[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [lojban] bi'i syntax
On Fri, Mar 02, 2001 at 07:42:23PM -0500, Pierre Abbat wrote:
> I tried these utterances on jbofi'e:
>
> mi klama le zdani ku bi'i le ckule
> (parses)
> mi klama le zdani bi'i le ckule
> --------------------
> SYNTAX ERROR IN TEXT
> --------------------
> Misparsed token :
> le [LE] (line 1, col 24)
> Latest successfully parsed tokens :
> bi'i [BIhI] (line 1, col 19)
> MARKER : PRIVATE_START_JOIK
> zdani [BRIVLA] (line 1, col 13)
> le [LE] (line 1, col 10)
> klama [BRIVLA] (line 1, col 4)
> mi [KOhA3] (line 1, col 1)
> --------------------
> What's wrong with the second one? I just dropped "ku" which is elidable...
>
The problem is that joiks are used as connectives for both selbri and
for sumti. Contrast the logical connectives, where selbri use jeks and
sumti use eks.
The problem comes down to how the Lojban grammar is specified in terms
of LALR1 parsing behaviour. In particular, whether or not an elidable
terminator can be elided in a particular case comes down to whether its
elision would change how the text is recognised by a yacc/bison style
LALR parsing automaton.
Because bi'i is a valid selbri connective, when it appears after le
zdani the parser will think that it's a selbri connective and shift it.
Now it finds 'le', which cannot be valid in this context => parse error.
By this stage it is too late to back up and reduce 'le zdani' as a sumti
so that bi'i can be applied as a sumti connective - yacc/bison generated
parsers simply don't work that way.
The jbofihe -sev option merely confirms the diagnosis :
echo "mi klama le zdani ku bi'i le ckule" | test_jbofihe -sev
Early shift hazards for words 'ku' (at line 1 col 19)
<- ku
X| <- <NEXT WORD>
The 'X' means that the next word (in this case bi'i) can be shifted in
some context which occurs prior to the 'ku' being shifted, i.e. removing
'ku' will change the meaning (or even, as in this case, the validity) of
the text.
--
----------------------------------------------------------------------
Richard P. Curnow rpc@myself.com
Weston-super-Mare
United Kingdom http://go.to/richard.curnow/