Re: Spirit X3 employee.cpp compile error

On 08/20/2016 05:20 AM, Larry Evans wrote:
> On 08/12/2016 08:29 AM, Larry Evans wrote:
>> On 08/12/2016 07:53 AM, Vincent Virgilio wrote:
>>>
>>>
>>> On Aug 12, 2016, at 08:27 AM, Larry Evans <[email protected]> wrote:
>>>
>>>> On 08/12/2016 06:40 AM, Larry Evans wrote:
>>>> [snip]
>>>>>
>>>>>> GitHub: [email protected] <mailto:[email protected]>:boostorg/spirit.git
>>>>>>
>>>>>> Use the develop branch.
>>>>>> --}--cut here--
>>>>>>
>>>>> The 2 template arg rule code is actually in the x3-devel branch instead
>>>>> of the develop branch:
>>>>>
>>>>> https://github.com/boostorg/spirit/blob/x3-devel/include/boost/spirit/home/x3/nonterminal/rule.hpp#L80
>>>>>
>>>> [snip]
>>>> However, if you decide to use x3-devel, then the employee.cpp code
>>>> referenced in your OP needs to be modified so that the arguments
>>>> to BOOST_SPIRIT_DEFINE contain assignments instead of just
>>>> rule names. For example, see:
>>>>
>>>> https://github.com/boostorg/spirit/blob/x3-devel/example/x3/calc4.cpp#L218
>>>>
>>>> HTH.
>>>>
>>>> -regards,
>>>> Larry
>>>
>>> Hi Larry,
>>>
>>> Thanks for all the help so far.
>>>
>>> Why is BOOST_SPIRIT_DEFINE required if the rule declarations and
>>> definitions are already associated via assignment? In another small
>>> project, I was able to parse with an X3 grammar without that macro. In
>>> that case, very schematically, I had "auto const rule_name = rule<...>()
>>> = rule_def;"
>>>
>> [snip]
>> Hi Vincent,
>>
>> I can only guess that the grammar in this small project had no
>> recursive rules like the calc4.cpp grammar I mentioned above.
>>
>> That's only a guess; however, I *think* the whole purpose of
>> BOOST_SPIRIT_DEFINE is to allow mutually recursive rules,
>
> Apparently that conclusion is wrong because here:
>
> https://github.com/boostorg/spirit/blob/master/test/x3/rule1.cpp#L48
>
> there's a single recursive rule and there's no BOOST_SPIRIT_DEFINE.
>
> Sorry for misleading you.  I'm still trying to learn how the code
> works :(
>

OOPS.  I should have re-looked at:

https://github.com/boostorg/spirit/blob/develop/workbench/x3/toy/toy.cpp#L134

which shows the rule_definition<>::parse function stores the
rule definition in the context dictionary.  Then, even
when BOOST_SPIRIT_DEFINE is not used, the rule rhs is looked up
by the rule<...>::parse function:

https://github.com/boostorg/spirit/blob/develop/workbench/x3/toy/toy.cpp#L154

The process is a little more complicated in the actual code.
There, rule_defintion<...>::parse:

https://github.com/boostorg/spirit/blob/master/include/boost/spirit/home/x3/nonterminal/rule.hpp#L64

*maybe* does something similar by putting the definition in the
context dictionary where the key is the rule ID.  Then, in subsequent
calls to rule<...>::parse:

https://github.com/boostorg/spirit/blob/master/include/boost/spirit/home/x3/nonterminal/rule.hpp#L113

the rule_parse<...> function looks up the definition in the context,
if it can:

https://github.com/boostorg/spirit/blob/master/include/boost/spirit/home/x3/nonterminal/rule.hpp#L37

Of course, I maybe missing something.  It's happened before :(

-regards,
Larry






------------------------------------------------------------------------------