ICONS identities and copying HOOK in parts without identifying it as a whole

Related to this, so the answer may be “don’t do it” :).

I created this type:

; OZ 2020-05-05 If there are embedded questions in the grammar,
; the complementizer needs to have its own opinion on the SF value.
raise-sem-no-sf-lex-item := no-hcons-lex-item &
  [ SYNSEM.LOCAL [ CONT [ RELS 0-alist,
        HOOK [ GTOP #gtop, LTOP #ltop, INDEX [ E #tam, SORT #sort, INSTLOC #instloc ],
                XARG #xarg, ICONS-KEY #icons-key, CLAUSE-KEY #clause-key ] ],
       CAT.VAL.COMPS < [ LOCAL.CONT.HOOK [ GTOP #gtop, LTOP #ltop, INDEX [ E #tam, SORT #sort, INSTLOC #instloc ],
                XARG #xarg, ICONS-KEY #icons-key, CLAUSE-KEY #clause-key ] ], ... > ] ].

And started using it instead of this one for some (not all) complementizers:

; SSH 2012-03-03

raise-sem-lex-item := no-hcons-lex-item &
  [ SYNSEM.LOCAL [ CONT [ RELS 0-alist,
			  HOOK #hook ],
		   CAT.VAL.COMPS < [ LOCAL.CONT.HOOK #hook ] , ... > ]].

Now I have the following diff in the regression tests (gold on top):

Screen Shot 2020-05-07 at 11.43.50 AM

The ICONS is no longer properly linked.

Assuming this is sensible to do at all (not identifying the entire HOOK sometimes), how do I fix this?

Don’t do it. This lex item creates a new event.

2 Likes