1 _____________________________________________________________________________
4 - If a top-level rule has labels but no head-tag, like this
6 then infer the name of the rule it belongs to
8 create( $c:{...}, class ) =
9 return create($c:{...})
11 create( h:{...}, class ) =
15 create( _:{...}, String) = treat as char[]
16 create( _:{...}, c[] ) = { create(.,c), create(.,c), ... }
21 - clean up the visualization (?)
23 - I still don't like Atom.Infer and Atom.Invert...
25 - better ambiguity debugging tools
27 - ParseFailed, GSS, Walk, Parser, Sequence, Forest
29 - Fix the metagrammar (really?)
30 - evil problems with (x y? z /ws)
35 ______________________________________________________________________________
38 - finalize metagrammar and rdp-op's
43 - RFC2822 (email message/headers)
44 - clean up the whole Walk situation (?)
46 - what if Tree<> could unwrap itself?
49 ______________________________________________________________________________
52 - serialization of parse tables
54 - "ambiguity modulo dropped fragments"?
55 - can this be checked statically?
56 - eliminated statically?
58 - substring parsing for better error messages
60 - right now I can only lift the last child in a forest... begs
61 the question of what the right representation for Forests is
62 if we need to be able to do lift operations on it.
65 - "Regular Right Part" grammars (NP Chapman, etc)
66 - Attribute unification
68 - inference of rejections for literals
69 - "prefer whitespace higher up" (?)
71 - Labeled edges on trees (associate a label with each slot in the
72 child array in Forest.Body? might make equality tough) --
73 equivalent to Feature Structures. Colon-labeling.
75 ______________________________________________________________________________
78 - Partly-Linear-PATR? (O(n^6) unification grammar)
80 - Implement a k-token peek buffer (for each state, see if it "dead
81 ends" during the next k Phases based solely on state -- ignoring
84 - Arrange for the SPPF corresponding to dropped subtrees to never be
85 generated (or merged, etc)
87 - Is there any way we can avoid creating a GSS.Node instance for
88 nodes which are transient in the sense that they have only one
91 - Re-read Rekers, particularly the stuff on optimal sharing
93 - Isolate the Element objects from Parse.Table/GSS so we can move
96 - consider allowing a Forest.Body to represent some other Tree whose
97 Body's should be [recursively] considered part of this Forest.
99 - perhaps not: right now we have a nice situation where
100 Forest.Ref instances become immutable once iterator()ed. This
101 also gives us a strong place to to culling with the certainty
102 that we won't throw out a Body which would later be salvaged
103 by some yet-to-be-added dependency.
105 - Figure out if there is a way to:
107 - allow unwrapping of children other than the very last one.
109 - fold repetitions into an array form in Forest, before
110 conversion to Tree. The major problem here is that multiple
111 tree-arrays are possible, all of different lengths. Worse,
112 even if they're all the same length, not all elements belong
113 in the same "possibility vector" as all others. You
114 essentially need a GSS to represent the array, which perhaps
115 is what the unfolded form was in the first place.
117 - Wikipedia grammar (needs to be both lexerless and boolean)
120 => Ordered Choice (";" operator)
122 - bring back in parse-table phase resolution of precedence (just
123 like associativity). This can be inferred from the use of ">"
124 when the rules are in one of these special forms:
135 where "_" is anything and "E" is the defining nonterminal.
136 Essentially what we're looking for is the situation where the
137 leftmost portion of one rule produces another rule, and the
138 rightmost portion of the latter produces the former.
140 I'm not 100% certain that this is as "strong" as the prefer/avoid
141 form (try to prove this, you probably can), but it's "what people
142 intend" most of the time.
144 - implement Johnstone's algorithm for "reduced, resolved LR
145 tables" to eliminate superfluous reductions on
148 ______________________________________________________________________________
151 - Rekers & Koorn note that GLR Substring Parsing can be used to do
152 really elegant and generalized "autocompletion".
155 ______________________________________________________________________________
158 - Incremental parse table construction
159 - "lazy GLR" and "lazy trees" -> language with first-class CF matching
160 - perhaps linear boolean grammars instead? (linear time, quad space)
161 - Forest parsing => chained parsers
162 - unification parsing, attributes, etc
164 - Take another stab at maximal-match? Nonterminal not-followed-by is
166 - Error recovery based on substring parsing