )? may also be written
  as [ exp ].
</li>
<li>
  An expansion of the form ( exp )*.  In this case, the generated parser
  must do the same thing as in the previous case, and furthermore, after
  each time a successful match of exp (if exp was chosen) is completed,
  this choice determination must be made again.
</li>
<li>
  An expansion of the form ( exp )+.  This is essentially similar to
  the previous case with a mandatory first match to exp.
</li>
</ol>
<p>
Remember that token specifications that occur within angular
brackets &lt;...&gt; also have choice points.  But these choices are made
in different ways and are the subject of a different tutorial.
</p>

<h2>THE DEFAULT CHOICE DETERMINATION ALGORITHM</h2>

<p>
The default choice determination algorithm looks ahead 1 token in the
input stream and uses this to help make its choice at choice points.
</p>
<p>
The following examples will describe the default algorithm fully:
</p>
<p>
Consider the following grammar (file Example2.jj):
</p>
<pre><tt>
	void basic_expr() :
	{}
	{
	  &lt;ID&gt; "(" expr() ")"	// Choice 1
	|
	  "(" expr() ")"	// Choice 2
	|
	  "new" &lt;ID&gt;		// Choice 3
	}
</tt></pre>
<p>
The choice determination algorithm works as follows:
</p>
<pre><tt>

	if (next token is &lt;ID&gt;) {
	  choose Choice 1
	} else if (next token is "(") {
	  choose Choice 2
	} else if (next token is "new") {
	  choose Choice 3
	} else {
	  produce an error message
	}
</tt></pre>
<hr />
<p>
In the above example, the grammar has been written such that the
default choice determination algorithm does the right thing.  Another
thing to note is that the choice determination algorithm works in a
top to bottom order - if Choice 1 was selected, the other choices are
not even considered.  While this is not an issue in this example
(except for performance), it will become important later below when
local ambiguities require the insertion of LOOKAHEAD hints.
</p>
<p>
Suppose the above grammar was modified to (file Example3.jj):
</p>
<pre><tt>
	void basic_expr() :
	{}
	{
	  &lt;ID&gt; "(" expr() ")"	// Choice 1
	|
	  "(" expr() ")"	// Choice 2
	|
	  "new" &lt;ID&gt;		// Choice 3
	|
	  &lt;ID&gt; "." &lt;ID&gt;		// Choice 4
	}

</tt></pre>
<p>
Then the default algorithm will always choose Choice 1 when the next
input token is &lt;ID&gt; and never choose Choice 4 even if the token
following &lt;ID&gt; is a ".".  More on this later.
</p>
<p>
You can try running the parser generated from Example3.jj on the input
"id1.id2".  It will complain that it encountered a "." when it was
expecting a "(".  Note - when you built the parser, it would have
given you the following warning message:
</p>
<pre><tt>
Warning: Choice conflict involving two expansions at
         line 25, column 3 and line 31, column 3 respectively.
         A common prefix is: &lt;ID&gt;
         Consider using a lookahead of 2 for earlier expansion.
</tt></pre>
<p>
Essentially, JavaCC is saying it has detected a situation in your
grammar which may cause the default lookahead algorithm to do strange
things.  The generated parser will still work using the default
lookahead algorithm - except that it may not do what you expect of it.
</p>
<hr />
<p>
Now consider the following example (file Example 4.jj):
</p>
<pre><tt>

	void identifier_list() :
	{}
	{
	  &lt;ID&gt; ( "," &lt;ID&gt; )*
	}

</tt></pre>
<p>
Suppose the first &lt;ID&gt; has already been matched and that the parser
has reached the choice point (the (...)* construct).  Here's how the
choice determination algorithm works:
</p>
<pre><tt>

	while (next token is ",") {
	  choose the nested expansion (i.e., go into the (...)* construct)
	  consume the "," token
	  if (next token is &lt;ID&gt;) consume it, otherwise report error
	}

</tt></pre>
<hr />
<p>
In the above example, note that the choice determination algorithm
does not look beyond the (...)* construct to make its decision.
Suppose there was another production in that same grammar as follows
(file Example5.jj):
</p>
<pre><tt>
	void funny_list() :
	{}
	{
	  identifier_list() "," &lt;INT&gt;
	}
</tt></pre>
<p>
When the default algorithm is making a choice at ( "," &lt;ID&gt; )*, it
will always go into the (...)* construct if the next token is a ",".
It will do this even when identifier_list was called from funny_list
and the token after the "," is an &lt;INT&gt;.  Intuitively, the right thing
to do in this situation is to skip the (...)* construct and return to
funny_list.  More on this later.
</p>
<p>
As a concrete example, suppose your input was "id1, id2, 5", the
parser will complain that it encountered a 5 when it was expecting an
&lt;ID&gt;.  Note - when you built the parser, it would have given you the
following warning message:
</p>
<pre><tt>

Warning: Choice conflict in (...)* construct at line 25, column 8.
         Expansion nested within construct and expansion following construct
         have common prefixes, one of which is: ","
         Consider using a lookahead of 2 or more for nested expansion.
</tt></pre>
<p>
Essentially, JavaCC is saying it has detected a situation in your
grammar which may cause the default lookahead algorithm to do strange
things.  The generated parser will still work using the default
lookahead algorithm - except that it may not do what you expect of it.
</p>
<hr />
<p>We have shown you examples of two kinds of choice points in the
examples above - "exp1 | exp2 | ...", and "(exp)*".  The other two
kinds of choice points - "(exp)+" and "(exp)?" - behave similarly to
(exp)* and we will not be providing examples of their use here.
</p>

<h3>MULTIPLE TOKEN LOOKAHEAD SPECIFICATIONS</h3>

<p>
So far, we have described the default lookahead algorithm of the
generated parsers.  In the majority of situations, the default
algorithm works just fine.  In situations where it does not work
well, Java Compiler Compiler provides you with warning messages like
the ones shown above.  If you have a grammar that goes through
Java Compiler Compiler without producing any warnings, then the
grammar is a LL(1) grammar.  Essentially, LL(1) grammars are those
that can be handled by top-down parsers (such as those generated
by Java Compiler Compiler) using at most one token of LOOKAHEAD.
</p>
<p>
When you get these warning messages, you can do one of two things.
</p>
<p><b>Option 1</b></p>
<p>
You can modify your grammar so that the warning messages go away.
That is, you can attempt to make your grammar LL(1) by making some
changes to it.
</p>
<p>
The following (file Example6.jj) shows how you may change Example3.jj
to make it LL(1):
</p>
<pre><tt>
	void basic_expr() :
	{}
	{
	  &lt;ID&gt; ( "(" expr() ")" | "." &lt;ID&gt; )
	|
	  "(" expr() ")"
	|
	  "new" &lt;ID&gt;
	}

</tt></pre>
<p>
What we have done here is to factor the fourth choice into the first
choice.  Note how we have placed their common first token &lt;ID&gt; outside
the parentheses, and then within the parentheses, we have yet another
choice which can now be performed by looking at only one token in the
input stream and comparing it with "(" and ".".  This process of
modifying grammars to make them LL(1) is called "left factoring".
</p>
<p>
The following (file Example7.jj) shows how Example5.jj may be changed
to make it LL(1):
</p>
<pre><tt>
	void funny_list() :
	{}
	{
	  &lt;ID&gt; "," ( &lt;ID&gt; "," )* &lt;INT&gt;
	}
</tt></pre>
<p>
Note that this change is somewhat more drastic.
</p>
<hr />
<p><b>Option 2</b></p>
<p>
You can provide the generated parser with some hints to help it out
in the non-LL(1) situations that the warning messages bring to your
attention.
</p>
<p>
All such hints are specified using either setting the global LOOKAHEAD
value to a larger value (see below) or by using the LOOKAHEAD(...)
construct to provide a local hint.
</p>
<p>
A design decision must be made to determine if Option 1 or Option 2 is
the right one to take.  The only advantage of choosing Option 1 is
that it makes your grammar perform better.  JavaCC generated parsers
can handle LL(1) constructs much faster than other constructs.
However, the advantage of choosing Option 2 is that you have a simpler
grammar - one that is easier to develop and maintain - one that
focuses on human-friendliness and not machine-friendliness.
</p>
<p>

Sometimes Option 2 is the only choice - especially in the presence of
user actions.  Suppose Example3.jj contained actions as shown below:
</p>
<pre><tt>

	void basic_expr() :
	{}
	{
	  { initMethodTables(); } &lt;ID&gt; "(" expr() ")"
	|
	  "(" expr() ")"
	|
	  "new" &lt;ID&gt;
	|
	  { initObjectTables(); } &lt;ID&gt; "." &lt;ID&gt;
	}
</tt></pre>
<p>
Since the actions are different, left-factoring cannot be performed.

</p>

<h3>SETTING A GLOBAL LOOKAHEAD SPECIFICATION</h3>

<p>
You can set a global LOOKAHEAD specification by using the option
"LOOKAHEAD" either from the command line, or at the beginning of the
grammar file in the options section.  The value of this option is an
integer which is the number of tokens to look ahead when making choice
decisions.  As you may have guessed, the default value of this option
is 1 - which derives the default LOOKAHEAD algorithm described above.
</p>
<p>
Suppose you set the value of this option to 2.  Then the LOOKAHEAD
algorithm derived from this looks at two tokens (instead of just one
token) before making a choice decision.  Hence, in Example3.jj, choice
1 will be taken only if the next two tokens are &lt;ID&gt; and "(", while
choice 4 will be taken only if the next two tokens are &lt;ID&gt; and ".".
Hence, the parser will now work properly for Example3.jj.  Similarly,
the problem with Example5.jj also goes away since the parser goes into
the (...)* construct only when the next two tokens are "," and &lt;ID&gt;.
</p>
<p>
By setting the global LOOKAHEAD to 2, the parsing algorithm
essentially becomes LL(2).  Since you can set the global LOOKAHEAD to
any value, parsers generated by Java Compiler Compiler are called
LL(k) parsers.
</p>

<h3>SETTING A LOCAL LOOKAHEAD SPECIFICATION</h3>

<p>
You can also set a local LOOKAHEAD specification that affects only a
specific choice point.  This way, the majority of the grammar can
remain LL(1) and hence perform better, while at the same time one gets
the flexibility of LL(k) grammars.  Here's how Example3.jj is modified
with local LOOKAHEAD to fix the choice ambiguity problem (file
Example8.jj):
</p>
<pre><tt>

	void basic_expr() :
	{}
	{
	  LOOKAHEAD(2)
	  &lt;ID&gt; "(" expr() ")"	// Choice 1
	|
	  "(" expr() ")"	// Choice 2
	|
	  "new" &lt;ID&gt;		// Choice 3
	|
	  &lt;ID&gt; "." &lt;ID&gt;		// Choice 4
	}
</tt></pre>
<p>
Only the first choice (the first condition in the translation below)
is affected by the LOOKAHEAD specification.  All others continue to
use a single token of LOOKAHEAD:
</p>
<pre><tt>

	if (next 2 tokens are &lt;ID&gt; and "(" ) {
	  choose Choice 1
	} else if (next token is "(") {
	  choose Choice 2
	} else if (next token is "new") {
	  choose Choice 3
	} else if (next token is &lt;ID&gt;) {
	  choose Choice 4
	} else {
	  produce an error message
	}
</tt></pre>
<p>
Similarily, Example5.jj can be modified as shown below (file
Example9.jj):
</p>
<pre><tt>

	void identifier_list() :
	{}
	{
	  &lt;ID&gt; ( LOOKAHEAD(2) "," &lt;ID&gt; )*
	}
</tt></pre>
<p>
Note, the LOOKAHEAD specification has to occur inside the (...)* which
is the choice is being made.  The translation for this construct is
shown below (after the first &lt;ID&gt; has been consumed):
</p>
<pre><tt>

	while (next 2 tokens are "," and &lt;ID&gt;) {
	  choose the nested expansion (i.e., go into the (...)* construct)
	  consume the "," token
	  consume the &lt;ID&gt; token
	}
</tt></pre>
<hr />
<p>
We strongly discourage you from modifying the global LOOKAHEAD
default.  Most grammars are predominantly LL(1), hence you will be
unnecessarily degrading performance by converting the entire grammar
to LL(k) to facilitate just some portions of the grammar that are not
LL(1).  If your grammar and input files being parsed are very small,
then this is okay.
</p>
<p>
You should also keep in mind that the warning messages JavaCC prints
when it detects ambiguities at choice points (such as the two messages
shown earlier) simply tells you that the specified choice points are
not LL(1).  JavaCC does not verify the correctness of your local
LOOKAHEAD specification - it assumes you know what you are doing, in
fact, it really cannot verify the correctness of local LOOKAHEAD's as
the following example of if statements illustrates (file
Example10.jj):
</p>
<pre><tt>

	void IfStm() :
	{}
	{
	 "if" C() S() [ "else" S() ]
	}

	void S() :
	{}
	{
	  ...
	|
	  IfStm()
	}

</tt></pre>
<p>
This example is the famous "dangling else" problem.  If you have a
program that looks like:
</p>
<pre><tt>

	"if C1 if C2 S1 else S2"

</tt></pre>
<p>
The "else S2" can be bound to either of the two if statements.  The
standard interpretation is that it is bound to the inner if statement
(the one closest to it).  The default choice determination algorithm
happens to do the right thing, but it still prints the following
warning message:
</p>
<pre><tt>

Warning: Choice conflict in [...] construct at line 25, column 15.
         Expansion nested within construct and expansion following construct
         have common prefixes, one of which is: "else"
         Consider using a lookahead of 2 or more for nested expansion.

</tt></pre>
<p>
To suppress the warning message, you could simply tell JavaCC that
you know what you are doing as follows:
</p>
<pre><tt>
	void IfStm() :
	{}
	{
	 "if" C() S() [ LOOKAHEAD(1) "else" S() ]
	}
</tt></pre>
<p>
To force lookahead ambiguity checking in such instances, set the option
FORCE_LA_CHECK to true.
</p>

<h3>SYNTACTIC LOOKAHEAD</h3>

<p>
Consider the following production taken from the Java grammar:
</p>
<pre><tt>

	void TypeDeclaration() :
	{}
	{
	  ClassDeclaration()
	|
	  InterfaceDeclaration()
	}
</tt></pre>
<p>
At the syntactic level, ClassDeclaration can start with any number of
"abstract"s, "final"s, and "public"s.  While a subsequent semantic
check will produce error messages for multiple uses of the same
modifier, this does not happen until parsing is completely over.
Similarly, InterfaceDeclaration can start with any number of
"abstract"s and "public"s.
</p>
<p>
What if the next tokens in the input stream are a very large number of
"abstract"s (say 100 of them) followed by "interface"?  It is clear
that a fixed amount of LOOKAHEAD (such as LOOKAHEAD(100) for example)
will not suffice.  One can argue that this is such a weird situation
that it does not warrant any reasonable error message and that it is
okay to make the wrong choice in some pathological situations.  But
suppose one wanted to be precise about this.
</p>
<p>
The solution here is to set the LOOKAHEAD to infinity - that is set no
bounds on the number of tokens to look ahead.  One way to do this is
to use a very large integer value (such as the largest possible
integer) as follows:
</p>
<pre><tt>
	void TypeDeclaration() :
	{}
	{
	  LOOKAHEAD(2147483647)
	  ClassDeclaration()
	|
	  InterfaceDeclaration()
	}
</tt></pre>
<p>
One can also achieve the same effect with "syntactic LOOKAHEAD".  In
syntactic LOOKAHEAD, you specify an expansion to try out and it that
succeeds, then the following choice is taken.  The above example is
rewritten using syntactic LOOKAHEAD below:
</p>
<pre><tt>

	void TypeDeclaration() :
	{}
	{
	  LOOKAHEAD(ClassDeclaration())
	  ClassDeclaration()
	|
	  InterfaceDeclaration()
	}
</tt></pre>
<p>
Essentially, what this is saying is:
</p>
<pre><tt>

	if (the tokens from the input stream match ClassDeclaration) {
	  choose ClassDeclaration()
	} else if (next token matches InterfaceDeclaration) {
	  choose InterfaceDeclaration()
	} else {
	  produce an error message
	}

</tt></pre>
<p>
The problem with the above syntactic LOOKAHEAD specification is that
the LOOKAHEAD calculation takes too much time and does a lot of
unnecessary checking.  In this case, the LOOKAHEAD calculation can
stop as soon as the token "class" is encountered, but the
specification forces the calculation to continue until the end of the
class declaration has been reached - which is rather time consuming.
This problem can be solved by placing a shorter expansion to try out
in the syntactic LOOKAHEAD specification as in the following example:
</p>
<pre><tt>

	void TypeDeclaration() :
	{}
	{
	  LOOKAHEAD( ( "abstract" | "final" | "public" )* "class" )
	  ClassDeclaration()
	|
	  InterfaceDeclaration()
	}
</tt></pre>
<p>
Essentially, what this is saying is:
</p>
<pre><tt>
	if (the nest set of tokens from the input stream are a sequence of
	    "abstract"s, "final"s, and "public"s followed by a "class") {
	  choose ClassDeclaration()
	} else if (next token matches InterfaceDeclaration) {
	  choose InterfaceDeclaration()
	} else {
	  produce an error message
	}
</tt></pre>
<p>
By doing this, you make the choice determination algorithm stop as
soon as it sees "class" - i.e., make its decision at the earliest
possible time.
</p>
<p>
You can place a bound on the number of tokens to consume during
syntactic lookahead as follows:
</p>
<pre><tt>
	void TypeDeclaration() :
	{}
	{
	  LOOKAHEAD(10, ( "abstract" | "final" | "public" )* "class" )
	  ClassDeclaration()
	|
	  InterfaceDeclaration()
	}
</tt></pre>
<p>
In this case, the LOOKAHEAD determination is not permitted to go beyond
10 tokens.  If it reaches this limit and is still successfully matching
( "abstract" | "final" | "public" )* "class", then ClassDeclaration is
selected.
</p>
<p>
Actually, when such a limit is not specified, it defaults to the largest
integer value (2147483647).
</p>

<h3>SEMANTIC LOOKAHEAD</h3>

<p>
Let us go back to Example1.jj:
</p>
<pre><tt>

	void Input() :
	{}
	{
	  "a" BC() "c"
	}

	void BC() :
	{}
	{
	  "b" [ "c" ]
	}
</tt></pre>
<p>
Let us suppose that there is a good reason for writing a grammar this
way (maybe the way actions are embedded).  As noted earlier, this
grammar recognizes two string "abc" and "abcc".  The problem here is
that the default LL(1) algorithm will choose the [ "c" ] every time
it sees a "c" and therefore "abc" will never be matched.  We need to
specify that this choice must be made only when the next token is a
"c", and the token following that is not a "c".  This is a negative
statement - one that cannot be made using syntactic LOOKAHEAD.
</p>
<p>
We can use semantic LOOKAHEAD for this purpose.  With semantic
LOOKAHEAD, you can specify any arbitrary boolean expression whose
evaluation determines which choice to take at a choice point.  The
above example can be instrumented with semantic LOOKAHEAD as follows:
</p>
<pre><tt>
	void BC() :
	{}
	{
	  "b"
	  [ LOOKAHEAD( { getToken(1).kind == C &amp;&amp; getToken(2).kind != C } )
	    &lt;C:"c"&gt;
	  ]
	}
</tt></pre>
<p>
First we give the token "c" a label C so that we can refer to it from
the semantic LOOKAHEAD.  The boolean expression essentially states the
desired property.  The choice determination decision is therefore:
</p>
<pre><tt>
	if (next token is "c" and following token is not "c") {
	  choose the nested expansion (i.e., go into the [...] construct)
	} else {
	  go beyond the [...] construct without entering it.
	}
</tt></pre>
<p>
This example can be rewritten to combine both syntactic and semantic
LOOKAHEAD as follows (recognize the first "c" using syntactic
LOOKAHEAD and the absence of the second using semantic LOOKAHEAD):
</p>
<pre><tt>
	void BC() :
	{}
	{
	  "b"
	  [ LOOKAHEAD( "c", { getToken(2).kind != C } )
	    &lt;C:"c"&gt;
	  ]
	}
</tt></pre>

<h3>GENERAL STRUCTURE OF LOOKAHEAD</h3>

<p>
We've pretty much covered the various aspects of LOOKAHEAD in the
previous sections.  A couple of advanced topics follow.  However,
we shall now present a formal language reference for LOOKAHEAD in
Java Compiler Compiler:
</p>
<p>
The general structure of a LOOKAHEAD specification is:
</p>
<pre><tt>

	LOOKAHEAD( amount,
	           expansion,
	           { boolean_expression }
	         )
</tt></pre>
<p>
"amount" specifies the number of tokens to LOOKAHEAD,"expansion"
specifies the expansion to use to perform syntactic LOOKAHEAD, and
"boolean_expression" is the expression to use for semantic
LOOKAHEAD.
</p>
<p>
At least one of the three entries must be present.  If more than
one are present, they are separated by commas.  The default values
for each of these entities is defined below:
</p>
<pre><tt>

"amount":
 - if "expansion is present, this defaults to 2147483647.
 - otherwise ("boolean_expression" must be present then) this
   defaults to 0.

Note: When "amount" is 0, no syntactic LOOKAHEAD is performed.  Also,
"amount" does not affect the semantic LOOKAHEAD.

"expansion":
- defaults to the expansion being considered.

"boolean_expression":

- defaults to true.
</tt></pre>

<h3>NESTED EVALUATION OF SEMANTIC LOOKAHEAD</h3>
<p>
To be done.
</p>

<h3>JAVACODE PRODUCTIONS</h3>
<p>
To be done. 
</p>

</body>
</html>