forked from kschiess/parslet
-
Notifications
You must be signed in to change notification settings - Fork 0
/
HISTORY.txt
208 lines (137 loc) · 7.54 KB
/
HISTORY.txt
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
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
= 2.0 / ?? (future release changes, like a reminder to self)
- prsnt? and absnt? are now finally banned into oblivion. Wasting vocals for
the win.
= 1.4.1 / ???
+ Handles unconsumed input at end of parse completely differently. Instead
of generating a toplevel error, it now raises an error in every branch
of the parse. More information in the resulting exception ensues! Thanks
again to John Mettraux for inspiration & acceptance specs.
NOTE that this means that the UnconsumedInput exception is gone, since the
unconsumed input case is nothing special anymore.
* This history now finally reads like the Changelog of the linux kernel.
Meaning that probably no one ever reads this.
= 1.4.0 / 25May2012
+ Revised documentation. A few new API features have finally made it into
the documentation. Examples in the documentation are now curated and
run against the current code so that they really really work.
Also, the website generation tools have been replaced with 2012-style
tools. Much less pain to update now.
+ Parslet::Source now doesn't hold a StringIO, it directly holds the
buffer to be parsed. The api of Source has changed a tiny bit. This change
has been made for speed optimisation reasons.
+ :reporter argument to parse, allowing to customize error reporting within
wide boundaries. See issue #64 for a discussion.
Included are two error reporters, one (default) with the existing error
tree functionality, one reporting deepest errors as defined by the above
ticket.
+ Optimistic parse: Parsing is two phase, with the first phase assuming
there will be no errors. This yields ~ 20% speed improvement in the
case where the parse succeeds.
Also, internal error handling is now using tuples. This and other
optimizations have yielded ~ 30% overall improvement.
! #error_tree and #cause removed from all of parslet. The
Parslet::ParseFailed exception now contains a #cause field that can
be asked for an #ascii_tree as before.
Cleaner internal error handling, not stateful in atoms anymore. Some
parsers will see correct error reporting for the first time. (issue #65)
+ Made it possible to pass a custom Parslet::Source implementor to #parse.
(see #63)
+ #parse has now a second argument that is an options hash. See
Parslet::Atoms::Base#parse for documentation.
- VM engine on the way out. No benefit except for the intellectual
challenge.
= 1.3.0 / 5Mar2012
! Parslet::Transform::Context is now much more well-behaved. It has
#respond_to? and #method_missing; it now looks like a plain old Ruby
object with instance variables and attribute readers.
- Grammar transforms turned out to be a dead end and have been removed.
! A few problems in error message generation have been fixed. This will
improve diagnostics further.
+ A VM driven parser engine: Removes the limitation that parsing needs a
lot of stack space, something dearly missing from Ruby 1.9.3 fibers.
This engine is experimental and might be removed in the future.
! Interaction with mathn fixed - Line number generation will terminate.
. Internal reorganisation, removing cruft and bit rot.
= 1.2.3 / 22Sep2011
+ Transform#apply can now be called with a hash as second argument. This
provides bindings and a way to inject context.
! Fixes a bug thar modified parslet atoms in place, defeating oop chaining.
(#50)
= 1.2.1 / 6Jun2011
! FIX: Input at the end of a parse raises Parslet::UnconsumedInput. (see
issue 18)
! FIX: Unicode parsing should now work as expected. (see issue 38)
! FIX: Slice#slice returned wrong bits at times (see issue 36).
= 1.2.0 / 4Feb2011
+ Parslet::Parser is now also a grammar atom, it can be composed freely with
other atoms. (str('f') >> MiniLispParser.new >> str('b'))
+ No strings, only slices are returned as part of the parser result.
Parslet::Slice is almost a string class, but one that remembers the
source offset. This has also bought us a slight speedup.
+ require 'parslet/convenience' now brings #parse_with_debug to all parslets.
This is a consequence of the above change.
+ Deprecates prsnt? and absnt? in favor of the more readable absent? and
prsnt?. Uses 3 bytes more RAM. The old variants will exist until we release
2.0.
INTERNALLY
+ Visitors now should have methods that all begin with 'visit_*'. #str
becomes #visit_str.
+ Parslet::Atoms::Entity now takes only a block argument instead of context
and block.
= 1.1.1 / 4Feb2011
! FIX: Line counting was broken by performance optimisations.
+ Squeezed out another few drops of performance.
= 1.1.0 / 2Feb2011
+ Uses return (fail/success), cached line counts, memoizing of parse results
and other tricks internally for at least an order of magnitude increase
in execution speed.
+ str('foo').maybe will now return an empty string again. Use .as(...) to
name things and get back [] from #repeat and nil from #maybe.
+ If you require 'parslet/atoms/visitor', you'll get an accept method on
all known Parslet::Atoms.
+ If you require 'parslet/export', you can call #to_citrus and #to_treetop
to produce string versions of your grammar in those dialects.
+ Requiring 'parslet/convenience' will given you a parse_with_debug on
your Parslet::Parser class. This prints some diagnostics on parse failure.
(Thanks to Florian Hanke)
= 1.0.1 / 17Jan2011
A happy new year!
! FIX: Parslet::Transform was wrongly fixed earlier - it now wont mangle
hashes anymore. (Blake Sweeney)
+ parslet/rig/rspec.rb contains useful rspec matchers. (R. Konstantin Haase)
= 1.0.0 / 29Dez2010
- #each_match was removed. There was some duplication of code that even
confused me - and we should not have 2 methods of achieving the same
goal.
+ Full documentation. Fixed sdoc.
= 0.11.0 / 25Nov2010
! Bugfixes to tree handling. Let's hope that was the last such significant
change to the core.
= 0.10.1 / 22Nov2010
+ Allow match['a-z'], shortcut for match('[a-z]')
! Fixed output inconsistencies (behaviour in connection to 'maybe')
= 0.10.0 / 22Nov2010
+ Parslet::Transform now takes a block on initialisation, wherein you can
define all the rules directly.
+ Parslet::Transform now only passes a hash to the block during transform
when its arity is 1. Otherwise all hash contents as bound as local
variables.
+ Both inline and other documentation have been improved.
+ You can now use 'subtree(:x)' to bind any subtree to x during tree pattern
matching.
+ Transform classes can now include rules into class definition. This makes
Parser and Transformer behave the same.
= 0.9.0 / 28Oct2010
* More of everything: Examples, documentation, etc...
* Breaking change: Ruby's binary or ('|') is now used for alternatives,
instead of the division sign ('/') - this reduces the amount of
parenthesis needed for a grammar overall.
* parslet.maybe now yields the result or nil in case of parse failure. This
is probably better than the array it did before; the jury is still out on
that.
* parslet.repeat(min, max) is now valid syntax
= 0.1.0 / not released.
* Initial version. Classes for parsing, matching in the resulting trees
and transforming the trees into something more useful.
* Parses and outputs intermediary trees
* Matching of single elements and sequences