1 _____________________________________________________________________________
4 - comparison test is probably chewing up most of the time
6 - Check if the only remaining stack is lame (hopeful/nothopeful)
7 - write a testcase for this
12 foo.add(y.andnot(x)) ==> this is broken
15 ..................................................
17 - serializable parse tables?
18 - Treewalker code compiler?
20 ______________________________________________________________________________
23 - precedes restrictions ("<-")
24 - MUST HAVE BETTER ERROR MESSAGES
25 - use for developing java15.g
27 - once this is ready, do big announcement
28 - broader regression testing (for stuff like error messages, etc)
29 - More topology untangling [later]
30 - grammar highlighting?
31 - Forest needs a "manual access" API
32 - the unwrap bit in Forest makes it really hard to expose an API for forests
36 ______________________________________________________________________________
39 - finalize metagrammar and rdp-op's
44 - RFC2822 (email message/headers)
45 - clean up the whole Walk situation (?)
48 ______________________________________________________________________________
51 - serialization of parse tables
53 - "ambiguity modulo dropped fragments"?
54 - can this be checked statically?
55 - eliminated statically?
57 - substring parsing for better error messages
60 - "Regular Right Part" grammars (NP Chapman, etc)
61 - Attribute unification
63 - inference of rejections for literals
64 - "prefer whitespace higher up" (?)
66 - Labeled edges on trees (associate a label with each slot in the
67 child array in Forest.Body? might make equality tough) --
68 equivalent to Feature Structures. Colon-labeling.
70 ______________________________________________________________________________
73 - understand and implement the RNGLR "kernel state" optimization.
74 The _Practical Early Parsing_ paper may help.
76 - Partly-Linear-PATR? (O(n^6) unification grammar)
78 - Implement a k-token peek buffer (for each state, see if it "dead
79 ends" during the next k Phases based solely on state -- ignoring
82 - Arrange for the SPPF corresponding to dropped subtrees to never be
83 generated (or merged, etc)
85 - Is there any way we can avoid creating a GSS.Node instance for
86 nodes which are transient in the sense that they have only one
89 - Re-read Rekers, particularly the stuff on optimal sharing
91 - Isolate the Element objects from Parse.Table/GSS so we can move
94 - consider allowing a Forest.Body to represent some other Tree whose
95 Body's should be [recursively] considered part of this Forest.
97 - perhaps not: right now we have a nice situation where
98 Forest.Ref instances become immutable once iterator()ed. This
99 also gives us a strong place to to culling with the certainty
100 that we won't throw out a Body which would later be salvaged
101 by some yet-to-be-added dependency.
103 - Figure out if there is a way to:
105 - allow unwrapping of children other than the very last one.
107 - fold repetitions into an array form in Forest, before
108 conversion to Tree. The major problem here is that multiple
109 tree-arrays are possible, all of different lengths. Worse,
110 even if they're all the same length, not all elements belong
111 in the same "possibility vector" as all others. You
112 essentially need a GSS to represent the array, which perhaps
113 is what the unfolded form was in the first place.
115 - Wikipedia grammar (needs to be both lexerless and boolean)
118 => Ordered Choice (";" operator)
120 - bring back in parse-table phase resolution of precedence (just
121 like associativity). This can be inferred from the use of ">"
122 when the rules are in one of these special forms:
133 where "_" is anything and "E" is the defining nonterminal.
134 Essentially what we're looking for is the situation where the
135 leftmost portion of one rule produces another rule, and the
136 rightmost portion of the latter produces the former.
138 I'm not 100% certain that this is as "strong" as the prefer/avoid
139 form (try to prove this, you probably can), but it's "what people
140 intend" most of the time.
142 - implement Johnstone's algorithm for "reduced, resolved LR
143 tables" to eliminate superfluous reductions on
146 ______________________________________________________________________________
149 - Rekers & Koorn note that GLR Substring Parsing can be used to do
150 really elegant and generalized "autocompletion".
153 ______________________________________________________________________________
156 - Incremental parse table construction
157 - "lazy GLR" and "lazy trees" -> language with first-class CF matching
158 - perhaps linear boolean grammars instead? (linear time, quad space)
159 - Forest parsing => chained parsers
160 - unification parsing, attributes, etc
162 - Take another stab at maximal-match? Nonterminal not-followed-by is
164 - Error recovery based on substring parsing