Cpp, Go, JavaScript, Python2/3: Template rendering error.

[AndreiPashkin]

Here is an example grammar:

grammar Test;

expression: op=NOT args+=expression
          | args+=expression (op=AND args+=expression)+
          | args+=expression (op=OR args+=expression)+
          | IDENTIFIER
          ;
AND : 'and' ;
OR : 'or' ;
NOT : 'not' ;
IDENTIFIER : [a-zA-Z_][a-zA-Z0-9_]* ;
WS : [ \t\r\n]+ -> skip ;

For Cpp, Go, JavaScript, Python2/3 target languages version 4.7 throws such errors while generating code:

error(32):  code generation template recRuleAltStartAction has missing, misnamed, or incomplete arg list; missing isListLabel

For the rest of the languages it works.

I believe that the error is in the templates since in the recRuleAltStartAction templates for the problematic languages argument isListLabel is not present while in the templates of the ok-languages it is present.

[ericvergnaud]

Hi, this grammar also fails in the Java reference implementation, although at a later stage: ./var/folders/7z/rdw_873x3_s5zr8pb81k9z7c0000gn/T/BaseJavaTest-main-95208223142060/TParser.java:173: error: incompatible types: ExpressionContext cannot be converted to List<ExpressionContext> _localctx.args = _prevctx; ^ /var/folders/7z/rdw_873x3_s5zr8pb81k9z7c0000gn/T/BaseJavaTest-main-95208223142060/TParser.java:206: error: incompatible types: ExpressionContext cannot be converted to List<ExpressionContext> _localctx.args = _prevctx; Not sure if this is bug in ANTLR or in the grammar, but I’d suggest getting this right with the reference implementation first. @parrt can you have a look (I’ve added a test descriptor, see enclosed) Eric

…

Le 19 sept. 2017 à 05:27, Andrew Pashkin ***@***.***> a écrit : Here is an example grammar: grammar Test; expression: op=NOT args+=expression | args+=expression (op=AND args+=expression)+ | args+=expression (op=OR args+=expression)+ | IDENTIFIER ; AND : 'and' ; OR : 'or' ; NOT : 'not' ; IDENTIFIER : [a-zA-Z_][a-zA-Z0-9_]* ; WS : [ \t\r\n]+ -> skip ; For Cpp, Go, JavaScript, Python2/3 target languages version 4.7 throws such errors while generating code: error(32): code generation template recRuleAltStartAction has missing, misnamed, or incomplete arg list; missing isListLabel — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2016>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADLYJCko1tuiC7EHc03BjAdNFbVWnNiLks5sjuA5gaJpZM4PbjXV>.

[parrt]

With the Java target, all is ok:

varmint:/tmp $ alias a4.7.1
alias a4.7.1='java -cp "/Users/parrt/.m2/repository/org/antlr/antlr4/4.7.1-SNAPSHOT/antlr4-4.7.1-SNAPSHOT-complete.jar:$CLASSPATH" org.antlr.v4.Tool'
varmint:/tmp $ a4.7.1 Test.g4
varmint:/tmp $ 

@ericvergnaud Looks like Swift, Java, C# are ok and have this argument. Probably should add to others.

It appears to be needed:

recRuleAltStartAction(ruleName, ctxName, label, isListLabel) ::= <<
_localctx = new <ctxName>Context(_parentctx, _parentState);
<if(label)>
<if(isListLabel)>
_localctx.<label>.add(_prevctx);
<else>
_localctx.<label> = _prevctx;
<endif>
<endif>
<if(label)>_localctx.<label> = _prevctx;<endif>
pushNewRecursionContext(_localctx, _startState, RULE_<ruleName>);
>>

[ericvergnaud]

But in Java the generated code won't compile Envoyé de mon iPhone

…

Le 20 sept. 2017 à 04:04, Terence Parr ***@***.***> a écrit : With the Java target, all is ok: varmint:/tmp $ alias a4.7.1 alias a4.7.1='java -cp "/Users/parrt/.m2/repository/org/antlr/antlr4/4.7.1-SNAPSHOT/antlr4-4.7.1-SNAPSHOT-complete.jar:$CLASSPATH" org.antlr.v4.Tool' varmint:/tmp $ a4.7.1 Test.g4 varmint:/tmp $ Looks like Swift, Java, C# are ok and have this argument. Probably should add to others. It appears to be needed: recRuleAltStartAction(ruleName, ctxName, label, isListLabel) ::= << _localctx = new <ctxName>Context(_parentctx, _parentState); <if(label)> <if(isListLabel)> _localctx.<label>.add(_prevctx); <else> _localctx.<label> = _prevctx; <endif> <endif> <if(label)>_localctx.<label> = _prevctx;<endif> pushNewRecursionContext(_localctx, _startState, RULE_<ruleName>); >> — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

[parrt]

Ah. Sorry ok I'll check

…

Sent from my iPhone

On Sep 19, 2017, at 4:02 PM, ericvergnaud ***@***.***> wrote: But in Java the generated code won't compile Envoyé de mon iPhone > Le 20 sept. 2017 à 04:04, Terence Parr ***@***.***> a écrit : > > With the Java target, all is ok: > > varmint:/tmp $ alias a4.7.1 > alias a4.7.1='java -cp "/Users/parrt/.m2/repository/org/antlr/antlr4/4.7.1-SNAPSHOT/antlr4-4.7.1-SNAPSHOT-complete.jar:$CLASSPATH" org.antlr.v4.Tool' > varmint:/tmp $ a4.7.1 Test.g4 > varmint:/tmp $ > Looks like Swift, Java, C# are ok and have this argument. Probably should add to others. > > > > It appears to be needed: > > recRuleAltStartAction(ruleName, ctxName, label, isListLabel) ::= << > _localctx = new <ctxName>Context(_parentctx, _parentState); > <if(label)> > <if(isListLabel)> > _localctx.<label>.add(_prevctx); > <else> > _localctx.<label> = _prevctx; > <endif> > <endif> > <if(label)>_localctx.<label> = _prevctx;<endif> > pushNewRecursionContext(_localctx, _startState, RULE_<ruleName>); > >> > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub, or mute the thread. > — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[parrt]

It's a bug in the Java target template.I believe that

<if(label)>_localctx.<label> = _prevctx;<endif>

was left in after creating the if statement right above it. I have created a unit test that should make everybody fail. I will see if I can figure out the other targets.

[parrt]

See PR #2022 I think i have it all fixed but it uncovered a C++ target prediction issue.

[jimidle]

PR #3834 submitted to correct this for the Go target.

[KvanTTT]

There is no error for Java runtime, but the generated parser is not compilable. In fact, the grammar does not work correctly for all runtimes.

[parrt]

Looks like I tried to fix it years ago but didn't finish? #2022

[KvanTTT]

Yes.

[jimidle]

Ah - it is because of reliance on OOP. Not sure why I did not spot that. Let me take another look.

[parrt]

Hoping to resolve with #3841

[parrt]

C++ and Go fail on this still but runtime now and not compilation.

[parrt]

Fixed-ish by #3841

[KvanTTT]

Please close since it's fixed.