1 _____________________________________________________________________________
3 - use 'a'-'z' or 'a-z' instead of [a-z]?
6 - better toString() methods all around...
8 foo.add(y.andnot(x)) ==> this is broken
9 - distinguish Conjunct from Sequence?
10 => !(Conjunct instanceof Reducible)
11 - document the assumption that Sequences that match epsilon
12 must have tag, and that ONLY that tag is returned
13 when the sequence matches epsilon
14 - try to avoid building the parts of the tree that end up getting
16 - double-check all the region logic
18 ..................................................
20 - paper/techreport opportunities
21 - interaction between RNGLR and follow restrictions
22 - "doomed node" optimization
24 - automatically collect time statistics and display
25 - serializable parse tables?
26 - better ambiguity reporting
27 - colorized tree-diffs?
30 ______________________________________________________________________________
33 - Treewalker code compiler?
38 - precedes restrictions ("<-")
39 - MUST HAVE BETTER ERROR MESSAGES
40 - use for developing java15.g
42 - once this is ready, do big announcement
43 - broader regression testing (for stuff like error messages, etc)
44 - More topology untangling [later]
45 - grammar highlighting?
46 - Forest needs a "manual access" API
47 - the unwrap bit in Forest makes it really hard
48 to expose an API for forests
49 - rewriting language? multiple passes?
51 ______________________________________________________________________________
54 - finalize metagrammar and rdp-op's
59 - RFC2822 (email message/headers)
60 - clean up the whole Walk situation (?)
63 ______________________________________________________________________________
66 - serialization of parse tables
68 - "ambiguity modulo dropped fragments"?
69 - can this be checked statically?
70 - eliminated statically?
72 - substring parsing for better error messages
75 - "Regular Right Part" grammars (NP Chapman, etc)
76 - Attribute unification
78 - inference of rejections for literals
79 - "prefer whitespace higher up" (?)
81 - Labeled edges on trees (associate a label with each slot in the
82 child array in Forest.Body? might make equality tough) --
83 equivalent to Feature Structures. Colon-labeling.
85 ______________________________________________________________________________
88 - understand and implement the RNGLR "kernel state" optimization.
89 The _Practical Early Parsing_ paper may help.
91 - Partly-Linear-PATR? (O(n^6) unification grammar)
93 - Implement a k-token peek buffer (for each state, see if it "dead
94 ends" during the next k Phases based solely on state -- ignoring
97 - Arrange for the SPPF corresponding to dropped subtrees to never be
98 generated (or merged, etc)
100 - Is there any way we can avoid creating a GSS.Node instance for
101 nodes which are transient in the sense that they have only one
104 - Re-read Rekers, particularly the stuff on optimal sharing
106 - Isolate the Element objects from Parse.Table/GSS so we can move
109 - consider allowing a Forest.Body to represent some other Tree whose
110 Body's should be [recursively] considered part of this Forest.
112 - perhaps not: right now we have a nice situation where
113 Forest.Ref instances become immutable once iterator()ed. This
114 also gives us a strong place to to culling with the certainty
115 that we won't throw out a Body which would later be salvaged
116 by some yet-to-be-added dependency.
118 - Figure out if there is a way to:
120 - allow unwrapping of children other than the very last one.
122 - fold repetitions into an array form in Forest, before
123 conversion to Tree. The major problem here is that multiple
124 tree-arrays are possible, all of different lengths. Worse,
125 even if they're all the same length, not all elements belong
126 in the same "possibility vector" as all others. You
127 essentially need a GSS to represent the array, which perhaps
128 is what the unfolded form was in the first place.
130 - Wikipedia grammar (needs to be both lexerless and boolean)
133 => Ordered Choice (";" operator)
135 - bring back in parse-table phase resolution of precedence (just
136 like associativity). This can be inferred from the use of ">"
137 when the rules are in one of these special forms:
148 where "_" is anything and "E" is the defining nonterminal.
149 Essentially what we're looking for is the situation where the
150 leftmost portion of one rule produces another rule, and the
151 rightmost portion of the latter produces the former.
153 I'm not 100% certain that this is as "strong" as the prefer/avoid
154 form (try to prove this, you probably can), but it's "what people
155 intend" most of the time.
157 - implement Johnstone's algorithm for "reduced, resolved LR
158 tables" to eliminate superfluous reductions on
161 ______________________________________________________________________________
164 - Rekers & Koorn note that GLR Substring Parsing can be used to do
165 really elegant and generalized "autocompletion".
168 ______________________________________________________________________________
171 - Incremental parse table construction
172 - "lazy GLR" and "lazy trees" -> language with first-class CF matching
173 - perhaps linear boolean grammars instead? (linear time, quad space)
174 - Forest parsing => chained parsers
175 - unification parsing, attributes, etc
177 - Take another stab at maximal-match? Nonterminal not-followed-by is
179 - Error recovery based on substring parsing