1 _____________________________________________________________________________
4 - simplify metagrammar => go to top-down rewriting => finalize metagrammar and rdp-op's
6 - What is our use model?
7 - Parse, attribute, unify (ag)
8 - Parse, transform tree, walk (rdp)
10 - cascading tree rewrites
12 ==> use the middle formalism "for now" and in meta.g; layer others
15 - Lay down the law on the different kinds of Sequence productions
21 - whitespace-in-braces?
22 - Deal with the problem of zero-rep productions and whitespace insertion
24 - switch maximal to not-followed-by (~/~)
26 - should Union.add() be there?
27 - should Atom.top() be there?
29 - fix the location stuff, it's broken
30 - decent/better error messages
36 - RFC2822 (email message/headers)
39 ______________________________________________________________________________
42 - clean up the whole Walk situation
44 - cleaner solution to "maximal"?
47 - right now I can only lift the last child in a forest... begs
48 the question of what the right representation for Forests is
49 if we need to be able to do lift operations on it.
52 - "Regular Right Part" grammars (NP Chapman, etc)
53 - Attribute unification
55 - serialization of parse tables
56 - inference of rejections for literals
57 - "prefer whitespace higher up" (?)
58 - "ambiguity modulo dropped fragments"?
59 - can this be checked statically?
60 - eliminated statically?
62 ______________________________________________________________________________
65 - Implement a k-token peek buffer (for each state, see if it "dead
66 ends" during the next k Phases based solely on state -- ignoring
69 - Arrange for the SPPF corresponding to dropped subtrees to never be
70 generated (or merged, etc)
72 - Is there any way we can avoid creating a GSS.Node instance for
73 nodes which are transient in the sense that they have only one
76 - Implement "GLR syntactic predicates" -- the ability to do
77 arbitrary lookahead (ie "followed-by" and "not-followed-by" for
78 arbitrary patterns). This enables generalized longest-match and
79 lets us drop the Maximal hack.
81 - Re-read Rekers, particularly the stuff on optimal sharing
83 - Isolate the Element objects from Parse.Table/GSS so we can move
86 - consider allowing a Forest.Body to represent some other Tree whose
87 Body's should be [recursively] considered part of this Forest.
89 - perhaps not: right now we have a nice situation where
90 Forest.Ref instances become immutable once iterator()ed. This
91 also gives us a strong place to to culling with the certainty
92 that we won't throw out a Body which would later be salvaged
93 by some yet-to-be-added dependency.
95 - Figure out if there is a way to:
97 - allow unwrapping of children other than the very last one.
99 - fold repetitions into an array form in Forest, before
100 conversion to Tree. The major problem here is that multiple
101 tree-arrays are possible, all of different lengths. Worse,
102 even if they're all the same length, not all elements belong
103 in the same "possibility vector" as all others. You
104 essentially need a GSS to represent the array, which perhaps
105 is what the unfolded form was in the first place.
107 - Wikipedia grammar (needs to be both lexerless and boolean)
110 => Ordered Choice (";" operator)
112 - bring back in parse-table phase resolution of precedence (just
113 like associativity). This can be inferred from the use of ">"
114 when the rules are in one of these special forms:
125 where "_" is anything and "E" is the defining nonterminal.
126 Essentially what we're looking for is the situation where the
127 leftmost portion of one rule produces another rule, and the
128 rightmost portion of the latter produces the former.
130 I'm not 100% certain that this is as "strong" as the prefer/avoid
131 form (try to prove this, you probably can), but it's "what people
132 intend" most of the time.
134 - implement Johnstone's algorithm for "reduced, resolved LR
135 tables" to eliminate superfluous reductions on
138 ______________________________________________________________________________
141 - Rekers & Koorn note that GLR Substring Parsing can be used to do
142 really elegant and generalized "autocompletion".