1 _____________________________________________________________________________
4 - keywordification (ie globally reject from all productions?
5 - generalized follow-by?)
7 - clean up util package
9 - currently we GC the doomed stack when the parent dies... but
10 [ideally] we should also GC the parent when the doomed stack dies
11 if it was a positive conjunct.
13 - single-tree-return assumption
14 - have a way to let question marks not be tagged?
15 - "flat" sequences (no subtrees contain "::"s?) -- stringifiable
16 - make it so that we can have multi-result nonterminals
17 so long as they always appear under double-colons?
18 auto-insert the unwrap?
20 - get rid of Sequence.Singleton if possible
22 - use 'a'-'z' or 'a-z' instead of [a-z]?
25 foo.add(y.andnot(x)) ==> this is broken
26 - distinguish Conjunct from Sequence?
27 => !(Conjunct instanceof Reducible)
28 - avoid building the parts of the tree that end up getting dropped
29 - is it worth adding an additional class of states for these?
30 - or perhaps just a runtime node marker (hasNonDroppedParent)
31 - "ambiguity modulo dropped fragments"?
32 - this may conceal highly inefficient grammars...
33 - double-check all the region logic
34 - automatically collect time statistics and display
36 ______________________________________________________________________________
39 - use regression/least-squares/trend-prof to look for reductions whose
40 behavior is O(n^2)? (ie performed a number of times proportional to
41 the input consumed so far).
46 - don't form result forests for negated productions
47 - early kills: (a &~ ... b ...) -> once "... b" is seen, "a" is dead
49 - MUST HAVE BETTER ERROR MESSAGES
50 - use for developing java15.g
51 - better ambiguity reporting
52 - colorized tree-diffs?
54 - better toString() methods all around...
56 - Treewalker code compiler?
57 - detect and reject circular gramars
59 - precedes restrictions ("<-")
60 - More topology untangling [later]
61 - grammar highlighting?
62 - Forest needs a "manual access" API (lifting makes this hard)
63 - rewriting language? multiple passes?
65 - Java grammar (java15.g)
67 - RFC2822 (email message/headers)
68 - Wikipedia grammar (needs to be both lexerless and boolean)
71 ______________________________________________________________________________
74 - try harder to "fuse" states together along two dimensions:
75 - identical (equivalent) states, or states that subsume each other
76 - unnecessary intermediate states ("short cut" GLR)
79 - better error messages
80 - Rekers & Koorn note that this can do really elegant and generalized "autocompletion".
82 - "Regular Right Part" grammars (NP Chapman, etc)
83 - Attribute unification
84 - Partly-Linear-PATR? (O(n^6) unification grammar)
86 - optional "prefer whitespace higher up" disambiguation heuristic
88 - Incremental parse table construction
90 - "lazy GLR" and "lazy trees" -> language with first-class CF matching
91 - perhaps linear boolean grammars instead? (linear time, quad space)
93 - Followed-by and not-followed-by predicates of arbitrary length
94 - expands the grammar beyond Boolean LR...
95 - requires *very* smart garbage collection
97 ______________________________________________________________________________
100 - understand and implement the RNGLR "kernel state" optimization.
101 The _Practical Early Parsing_ paper may help.
103 - implement Johnstone's algorithm for "reduced, resolved LR
104 tables" to eliminate superfluous reductions on
107 - Implement a k-token peek buffer (for each state, see if it "dead
108 ends" during the next k Phases based solely on state -- ignoring
111 - Is there any way we can avoid creating a GSS.Node instance for
112 nodes which are transient in the sense that they have only one
115 - Re-read Rekers, particularly the stuff on optimal sharing
117 - bring back in parse-table phase resolution of precedence (just
118 like associativity). This can be inferred from the use of ">"
119 when the rules are in one of these special forms:
130 where "_" is anything and "E" is the defining nonterminal.
131 Essentially what we're looking for is the situation where the
132 leftmost portion of one rule produces another rule, and the
133 rightmost portion of the latter produces the former.
135 I'm not 100% certain that this is as "strong" as the prefer/avoid
136 form (try to prove this, you probably can), but it's "what people
137 intend" most of the time.