[antlr-interest] OPEN, EOF ambiguity

[Cliff Hudson]

Well, it's not always true that grammar specs are, for instance, LALR(1) or
LL(k).  In such cases, you have to rejigger the grammar to make it work.
 The important thing is (or should be) that the grammar you do produce,
regardless of the technicalities, will parse what you intend and nothing
more.

On Fri, Apr 9, 2010 at 5:03 AM, Cameron Ross <cross at kojeware.com> wrote:

> Hi Cliff,
>
> What you've suggested is how I have the production version of the grammar
> implemented now.  However, the language spec that I must conform to
> specifically states that a text can have zero or more phrases (which is the
> crux of the problem).  I can leave my implementation the way it is and try
> to have the spec changed, but it would be a heck of a lot easier to just
> implement to the current spec.
>
> Thanks for the help,
> Cameron.
>
> On Thu, Apr 8, 2010 at 8:49 PM, Cliff Hudson <cliff.s.hudson at gmail.com>wrote:
>
>> I think you might just have to pop up one level to think about how to
>> express this - no magic needed.  The way that grammar is written, your file
>> is composed of one or more cltexts, which themselves may be composed either
>> of a module or any number of phrases (including none at all).  It seems to
>> me this can be re-specified to say that your file may be composed of any
>> combination of modules and texts, but may also be empty.  The empty file
>> condition should *not* be specified by the closure on phrase but rather as a
>> top-level rule.  For example:
>>
>> start_rule
>>   : clif_file
>>   |
>>   ;
>>
>> clif_file:
>>   cltext+ ;
>>
>> cltext
>>   : module
>>   | text
>>   ;
>>
>> text:
>>   phrase+;
>>
>> This ensures that each cltext will produce at least something, and no
>> production from that rule will lead to the EOF condition so there will be no
>> confusion by ANTLR.
>>
>> NOTE: I don't have ANTLR handy to test this...
>>
>>
>> On Thu, Apr 8, 2010 at 1:11 PM, Cameron Ross <cross at kojeware.com> wrote:
>>
>>> Yes, this is the way I have it implemented in my production version.
>>>  However, I have to conform to a language specification which allows text
>>> to
>>> have zero or more phrases.  Is there some ANLTR magic that I can use to
>>> support this?
>>>
>>>
>>> On Thu, Apr 8, 2010 at 4:07 PM, Jim Idle <jimi at temporal-wave.com> wrote:
>>>
>>> > Your text rule can be empty, you need to rejig your use of + and * or
>>> you
>>> > end up with the empty match in more than one place.
>>> >
>>> >
>>> >
>>> > Jim
>>> >
>>> >
>>> >
>>> > From: Cameron Ross [mailto:cross at kojeware.com]
>>> > Sent: Thursday, April 08, 2010 1:05 PM
>>> > To: Jim Idle
>>> > Subject: Re: [antlr-interest] OPEN, EOF ambiguity
>>> >
>>> >
>>> >
>>> > These rules give the gist of it...
>>> >
>>> >
>>> >
>>> > start_rule
>>> >
>>> >          : clif_file EOF
>>> >
>>> >          ;
>>> >
>>> >
>>> >
>>> > clif_file
>>> >
>>> >          : cltext+
>>> >
>>> >          ;
>>> >
>>> >
>>> >
>>> > cltext
>>> >
>>> >          : (OPEN CL_MODULE) => module -> ^(CLTEXT module)
>>> >
>>> >          | text -> ^(CLTEXT text)
>>> >
>>> >          ;
>>> >
>>> >
>>> >
>>> > module
>>> >
>>> >          : OPEN CL_MODULE interpretablename exclusion_set? cltext CLOSE
>>> ->
>>> > ^(MODULE ^(MODULE_IDENTIFIER interpretablename) exclusion_set? cltext)
>>> >
>>> >          ;
>>> >
>>> >
>>> >
>>> > text
>>> >
>>> >          : phrase* -> ^(TEXT phrase*)
>>> >
>>> >          ;
>>> >
>>> > ...
>>> >
>>> >
>>> >
>>> >
>>> >
>>> >
>>> >
>>> > On Thu, Apr 8, 2010 at 3:57 PM, Jim Idle <jimi at temporal-wave.com>
>>> wrote:
>>> >
>>> > You probably need to send the grammar rather than that graphic. It is
>>> > likely something more fundamental and we cannot see what you predicates
>>> are
>>> > doing from the graphic.
>>> >
>>> > Jim
>>> >
>>> >
>>> > > -----Original Message-----
>>> > > From: antlr-interest-bounces at antlr.org [mailto:antlr-interest-
>>> > > bounces at antlr.org] On Behalf Of Cameron Ross
>>> > > Sent: Thursday, April 08, 2010 12:34 PM
>>> > > To: antlr-interest at antlr.org
>>> > > Subject: [antlr-interest] OPEN, EOF ambiguity
>>> > >
>>> > > Hi,
>>> > >
>>> > > I have a grammar with an ambiguity that I just can't resolve.  When I
>>> > > check the grammar using AntlrWorks I get the following: "Decision can
>>> > > match input such as "{EOF, OPEN}" using multiple alternatives: 1, 2".
>>> > > I've attached a screen grab of the syntax diagram generated by
>>> > > AntrlWorks.  I don't understand why ANLTR is thinking that EOF is
>>> > > ambiguous with OPEN.  I've checked the grammar and clif_file is the
>>> > > only rule that is not referenced from within the grammar, so I don't
>>> > > think that ANLTR should see module as a start rule.  Any help is much
>>> > > appreciated.
>>> > >
>>> > > Cameron
>>> >
>>> >
>>> >
>>> >
>>> >
>>> > List: http://www.antlr.org/mailman/listinfo/antlr-interest
>>> > Unsubscribe:
>>> > http://www.antlr.org/mailman/options/antlr-interest/your-email-address
>>> >
>>> >
>>> >
>>> >
>>> > --
>>> > Kojeware Corporation
>>> >
>>> >
>>> >
>>> >
>>> > List: http://www.antlr.org/mailman/listinfo/antlr-interest
>>> > Unsubscribe:
>>> > http://www.antlr.org/mailman/options/antlr-interest/your-email-address
>>> >
>>>
>>>
>>>
>>> --
>>> Kojeware Corporation
>>>
>>> List: http://www.antlr.org/mailman/listinfo/antlr-interest
>>> Unsubscribe:
>>> http://www.antlr.org/mailman/options/antlr-interest/your-email-address
>>>
>>
>>
>
>
> --
> Kojeware Corporation
>