Re: IPv6 address parsing

On Sat, Dec 16, 2017 at 1:08 PM, Michael Powell <[email protected]> wrote:
> On Fri, Dec 15, 2017 at 1:59 PM, Michael Powell <[email protected]> wrote:
>> On Fri, Dec 15, 2017 at 12:22 PM, Michael Powell <[email protected]> wrote:
>>> On Fri, Dec 15, 2017 at 11:31 AM, Michael Powell <[email protected]> wrote:
>>>> On Thu, Dec 7, 2017 at 1:21 PM, Michael Powell <[email protected]> wrote:
>>>>> Again, I am moving beyond IPv4 into IPv6. Do I need to re-post that
>>>>> grammar attempt? Maybe I do at this point.
>>>>>
>>>>> _dot = ".";
>>>>> _col = char_(':');
>>>>> _col_x2 = _col >> _col;
>>>>>
>>>>> // _dec_octet = uint_parser<std::uint8_t, 10, 1, 3>();
>>>>>
>>>>> qi::rule<Iterator> _dec_octet = ... ; // Approach TBD; probably
>>>>> running with the IPv4 proposal discussed in earlier thread(s).
>>>>>
>>>>> qi::rule<Iterator> _ipv4_addr = _dec_octet >> repeat(3)[_dot >> _dec_octet];
>>>>> // or, qi::rule<Iterator> _ipv4_addr = raw[_dec_octet >> _dot >>
>>>>> _dec_octet >> _dot >> _dec_octet >> _dot >> _dec_octet];
>>>>>
>>>>> qi::rule<Iterator> _h16 = raw[repeat(1, 4)[hex]];
>>>>> // Or (?): _h16 = hex >> -(hex >> -(hex >> -hex));
>>>>> qi::rule<Iterator> _h16_col = _h16 >> _col;
>>>>> qi::rule<Iterator> _ls32 = _h16_col >> _h16 | _ipv4_addr;
>>>>>
>>>>> qi::rule<Iterator, std::string()> _ipv6_addr
>>>>>     = repeat(6)[_h16_col] >> _ls32
>>>>>     | _col_x2 >> repeat(5)[_h16_col] >> _ls32
>>>>>     | -(_h16) >> _col_x2 >> repeat(4)[_h16_col] >> _ls32
>>>>>     | -(_h16_col >> _h16) >> _col_x2 >> repeat(3)[_h16_col] >> _ls32
>>>>>     | -(repeat(1, 2)[_h16_col] >> _h16) >> _col_x2 >>
>>>>> repeat(2)[_h16_col] >> _ls32
>>>>>     | -(repeat(1, 3)[_h16_col] >> _h16) >> _col_x2 >> _h16_col >> _ls32
>>>>>     | -(repeat(1, 4)[_h16_col] >> _h16) >> _col_x2 >> _ls32
>>>>>     | -(repeat(1, 5)[_h16_col] >> _h16) >> _col_x2 >> _h16
>>>>>     | -(repeat(1, 6)[_h16_col] >> _h16) >> _col_x2
>>>>>     ;
>>>>
>>>> I am testing the grammar now, and it seems like to me that the
>>>> optional and/or repeat are not working correctly.
>>>>
>>>> For test case: prefix = pf_max_4|pf_count_1, suffix = pf_none,
>>>> appendix = bo_hex_2, which yields address: "6a80:9024::c7e3:4150".
>>>>
>>>> Should be matching the alternative, from the RFC grammar:
>>>>   / [*4(h16 ":") h16] "::"            ls32
>>>>
>>>> Or from the Qi grammar:
>>>>      | -(repeat(1, 4)[_h16_col] >> _h16) >> _col_x2 >> _ls32
>>>>
>>>> But gets rejected. Did not parse, or rejected the entire string.
>>>>
>>>> Perhaps repeat does not work quite as advertised?
>>>>
>>>> Thoughts? Suggestions?
>>>
>>>
>>> qi::rule<Iterator, std::string()> _ipv6_addr
>>>     = repeat(6)[_h16_col] >> _ls32
>>>     | _col_x2 >> repeat(5)[_h16_col] >> _ls32
>>>     | -(-_h16) >> _col_x2 >> repeat(4)[_h16_col] >> _ls32
>>>     | -(-_h16_col >> _h16) >> _col_x2 >> repeat(3)[_h16_col] >> _ls32
>>>     | -(-repeat(1, 2)[_h16_col] >> _h16) >> _col_x2 >>
>>> repeat(2)[_h16_col] >> _ls32
>>>     | -(-repeat(1, 3)[_h16_col] >> _h16) >> _col_x2 >> _h16_col >> _ls32
>>>     | -(-repeat(1, 4)[_h16_col] >> _h16) >> _col_x2 >> _ls32
>>>     | -(-repeat(1, 5)[_h16_col] >> _h16) >> _col_x2 >> _h16
>>>     | -(-repeat(1, 6)[_h16_col] >> _h16) >> _col_x2
>>>     ;
>>>
>>> I did identify that possibly the prefix phrase needs to be built from
>>> optional elements within itself.
>>>
>>> However, still not parsing.
>>>
>>> I am concerned that either Qi optional, repeat, etc, perhaps even
>>> alternatives, are not working quite right in this instance.
>>>
>>> Do I need to separate the alternatives here into "micro" grammars
>>> within themselves for this to possibly work?
>>
>> Thus far, separating the more complex grammar into micro-grammars
>> seems to be the correct answer, the one that will work.
>
> I removed the repeats, and I think I've isolated an issue that nested
> optional operators do not work. i.e. -(-(-(...) ...) ...) >> ...
>
> Which does not match the phrase I expect it to.

Moving on, a little...

I now have a test case "8d5d:5be71c85:7c0e:d86f:540b:d0ec:f177:2316"
which should not be valid. I expect it to be invalid and rejected.
However, the grammar

_col = char_(':');

_h16 = raw[
  hex >> hex >> hex >> hex
    | hex >> hex >> hex
    | hex >> hex
    | hex
  ];
//// I was using repeat(...), but I have concerns whether REPEAT
(and/or OPTIONAL) are working properly, not least of all in this
context, never mind concerns over nested repeat, optional, etc.
//_h16 = repeat(1, 4)[hex];

_h16_col = _h16 >> _col;

_ls32 = _h16_col >> _h16;

_ipv6_addr = lexeme[
    _h16_col >> _h16_col >> _h16_col >> _h16_col >> _h16_col >>
_h16_col >> _ls32
];

_start = _ipv6_addr >> eoi;

Which leads me to believe that boost::spirit::hex
(boost::spirit::hex_type) is excessively greedy. Would it be more
appropriate if I simply said this: char_("0123456789abcdef"),
notwithstanding case?

I am also chewing on the test case parameters, whether I should drive
the test data to yield "8d5d:5be7::1c85:7c0e:d86f:540b:d0ec:f177:2316"
instead, but this is beyond the scope of my grammar concerns.

Any thoughts? Insights?

Thanks!

>> Not sure why the broader parent grammar doesn't work, however. Apart
>> from "greediness" suspicions, that is; the grammar is what it is, no?
>>
>>>> Thank you!
>>>>
>>>>> qi::rule<Iterator> _start = _ipv6_addr >> eoi;
>>>>>
>>>>> Q: Notwithstanding concern over raw IPv4, does the overall Qi-based
>>>>> IPv6 grammar look reasonable? If not, why not?
>>>>>
>>>>> I'd like to land in a Fusion data structure something like (string >>
>>>>> uint16_t >> eoi), but if I can land here, that's acceptable, too:
>>>>> (string >> -(uint8_t >> uint8_t >>
>>>>> uint8_t >> uint8_t) >> uint16_t >> eoi).
>>>>>
>>>>> From what I can determine, then, raw[...] is the difference between
>>>>> parsed types like uint8_t and string?
>>>>>
>>>>> ref: http://www.ietf.org/rfc/rfc5954.txt, section 4.1 Resolution for
>>>>> Extra Colon in IPv4-Mapped IPv6 Address
>>>>>
>>>>> Constructive feedback appreciated.
>>>>>
>>>>> Untested ATM, to be fair.
>>>>>
>>>>> On Thu, Dec 7, 2017 at 1:09 PM, Seth <[email protected]> wrote:
>>>>>> On 07-12-17 18:21, Michael Powell wrote:
>>>>>>> Well, you remembered incorrectly.
>>>>>>
>>>>>> I was asking.
>>>>>>
>>>>>>> If there's a doable intermediate step of parsed individual octets, so
>>>>>>> be it; no problem.
>>>>>>
>>>>>> Please per-use all the parts of my answer.
>>>>>>
>>>>>> On 07-12-17 17:13, Seth wrote:
>>>>>>> Maybe you'll be more convinced when it's demonstrated (I think I did
>>>>>>> this before):
>>>>>>>
>>>>>>> https://wandbox.org/permlink/M12U5PEYcB8vnxIs
>>>>>>>
>>>>>>> It's really just ~3 lines of code changed:
>>>>>>>
>>>>>>>     using ipv4_address = std::string; // much less useful representation
>>>>>>>     // ....
>>>>>>>     addr_  = raw[ octet_ >> '.' >> octet_ >> '.' >> octet_ >> '.' >>
>>>>>>> octet_ ];
>>>>>>>
>>>>>>> Note now that test fails:
>>>>>>
>>>>>> It's all there.
>>>>>>

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot