2 ------------------------------------------------------------
3 This is the second part of a two part file.
4 This is a list of changes to pccts 1.33 prior to MR13
5 For more recent information see CHANGES_FROM_133.txt
6 ------------------------------------------------------------
10 The software and these notes are provided "as is". They may include
11 typographical or technical errors and their authors disclaims all
12 liability of any kind or nature for damages due to error, fault,
13 defect, or deficiency regardless of cause. All warranties of any
14 kind, either express or implied, including, but not limited to, the
15 implied warranties of merchantability and fitness for a particular
16 purpose are disclaimed.
19 #153. (Changed in MR12b) Bug in computation of -mrhoist suppression set
21 Consider the following grammar with k=1 and "-mrhoist on":
23 r1 : (A)? => ((p>>? x /* l1 */
27 | (B)? => <<q>>? y /* l5 */
30 In earlier versions the mrhoist routine would see that both l1 and
31 l2 contained predicates and would assume that this prevented either
32 from acting to suppress the other predicate. In the example above
33 it didn't realize the A at line l4 is capable of suppressing the
34 predicate at l1 even though alt l2 contains (indirectly) a predicate.
36 This is fixed in MR12b.
38 Reported by Reinier van den Born (reinier@vnet.ibm.com)
40 #153. (Changed in MR12a) Bug in computation of -mrhoist suppression set
42 An oversight similar to that described in Item #152 appeared in
43 the computation of the set that "covered" a predicate. If a
44 predicate expression included a term such as p=AND(q,r) the context
45 of p was taken to be context(q) & context(r), when it should have
46 been context(q) | context(r). This is fixed in MR12a.
48 #152. (Changed in MR12) Bug in generation of predicate expressions
50 The primary purpose for MR12 is to make quite clear that MR11 is
51 obsolete and to fix the bug related to predicate expressions.
53 In MR10 code was added to optimize the code generated for
54 predicate expression tests. Unfortunately, there was a
55 significant oversight in the code which resulted in a bug in
56 the generation of code for predicate expression tests which
57 contained predicates combined using AND:
60 r1 : (AAA)? => <<p LATEXT(1)>>? r2 ;
61 r2 : (BBB)? => <<q LATEXT(1)>>? Q
62 | (BBB)? => <<r LATEXT(1)>>? Q
65 In MR11 (and MR10 when using "-mrhoist on") the code generated
66 for r0 to predict r1 would be equivalent to:
69 (LA(1)==AAA && LA(1)==BBB) &&
70 ( p && ( q || r )) ) {
72 This is incorrect because it expresses the idea that LA(1)
73 *must* be AAA in order to attempt r1, and *must* be BBB to
74 attempt r2. The result was that r1 became unreachable since
75 both condition can not be simultaneously true.
77 The general philosophy of code generation for predicates
78 can be summarized as follows:
80 a. If the context is true don't enter an alt
81 for which the corresponding predicate is false.
83 If the context is false then it is okay to enter
84 the alt without evaluating the predicate at all.
86 b. A predicate created by ORing of predicates has
87 context which is the OR of their individual contexts.
89 c. A predicate created by ANDing of predicates has
90 (surprise) context which is the OR of their individual
93 d. Apply these rules recursively.
97 The correct code should express the idea that *if* LA(1) is
98 AAA then p must be true to attempt r1, but if LA(1) is *not*
99 AAA then it is okay to attempt r1, provided that *if* LA(1) is
100 BBB then one of q or r must be true.
103 ( !(LA(1)==AAA || LA(1)==BBB) ||
104 ( ! LA(1) == AAA || p) &&
105 ( ! LA(1) == BBB || q || r ) ) ) {
107 I believe this is fixed in MR12.
109 Reported by Reinier van den Born (reinier@vnet.ibm.com)
111 #151a. (Changed in MR12) ANTLRParser::getLexer()
113 As a result of several requests, I have added public methods to
114 get a pointer to the lexer belonging to a parser.
116 ANTLRTokenStream *ANTLRParser::getLexer() const
118 Returns a pointer to the lexer being used by the
119 parser. ANTLRTokenStream is the base class of
122 ANTLRTokenStream *ANTLRTokenBuffer::getLexer() const
124 Returns a pointer to the lexer being used by the
125 ANTLRTokenBuffer. ANTLRTokenStream is the base
128 You must manually cast the ANTLRTokenStream to your program's
129 lexer class. Because the name of the lexer's class is not fixed.
130 Thus it is impossible to incorporate it into the DLGLexerBase
133 #151b.(Changed in MR12) ParserBlackBox member getLexer()
135 The template class ParserBlackBox now has a member getLexer()
136 which returns a pointer to the lexer.
138 #150. (Changed in MR12) syntaxErrCount and lexErrCount now public
140 See Item #127 for more information.
142 #149. (Changed in MR12) antlr option -info o (letter o for orphan)
144 If there is more than one rule which is not referenced by any
145 other rule then all such rules are listed. This is useful for
146 alerting one to rules which are not used, but which can still
147 contribute to ambiguity. For example:
153 will cause an ambiguity report for rule "a" which will be
154 difficult to understand if the user forgets about rule "unused"
155 simply because it is not used in the grammar.
157 #148. (Changed in MR11) #token names appearing in zztokens,token_tbl
159 In a #token statement like the following:
163 the string "Plus" appears in the zztokens array (C mode) and
164 token_tbl (C++ mode). This string is used in most error
165 messages. In MR11 one has the option of using some other string,
166 (e.g. "+") in those tables.
168 In MR11 one can write:
170 #token Plus ("+") "\+"
172 #token COM ("comment begin") "/\*"
174 A #token statement is allowed to appear in more than one #lexclass
175 with different regular expressions. However, the token name appears
176 only once in the zztokens/token_tbl array. This means that only
177 one substitute can be specified for a given #token name. The second
178 attempt to define a substitute name (different from the first) will
179 result in an error message.
181 #147. (Changed in MR11) Bug in follow set computation
183 There is a bug in 1.33 vanilla and all maintenance releases
184 prior to MR11 in the computation of the follow set. The bug is
185 different than that described in Item #82 and probably more
186 common. It was discovered in the ansi.g grammar while testing
187 the "ambiguity aid" (Item #119). The search for a bug started
188 when the ambiguity aid was unable to discover the actual source
189 of an ambiguity reported by antlr.
191 The problem appears when an optimization of the follow set
192 computation is used inappropriately. The result is that the
193 follow set used is the "worst case". In other words, the error
194 can lead to false reports of ambiguity. The good news is that
195 if you have a grammar in which you have addressed all reported
196 ambiguities you are ok. The bad news is that you may have spent
197 time fixing ambiguities that were not real, or used k=2 when
198 ck=2 might have been sufficient, and so on.
200 The following grammar demonstrates the problem:
202 ------------------------------------------------------------
207 stmt : CASE expr COLON
212 plain_stmt : ID COLON ;
213 ------------------------------------------------------------
215 When compiled with k=1 and ck=2 it will report:
217 warning: alts 2 and 3 of the rule itself ambiguous upon
218 { IDENTIFIER }, { COLON }
220 When antlr analyzes "stmt" it computes the first[1] set of all
221 alternatives. It finds an ambiguity between alts 2 and 3 for ID.
222 It then computes the first[2] set for alternatives 2 and 3 to resolve
223 the ambiguity. In computing the first[2] set of "expr" (which is
224 only one token long) it needs to determine what could follow "expr".
225 Under a certain combination of circumstances antlr forgets that it
226 is trying to analyze "stmt" which can only be followed by SEMI and
227 adds to the first[2] set of "expr" the "global" follow set (including
228 "COLON") which could follow "expr" (under other conditions) in the
229 phrase "CASE expr COLON".
231 #146. (Changed in MR11) Option -treport for locating "difficult" alts
233 It can be difficult to determine which alternatives are causing
234 pccts to work hard to resolve an ambiguity. In some cases the
235 ambiguity is successfully resolved after much CPU time so there
236 is no message at all.
238 A rough measure of the amount of work being peformed which is
239 independent of the CPU speed and system load is the number of
240 tnodes created. Using "-info t" gives information about the
241 total number of tnodes created and the peak number of tnodes.
243 Tree Nodes: peak 1300k created 1416k lost 0
245 It also puts in the generated C or C++ file the number of tnodes
246 created for a rule (at the end of the rule). However this
247 information is not sufficient to locate the alternatives within
248 a rule which are causing the creation of tnodes.
252 antlr -treport 100000 ....
254 causes antlr to list on stdout any alternatives which require the
255 creation of more than 100,000 tnodes, along with the lookahead sets
256 for those alternatives.
258 The following is a trivial case from the ansi.g grammar which shows
259 the format of the report. This report might be of more interest
260 in cases where 1,000,000 tuples were created to resolve the ambiguity.
262 -------------------------------------------------------------------------
263 There were 0 tuples whose ambiguity could not be resolved
265 There were 157 tnodes created to resolve ambiguity between:
267 Choice 1: statement/2 line 475 file ansi.g
268 Choice 2: statement/3 line 476 file ansi.g
270 Intersection of lookahead[1] sets:
274 Intersection of lookahead[2] sets:
276 LPARENTHESIS COLON AMPERSAND MINUS
277 STAR PLUSPLUS MINUSMINUS ONESCOMPLEMENT
278 NOT SIZEOF OCTALINT DECIMALINT
279 HEXADECIMALINT FLOATONE FLOATTWO IDENTIFIER
281 -------------------------------------------------------------------------
283 #145. (Documentation) Generation of Expression Trees
285 Item #99 was misleading because it implied that the optimization
286 for tree expressions was available only for trees created by
287 predicate expressions and neglected to mention that it required
288 the use of "-mrhoist on". The optimization applies to tree
289 expressions created for grammars with k>1 and for predicates with
292 In MR11 the optimized version is always used so the -mrhoist on
293 option need not be specified.
295 #144. (Changed in MR11) Incorrect test for exception group
297 In testing for a rule's exception group the label a pointer
298 is compared against '\0'. The intention is "*pointer".
300 Reported by Jeffrey C. Fried (Jeff@Fried.net).
302 #143. (Changed in MR11) Optional ";" at end of #token statement
314 #token X "x" <<lexical action>>
316 #142. (Changed in MR11) class BufFileInput subclass of DLGInputStream
318 Alexey Demakov (demakov@kazbek.ispras.ru) has supplied class
319 BufFileInput derived from DLGInputStream which provides a
320 function lookahead(char *string) to test characters in the
321 input stream more than one character ahead.
323 The default amount of lookahead is specified by the constructor
324 and defaults to 8 characters. This does *not* include the one
325 character of lookahead maintained internally by DLG in member "ch"
326 and which is not available for testing via BufFileInput::lookahead().
328 This is a useful class for overcoming the one-character-lookahead
329 limitation of DLG without resorting to a lexer capable of
330 backtracking (like flex) which is not integrated with antlr as is
333 There are no restrictions on copying or using BufFileInput.* except
334 that the authorship and related information must be retained in the
337 The class is located in pccts/h/BufFileInput.* of the kit.
339 #141. (Changed in MR11) ZZDEBUG_CONSUME for ANTLRParser::consume()
341 A debug aid has been added to file ANTLRParser::consume() in
344 #ifdef ZZDEBUG_CONSUME_ACTION
345 zzdebug_consume_action();
348 Suggested by Sramji Ramanathan (ps@kumaran.com).
350 #140. (Changed in MR11) #pred to define predicates
352 +---------------------------------------------------+
353 | Note: Assume "-prc on" for this entire discussion |
354 +---------------------------------------------------+
356 A problem with predicates is that each one is regarded as
357 unique and capable of disambiguating cases where two
358 alternatives have identical lookahead. For example:
360 rule : <<pred(LATEXT(1))>>? A
361 | <<pred(LATEXT(1))>>? A
364 will not cause any error messages or warnings to be issued
365 by earlier versions of pccts. To compare the text of the
366 predicates is an incomplete solution.
368 In 1.33MR11 I am introducing the #pred statement in order to
369 solve some problems with predicates. The #pred statement allows
370 one to give a symbolic name to a "predicate literal" or a
371 "predicate expression" in order to refer to it in other predicate
372 expressions or in the rules of the grammar.
374 The predicate literal associated with a predicate symbol is C
375 or C++ code which can be used to test the condition. A
376 predicate expression defines a predicate symbol in terms of other
377 predicate symbols using "!", "&&", and "||". A predicate symbol
378 can be defined in terms of a predicate literal, a predicate
379 expression, or *both*.
381 When a predicate symbol is defined with both a predicate literal
382 and a predicate expression, the predicate literal is used to generate
383 code, but the predicate expression is used to check for two
384 alternatives with identical predicates in both alternatives.
386 Here are some examples of #pred statements:
388 #pred IsLabel <<isLabel(LATEXT(1))>>?
389 #pred IsLocalVar <<isLocalVar(LATEXT(1))>>?
390 #pred IsGlobalVar <<isGlobalVar(LATEXT(1)>>?
391 #pred IsVar <<isVar(LATEXT(1))>>? IsLocalVar || IsGlobalVar
392 #pred IsScoped <<isScoped(LATEXT(1))>>? IsLabel || IsLocalVar
394 I hope that the use of EBNF notation to describe the syntax of the
395 #pred statement will not cause problems for my readers (joke).
397 predStatement : "#pred"
400 "<<predicate_literal>>?"
401 | "<<predicate_literal>>?" predOrExpr
406 predOrExpr : predAndExpr ( "||" predAndExpr ) * ;
408 predAndExpr : predPrimary ( "&&" predPrimary ) * ;
410 predPrimary : CapitalizedName
415 What is the purpose of this nonsense ?
417 To understand how predicate symbols help, you need to realize that
418 predicate symbols are used in two different ways with two different
421 a. Allow simplification of predicates which have been combined
422 during predicate hoisting.
424 b. Allow recognition of identical predicates which can't disambiguate
425 alternatives with common lookahead.
427 First we will discuss goal (a). Consider the following rule:
438 rule2: <<isX(LATEXT(1))>>? ID ;
439 rule3: <<!isX(LATEXT(1)>>? ID ;
441 When the predicates in rule2 and rule3 are combined by hoisting
442 to create a prediction expression for rule1 the result is:
445 && ( isX(LATEXT(1) || !isX(LATEXT(1) ) ) { rule1(); ...
447 This is inefficient, but more importantly, can lead to false
448 assumptions that the predicate expression distinguishes the rule1
449 alternative with some other alternative with lookahead ID. In
452 #pred IsX <<isX(LATEXT(1))>>?
457 rule3: <<!IsX>>? ID ;
459 During hoisting MR11 recognizes this as a special case and
460 eliminates the predicates. The result is a prediction
461 expression like the following:
463 if ( LA(1)==ID ) { rule1(); ...
465 Please note that the following cases which appear to be equivalent
466 *cannot* be simplified by MR11 during hoisting because the hoisting
467 logic only checks for a "!" in the predicate action, not in the
468 predicate expression for a predicate symbol.
470 *Not* equivalent and is not simplified during hoisting:
472 #pred IsX <<isX(LATEXT(1))>>?
473 #pred NotX <<!isX(LATEXT(1))>>?
476 rule3: <<NotX>>? ID ;
478 *Not* equivalent and is not simplified during hoisting:
480 #pred IsX <<isX(LATEXT(1))>>?
484 rule3: <<NotX>>? ID ;
486 Now we will discuss goal (b).
488 When antlr discovers that there is a lookahead ambiguity between
489 two alternatives it attempts to resolve the ambiguity by searching
490 for predicates in both alternatives. In the past any predicate
491 would do, even if the same one appeared in both alternatives:
493 rule: <<p(LATEXT(1))>>? X
494 | <<p(LATEXT(1))>>? X
497 The #pred statement is a start towards solving this problem.
498 During ambiguity resolution (*not* predicate hoisting) the
499 predicates for the two alternatives are expanded and compared.
500 Consider the following example:
502 #pred Upper <<isUpper(LATEXT(1))>>?
503 #pred Lower <<isLower(LATEXT(1))>>?
504 #pred Alpha <<isAlpha(LATEXT(1))>>? Upper || Lower
516 rule2: <<Upper>>? ID;
517 rule3: <<Lower>>? ID;
519 The definition of #pred Alpha expresses:
521 a. to test the predicate use the C code "isAlpha(LATEXT(1))"
523 b. to analyze the predicate use the information that
524 Alpha is equivalent to the union of Upper and Lower,
526 During ambiguity resolution the definition of Alpha is expanded
527 into "Upper || Lower" and compared with the predicate in the other
528 alternative, which is also "Upper || Lower". Because they are
529 identical MR11 will report a problem.
531 -------------------------------------------------------------------------
532 t10.g, line 5: warning: the predicates used to disambiguate rule rule0
533 (file t10.g alt 1 line 5 and alt 2 line 6)
534 are identical when compared without context and may have no
535 resolving power for some lookahead sequences.
536 -------------------------------------------------------------------------
538 If you use the "-info p" option the output file will contain:
540 +----------------------------------------------------------------------+
543 |The following predicates are identical when compared without |
544 | lookahead context information. For some ambiguous lookahead |
545 | sequences they may not have any power to resolve the ambiguity. |
547 |Choice 1: rule0/1 alt 1 line 5 file t10.g |
549 | The original predicate for choice 1 with available context |
555 | depth=k=1 rule rule2 line 14 t10.g |
560 | depth=k=1 rule rule3 line 15 t10.g |
564 | The predicate for choice 1 after expansion (but without context |
569 | pred << isUpper(LATEXT(1))>>? |
570 | depth=k=1 rule line 1 t10.g |
572 | pred << isLower(LATEXT(1))>>? |
573 | depth=k=1 rule line 2 t10.g |
576 |Choice 2: rule0/2 alt 2 line 6 file t10.g |
578 | The original predicate for choice 2 with available context |
582 | depth=k=1 rule rule0 line 6 t10.g |
586 | The predicate for choice 2 after expansion (but without context |
591 | pred << isUpper(LATEXT(1))>>? |
592 | depth=k=1 rule line 1 t10.g |
594 | pred << isLower(LATEXT(1))>>? |
595 | depth=k=1 rule line 2 t10.g |
599 +----------------------------------------------------------------------+
601 The comparison of the predicates for the two alternatives takes
602 place without context information, which means that in some cases
603 the predicates will be considered identical even though they operate
604 on disjoint lookahead sets. Consider:
612 Because the comparison of predicates takes place without context
613 these will be considered identical. The reason for comparing
614 without context is that otherwise it would be necessary to re-evaluate
615 the entire predicate expression for each possible lookahead sequence.
616 This would require more code to be written and more CPU time during
617 grammar analysis, and it is not yet clear whether anyone will even make
618 use of the new #pred facility.
620 A temporary workaround might be to use different #pred statements
621 for predicates you know have different context. This would avoid
624 The above example might be termed a "false positive". Comparison
625 without context will also lead to "false negatives". Consider the
639 The predicate used for alt 2 of rule1 is (Alpha || Beta). This
640 appears to be different than the predicate Alpha used for alt1.
641 However, the context of Beta is B. Thus when the lookahead is A
642 Beta will have no resolving power and Alpha will be used for both
643 alternatives. Using the same predicate for both alternatives isn't
644 very helpful, but this will not be detected with 1.33MR11.
646 To properly handle this the predicate expression would have to be
647 evaluated for each distinct lookahead context.
649 To determine whether two predicate expressions are identical is
650 difficult. The routine may fail to identify identical predicates.
652 The #pred feature also compares predicates to see if a choice between
653 alternatives which is resolved by a predicate which makes the second
654 choice unreachable. Consider the following example:
656 #pred A <<A(LATEXT(1)>>?
657 #pred B <<B(LATEXT(1)>>?
668 ----------------------------------------------------------------------------
669 t11.g, line 5: warning: the predicate used to disambiguate the
670 first choice of rule r
671 (file t11.g alt 1 line 5 and alt 2 line 6)
672 appears to "cover" the second predicate when compared without context.
673 The second predicate may have no resolving power for some lookahead
675 ----------------------------------------------------------------------------
677 #139. (Changed in MR11) Problem with -gp in C++ mode
679 The -gp option to add a prefix to rule names did not work in
680 C++ mode. This has been fixed.
682 Reported by Alexey Demakov (demakov@kazbek.ispras.ru).
684 #138. (Changed in MR11) Additional makefiles for non-MSVC++ MS systems
686 Sramji Ramanathan (ps@kumaran.com) has supplied makefiles for
687 building antlr and dlg with Win95/NT development tools that
688 are not based on MSVC5. They are pccts/antlr/AntlrMS.mak and
691 The first line of the makefiles require a definition of PCCTS_HOME.
693 These are in additiion to the AntlrMSVC50.* and DlgMSVC50.*
694 supplied by Jeff Vincent (JVincent@novell.com).
696 #137. (Changed in MR11) Token getType(), getText(), getLine() const members
698 --------------------------------------------------------------------
699 If you use ANTLRCommonToken this change probably does not affect you.
700 --------------------------------------------------------------------
702 For a long time it has bothered me that these accessor functions
703 in ANTLRAbstractToken were not const member functions. I have
704 refrained from changing them because it require users to modify
705 existing token class definitions which are derived directly
706 from ANTLRAbstractToken. I think it is now time.
708 For those who are not used to C++, a "const member function" is a
709 member function which does not modify its own object - the thing
710 to which "this" points. This is quite different from a function
711 which does not modify its arguments
713 Most token definitions based on ANTLRAbstractToken have something like
714 the following in order to create concrete definitions of the pure
715 virtual methods in ANTLRAbstractToken:
717 class MyToken : public ANTLRAbstractToken {
719 ANTLRTokenType getType() {return _type; }
720 int getLine() {return _line; }
721 ANTLRChar * getText() {return _text; }
725 The required change is simply to put "const" following the function
726 prototype in the header (.h file) and the definition file (.cpp if
729 class MyToken : public ANTLRAbstractToken {
731 ANTLRTokenType getType() const {return _type; }
732 int getLine() const {return _line; }
733 ANTLRChar * getText() const {return _text; }
737 This was originally proposed a long time ago by Bruce
738 Guenter (bruceg@qcc.sk.ca).
740 #136. (Changed in MR11) Added getLength() to ANTLRCommonToken
742 Classes ANTLRCommonToken and ANTLRCommonTokenNoRefCountToken
743 now have a member function:
745 int getLength() const { return strlen(getText()) }
747 Suggested by Sramji Ramanathan (ps@kumaran.com).
749 #135. (Changed in MR11) Raised antlr's own default ZZLEXBUFSIZE to 8k
751 #134a. (ansi_mr10.zip) T.J. Parr's ANSI C grammar made 1.33MR11 compatible
753 There is a typographical error in the definition of BITWISEOREQ:
755 #token BITWISEOREQ "!=" should be "\|="
757 When this change is combined with the bugfix to the follow set cache
758 problem (Item #147) and a minor rearrangement of the grammar
759 (Item #134b) it becomes a k=1 ck=2 grammar.
761 #134b. (ansi_mr10.zip) T.J. Parr's ANSI C grammar made 1.33MR11 compatible
763 The following changes were made in the ansi.g grammar (along with
768 void tracein(char *) ====> void tracein(const char *)
769 void traceout(char *) ====> void traceout(const char *)
771 <LT(1)->getType()==IDENTIFIER ? isTypeName(LT(1)->getText()) : 1>>?
772 ====> <<isTypeName(LT(1)->getText())>>?
774 <<(LT(1)->getType()==LPARENTHESIS && LT(2)->getType()==IDENTIFIER) ? \
775 isTypeName(LT(2)->getText()) : 1>>?
776 ====> (LPARENTHESIS IDENTIFIER)? => <<isTypeName(LT(2)->getText())>>?
778 <<(LT(1)->getType()==LPARENTHESIS && LT(2)->getType()==IDENTIFIER) ? \
779 isTypeName(LT(2)->getText()) : 1>>?
780 ====> (LPARENTHESIS IDENTIFIER)? => <<isTypeName(LT(2)->getText())>>?
782 added to init(): traceOptionValueDefault=0;
783 added to init(): traceOption(-1);
785 change rule "statement":
788 : plain_label_statement
789 | case_label_statement
790 | <<;>> expression SEMICOLON
792 | selection_statement
793 | iteration_statement
798 plain_label_statement
799 : IDENTIFIER COLON statement
803 : CASE constant_expression COLON statement
804 | DEFAULT COLON statement
809 void tracein(char *) ====> void tracein(const char *)
810 void traceout(char *) ====> void traceout(const char *)
812 added to tracein(): ANTLRParser::tracein(r); // call superclass method
813 added to traceout(): ANTLRParser::traceout(r); // call superclass method
817 added to AFLAGS: -mrhoist on -prc on
819 #133. (Changed in 1.33MR11) Make trace options public in ANTLRParser
821 In checking T.J. Parr's ANSI C grammar for compatibility with
822 1.33MR11 discovered that it was inconvenient to have the
823 trace facilities with protected access.
825 #132. (Changed in 1.33MR11) Recognition of identical predicates in alts
827 Prior to 1.33MR11, there would be no ambiguity warning when the
828 very same predicate was used to disambiguate both alternatives:
834 ref : <<pred(LATEXT(1)>>? A
836 In 1.33MR11 this will cause the warning:
838 warning: the predicates used to disambiguate rule test
839 (file v98.g alt 1 line 1 and alt 2 line 2)
840 are identical and have no resolving power
842 ----------------- Note -----------------
844 This is different than the following case
846 test: <<pred(LATEXT(1))>>? A B
847 | <<pred(LATEXT(1)>>? A C
850 In this case there are two distinct predicates
851 which have exactly the same text. In the first
852 example there are two references to the same
853 predicate. The problem represented by this
854 grammar will be addressed later.
856 #131. (Changed in 1.33MR11) Case insensitive command line options
858 Command line switches like "-CC" and keywords like "on", "off",
859 and "stdin" are no longer case sensitive in antlr, dlg, and sorcerer.
861 #130. (Changed in 1.33MR11) Changed ANTLR_VERSION to int from string
863 The ANTLR_VERSION was not an integer, making it difficult to
864 perform conditional compilation based on the antlr version.
866 Henceforth, ANTLR_VERSION will be:
868 (base_version * 10000) + release number
870 thus 1.33MR11 will be: 133*100+11 = 13311
872 Suggested by Rainer Janssen (Rainer.Janssen@Informatik.Uni-Oldenburg.DE).
874 #129. (Changed in 1.33MR11) Addition of ANTLR_VERSION to <parserName>.h
876 The following code is now inserted into <parserName>.h amd
879 #ifndef ANTLR_VERSION
880 #define ANTLR_VERSION 13311
883 Suggested by Rainer Janssen (Rainer.Janssen@Informatik.Uni-Oldenburg.DE)
885 #128. (Changed in 1.33MR11) Redundant predicate code in (<<pred>>? ...)+
887 Prior to 1.33MR11, the following grammar would generate
888 redundant tests for the "while" condition.
890 rule2 : (<<pred>>? X)+ X
894 The code would resemble:
899 if (!pred) {zzfailed_pred(" pred");}
900 zzmatch(X); zzCONSUME;
901 } while (LA(1)==X && pred && pred);
904 With 1.33MR11 the redundant predicate test is omitted.
906 #127. (Changed in 1.33MR11)
908 Count Syntax Errors Count DLG Errors
909 ------------------- ----------------
911 C++ mode ANTLRParser:: DLGLexerBase::
912 syntaxErrCount lexErrCount
913 C mode zzSyntaxErrCount zzLexErrCount
915 The C mode variables are global and initialized to 0.
916 They are *not* reset to 0 automatically when antlr is
919 The C++ mode variables are public. They are initialized
920 to 0 by the constructors. They are *not* reset to 0 by the
921 ANTLRParser::init() method.
923 Suggested by Reinier van den Born (reinier@vnet.ibm.com).
925 #126. (Changed in 1.33MR11) Addition of #first <<...>>
927 The #first <<...>> inserts the specified text in the output
928 files before any other #include statements required by pccts.
929 The only things before the #first text are comments and
930 a #define ANTLR_VERSION.
932 Requested by and Esa Pulkkinen (esap@cs.tut.fi) and Alexin
933 Zoltan (alexin@inf.u-szeged.hu).
935 #125. (Changed in 1.33MR11) Lookahead for (guard)? && <<p>>? predicates
937 When implementing the new style of guard predicate (Item #113)
938 in 1.33MR10 I decided to temporarily ignore the problem of
939 computing the "narrowest" lookahead context.
941 Consider the following k=1 grammar:
947 a : (A)? && <<pred1(LATEXT(1))>>? ab ;
948 b : (B)? && <<pred2(LATEXT(1))>>? ab ;
952 In MR10 the context for both "a" and "b" was {A B} because this is
953 the first set of rule "ab". Normally, this is not a problem because
954 the predicate which follows the guard inhibits any ambiguity report
957 In MR11 the first set for rule "a" is {A} and for rule "b" it is {B}.
959 #124. A Note on the New "&&" Style Guarded Predicates
961 I've been asked several times, "What is the difference between
962 the old "=>" style guard predicates and the new style "&&" guard
963 predicates, and how do you choose one over the other" ?
965 The main difference is that the "=>" does not apply the
966 predicate if the context guard doesn't match, whereas
967 the && form always does. What is the significance ?
969 If you have a predicate which is not on the "leading edge"
970 it cannot be hoisted. Suppose you need a predicate that
971 looks at LA(2). You must introduce it manually. The
979 typeName : <<isTypeName(LATEXT(1))>>? ID
983 The problem is that typeName isn't on the leading edge
984 of castExpr, so the predicate isTypeName won't be hoisted into
985 castExpr to help make a decision on which production to choose.
987 The *first* attempt to fix it is this:
990 <<isTypeName(LATEXT(2))>>?
995 Unfortunately, this won't work because it ignores
996 the problem of STRUCT. The solution is to apply
997 isTypeName() in castExpr if LA(2) is an ID and
998 don't apply it when LA(2) is STRUCT:
1001 (LP ID)? => <<isTypeName(LATEXT(2))>>?
1006 In conclusion, the "=>" style guarded predicate is
1009 a. the tokens required for the predicate
1010 are not on the leading edge
1011 b. there are alternatives in the expression
1012 selected by the predicate for which the
1013 predicate is inappropriate
1015 If (b) were false, then one could use a simple
1016 predicate (assuming "-prc on"):
1019 <<isTypeName(LATEXT(2))>>?
1024 typeName : <<isTypeName(LATEXT(1))>>? ID
1027 So, when do you use the "&&" style guarded predicate ?
1029 The new-style "&&" predicate should always be used with
1030 predicate context. The context guard is in ADDITION to
1031 the automatically computed context. Thus it useful for
1032 predicates which depend on the token type for reasons
1035 The following example is contributed by Reinier van den Born
1036 (reinier@vnet.ibm.com).
1038 +-------------------------------------------------------------------------+
1039 | This grammar has two ways to call functions: |
1041 | - a "standard" call syntax with parens and comma separated args |
1042 | - a shell command like syntax (no parens and spacing separated args) |
1044 | The former also allows a variable to hold the name of the function, |
1045 | the latter can also be used to call external commands. |
1047 | The grammar (simplified) looks like this: |
1049 | fun_call : ID "(" { expr ("," expr)* } ")" |
1050 | /* ID is function name */ |
1051 | | "@" ID "(" { expr ("," expr)* } ")" |
1052 | /* ID is var containing fun name */ |
1055 | command : ID expr* /* ID is function name */ |
1056 | | path expr* /* path is external command name */ |
1059 | path : ID /* left out slashes and such */ |
1060 | | "@" ID /* ID is environment var */ |
1070 | Obviously the call is wildly ambiguous. This is more or less how this |
1071 | is to be resolved: |
1073 | A call begins with an ID or an @ followed by an ID. |
1075 | If it is an ID and if it is an ext. command name -> command |
1076 | if followed by a paren -> fun_call |
1077 | otherwise -> command |
1079 | If it is an @ and if the ID is a var name -> fun_call |
1080 | otherwise -> command |
1082 | One can implement these rules quite neatly using && predicates: |
1084 | call : ("@" ID)? && <<isVarName(LT(2))>>? fun_call |
1085 | | (ID)? && <<isExtCmdName>>? command |
1086 | | (ID "(")? fun_call |
1090 | This can be done better, so it is not an ideal example, but it |
1091 | conveys the principle. |
1092 +-------------------------------------------------------------------------+
1094 #123. (Changed in 1.33MR11) Correct definition of operators in ATokPtr.h
1096 The return value of operators in ANTLRTokenPtr:
1098 changed: unsigned ... operator !=(...)
1099 to: int ... operator != (...)
1100 changed: unsigned ... operator ==(...)
1101 to: int ... operator == (...)
1103 Suggested by R.A. Nelson (cowboy@VNET.IBM.COM)
1105 #122. (Changed in 1.33MR11) Member functions to reset DLG in C++ mode
1107 void DLGFileReset(FILE *f) { input = f; found_eof = 0; }
1108 void DLGStringReset(DLGChar *s) { input = s; p = &input[0]; }
1110 Supplied by R.A. Nelson (cowboy@VNET.IBM.COM)
1112 #121. (Changed in 1.33MR11) Another attempt to fix -o (output dir) option
1114 Another attempt is made to improve the -o option of antlr, dlg,
1115 and sorcerer. This one by JVincent (JVincent@novell.com).
1119 a. If -o is not specified than any explicit directory
1122 b. If -o is specified than the -o directory name overrides any
1123 explicit directory names.
1125 c. The directory name of the grammar file is *not* stripped
1126 to create the main output file. However it is stil subject
1127 to override by the -o directory name.
1129 #120. (Changed in 1.33MR11) "-info f" output to stdout rather than stderr
1131 Added option 0 (e.g. "-info 0") which is a noop.
1133 #119. (Changed in 1.33MR11) Ambiguity aid for grammars
1135 The user can ask for additional information on ambiguities reported
1136 by antlr to stdout. At the moment, only one ambiguity report can
1137 be created in an antlr run.
1139 This feature is enabled using the "-aa" (Ambiguity Aid) option.
1141 The following options control the reporting of ambiguities:
1143 -aa ruleName Selects reporting by name of rule
1144 -aa lineNumber Selects reporting by line number
1145 (file name not compared)
1147 -aam Selects "multiple" reporting for a token
1148 in the intersection set of the
1151 For instance, the token ID may appear dozens
1152 of times in various paths as the program
1153 explores the rules which are reachable from
1154 the point of an ambiguity. With option -aam
1155 every possible path the search program
1156 encounters is reported.
1158 Without -aam only the first encounter is
1159 reported. This may result in incomplete
1160 information, but the information may be
1161 sufficient and much shorter.
1163 -aad depth Selects the depth of the search.
1164 The default value is 1.
1166 The number of paths to be searched, and the
1167 size of the report can grow geometrically
1168 with the -ck value if a full search for all
1169 contributions to the source of the ambiguity
1172 The depth represents the number of tokens
1173 in the lookahead set which are matched against
1174 the set of ambiguous tokens. A depth of 1
1175 means that the search stops when a lookahead
1176 sequence of just one token is matched.
1178 A k=1 ck=6 grammar might generate 5,000 items
1179 in a report if a full depth 6 search is made
1180 with the Ambiguity Aid. The source of the
1181 problem may be in the first token and obscured
1182 by the volume of data - I hesitate to call
1185 When the user selects a depth > 1, the search
1186 is first performed at depth=1 for both
1187 alternatives, then depth=2 for both alternatives,
1190 Sample output for rule grammar in antlr.g itself:
1192 +---------------------------------------------------------------------+
1195 | Choice 1: grammar/70 line 632 file a.g |
1196 | Choice 2: grammar/82 line 644 file a.g |
1198 | Intersection of lookahead[1] sets: |
1200 | "\}" "class" "#errclass" "#tokclass" |
1202 | Choice:1 Depth:1 Group:1 ("#errclass") |
1203 | 1 in (...)* block grammar/70 line 632 a.g |
1204 | 2 to error grammar/73 line 635 a.g |
1205 | 3 error error/1 line 894 a.g |
1206 | 4 #token "#errclass" error/2 line 895 a.g |
1208 | Choice:1 Depth:1 Group:2 ("#tokclass") |
1209 | 2 to tclass grammar/74 line 636 a.g |
1210 | 3 tclass tclass/1 line 937 a.g |
1211 | 4 #token "#tokclass" tclass/2 line 938 a.g |
1213 | Choice:1 Depth:1 Group:3 ("class") |
1214 | 2 to class_def grammar/75 line 637 a.g |
1215 | 3 class_def class_def/1 line 669 a.g |
1216 | 4 #token "class" class_def/3 line 671 a.g |
1218 | Choice:1 Depth:1 Group:4 ("\}") |
1219 | 2 #token "\}" grammar/76 line 638 a.g |
1221 | Choice:2 Depth:1 Group:5 ("#errclass") |
1222 | 1 in (...)* block grammar/83 line 645 a.g |
1223 | 2 to error grammar/93 line 655 a.g |
1224 | 3 error error/1 line 894 a.g |
1225 | 4 #token "#errclass" error/2 line 895 a.g |
1227 | Choice:2 Depth:1 Group:6 ("#tokclass") |
1228 | 2 to tclass grammar/94 line 656 a.g |
1229 | 3 tclass tclass/1 line 937 a.g |
1230 | 4 #token "#tokclass" tclass/2 line 938 a.g |
1232 | Choice:2 Depth:1 Group:7 ("class") |
1233 | 2 to class_def grammar/95 line 657 a.g |
1234 | 3 class_def class_def/1 line 669 a.g |
1235 | 4 #token "class" class_def/3 line 671 a.g |
1237 | Choice:2 Depth:1 Group:8 ("\}") |
1238 | 2 #token "\}" grammar/96 line 658 a.g |
1239 +---------------------------------------------------------------------+
1241 For a linear lookahead set ambiguity (where k=1 or for k>1 but
1242 when all lookahead sets [i] with i<k all have degree one) the
1243 reports appear in the following order:
1245 for (depth=1 ; depth <= "-aad depth" ; depth++) {
1246 for (alternative=1; alternative <=2 ; alternative++) {
1247 while (matches-are-found) {
1254 For reporting a k-tuple ambiguity, the reports appear in the
1257 for (depth=1 ; depth <= "-aad depth" ; depth++) {
1258 while (matches-are-found) {
1259 for (alternative=1; alternative <=2 ; alternative++) {
1266 This is because matches are generated in different ways for
1267 linear lookahead and k-tuples.
1269 #118. (Changed in 1.33MR11) DEC VMS makefile and VMS related changes
1271 Revised makefiles for DEC/VMS operating system for antlr, dlg,
1274 Reduced names of routines with external linkage to less than 32
1275 characters to conform to DEC/VMS linker limitations.
1277 Jean-Francois Pieronne discovered problems with dlg and antlr
1278 due to the VMS linker not being case sensitive for names with
1279 external linkage. In dlg the problem was with "className" and
1280 "ClassName". In antlr the problem was with "GenExprSets" and
1283 Added genmms, a version of genmk for the DEC/VMS version of make.
1284 The source is in directory pccts/support/DECmms.
1286 All VMS contributions by Jean-Francois Pieronne (jfp@iname.com).
1288 #117. (Changed in 1.33MR10) new EXPERIMENTAL predicate hoisting code
1290 The hoisting of predicates into rules to create prediction
1291 expressions is a problem in antlr. Consider the following
1292 example (k=1 with -prc on):
1296 b : <<isUpper(LATEXT(1))>>? A ;
1299 Prior to 1.33MR10 the code generated for "start" would resemble:
1303 (!LA(1)==A || isUpper())) {
1308 This code is wrong because it makes rule "c" unreachable from
1309 "start". The essence of the problem is that antlr fails to
1310 recognize that there can be a valid alternative within "a" even
1311 when the predicate <<isUpper(LATEXT(1))>>? is false.
1313 In 1.33MR10 with -mrhoist the hoisting of the predicate into
1314 "start" is suppressed because it recognizes that "c" can
1315 cover all the cases where the predicate is false:
1323 With the antlr "-info p" switch the user will receive information
1324 about the predicate suppression in the generated file:
1326 --------------------------------------------------------------
1329 Hoisting of predicate suppressed by alternative without predicate.
1330 The alt without the predicate includes all cases where
1331 the predicate is false.
1333 WITH predicate: line 7 v1.g
1334 WITHOUT predicate: line 7 v1.g
1336 The context set for the predicate:
1340 The lookahead set for the alt WITHOUT the semantic predicate:
1346 pred << isUpper(LATEXT(1))>>?
1347 depth=k=1 rule b line 9 v1.g
1352 Chain of referenced rules:
1354 #0 in rule start (line 5 v1.g) to rule a
1355 #1 in rule a (line 7 v1.g)
1358 --------------------------------------------------------------
1360 A predicate can be suppressed by a combination of alternatives
1361 which, taken together, cover a predicate:
1365 a : b | ca | cb | cc ;
1367 b : <<isUpper(LATEXT(1))>>? ( A | B | C ) ;
1373 Consider a more complex example in which "c" covers only part of
1382 b : <<isUpper(LATEXT(1))>>?
1390 Prior to 1.33MR10 the code generated for "start" would resemble:
1393 if ( (LA(1)==A || LA(1)==X) &&
1394 (! (LA(1)==A || LA(1)==X) || isUpper()) {
1399 With 1.33MR10 and -mrhoist the predicate context is restricted to
1400 the non-covered lookahead. The code resembles:
1403 if ( (LA(1)==A || LA(1)==X) &&
1404 (! (LA(1)==X) || isUpper()) {
1409 With the antlr "-info p" switch the user will receive information
1410 about the predicate restriction in the generated file:
1412 --------------------------------------------------------------
1415 Restricting the context of a predicate because of overlap
1416 in the lookahead set between the alternative with the
1417 semantic predicate and one without
1418 Without this restriction the alternative without the predicate
1419 could not be reached when input matched the context of the
1420 predicate and the predicate was false.
1422 WITH predicate: line 11 v4.g
1423 WITHOUT predicate: line 12 v4.g
1425 The original context set for the predicate:
1429 The lookahead set for the alt WITHOUT the semantic predicate:
1433 The intersection of the two sets
1437 The original predicate:
1439 pred << isUpper(LATEXT(1))>>?
1440 depth=k=1 rule b line 15 v4.g
1445 The new (modified) form of the predicate:
1447 pred << isUpper(LATEXT(1))>>?
1448 depth=k=1 rule b line 15 v4.g
1454 --------------------------------------------------------------
1456 The bad news about -mrhoist:
1458 (a) -mrhoist does not analyze predicates with lookahead
1461 (b) -mrhoist does not look past a guarded predicate to
1462 find context which might cover other predicates.
1464 For these cases you might want to use syntactic predicates.
1465 When a semantic predicate fails during guess mode the guess
1466 fails and the next alternative is tried.
1468 Limitation (a) is illustrated by the following example:
1470 start : (stmt)* EOF ;
1475 cast : <<isTypename(LATEXT(2))>>? LP ID RP ;
1479 This is not much different from the first example, except that
1480 it requires two tokens of lookahead context to determine what
1481 to do. This predicate is NOT suppressed because the current version
1482 is unable to handle predicates with depth > 1.
1484 A predicate can be combined with other predicates during hoisting.
1485 In those cases the depth=1 predicates are still handled. Thus,
1486 in the following example the isUpper() predicate will be suppressed
1487 by line #4 when hoisted from "bizarre" into "start", but will still
1488 be present in "bizarre" in order to predict "stmt".
1490 start : (bizarre)* EOF ; // #1
1492 bizarre : stmt // #3
1500 cast : <<isTypename(LATEXT(2))>>? LP ID RP ;
1503 | <<isUpper(LATEXT(1))>>? A
1505 Limitation (b) is illustrated by the following example of a
1506 context guarded predicate:
1508 rule : (A)? <<p>>? // #1
1515 Recall that this means that when the lookahead is NOT A then
1516 the predicate "p" is ignored and it attempts to match "A|B".
1517 Ideally, the "B" at line #3 should suppress predicate "q".
1518 However, the current version does not attempt to look past
1519 the guard predicate to find context which might suppress other
1522 In some cases -mrhoist will lead to the reporting of ambiguities
1523 which were not visible before:
1529 b : <<isUpper(LATEXT(1))>>? A;
1534 In this case there is a true ambiguity in "a" between "bc" and "d"
1535 which can both match "A". Without -mrhoist the predicate in "b"
1536 is hoisted into "a" and there is no ambiguity reported. However,
1537 with -mrhoist, the predicate in "b" is suppressed by "c" (as it
1538 should be) making the ambiguity in "a" apparent.
1540 The motivations for these changes were hoisting problems reported
1541 by Reinier van den Born (reinier@vnet.ibm.com) and several others.
1543 #116. (Changed in 1.33MR10) C++ mode: tracein/traceout rule name is (const char *)
1545 The prototype for C++ mode routine tracein (and traceout) has changed from
1546 "char *" to "const char *".
1548 #115. (Changed in 1.33MR10) Using guess mode with exception handlers in C mode
1550 The definition of the C mode macros zzmatch_wsig and zzsetmatch_wsig
1551 neglected to consider guess mode. When control passed to the rule's
1552 parse exception handler the routine would exit without ever closing the
1553 guess block. This would lead to unpredictable behavior.
1555 In 1.33MR10 the behavior of exceptions in C mode and C++ mode should be
1558 #114. (Changed in 1.33MR10) difference in [zz]resynch() between C and C++ modes
1560 There was a slight difference in the way C and C++ mode resynchronized
1561 following a parsing error. The C routine would sometimes skip an extra
1562 token before attempting to resynchronize.
1564 The C routine was changed to match the C++ routine.
1566 #113. (Changed in 1.33MR10) new context guarded pred: (g)? && <<p>>? expr
1568 The existing context guarded predicate:
1570 rule : (guard)? => <<p>>? expr
1574 generates code which resembles:
1576 if (lookahead(expr) && (!guard || pred)) {
1580 This is not suitable for some applications because it allows
1581 expr() to be invoked when the predicate is false. This is
1582 intentional because it is meant to mimic automatically computed
1585 The new context guarded predicate uses the guard information
1586 differently because it has a different goal. Consider:
1588 rule : (guard)? && <<p>>? expr
1592 The new style of context guarded predicate is equivalent to:
1594 rule : <<guard==true && pred>>? expr
1598 It generates code which resembles:
1600 if (lookahead(expr) && guard && pred) {
1604 Both forms of guarded predicates severely restrict the form of
1605 the context guard: it can contain no rule references, no
1606 (...)*, no (...)+, and no {...}. It may contain token and
1607 token class references, and alternation ("|").
1609 Addition for 1.33MR11: in the token expression all tokens must
1610 be at the same height of the token tree:
1612 (A ( B | C))? && ... is ok (all height 2)
1613 (A ( B | ))? && ... is not ok (some 1, some 2)
1614 (A B C D | E F G H)? && ... is ok (all height 4)
1615 (A B C D | E )? && ... is not ok (some 4, some 1)
1617 This restriction is required in order to properly compute the lookahead
1618 set for expressions like:
1620 rule1 : (A B C)? && <<pred>>? rule2 ;
1621 rule2 : (A|X) (B|Y) (C|Z);
1623 This addition was suggested by Rienier van den Born (reinier@vnet.ibm.com)
1625 #112. (Changed in 1.33MR10) failed validation predicate in C guess mode
1627 John Lilley (jlilley@empathy.com) suggested that failed validation
1628 predicates abort a guess rather than reporting a failed error.
1629 This was installed in C++ mode (Item #4). Only now was it noticed
1630 that the fix was never installed for C mode.
1632 #111. (Changed in 1.33MR10) moved zzTRACEIN to before init action
1634 When the antlr -gd switch is present antlr generates calls to
1635 zzTRACEIN at the start of a rule and zzTRACEOUT at the exit
1636 from a rule. Prior to 1.33MR10 Tthe call to zzTRACEIN was
1637 after the init-action, which could cause confusion because the
1638 init-actions were reported with the name of the enclosing rule,
1639 rather than the active rule.
1641 #110. (Changed in 1.33MR10) antlr command line copied to generated file
1643 The antlr command line is now copied to the generated file near
1646 #109. (Changed in 1.33MR10) improved trace information
1648 The quality of the trace information provided by the "-gd"
1649 switch has been improved significantly. Here is an example
1650 of the output from a test program. It shows the rule name,
1651 the first token of lookahead, the call depth, and the guess
1654 exit rule gusxx {"?"} depth 2
1655 enter rule gusxx {"?"} depth 2
1656 enter rule gus1 {"o"} depth 3 guessing
1657 guess done - returning to rule gus1 {"o"} at depth 3
1658 (guess mode continues - an enclosing guess is still active)
1659 guess done - returning to rule gus1 {"Z"} at depth 3
1660 (guess mode continues - an enclosing guess is still active)
1661 exit rule gus1 {"Z"} depth 3 guessing
1662 guess done - returning to rule gusxx {"o"} at depth 2 (guess mode ends)
1663 enter rule gus1 {"o"} depth 3
1664 guess done - returning to rule gus1 {"o"} at depth 3 (guess mode ends)
1665 guess done - returning to rule gus1 {"Z"} at depth 3 (guess mode ends)
1666 exit rule gus1 {"Z"} depth 3
1667 line 1: syntax error at "Z" missing SC
1670 Rule trace reporting is controlled by the value of the integer
1671 [zz]traceOptionValue: when it is positive tracing is enabled,
1672 otherwise it is disabled. Tracing during guess mode is controlled
1673 by the value of the integer [zz]traceGuessOptionValue. When
1674 it is positive AND [zz]traceOptionValue is positive rule trace
1675 is reported in guess mode.
1677 The values of [zz]traceOptionValue and [zz]traceGuessOptionValue
1678 can be adjusted by subroutine calls listed below.
1680 Depending on the presence or absence of the antlr -gd switch
1681 the variable [zz]traceOptionValueDefault is set to 0 or 1. When
1682 the parser is initialized or [zz]traceReset() is called the
1683 value of [zz]traceOptionValueDefault is copied to [zz]traceOptionValue.
1684 The value of [zz]traceGuessOptionValue is always initialzed to 1,
1685 but, as noted earlier, nothing will be reported unless
1686 [zz]traceOptionValue is also positive.
1688 When the parser state is saved/restored the value of the trace
1689 variables are also saved/restored. If a restore causes a change in
1690 reporting behavior from on to off or vice versa this will be reported.
1692 When the -gd option is selected, the macro "#define zzTRACE_RULES"
1693 is added to appropriate output files.
1697 int traceOption(int delta)
1698 int traceGuessOption(int delta)
1700 int traceOptionValueDefault
1704 int zzTraceOption(int delta)
1705 int zzTraceGuessOption(int delta)
1707 int zzTraceOptionValueDefault
1709 The argument "delta" is added to the traceOptionValue. To
1710 turn on trace when inside a particular rule one:
1712 rule : <<traceOption(+1);>>
1716 <<traceOption(-1);>>
1717 ; /* fail clause */ <<traceOption(-1);>>
1719 One can use the same idea to turn *off* tracing within a
1720 rule by using a delta of (-1).
1722 An improvement in the rule trace was suggested by Sramji
1723 Ramanathan (ps@kumaran.com).
1725 #108. A Note on Deallocation of Variables Allocated in Guess Mode
1728 ------------------------------------------------------
1729 This mechanism only works for heap allocated variables
1730 ------------------------------------------------------
1732 The rewrite of the trace provides the machinery necessary
1733 to properly free variables or undo actions following a
1736 The macro zzUSER_GUESS_HOOK(guessSeq,zzrv) is expanded
1737 as part of the zzGUESS macro. When a guess is opened
1738 the value of zzrv is 0. When a longjmp() is executed to
1739 undo the guess, the value of zzrv will be 1.
1741 The macro zzUSER_GUESS_DONE_HOOK(guessSeq) is expanded
1742 as part of the zzGUESS_DONE macro. This is executed
1743 whether the guess succeeds or fails as part of closing
1746 The guessSeq is a sequence number which is assigned to each
1747 guess and is incremented by 1 for each guess which becomes
1748 active. It is needed by the user to associate the start of
1749 a guess with the failure and/or completion (closing) of a
1752 Guesses are nested. They must be closed in the reverse
1753 of the order that they are opened.
1755 In order to free memory used by a variable during a guess
1756 a user must write a routine which can be called to
1757 register the variable along with the current guess sequence
1758 number provided by the zzUSER_GUESS_HOOK macro. If the guess
1759 fails, all variables tagged with the corresponding guess
1760 sequence number should be released. This is ugly, but
1761 it would require a major rewrite of antlr 1.33 to use
1762 some mechanism other than setjmp()/longjmp().
1764 The order of calls for a *successful* guess would be:
1766 zzUSER_GUESS_HOOK(guessSeq,0);
1767 zzUSER_GUESS_DONE_HOOK(guessSeq);
1769 The order of calls for a *failed* guess would be:
1771 zzUSER_GUESS_HOOK(guessSeq,0);
1772 zzUSER_GUESS_HOOK(guessSeq,1);
1773 zzUSER_GUESS_DONE_HOOK(guessSeq);
1775 The default definitions of these macros are empty strings.
1777 Here is an example in C++ mode. The zzUSER_GUESS_HOOK and
1778 zzUSER_GUESS_DONE_HOOK macros and myGuessHook() routine
1779 can be used without change in both C and C++ versions.
1781 ----------------------------------------------------------------------
1786 typedef ANTLRCommonToken ANTLRToken;
1788 #include "DLGLexer.h"
1793 DLGFileInput in(stdin);
1794 DLGLexer lexer(&in,2000);
1795 ANTLRTokenBuffer pipe(&lexer,1);
1796 ANTLRCommonToken aToken;
1799 lexer.setToken(&aToken);
1814 #undef zzUSER_GUESS_HOOK
1815 #define zzUSER_GUESS_HOOK(guessSeq,zzrv) myGuessHook(guessSeq,zzrv);
1816 #undef zzUSER_GUESS_DONE_HOOK
1817 #define zzUSER_GUESS_DONE_HOOK(guessSeq) myGuessHook(guessSeq,2);
1819 void myGuessHook(int guessSeq,int zzrv) {
1821 fprintf(stderr,"User hook: starting guess #%d\n",guessSeq);
1822 } else if (zzrv == 1) {
1825 fprintf(stderr,"User hook: failed guess #%d\n",guessSeq);
1826 } else if (zzrv == 2) {
1829 fprintf(stderr,"User hook: ending guess #%d\n",guessSeq);
1836 #token "[\t \ \n]" <<skip();>>
1843 top : (which) ? <<fprintf(stderr,"%s is a which\n",s); free(s); s=NULL; >>
1844 | other <<fprintf(stderr,"%s is an other\n",s); free(s); s=NULL; >>
1845 ; <<if (s != NULL) free(s); s=NULL; >>
1853 : (label)? <<fprintf(stderr,"%s is a label\n",s);>>
1854 | (global)? <<fprintf(stderr,"%s is a global\n",s);>>
1855 | (exclamation)? <<fprintf(stderr,"%s is an exclamation\n",s);>>
1858 label : <<s=strdup(LT(1)->getText());>> A ":" ;
1860 global : <<s=strdup(LT(1)->getText());>> A "::" ;
1862 exclamation : <<s=strdup(LT(1)->getText());>> A "!" ;
1864 other : <<s=strdup(LT(1)->getText());>> "other" ;
1867 ----------------------------------------------------------------------
1869 This is a silly example, but illustrates the idea. For the input
1870 "a ::" with tracing enabled the output begins:
1872 ----------------------------------------------------------------------
1873 enter rule "start" depth 1
1874 enter rule "top" depth 2
1875 User hook: starting guess #1
1876 enter rule "which" depth 3 guessing
1877 enter rule "which2" depth 4 guessing
1878 enter rule "which3" depth 5 guessing
1879 User hook: starting guess #2
1880 enter rule "label" depth 6 guessing
1882 User hook: failed guess #2
1883 guess done - returning to rule "which3" at depth 5 (guess mode continues
1884 - an enclosing guess is still active)
1885 User hook: ending guess #2
1886 User hook: starting guess #3
1887 enter rule "global" depth 6 guessing
1888 exit rule "global" depth 6 guessing
1889 guess done - returning to rule "which3" at depth 5 (guess mode continues
1890 - an enclosing guess is still active)
1891 User hook: ending guess #3
1892 enter rule "global" depth 6 guessing
1893 exit rule "global" depth 6 guessing
1894 exit rule "which3" depth 5 guessing
1895 exit rule "which2" depth 4 guessing
1896 exit rule "which" depth 3 guessing
1897 guess done - returning to rule "top" at depth 2 (guess mode ends)
1898 User hook: ending guess #1
1899 enter rule "which" depth 3
1901 ----------------------------------------------------------------------
1905 (a) Only init-actions are executed during guess mode.
1906 (b) A rule can be invoked multiple times during guess mode.
1907 (c) If the guess succeeds the rule will be called once more
1908 without guess mode so that normal actions will be executed.
1909 This means that the init-action might need to distinguish
1910 between guess mode and non-guess mode using the variable
1913 #107. (Changed in 1.33MR10) construction of ASTs in guess mode
1915 Prior to 1.33MR10, when using automatic AST construction in C++
1916 mode for a rule, an AST would be constructed for elements of the
1917 rule even while in guess mode. In MR10 this no longer occurs.
1919 #106. (Changed in 1.33MR10) guess variable confusion
1921 In C++ mode a guess which failed always restored the parser state
1922 using zzGUESS_DONE as part of zzGUESS_FAIL. Prior to 1.33MR10,
1923 C mode required an explicit call to zzGUESS_DONE after the
1924 call to zzGUESS_FAIL.
1928 rule : (alpha)? beta
1932 The generated code resembles:
1935 if (!zzrv && LA(1)==ID) { <==== line #1
1940 if (! zzrv) zzGUESS_DONE <==== line #2a
1943 However, in some cases line #2 was rendered:
1945 if (guessing) zzGUESS_DONE <==== line #2b
1947 This would work for simple test cases, but would fail in
1948 some cases where there was a guess while another guess was active.
1949 One kind of failure would be to match up the zzGUESS_DONE at line
1950 #2b with the "outer" guess which was still active. The outer
1951 guess would "succeed" when only the inner guess should have
1954 In 1.33MR10 the behavior of zzGUESS and zzGUESS_FAIL in C and
1955 and C++ mode should be identical.
1957 The same problem appears in 1.33 vanilla in some places. For
1960 start : { (sub)? } ;
1971 generates incorrect code.
1973 The general principle is:
1975 (a) use [zz]guessing only when deciding between a call to zzFAIL
1978 (b) use zzrv in all other cases
1980 This problem was discovered while testing changes to item #105.
1981 I believe this is now fixed. My apologies.
1983 #105. (Changed in 1.33MR10) guess block as single alt of (...)+
1985 Prior to 1.33MR10 the following constructs:
1997 generated incorrect code for the guess block (which could result
1998 in runtime errors) because of an incorrect optimization of a
1999 block with only a single alternative.
2001 The fix caused some changes to the fix described in Item #49
2002 because there are now three code generation sequences for (...)+
2003 blocks containing a guess block:
2005 a. single alternative which is a guess block
2006 b. multiple alternatives in which the last is a guess block
2009 Forms like "rule_star" can have unexpected behavior when there
2010 is a syntax error: if the subrule "sub" is not matched *exactly*
2011 then "rule_star" will consume no tokens.
2013 Reported by Esa Pulkkinen (esap@cs.tut.fi).
2015 #104. (Changed in 1.33MR10) -o option for dlg
2017 There was problem with the code added by item #74 to handle the
2018 -o option of dlg. This should fix it.
2020 #103. (Changed in 1.33MR10) ANDed semantic predicates
2024 The optimization was a mistake.
2025 The resulting problem is described in Item #150.
2027 #102. (Changed in 1.33MR10) allow "class parser : .... {"
2029 The syntax of the class statement ("class parser-name {")
2030 has been extended to allow for the specification of base
2031 classes. An arbirtrary number of tokens may now appear
2032 between the class name and the "{". They are output
2033 again when the class declaration is generated. For
2036 class Parser : public MyBaseClassANTLRparser {
2038 This was suggested by a user, but I don't have a record
2041 #101. (Changed in 1.33MR10) antlr -info command line switch
2045 p - extra predicate information in generated file
2047 t - information about tnode use:
2048 at the end of each rule in generated file
2049 summary on stderr at end of program
2051 m - monitor progress
2052 prints name of each rule as it is started
2053 flushes output at start of each rule
2055 f - first/follow set information to stdout
2057 0 - no operation (added in 1.33MR11)
2059 The options may be combined and may appear in any order.
2062 antlr -info ptm -CC -gt -mrhoist on mygrammar.g
2064 #100a. (Changed in 1.33MR10) Predicate tree simplification
2066 When the same predicates can be referenced in more than one
2067 alternative of a block large predicate trees can be formed.
2069 The difference that these optimizations make is so dramatic
2070 that I have decided to use it even when -mrhoist is not selected.
2072 Consider the following grammar:
2086 c : <<AAA(LATEXT(2))>>?
2089 d : <<BBB(LATEXT(2))>>? B C
2092 e : <<CCC(LATEXT(2))>>? B C
2098 In rule "a" there is a reference to rule "c" in both alternatives.
2099 The length of the predicate AAA is k=2 and it can be followed in
2100 alternative 1 only by (A B) while in alternative 2 it can be
2101 followed only by (A C). Thus they do not have identical context.
2103 In rule "all" the alternatives which refer to rules "e" and "f" allow
2104 elimination of the duplicate reference to predicate CCC.
2106 The table below summarized the kind of simplification performed by
2107 1.33MR10. In the table, X and Y stand for single predicates
2110 (OR X (OR Y (OR Z))) => (OR X Y Z)
2111 (AND X (AND Y (AND Z))) => (AND X Y Z)
2113 (OR X (... (OR X Y) ... )) => (OR X (... Y ... ))
2114 (AND X (... (AND X Y) ... )) => (AND X (... Y ... ))
2115 (OR X (... (AND X Y) ... )) => (OR X (... ... ))
2116 (AND X (... (OR X Y) ... )) => (AND X (... ... ))
2121 In a test with a complex grammar for a real application, a predicate
2122 tree with six OR nodes and 12 leaves was reduced to "(OR X Y Z)".
2124 In 1.33MR10 there is a greater effort to release memory used
2125 by predicates once they are no longer in use.
2127 #100b. (Changed in 1.33MR10) Suppression of extra predicate tests
2129 The following optimizations require that -mrhoist be selected.
2131 It is relatively easy to optimize the code generated for predicate
2132 gates when they are of the form:
2137 where X, Y, Z, and "..." represent individual predicates (leaves) not
2140 If the predicate is an AND the contexts of the X, Y, Z, etc. are
2141 ANDed together to create a single Tree context for the group and
2142 context tests for the individual predicates are suppressed:
2144 --------------------------------------------------
2145 Note: This was incorrect. The contexts should be
2146 ORed together. This has been fixed. A more
2147 complete description is available in item #152.
2148 ---------------------------------------------------
2150 Optimization 1: (AND X Y Z ...)
2152 Suppose the context for Xtest is LA(1)==LP and the context for
2153 Ytest is LA(1)==LP && LA(2)==ID.
2155 Without the optimization the code would resemble:
2157 if (lookaheadContext &&
2158 !(LA(1)==LP && LA(1)==LP && LA(2)==ID) ||
2159 ( (! LA(1)==LP || Xtest) &&
2160 (! (LA(1)==LP || LA(2)==ID) || Xtest)
2163 With the -mrhoist optimization the code would resemble:
2165 if (lookaheadContext &&
2166 ! (LA(1)==LP && LA(2)==ID) || (Xtest && Ytest) {...
2168 Optimization 2: (OR X Y Z ...) with identical contexts
2170 Suppose the context for Xtest is LA(1)==ID and for Ytest
2171 the context is also LA(1)==ID.
2173 Without the optimization the code would resemble:
2175 if (lookaheadContext &&
2176 ! (LA(1)==ID || LA(1)==ID) ||
2177 (LA(1)==ID && Xtest) ||
2178 (LA(1)==ID && Ytest) {...
2180 With the -mrhoist optimization the code would resemble:
2182 if (lookaheadContext &&
2183 (! LA(1)==ID) || (Xtest || Ytest) {...
2185 Optimization 3: (OR X Y Z ...) with distinct contexts
2187 Suppose the context for Xtest is LA(1)==ID and for Ytest
2188 the context is LA(1)==LP.
2190 Without the optimization the code would resemble:
2192 if (lookaheadContext &&
2193 ! (LA(1)==ID || LA(1)==LP) ||
2194 (LA(1)==ID && Xtest) ||
2195 (LA(1)==LP && Ytest) {...
2197 With the -mrhoist optimization the code would resemble:
2199 if (lookaheadContext &&
2201 (LA(1)==ID && (zzpf=1) && Xtest) ||
2202 (LA(1)==LP && (zzpf=1) && Ytest) ||
2205 These may appear to be of similar complexity at first,
2206 but the non-optimized version contains two tests of each
2207 context while the optimized version contains only one
2208 such test, as well as eliminating some of the inverted
2209 logic (" !(...) || ").
2211 Optimization 4: Computation of predicate gate trees
2213 When generating code for the gates of predicate expressions
2214 antlr 1.33 vanilla uses a recursive procedure to generate
2215 "&&" and "||" expressions for testing the lookahead. As each
2216 layer of the predicate tree is exposed a new set of "&&" and
2217 "||" expressions on the lookahead are generated. In many
2218 cases the lookahead being tested has already been tested.
2220 With -mrhoist a lookahead tree is computed for the entire
2221 lookahead expression. This means that predicates with identical
2222 context or context which is a subset of another predicate's
2225 This is especially important for predicates formed by rules
2228 uppperCaseVowel : <<isUpperCase(LATEXT(1))>>? vowel;
2229 vowel: : <<isVowel(LATEXT(1))>>? LETTERS;
2231 These predicates are combined using AND since both must be
2232 satisfied for rule upperCaseVowel. They have identical
2233 context which makes this optimization very effective.
2235 The affect of Items #100a and #100b together can be dramatic. In
2236 a very large (but real world) grammar one particular predicate
2237 expression was reduced from an (unreadable) 50 predicate leaves,
2238 195 LA(1) terms, and 5500 characters to an (easily comprehensible)
2239 3 predicate leaves (all different) and a *single* LA(1) term.
2241 #99. (Changed in 1.33MR10) Code generation for expression trees
2243 Expression trees are used for k>1 grammars and predicates with
2244 lookahead depth >1. This optimization must be enabled using
2245 "-mrhoist on". (Clarification added for 1.33MR11).
2247 In the processing of expression trees, antlr can generate long chains
2248 of token comparisons. Prior to 1.33MR10 there were many redundant
2249 parenthesis which caused problems for compilers which could handle
2250 expressions of only limited complexity. For example, to test an
2251 expression tree (root R A B C D), antlr would generate something
2254 (LA(1)==R && (LA(2)==A || (LA(2)==B || (LA(2)==C || LA(2)==D)))))
2256 If there were twenty tokens to test then there would be twenty
2257 parenthesis at the end of the expression.
2259 In 1.33MR10 the generated code for tree expressions resembles:
2261 (LA(1)==R && (LA(2)==A || LA(2)==B || LA(2)==C || LA(2)==D))
2263 For "complex" expressions the output is indented to reflect the LA
2264 number being tested:
2267 && (LA(2)==A || LA(2)==B || LA(2)==C || LA(2)==D
2268 || LA(2)==E || LA(2)==F)
2270 && (LA(2)==G || LA(2)==H))
2273 Suggested by S. Bochnak (S.Bochnak@@microTool.com.pl),
2275 #98. (Changed in 1.33MR10) Option "-info p"
2277 When the user selects option "-info p" the program will generate
2278 detailed information about predicates. If the user selects
2279 "-mrhoist on" additional detail will be provided explaining
2280 the promotion and suppression of predicates. The output is part
2281 of the generated file and sandwiched between #if 0/#endif statements.
2283 Consider the following k=1 grammar:
2299 b : <<LATEXT(1)>>? X
2302 Below is an excerpt of the output for rule "start" for the three
2303 predicate options (off, on, and maintenance release style hoisting).
2305 For those who do not wish to use the "-mrhoist on" option for code
2306 generation the option can be used in a "diagnostic" mode to provide
2307 valuable information:
2309 a. where one should insert null actions to inhibit hoisting
2310 b. a chain of rule references which shows where predicates are
2313 ======================================================================
2314 Example of "-info p" with "-mrhoist on"
2315 ======================================================================
2318 Hoisting of predicate suppressed by alternative without predicate.
2319 The alt without the predicate includes all cases where the
2322 WITH predicate: line 11 v36.g
2323 WITHOUT predicate: line 12 v36.g
2325 The context set for the predicate:
2329 The lookahead set for alt WITHOUT the semantic predicate:
2335 pred << LATEXT(1)>>? depth=k=1 rule c line 11 v36.g
2341 Chain of referenced rules:
2343 #0 in rule start (line 1 v36.g) to rule all
2344 #1 in rule all (line 3 v36.g) to rule a
2345 #2 in rule a (line 8 v36.g) to rule c
2346 #3 in rule c (line 11 v36.g)
2352 pred << LATEXT(1)>>? depth=k=1 rule b line 15 v36.g
2359 ======================================================================
2360 Example of "-info p" with the default -prc setting ( "-prc off")
2361 ======================================================================
2365 pred << LATEXT(1)>>? depth=k=1 rule c line 11 v36.g
2371 pred << LATEXT(1)>>? depth=k=1 rule b line 15 v36.g
2378 ======================================================================
2379 Example of "-info p" with "-prc on" and "-mrhoist off"
2380 ======================================================================
2384 pred << LATEXT(1)>>? depth=k=1 rule c line 11 v36.g
2390 pred << LATEXT(1)>>? depth=k=1 rule b line 15 v36.g
2397 ======================================================================
2399 #97. (Fixed in 1.33MR10) "Predicate applied for more than one ... "
2401 In 1.33 vanilla, the grammar listed below produced this message for
2402 the first alternative (only) of rule "b":
2404 warning: predicate applied for >1 lookahead 1-sequences
2405 [you may only want one lookahead 1-sequence to apply.
2406 Try using a context guard '(...)? =>'
2408 In 1.33MR10 the message is issued for both alternatives.
2413 b : <<PPP(LATEXT(1))>>? ( AAA | BBB )
2414 | <<QQQ(LATEXT(1))>>? ( XXX | YYY )
2419 #96. (Fixed in 1.33MR10) Guard predicates ignored when -prc off
2421 Prior to 1.33MR10, guard predicate code was not generated unless
2422 "-prc on" was selected.
2424 This was incorrect, since "-prc off" (the default) is supposed to
2425 disable only AUTOMATIC computation of predicate context, not the
2426 programmer specified context supplied by guard predicates.
2428 #95. (Fixed in 1.33MR10) Predicate guard context length was k, not max(k,ck)
2430 Prior to 1.33MR10, predicate guards were computed to k tokens rather
2431 than max(k,ck). Consider the following grammar:
2433 a : ( A B C)? => <<AAA(LATEXT(1))>>? (A|X) (B|Y) (C|Z) ;
2435 The code generated by 1.33 vanilla with "-k 1 -ck 3 -prc on"
2436 for the predicate in "a" resembles:
2438 if ( (! LA(1)==A) || AAA(LATEXT(1))) {...
2440 With 1.33MR10 and the same options the code resembles:
2442 if ( (! (LA(1)==A && LA(2)==B && LA(3)==C) || AAA(LATEXT(1))) {...
2444 #94. (Fixed in 1.33MR10) Predicates followed by rule references
2446 Prior to 1.33MR10, a semantic predicate which referenced a token
2447 which was off the end of the rule caused an incomplete context
2448 to be computed (with "-prc on") for the predicate under some circum-
2449 stances. In some cases this manifested itself as illegal C code
2450 (e.g. "LA(2)==[Ep](1)" in the k=2 examples below:
2454 a : <<AAA(LATEXT(2))>>? ID X
2455 | <<BBB(LATEXT(2))>>? Y
2459 This might also occur when the semantic predicate was followed
2460 by a rule reference which was shorter than the length of the
2465 a : <<AAA(LATEXT(2))>>? ID X
2466 | <<BBB(LATEXT(2))>>? y
2472 Depending on circumstance, the resulting context might be too
2473 generous because it was too short, or too restrictive because
2474 of missing alternatives.
2476 #93. (Changed in 1.33MR10) Definition of Purify macro
2478 Ofer Ben-Ami (gremlin@cs.huji.ac.il) has supplied a definition
2479 for the Purify macro:
2481 #define PURIFY(r, s) memset((char *) &(r), '\0', (s));
2483 Note: This may not be the right thing to do for C++ objects that
2484 have constructors. Reported by Bonny Rais (bonny@werple.net.au).
2486 For those cases one should #define PURIFY to an empty macro in the
2487 #header or #first actions.
2489 #92. (Fixed in 1.33MR10) Guarded predicates and hoisting
2491 When a guarded predicate participates in hoisting it is linked into
2492 a predicate expression tree. Prior to 1.33MR10 this link was never
2493 cleared and the next time the guard was used to construct a new
2494 tree the link could contain a spurious reference to another element
2495 which had previosly been joined to it in the semantic predicate tree.
2502 start2 : ( all2 ) *;
2505 a : (A)? => <<AAA(LATEXT(1))>>? A ;
2506 b : (B)? => <<BBB(LATEXT(1))>>? B ;
2508 Prior to 1.33MR10 the code for "start2" would include a spurious
2509 reference to the BBB predicate which was left from constructing
2510 the predicate tree for rule "start" (i.e. or(AAA,BBB) ).
2512 In 1.33MR10 this problem is avoided by cloning the original guard
2513 each time it is linked into a predicate tree.
2515 #91. (Changed in 1.33MR10) Extensive changes to semantic pred hoisting
2517 ============================================
2518 This has been rendered obsolete by Item #117
2519 ============================================
2521 #90. (Fixed in 1.33MR10) Semantic pred with LT(i) and i>max(k,ck)
2523 There is a bug in antlr 1.33 vanilla and all maintenance releases
2524 prior to 1.33MR10 which allows semantic predicates to reference
2525 an LT(i) or LATEXT(i) where i is larger than max(k,ck). When
2526 this occurs antlr will attempt to mark the ith element of an array
2527 in which there are only max(k,ck) elements. The result cannot
2530 Using LT(i) or LATEXT(i) for i>max(k,ck) is reported as an error
2535 #88. (Fixed in 1.33MR10) Tokens used in semantic predicates in guess mode
2537 Consider the behavior of a semantic predicate during guess mode:
2544 Prior to MR10 the assignment of the token or attribute to
2545 $a did not occur during guess mode, which would cause the
2546 semantic predicate to misbehave because $a would be null.
2548 In 1.33MR10 a semantic predicate with a reference to an
2549 element label (such as $a) forces the assignment to take
2550 place even in guess mode.
2552 In order to work, this fix REQUIRES use of the $label format
2553 for token pointers and attributes referenced in semantic
2556 The fix does not apply to semantic predicates using the
2557 numeric form to refer to attributes (e.g. <<test($1)>>?).
2558 The user will receive a warning for this case.
2560 Reported by Rob Trout (trout@mcs.cs.kent.edu).
2562 #87. (Fixed in 1.33MR10) Malformed guard predicates
2564 Context guard predicates may contain only references to
2565 tokens. They may not contain references to (...)+ and
2566 (...)* blocks. This is now checked. This replaces the
2567 fatal error message in item #78 with an appropriate
2568 (non-fatal) error messge.
2570 In theory, context guards should be allowed to reference
2571 rules. However, I have not had time to fix this.
2572 Evaluation of the guard takes place before all rules have
2573 been read, making it difficult to resolve a forward reference
2574 to rule "zzz" - it hasn't been read yet ! To postpone evaluation
2575 of the guard until all rules have been read is too much
2578 #86. (Fixed in 1.33MR10) Unequal set size in set_sub
2580 Routine set_sub() in pccts/support/set/set.h did not work
2581 correctly when the sets were of unequal sizes. Rewrote
2582 set_equ to make it simpler and remove unnecessary and
2583 expensive calls to set_deg(). This routine was not used
2586 #85. (Changed in 1.33MR10) Allow redefinition of MaxNumFiles
2588 Raised the maximum number of input files to 99 from 20.
2589 Put a #ifndef/#endif around the "#define MaxNumFiles 99".
2591 #84. (Fixed in 1.33MR10) Initialize zzBadTok in macro zzRULE
2593 Initialize zzBadTok to NULL in zzRULE macro of AParser.h.
2594 in order to get rid of warning messages.
2596 #83. (Fixed in 1.33MR10) False warnings with -w2 for #tokclass
2598 When -w2 is selected antlr gives inappropriate warnings about
2599 #tokclass names not having any associated regular expressions.
2600 Since a #tokclass is not a "real" token it will never have an
2601 associated regular expression and there should be no warning.
2603 Reported by Derek Pappas (derek.pappas@eng.sun.com)
2605 #82. (Fixed in 1.33MR10) Computation of follow sets with multiple cycles
2607 Reinier van den Born (reinier@vnet.ibm.com) reported a problem
2608 in the computation of follow sets by antlr. The problem (bug)
2609 exists in 1.33 vanilla and all maintenance releases prior to 1.33MR10.
2611 The problem involves the computation of follow sets when there are
2612 cycles - rules which have mutual references. I believe the problem
2613 is restricted to cases where there is more than one cycle AND
2614 elements of those cycles have rules in common. Even when this
2615 occurs it may not affect the code generated - but it might. It
2616 might also lead to undetected ambiguities.
2618 There were no changes in antlr or dlg output from the revised version.
2620 The following fragment demonstates the problem by giving different
2621 follow sets (option -pa) for var_access when built with k=1 and ck=2 on
2622 1.33 vanilla and 1.33MR10:
2624 echo_statement : ECHO ( echo_expr )*
2627 echo_expr : ( command )?
2631 command : IDENTIFIER
2635 expression : operand ( OPERATOR operand )*
2646 concat : var_access { CONCAT value }
2649 var_access : IDENTIFIER { INDEX }
2652 #81. (Changed in 1.33MR10) C mode use of attributes and ASTs
2654 Reported by Isaac Clark (irclark@mindspring.com).
2656 C mode code ignores attributes returned by rules which are
2657 referenced using element labels when ASTs are enabled (-gt option).
2659 1. start : r:rule t:Token <<$start=$r;>>
2661 The $r refrence will not work when combined with
2664 2. start : t:Token <<$start=$t;>>
2666 The $t reference works in all cases.
2668 3. start : rule <<$0=$1;>>
2670 Numeric labels work in all cases.
2672 With MR10 the user will receive an error message for case 1 when
2673 the -gt option is used.
2675 #80. (Fixed in 1.33MR10) (...)? as last alternative of block
2677 A construct like the following:
2683 does not make sense because there is no alternative when
2684 the guess block fails. This is now reported as a warning
2687 Previously, there was a code generation error for this case:
2688 the guess block was not "closed" when the guess failed.
2689 This could cause an infinite loop or other problems. This
2696 #include "charptr.h"
2700 #include "charptr.c"
2703 ANTLR(start(),stdin);
2707 #token "[\ \t]+" << zzskip(); >>
2708 #token "[\n]" << zzline++; zzskip(); >>
2710 #token Word "[a-z]+"
2711 #token Number "[0-9]+"
2717 test1 : (Word Word Word Word)?
2718 | (Word Word Word Number)?
2720 test2 : (Word Word Number Word)?
2721 | (Word Word Number Number)?
2724 Test data which caused infinite loop:
2728 #79. (Changed in 1.33MR10) Use of -fh with multiple parsers
2730 Previously, antlr always used the pre-processor symbol
2731 STDPCCTS_H as a gate for the file stdpccts.h. This
2732 caused problems when there were multiple parsers defined
2733 because they used the same gate symbol.
2735 In 1.33MR10, the -fh filename is used to generate the
2736 gate file for stdpccts.h. For instance:
2738 antlr -fh std_parser1.h
2740 generates the pre-processor symbol "STDPCCTS_std_parser1_H".
2742 Reported by Ramanathan Santhanam (ps@kumaran.com).
2744 #78. (Changed in 1.33MR9) Guard predicates that refer to rules
2746 ------------------------
2747 Please refer to Item #87
2748 ------------------------
2750 Guard predicates are processed during an early phase
2751 of antlr (during parsing) before all data structures
2754 There is an apparent bug in earlier versions of 1.33
2755 which caused guard predicates which contained references
2756 to rules (rather than tokens) to reference a structure
2757 which hadn't yet been initialized.
2759 In some cases (perhaps all cases) references to rules
2760 in guard predicates resulted in the use of "garbage".
2762 #79. (Changed in 1.33MR9) Jeff Vincent (JVincent@novell.com)
2764 Previously, the maximum length file name was set
2765 arbitrarily to 300 characters in antlr, dlg, and sorcerer.
2767 The config.h file now attempts to define the maximum length
2768 filename using _MAX_PATH from stdlib.h before falling back
2769 to using the value 300.
2771 #78. (Changed in 1.33MR9) Jeff Vincent (JVincent@novell.com)
2773 Put #ifndef/#endif around definition of ZZLEXBUFSIZE in
2776 #77. (Changed in 1.33MR9) Arithmetic overflow for very large grammars
2778 In routine HandleAmbiguities() antlr attempts to compute the
2779 number of possible elements in a set that is order of
2780 number-of-tokens raised to the number-of-lookahead-tokens power.
2781 For large grammars or large lookahead (e.g. -ck 7) this can
2782 cause arithmetic overflow.
2784 With 1.33MR9, arithmetic overflow in this computation is reported
2785 the first time it happens. The program continues to run and
2786 the program branches based on the assumption that the computed
2787 value is larger than any number computed by counting actual cases
2788 because 2**31 is larger than the number of bits in most computers.
2790 Before 1.33MR9 overflow was not reported. The behavior following
2791 overflow is not predictable by anyone but the original author.
2795 In 1.33MR10 the warning message is suppressed.
2796 The code which detects the overflow allows the
2797 computation to continue without an error. The
2798 error message itself made made users worry.
2800 #76. (Changed in 1.33MR9) Jeff Vincent (JVincent@novell.com)
2802 Jeff Vincent has convinced me to make ANTLRCommonToken and
2803 ANTLRCommonNoRefCountToken use variable length strings
2804 allocated from the heap rather than fixed length strings.
2805 By suitable definition of setText(), the copy constructor,
2806 and operator =() it is possible to maintain "copy" semantics.
2807 By "copy" semantics I mean that when a token is copied from
2808 an existing token it receives its own, distinct, copy of the
2809 text allocated from the heap rather than simply a pointer
2810 to the original token's text.
2812 ============================================================
2813 W * A * R * N * I * N * G
2814 ============================================================
2816 It is possible that this may cause problems for some users.
2817 For those users I have included the old version of AToken.h as
2818 pccts/h/AToken_traditional.h.
2820 #75. (Changed in 1.33MR9) Bruce Guenter (bruceg@qcc.sk.ca)
2822 Make DLGStringInput const correct. Since this is infrequently
2823 subclassed, it should affect few users, I hope.
2825 #74. (Changed in 1.33MR9) -o (output directory) option
2827 Antlr does not properly handle the -o output directory option
2828 when the filename of the grammar contains a directory part. For
2831 antlr -o outdir pccts_src/myfile.g
2833 causes antlr create a file called "outdir/pccts_src/myfile.cpp.
2834 It SHOULD create outdir/myfile.cpp
2836 The suggested code fix has been installed in antlr, dlg, and
2839 #73. (Changed in 1.33MR9) Hoisting of semantic predicates and -mrhoist
2841 ============================================
2842 This has been rendered obsolete by Item #117
2843 ============================================
2845 #72. (Changed in 1.33MR9) virtual saveState()/restoreState()/guess_XXX
2847 The following methods in ANTLRParser were made virtual at
2848 the request of S. Bochnak (S.Bochnak@microTool.com.pl):
2850 saveState() and restoreState()
2851 guess(), guess_fail(), and guess_done()
2853 #71. (Changed in 1.33MR9) Access to omitted command line argument
2855 If a switch requiring arguments is the last thing on the
2856 command line, and the argument is omitted, antlr would core.
2862 antlr test.g -prc off
2864 #70. (Changed in 1.33MR9) Addition of MSVC .dsp and .mak build files
2866 The following MSVC .dsp and .mak files for pccts and sorcerer
2867 were contributed by Stanislaw Bochnak (S.Bochnak@microTool.com.pl)
2868 and Jeff Vincent (JVincent@novell.com)
2870 PCCTS Distribution Kit
2871 ----------------------
2872 pccts/PCCTSMSVC50.dsw
2874 pccts/antlr/AntlrMSVC50.dsp
2875 pccts/antlr/AntlrMSVC50.mak
2877 pccts/dlg/DlgMSVC50.dsp
2878 pccts/dlg/DlgMSVC50.mak
2880 pccts/support/msvc.dsp
2882 Sorcerer Distribution Kit
2883 -------------------------
2884 pccts/sorcerer/SorcererMSVC50.dsp
2885 pccts/sorcerer/SorcererMSVC50.mak
2887 pccts/sorcerer/lib/msvc.dsp
2889 #69. (Changed in 1.33MR9) Change "unsigned int" to plain "int"
2891 Declaration of max_token_num in misc.c as "unsigned int"
2892 caused comparison between signed and unsigned ints giving
2893 warning message without any special benefit.
2895 #68. (Changed in 1.33MR9) Add void return for dlg internal_error()
2897 Get rid of "no return value" message in internal_error()
2898 in file dlg/support.c and dlg/dlg.h.
2900 #67. (Changed in Sor) sor.g: lisp() has no return value
2902 Added a "void" for the return type.
2904 #66. (Added to Sor) sor.g: ZZLEXBUFSIZE enclosed in #ifndef/#endif
2906 A user needed to be able to change the ZZLEXBUFSIZE for
2907 sor. Put the definition of ZZLEXBUFSIZE inside #ifndef/#endif
2909 #65. (Changed in 1.33MR9) PCCTSAST::deepCopy() and ast_dup() bug
2911 Jeff Vincent (JVincent@novell.com) found that deepCopy()
2912 made new copies of only the direct descendents. No new
2913 copies were made of sibling nodes, Sibling pointers are
2914 set to zero by shallowCopy().
2916 PCCTS_AST::deepCopy() has been changed to make a
2917 deep copy in the traditional sense.
2919 The deepCopy() routine depends on the behavior of
2920 shallowCopy(). In all sor examples I've found,
2921 shallowCopy() zeroes the right and down pointers.
2923 Original Tree Original deepCopy() Revised deepCopy
2924 ------------- ------------------- ----------------
2933 While comparing deepCopy() for C++ mode with ast_dup for
2934 C mode I found a problem with ast_dup().
2936 Routine ast_dup() has been changed to make a deep copy
2937 in the traditional sense.
2939 Original Tree Original ast_dup() Revised ast_dup()
2940 ------------- ------------------- ----------------
2943 d->e->f D->E->F D->E->F
2945 g->h->i G->H->I G->H->I
2950 I believe this affects transform mode sorcerer programs only.
2952 #64. (Changed in 1.33MR9) anltr/hash.h prototype for killHashTable()
2954 #63. (Changed in 1.33MR8) h/charptr.h does not zero pointer after free
2956 The charptr.h routine now zeroes the pointer after free().
2958 Reported by Jens Tingleff (jensting@imaginet.fr)
2960 #62. (Changed in 1.33MR8) ANTLRParser::resynch had static variable
2962 The static variable "consumed" in ANTLRParser::resynch was
2963 changed into an instance variable of the class with the
2964 name "resynchConsumed".
2966 Reported by S.Bochnak@microTool.com.pl
2968 #61. (Changed in 1.33MR8) Using rule>[i,j] when rule has no return values
2970 Previously, the following code would cause antlr to core when
2971 it tried to generate code for rule1 because rule2 had no return
2972 values ("upward inheritance"):
2974 rule1 : <<int i; int j>>
2980 Reported by S.Bochnak@microTool.com.pl
2982 Verified correct operation of antlr MR8 when missing or extra
2983 inheritance arguments for all combinations. When there are
2984 missing or extra arguments code will still be generated even
2985 though this might cause the invocation of a subroutine with
2986 the wrong number of arguments.
2988 #60. (Changed in 1.33MR7) Major changes to exception handling
2990 There were significant problems in the handling of exceptions
2991 in 1.33 vanilla. The general problem is that it can only
2992 process one level of exception handler. For example, a named
2993 exception handler, an exception handler for an alternative, or
2994 an exception for a subrule always went to the rule's exception
2995 handler if there was no "catch" which matched the exception.
2997 In 1.33MR7 the exception handlers properly "nest". If an
2998 exception handler does not have a matching "catch" then the
2999 nextmost outer exception handler is checked for an appropriate
3000 "catch" clause, and so on until an exception handler with an
3001 appropriate "catch" is found.
3003 There are still undesirable features in the way exception
3004 handlers are implemented, but I do not have time to fix them
3007 The exception handlers for alternatives are outside the
3008 block containing the alternative. This makes it impossible
3009 to access variables declared in a block or to resume the
3010 parse by "falling through". The parse can still be easily
3011 resumed in other ways, but not in the most natural fashion.
3013 This results in an inconsistentcy between named exception
3014 handlers and exception handlers for alternatives. When
3015 an exception handler for an alternative "falls through"
3016 it goes to the nextmost outer handler - not the "normal
3019 A major difference between 1.33MR7 and 1.33 vanilla is
3020 the default action after an exception is caught:
3024 In 1.33 vanilla the signal value is set to zero ("NoSignal")
3025 and the code drops through to the code following the exception.
3026 For named exception handlers this is the "normal action".
3027 For alternative exception handlers this is the rule's handler.
3031 In 1.33MR7 the signal value is NOT automatically set to zero.
3033 There are two cases:
3035 For named exception handlers: if the signal value has been
3036 set to zero the code drops through to the "normal action".
3038 For all other cases the code branches to the nextmost outer
3039 exception handler until it reaches the handler for the rule.
3041 The following macros have been defined for convenience:
3044 --------------------
3046 set signal & return signal arg to 0 ("NoSignal")
3047 (zz)setSignal(intValue)
3048 set signal & return signal arg to some value
3050 copy the signal value to the return signal arg
3052 I'm not sure why PCCTS make a distinction between the local
3053 signal value and the return signal argument, but I'm loathe
3054 to change the code. The burden of copying the local signal
3055 value to the return signal argument can be given to the
3056 default signal handler, I suppose.
3058 #59. (Changed in 1.33MR7) Prototypes for some functions
3060 Added prototypes for the following functions to antlr.h
3063 zzconsumeUntilToken()
3065 #58. (Changed in 1.33MR7) Added defintion of zzbufsize to dlgauto.h
3067 #57. (Changed in 1.33MR7) Format of #line directive
3069 Previously, the -gl directive for line 1234 would
3070 resemble: "# 1234 filename.g". This caused problems
3071 for some compilers/pre-processors. In MR7 it generates
3072 "#line 1234 filename.g".
3074 #56. (Added in 1.33MR7) Jan Mikkelsen <janm@zeta.org.au>
3076 Move PURIFY macro invocaton to after rule's init action.
3078 #55. (Fixed in 1.33MR7) Unitialized variables in ANTLRParser
3080 Member variables inf_labase and inf_last were not initialized.
3083 #54. (Fixed in 1.33MR6) Brad Schick (schick@interacess.com)
3085 Previously, the following constructs generated the same
3096 In all versions of pccts rule1 guesses (A B C) and then
3097 consume all three tokens if the guess succeeds. In MR6
3098 rule2 guesses (A B C) but consumes NONE of the tokens
3099 when the guess succeeds because "()" matches epsilon.
3101 #53. (Explanation for 1.33MR6) What happens after an exception is caught ?
3103 The Book is silent about what happens after an exception
3106 The following code fragment prints "Error Action" followed
3109 test : Word ex:Number <<printf("Normal Action\n");>>
3112 <<printf("Error Action\n");>>
3115 The reason for "Normal Action" is that the normal flow of the
3116 program after a user-written exception handler is to "drop through".
3117 In the case of an exception handler for a rule this results in
3118 the exection of a "return" statement. In the case of an
3119 exception handler attached to an alternative, rule, or token
3120 this is the code that would have executed had there been no
3123 The user can achieve the desired result by using a "return"
3126 test : Word ex:Number <<printf("Normal Action\n");>>
3129 <<printf("Error Action\n"); return;>>
3132 The most powerful mechanism for recovery from parse errors
3133 in pccts is syntactic predicates because they provide
3134 backtracking. Exceptions allow "return", "break",
3135 "consumeUntil(...)", "goto _handler", "goto _fail", and
3136 changing the _signal value.
3138 #52. (Fixed in 1.33MR6) Exceptions without syntactic predicates
3140 The following generates bad code in 1.33 if no syntactic
3141 predicates are present in the grammar.
3143 test : Word ex:Number <<printf("Normal Action\n");>>
3146 <<printf("Error Action\n");>>
3148 There is a reference to a guess variable. In C mode
3149 this causes a compiler error. In C++ mode it generates
3150 an extraneous check on member "guessing".
3152 In MR6 correct code is generated for both C and C++ mode.
3154 #51. (Added to 1.33MR6) Exception operator "@" used without exceptions
3156 In MR6 added a warning when the exception operator "@" is
3157 used and no exception group is defined. This is probably
3158 a case where "\@" or "@" is meant.
3160 #50. (Fixed in 1.33MR6) Gunnar Rxnning (gunnar@candleweb.no)
3161 http://www.candleweb.no/~gunnar/
3163 Routines zzsave_antlr_state and zzrestore_antlr_state don't
3164 save and restore all the data needed when switching states.
3166 Suggested patch applied to antlr.h and err.h for MR6.
3168 #49. (Fixed in 1.33MR6) Sinan Karasu (sinan@boeing.com)
3170 Generated code failed to turn off guess mode when leaving a
3171 (...)+ block which contained a guess block. The result was
3172 an infinite loop. For example:
3179 Suggested code fix implemented in MR6. Replaced
3181 ... else if (zzcnt>1) break;
3186 ... else if (zzcnt>1) {if (!zzrv) zzGUESS_DONE; break;};
3188 ... else if (zzcnt>1) {if (zzguessing) zzGUESS_DONE; break;};
3190 #48. (Fixed in 1.33MR6) Invalid exception element causes core
3192 A label attached to an invalid construct can cause
3193 pccts to crash while processing the exception associated
3194 with the label. For example:
3197 exception[t] catch MismatchedToken: <<printf(...);>>
3199 Version MR6 generates the message:
3201 reference in exception handler to undefined label 't'
3203 #47. (Fixed in 1.33MR6) Manuel Ornato
3205 Under some circumstances involving a k >1 or ck >1
3206 grammar and a loop block (i.e. (...)* ) pccts will
3207 fail to detect a syntax error and loop indefinitely.
3208 The problem did not exist in 1.20, but has existed
3209 from 1.23 to the present.
3213 ---------------------------------------------------
3214 Complete test program
3215 ---------------------------------------------------
3218 #include "charptr.h"
3222 #include "charptr.c"
3225 ANTLR(global(),stdin);
3229 #token "[\ \t]+" << zzskip(); >>
3230 #token "[\n]" << zzline++; zzskip(); >>
3248 listeb : LP ( B ( B | C )* ) RP ;
3249 listec : LP ( C ( B | C )* ) RP ;
3250 listed : LP ( D ( B | C )* ) RP ;
3251 liste : ( listeb | listec )* ;
3253 ---------------------------------------------------
3254 Sample data causing infinite loop
3255 ---------------------------------------------------
3257 ---------------------------------------------------
3259 #46. (Fixed in 1.33MR6) Robert Richter
3260 (Robert.Richter@infotech.tu-chemnitz.de)
3262 This item from the list of known problems was
3263 fixed by item #18 (below).
3265 #45. (Fixed in 1.33MR6) Brad Schick (schick@interaccess.com)
3267 The dependency scanner in VC++ mistakenly sees a
3268 reference to an MPW #include file even though properly
3269 #ifdef/#endif in config.h. The suggested workaround
3270 has been implemented:
3274 #define MPW_CursorCtl_Header <CursorCtl.h>
3275 #include MPW_CursorCtl_Header
3279 #44. (Fixed in 1.33MR6) cast malloc() to (char *) in charptr.c
3281 Added (char *) cast for systems where malloc returns "void *".
3283 #43. (Added to 1.33MR6) Bruce Guenter (bruceg@qcc.sk.ca)
3285 Add setLeft() and setUp methods to ASTDoublyLinkedBase
3286 for symmetry with setRight() and setDown() methods.
3288 #42. (Fixed in 1.33MR6) Jeff Katcher (jkatcher@nortel.ca)
3290 C++ style comment in antlr.c corrected.
3292 #41. (Added in 1.33MR6) antlr -stdout
3294 Using "antlr -stdout ..." forces the text that would
3295 normally go to the grammar.c or grammar.cpp file to
3298 #40. (Added in 1.33MR6) antlr -tab to change tab stops
3300 Using "antlr -tab number ..." changes the tab stops
3301 for the grammar.c or grammar.cpp file. The number
3302 must be between 0 and 8. Using 0 gives tab characters,
3303 values between 1 and 8 give the appropriate number of
3306 #39. (Fixed in 1.33MR5) Jan Mikkelsen <janm@zeta.org.au>
3308 Commas in function prototype still not correct under
3309 some circumstances. Suggested code fix installed.
3311 #38. (Fixed in 1.33MR5) ANTLRTokenBuffer constructor
3313 Have ANTLRTokenBuffer ctor initialize member "parser" to null.
3315 #37. (Fixed in 1.33MR4) Bruce Guenter (bruceg@qcc.sk.ca)
3317 In ANTLRParser::FAIL(int k,...) released memory pointed to by
3318 f[i] (as well as f itself. Should only free f itself.
3320 #36. (Fixed in 1.33MR3) Cortland D. Starrett (cort@shay.ecn.purdue.edu)
3322 Neglected to properly declare isDLGmaxToken() when fixing problem
3323 reported by Andreas Magnusson.
3325 Undo "_retv=NULL;" change which caused problems for return values
3326 from rules whose return values weren't pointers.
3328 Failed to create bin directory if it didn't exist.
3330 #35. (Fixed in 1.33MR2) Andreas Magnusson
3331 (Andreas.Magnusson@mailbox.swipnet.se)
3333 Repair bug introduced by 1.33MR1 for #tokdefs. The original fix
3334 placed "DLGmaxToken=9999" and "DLGminToken=0" in the TokenType enum
3335 in order to fix a problem with an aggresive compiler assigning an 8
3336 bit enum which might be too narrow. This caused #tokdefs to assume
3337 that there were 9999 real tokens. The repair to the fix causes antlr to
3338 ignore TokenTypes "DLGmaxToken" and "DLGminToken" in a #tokdefs file.
3340 #34. (Added to 1.33MR1) Add public DLGLexerBase::set_line(int newValue)
3342 Previously there was no public function for changing the line
3343 number maintained by the lexer.
3345 #33. (Fixed in 1.33MR1) Franklin Chen (chen@adi.com)
3347 Accidental use of EXIT_FAILURE rather than PCCTS_EXIT_FAILURE
3348 in pccts/h/AParser.cpp.
3350 #32. (Fixed in 1.33MR1) Franklin Chen (chen@adi.com)
3352 In PCCTSAST.cpp lines 405 and 466: Change
3360 #31. (Added to 1.33MR1) Pointer to parser in ANTLRTokenBuffer
3361 Pointer to parser in DLGLexerBase
3363 The ANTLRTokenBuffer class now contains a pointer to the
3364 parser which is using it. This is established by the
3365 ANTLRParser constructor calling ANTLRTokenBuffer::
3366 setParser(ANTLRParser *p).
3368 When ANTLRTokenBuffer::setParser(ANTLRParser *p) is
3369 called it saves the pointer to the parser and then
3370 calls ANTLRTokenStream::setParser(ANTLRParser *p)
3371 so that the lexer can also save a pointer to the
3374 There is also a function getParser() in each class
3375 with the obvious purpose.
3377 It is possible that these functions will return NULL
3378 under some circumstances (e.g. a non-DLG lexer is used).
3380 #30. (Added to 1.33MR1) function tokenName(int token) standard
3382 The generated parser class now includes the
3385 static const ANTLRChar * tokenName(int token)
3387 which returns a pointer to the "name" corresponding
3390 The base class (ANTLRParser) always includes the
3393 const ANTLRChar * parserTokenName(int token)
3395 which can be accessed by objects which have a pointer
3396 to an ANTLRParser, but do not know the name of the
3397 parser class (e.g. ANTLRTokenBuffer and DLGLexerBase).
3399 #29. (Added to 1.33MR1) Debugging DLG lexers
3401 If the pre-processor symbol DEBUG_LEXER is defined
3402 then DLexerBase will include code for printing out
3403 key information about tokens which are recognized.
3405 The debug feature of the lexer is controlled by:
3407 int previousDebugValue=lexer.debugLexer(newValue);
3409 a value of 0 disables output
3410 a value of 1 enables output
3412 Even if the lexer debug code is compiled into DLexerBase
3413 it must be enabled before any output is generated. For
3416 DLGFileInput in(stdin);
3417 MyDLG lexer(&in,2000);
3419 lexer.setToken(&aToken);
3422 lexer.debugLexer(1); // enable debug information
3425 #28. (Added to 1.33MR1) More control over DLG header
3427 Version 1.33MR1 adds the following directives to PCCTS
3430 #lexprefix <<source code>>
3432 Adds source code to the DLGLexer.h file
3433 after the #include "DLexerBase.h" but
3434 before the start of the class definition.
3436 #lexmember <<source code>>
3438 Adds source code to the DLGLexer.h file
3439 as part of the DLGLexer class body. It
3440 appears immediately after the start of
3441 the class and a "public: statement.
3443 #27. (Fixed in 1.33MR1) Comments in DLG actions
3445 Previously, DLG would not recognize comments as a special case.
3446 Thus, ">>" in the comments would cause errors. This is fixed.
3448 #26. (Fixed in 1.33MR1) Removed static variables from error routines
3450 Previously, the existence of statically allocated variables
3451 in some of the parser's member functions posed a danger when
3452 there was more than one parser active.
3454 Replaced with dynamically allocated/freed variables in 1.33MR1.
3456 #25. (Fixed in 1.33MR1) Use of string literals in semantic predicates
3458 Previously, it was not possible to place a string literal in
3459 a semantic predicate because it was not properly "stringized"
3460 for the report of a failed predicate.
3462 #24. (Fixed in 1.33MR1) Continuation lines for semantic predicates
3464 Previously, it was not possible to continue semantic
3465 predicates across a line because it was not properly
3466 "stringized" for the report of a failed predicate.
3468 rule : <<ifXYZ()>>?[ a very
3471 #23. (Fixed in 1.33MR1) {...} envelope for failed semantic predicates
3473 Previously, there was a code generation error for failed
3474 semantic predicates:
3476 rule : <<xyz()>>?[ stmt1; stmt2; ]
3478 which generated code which resembled:
3480 if (! xyz()) stmt1; stmt2;
3482 It now puts the statements in a {...} envelope:
3484 if (! xyz()) { stmt1; stmt2; };
3486 #22. (Fixed in 1.33MR1) Continuation of #token across lines using "\"
3488 Previously, it was not possible to continue a #token regular
3489 expression across a line. The trailing "\" and newline caused
3490 a newline to be inserted into the regular expression by DLG.
3494 #21. (Fixed in 1.33MR1) Use of ">>" (right shift operator in DLG actions
3496 It is now possible to use the C++ right shift operator ">>"
3497 in DLG actions by using the normal escapes:
3499 #token "shift-right" << value=value \>\> 1;>>
3501 #20. (Version 1.33/19-Jan-97 Karl Eccleson <karle@microrobotics.co.uk>
3502 P.A. Keller (P.A.Keller@bath.ac.uk)
3504 There is a problem due to using exceptions with the -gh option.
3506 Suggested fix now in 1.33MR1.
3508 #19. (Fixed in 1.33MR1) Tom Piscotti and John Lilley
3510 There were problems suppressing messages to stdin and stdout
3511 when running in a window environment because some functions
3512 which uses fprint were not virtual.
3514 Suggested change now in 1.33MR1.
3516 I believe all functions containing error messages (excluding those
3517 indicating internal inconsistency) have been placed in functions
3520 #18. (Version 1.33/ 22-Nov-96) John Bair (jbair@iftime.com)
3522 Under some combination of options a required "return _retv" is
3525 Suggested fix now in 1.33MR1.
3527 #17. (Version 1.33/3-Sep-96) Ron House (house@helios.usq.edu.au)
3529 The routine ASTBase::predorder_action omits two "tree->"
3530 prefixes, which results in the preorder_action belonging
3531 to the wrong node to be invoked.
3533 Suggested fix now in 1.33MR1.
3535 #16. (Version 1.33/7-Jun-96) Eli Sternheim <eli@interhdl.com>
3537 Routine consumeUntilToken() does not check for end-of-file
3540 Suggested fix now in 1.33MR1.
3542 #15. (Version 1.33/8 Apr 96) Asgeir Olafsson <olafsson@cstar.ac.com>
3544 Problem with tree duplication of doubly linked ASTs in ASTBase.cpp.
3546 Suggested fix now in 1.33MR1.
3548 #14. (Version 1.33/28-Feb-96) Andreas.Magnusson@mailbox.swipnet.se
3550 Problem with definition of operator = (const ANTLRTokenPtr rhs).
3552 Suggested fix now in 1.33MR1.
3554 #13. (Version 1.33/13-Feb-96) Franklin Chen (chen@adi.com)
3556 Sun C++ Compiler 3.0.1 can't compile testcpp/1 due to goto in
3557 block with destructors.
3559 Apparently fixed. Can't locate "goto".
3561 #12. (Version 1.33/10-Nov-95) Minor problems with 1.33 code
3563 The following items have been fixed in 1.33MR1:
3565 1. pccts/antlr/main.c line 142
3567 "void" appears in classic C code
3569 2. no makefile in support/genmk
3571 3. EXIT_FAILURE/_SUCCESS instead of PCCTS_EXIT_FAILURE/_SUCCESS
3573 pccts/h/PCCTSAST.cpp
3574 pccts/h/DLexerBase.cpp
3575 pccts/testcpp/6/test.g
3577 4. use of "signed int" isn't accepted by AT&T cfront
3579 pccts/h/PCCTSAST.h line 42
3581 5. in call to ANTLRParser::FAIL the var arg err_k is passed as
3582 "int" but is declared "unsigned int".
3584 6. I believe that a failed validation predicate still does not
3585 get put in a "{...}" envelope, despite the release notes.
3587 7. The #token ">>" appearing in the DLG grammar description
3588 causes DLG to generate the string literal "\>\>" which
3589 is non-conforming and will cause some compilers to
3590 complain (scan.c function act10 line 143 of source code).
3592 #11. (Version 1.32b6) Dave Kuhlman (dkuhlman@netcom.com)
3594 Problem with file close in gen.c. Already fixed in 1.33.
3596 #10. (Version 1.32b6/29-Aug-95)
3598 pccts/antlr/main.c contains a C++ style comments on lines 149
3599 and 176 which causes problems for most C compilers.
3601 Already fixed in 1.33.
3603 #9. (Version 1.32b4/14-Mar-95) dlgauto.h #include "config.h"
3605 The file pccts/h/dlgauto.h should probably contain a #include
3606 "config.h" as it uses the #define symbol __USE_PROTOS.
3610 #8. (Version 1.32b4/6-Mar-95) Michael T. Richter (mtr@igs.net)
3612 In C++ output mode anonymous tokens from in-line regular expressions
3613 can create enum values which are too wide for the datatype of the enum
3614 assigned by the C++ compiler.
3618 #7. (Version 1.32b4/6-Mar-95) C++ does not imply __STDC__
3620 In err.h the combination of # directives assumes that a C++
3621 compiler has __STDC__ defined. This is not necessarily true.
3623 This problem also appears in the use of __USE_PROTOS which
3624 is appropriate for both Standard C and C++ in antlr/gen.c
3629 #6. (Version 1.32 ?/15-Feb-95) Name conflict for "TokenType"
3631 Already fixed in 1.33.
3633 #5. (23-Jan-95) Douglas_Cuthbertson.JTIDS@jtids_qmail.hanscom.af.mil
3635 The fail action following a semantic predicate is not enclosed in
3636 "{...}". This can lead to problems when the fail action contains
3637 more than one statement.
3641 #4 . (Version 1.33/31-Mar-96) jlilley@empathy.com (John Lilley)
3643 Put briefly, a semantic predicate ought to abort a guess if it fails.
3645 Correction suggested by J. Lilley has been added to 1.33MR1.
3647 #3 . (Version 1.33) P.A.Keller@bath.ac.uk
3649 Extra commas are placed in the K&R style argument list for rules
3650 when using both exceptions and ASTs.
3654 #2. (Version 1.32b6/2-Oct-95) Brad Schick <schick@interaccess.com>
3656 Construct #[] generates zzastnew() in C++ mode.
3658 Already fixed in 1.33.
3660 #1. (Version 1.33) Bob Bailey (robert@oakhill.sps.mot.com)
3662 Previously, config.h assumed that all PC systems required
3663 "short" file names. The user can now override that
3664 assumption with "#define LONGFILENAMES".