1 _____________________________________________________________________________
3 - check ability to use epsilon as a conjunct
4 - use 'a'-'z' or 'a-z' instead of [a-z]?
7 - better toString() methods all around...
9 foo.add(y.andnot(x)) ==> this is broken
10 - distinguish Conjunct from Sequence?
11 => !(Conjunct instanceof Reducible)
12 - document the assumption that Sequences that match epsilon
13 must have tag, and that ONLY that tag is returned
14 when the sequence matches epsilon
15 - try to avoid building the parts of the tree that end up getting
17 - double-check all the region logic
19 ..................................................
21 - paper/techreport opportunities
22 - interaction between RNGLR and follow restrictions
23 - "doomed node" optimization
25 - automatically collect time statistics and display
26 - serializable parse tables?
27 - better ambiguity reporting
28 - colorized tree-diffs?
31 ______________________________________________________________________________
34 - Treewalker code compiler?
39 - precedes restrictions ("<-")
40 - MUST HAVE BETTER ERROR MESSAGES
41 - use for developing java15.g
43 - once this is ready, do big announcement
44 - broader regression testing (for stuff like error messages, etc)
45 - More topology untangling [later]
46 - grammar highlighting?
47 - Forest needs a "manual access" API
48 - the unwrap bit in Forest makes it really hard
49 to expose an API for forests
50 - rewriting language? multiple passes?
52 ______________________________________________________________________________
55 - finalize metagrammar and rdp-op's
60 - RFC2822 (email message/headers)
61 - clean up the whole Walk situation (?)
64 ______________________________________________________________________________
67 - serialization of parse tables
69 - "ambiguity modulo dropped fragments"?
70 - can this be checked statically?
71 - eliminated statically?
73 - substring parsing for better error messages
76 - "Regular Right Part" grammars (NP Chapman, etc)
77 - Attribute unification
79 - inference of rejections for literals
80 - "prefer whitespace higher up" (?)
82 - Labeled edges on trees (associate a label with each slot in the
83 child array in Forest.Body? might make equality tough) --
84 equivalent to Feature Structures. Colon-labeling.
86 ______________________________________________________________________________
89 - understand and implement the RNGLR "kernel state" optimization.
90 The _Practical Early Parsing_ paper may help.
92 - Partly-Linear-PATR? (O(n^6) unification grammar)
94 - Implement a k-token peek buffer (for each state, see if it "dead
95 ends" during the next k Phases based solely on state -- ignoring
98 - Arrange for the SPPF corresponding to dropped subtrees to never be
99 generated (or merged, etc)
101 - Is there any way we can avoid creating a GSS.Node instance for
102 nodes which are transient in the sense that they have only one
105 - Re-read Rekers, particularly the stuff on optimal sharing
107 - Isolate the Element objects from Parse.Table/GSS so we can move
110 - consider allowing a Forest.Body to represent some other Tree whose
111 Body's should be [recursively] considered part of this Forest.
113 - perhaps not: right now we have a nice situation where
114 Forest.Ref instances become immutable once iterator()ed. This
115 also gives us a strong place to to culling with the certainty
116 that we won't throw out a Body which would later be salvaged
117 by some yet-to-be-added dependency.
119 - Figure out if there is a way to:
121 - allow unwrapping of children other than the very last one.
123 - fold repetitions into an array form in Forest, before
124 conversion to Tree. The major problem here is that multiple
125 tree-arrays are possible, all of different lengths. Worse,
126 even if they're all the same length, not all elements belong
127 in the same "possibility vector" as all others. You
128 essentially need a GSS to represent the array, which perhaps
129 is what the unfolded form was in the first place.
131 - Wikipedia grammar (needs to be both lexerless and boolean)
134 => Ordered Choice (";" operator)
136 - bring back in parse-table phase resolution of precedence (just
137 like associativity). This can be inferred from the use of ">"
138 when the rules are in one of these special forms:
149 where "_" is anything and "E" is the defining nonterminal.
150 Essentially what we're looking for is the situation where the
151 leftmost portion of one rule produces another rule, and the
152 rightmost portion of the latter produces the former.
154 I'm not 100% certain that this is as "strong" as the prefer/avoid
155 form (try to prove this, you probably can), but it's "what people
156 intend" most of the time.
158 - implement Johnstone's algorithm for "reduced, resolved LR
159 tables" to eliminate superfluous reductions on
162 ______________________________________________________________________________
165 - Rekers & Koorn note that GLR Substring Parsing can be used to do
166 really elegant and generalized "autocompletion".
169 ______________________________________________________________________________
172 - Incremental parse table construction
173 - "lazy GLR" and "lazy trees" -> language with first-class CF matching
174 - perhaps linear boolean grammars instead? (linear time, quad space)
175 - Forest parsing => chained parsers
176 - unification parsing, attributes, etc
178 - Take another stab at maximal-match? Nonterminal not-followed-by is
180 - Error recovery based on substring parsing