File: NOTES

package info (click to toggle)
yapps2 2.1.1-17.1
  • links: PTS
  • area: main
  • in suites: etch, etch-m68k, lenny
  • size: 328 kB
  • ctags: 184
  • sloc: python: 1,008; makefile: 63; sh: 14
file content (78 lines) | stat: -rw-r--r-- 3,019 bytes parent folder | download | duplicates (9)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
[Last updated August 11, 2003]

Notes for myself:

Document the LINENO trick

Add a way to have a self-contained mode that doesn't require yappsrt?

Add a debugging mode that helps you understand how the grammar
  is constructed and how things are being parsed

Optimize (remove) unused variables

Yapps produces a bunch of inline list literals.  We should be able to
  instead create these lists as class variables (but this makes it
  harder to read the code).  Also, 'A in X' could be written
  'X.has_key(A)' if we can convert the lists into dictionaries ahead
  of time.

Add a convenience to automatically gather up the values returned 
  from subpatterns, put them into a list, and return them

"Gather" mode that simply outputs the return values for certain nodes.
  For example, if you just want all expressions, you could ask yapps
  to gather the results of the 'expr' rule into a list.  This would 
  ignore all the higher level structure.

Improve the documentation

Write some larger examples (probably XML/HTML)

EOF needs to be dealt with.  It's probably a token that can match anywhere.

Get rid of old-style regex support

Use SRE's lex support to speed up lexing (this may be hard given that
  yapps allows for context-sensitive lexers)

Look over Dan Connoly's experience with Yapps (bugs, frustrations, etc.)
  and see what improvements could be made

Add something to pretty-print the grammar (without the actions)

Maybe conditionals?  Follow this rule only if <condition> holds.
  But this would be useful mainly when multiple rules match, and we
  want the first matching rule.  The conditional would mean we skip to
  the next rule.  Maybe this is part of the attribute grammar system,
  where rule X<0> can be specified separately from X<N>.

Convenience functions that could build return values for all rules
  without specifying the code for each rule individually

Patterns (abstractions over rules) -- for example, comma separated values
  have a certain rule pattern that gets replicated all over the place

  These are rules that take other rules as parameters.

  rule list<separator,element>:            {{ result = [] }}
                [ element                  {{ result.append(element) }}
                  ( separator element      {{ result.append(element) }}
                  )*
                ]                          {{ return result }}
      
Inheritance of parser and scanner classes.  The base class (Parser)
  may define common tokens like ID, STR, NUM, space, comments, EOF,
  etc., and common rules/patterns like optional, sequence,
  delimiter-separated sequence.

Why do A? and (A | ) produce different code?  It seems that they
should produce the very same code.

Look at everyone's Yapps grammars, and come up with larger examples
  http://www.w3.org/2000/10/swap/SemEnglish.g
  http://www.w3.org/2000/10/swap/kifExpr.g
  http://www.w3.org/2000/10/swap/rdfn3.g

Construct lots of erroneous grammars and see what Yapps does with them
  (improve error reporting)