cataract 0.2.3 → 0.2.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (28) hide show
  1. checksums.yaml +4 -4
  2. data/.rubocop.yml +7 -3
  3. data/BENCHMARKS.md +32 -32
  4. data/CHANGELOG.md +10 -0
  5. data/Gemfile +3 -0
  6. data/ext/cataract/cataract.c +219 -112
  7. data/ext/cataract/cataract.h +5 -1
  8. data/ext/cataract/css_parser.c +523 -33
  9. data/ext/cataract/flatten.c +233 -91
  10. data/ext/cataract/shorthand_expander.c +7 -0
  11. data/lib/cataract/at_rule.rb +2 -1
  12. data/lib/cataract/constants.rb +10 -0
  13. data/lib/cataract/import_resolver.rb +18 -87
  14. data/lib/cataract/import_statement.rb +29 -5
  15. data/lib/cataract/media_query.rb +98 -0
  16. data/lib/cataract/pure/byte_constants.rb +11 -0
  17. data/lib/cataract/pure/flatten.rb +127 -15
  18. data/lib/cataract/pure/parser.rb +637 -270
  19. data/lib/cataract/pure/serializer.rb +216 -115
  20. data/lib/cataract/pure.rb +6 -7
  21. data/lib/cataract/rule.rb +9 -5
  22. data/lib/cataract/stylesheet.rb +321 -99
  23. data/lib/cataract/stylesheet_scope.rb +10 -7
  24. data/lib/cataract/version.rb +1 -1
  25. data/lib/cataract.rb +4 -8
  26. data/lib/tasks/profile.rake +210 -0
  27. metadata +4 -2
  28. data/lib/cataract/pure/imports.rb +0 -268
checksums.yaml CHANGED
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  SHA256:
3
- metadata.gz: a5945042f40d16bf57bcc0c5479221959e231eb2742c51019f907d1d3ab1110c
4
- data.tar.gz: 92734f20818a3c0e94b8f7adaccf5e6a37d65325223f0dbbafae2efc8ddc11d8
3
+ metadata.gz: 90e5adcec05553f73e5829b769d010ecda9ca5bf5eb1b2cc621b74707758c913
4
+ data.tar.gz: 4a82a843a72f2cc3c944798af3b7e7a8271b5ad696c1ed53247854d5ecdda7bb
5
5
  SHA512:
6
- metadata.gz: b1c8dc3e693f042549bdaf2561599c76cf2f5940809f3c5005b17ee90a04cab9a99f35598f55f9d422a3f00ab04d493c76a0f291ce636711c9dabcb2879aceb2
7
- data.tar.gz: 6d239fd54239b01ec42215296964d0140e40ae1841eae2d4af12e0982d5478ad2ee49c5f6628f067ef5168bef5845f07780e63fc6e6be4875c06c21f8f0afb20
6
+ metadata.gz: afb9718d653c7bc0c905eb3d273afec8a364ae37de2952ea3285ecf3ef7f5ff29b40887348ef3646adedfab0baeff873211a883750e7c9b992f3e6a25b14cbb0
7
+ data.tar.gz: c40ba4cefa95024f34c7d9306d93ad07b8f448e847c51231381fa83833f2b97e314e5fbab9d7fb2a8edd1fe45fea0760c6db514c79225829e397c339fd5951a6
data/.rubocop.yml CHANGED
@@ -103,10 +103,12 @@ Naming/PredicatePrefix:
103
103
  # These cosmetic style rules add overhead via method dispatch or allocation
104
104
 
105
105
  # Disable numeric predicates in hot path - method dispatch overhead
106
- # Benchmark shows `> 0` is faster than `.positive?` due to avoiding method call
106
+ # Benchmark shows `length > 0` is 30% faster than `.positive?` (method dispatch overhead)
107
+ # `!empty?` has no perf penalty vs `length > 0` but `.positive?` is significantly slower
107
108
  Style/NumericPredicate:
108
109
  Exclude:
109
110
  - 'lib/cataract/pure/**/*.rb'
111
+ - 'lib/cataract/stylesheet.rb'
110
112
 
111
113
  # Disable multiple comparison style - avoid array allocation
112
114
  # `||` chain is faster than `Array#include?` which allocates an array
@@ -114,11 +116,13 @@ Style/MultipleComparison:
114
116
  Exclude:
115
117
  - 'lib/cataract/pure/**/*.rb'
116
118
 
117
- # Disable zero-length predicate in hot path - method dispatch overhead
118
- # Direct comparison `.length > 0` is faster than `.empty?` method call
119
+ # Disable zero-length predicate - prefer `length > 0` for consistency with NumericPredicate
120
+ # While `!empty?` has no perf penalty vs `length > 0`, we use `length > 0` consistently
121
+ # to avoid mixing styles (since `.positive?` is 30% slower and must be avoided)
119
122
  Style/ZeroLengthPredicate:
120
123
  Exclude:
121
124
  - 'lib/cataract/pure/**/*.rb'
125
+ - 'lib/cataract/stylesheet.rb'
122
126
 
123
127
  # Benchmarked range vs offsets. Range has to allocated
124
128
  # so not ideal in the hot path
data/BENCHMARKS.md CHANGED
@@ -20,17 +20,17 @@ Time to parse CSS into internal data structures
20
20
 
21
21
  | Test Case | Native | Pure (no YJIT) | Pure (YJIT) |
22
22
  |-----------|--------|----------------|-------------|
23
- | Small CSS (64 lines, 1.0KB) | 36.46K i/s | 3.46K i/s | 14.1K i/s |
24
- | Medium CSS with @media (139 lines, 1.6KB) | 37.11K i/s | 2.1K i/s | 9.26K i/s |
25
- | Selector lists (3500 lines, 62.5KB, 500 lists) | 447.6 i/s | 57.9 i/s | 212.5 i/s |
23
+ | Small CSS (64 lines, 1.0KB) | 37.4K i/s | 3.25K i/s | 13.43K i/s |
24
+ | Medium CSS with @media (139 lines, 1.6KB) | 34.71K i/s | 2.04K i/s | 9.15K i/s |
25
+ | Selector lists (3500 lines, 62.5KB, 500 lists) | 471.8 i/s | 56.7 i/s | 214.1 i/s |
26
26
 
27
27
  ### Speedups
28
28
 
29
29
  | Comparison | Speedup |
30
30
  |------------|---------|
31
- | Native vs Pure (no YJIT) | 12.14x faster (avg) |
32
- | Native vs Pure (YJIT) | 2.89x faster (avg) |
33
- | YJIT impact on Pure Ruby | 4.14x faster (avg) |
31
+ | Native vs Pure (no YJIT) | 12.76x faster (avg) |
32
+ | Native vs Pure (YJIT) | 2.99x faster (avg) |
33
+ | YJIT impact on Pure Ruby | 4.22x faster (avg) |
34
34
 
35
35
  ---
36
36
 
@@ -40,19 +40,19 @@ Time to convert parsed CSS back to string format
40
40
 
41
41
  | Test Case | Native | Pure (no YJIT) | Pure (YJIT) |
42
42
  |-----------|--------|----------------|-------------|
43
- | to_s (Bootstrap - 191KB) | 1.16K i/s | 34.9 i/s | 49.7 i/s |
44
- | to_s (Compact utilities - 2.9KB) | 1.16K i/s | 34.9 i/s | 49.7 i/s |
45
- | to_formatted_s (Nested CSS - 1.3KB) | 158.75K i/s | 63.62K i/s | 114.79K i/s |
46
- | to_s with selector_lists (3.4KB) | 17.95K i/s | 1.43K i/s | 1.99K i/s |
47
- | Media filtering (Bootstrap print only) | 235.05K i/s | 98.95K i/s | 149.3K i/s |
43
+ | to_s (Bootstrap - 191KB) | 1.32K i/s | 360.2 i/s | 656.6 i/s |
44
+ | to_s (Compact utilities - 2.9KB) | 1.32K i/s | 360.2 i/s | 656.6 i/s |
45
+ | to_formatted_s (Nested CSS - 1.3KB) | 176.13K i/s | 60.97K i/s | 121.5K i/s |
46
+ | to_s with selector_lists (3.4KB) | 19.03K i/s | 8.01K i/s | 14.2K i/s |
47
+ | Media filtering (Bootstrap print only) | 283.24K i/s | 97.28K i/s | 161.56K i/s |
48
48
 
49
49
  ### Speedups
50
50
 
51
51
  | Comparison | Speedup |
52
52
  |------------|---------|
53
- | Native vs Pure (no YJIT) | 10.97x faster (avg) |
54
- | Native vs Pure (YJIT) | 7.56x faster (avg) |
55
- | YJIT impact on Pure Ruby | 1.57x faster (avg) |
53
+ | Native vs Pure (no YJIT) | 3.24x faster (avg) |
54
+ | Native vs Pure (YJIT) | 1.73x faster (avg) |
55
+ | YJIT impact on Pure Ruby | 1.86x faster (avg) |
56
56
 
57
57
  ---
58
58
 
@@ -62,20 +62,20 @@ Time to calculate CSS selector specificity values
62
62
 
63
63
  | Test Case | Native | Pure (no YJIT) | Pure (YJIT) |
64
64
  |-----------|--------|----------------|-------------|
65
- | Simple Selectors | 8.42M i/s | 508.29K i/s | 2.6M i/s |
66
- | Compound Selectors | 6.6M i/s | 186.16K i/s | 400.7K i/s |
67
- | Combinators | 5.21M i/s | 147.76K i/s | 255.26K i/s |
68
- | Pseudo-classes & Pseudo-elements | 5.39M i/s | 117.25K i/s | 196.91K i/s |
69
- | :not() Pseudo-class (CSS3) | 3.48M i/s | 103.11K i/s | 164.66K i/s |
70
- | Complex Real-world Selectors | 4.05M i/s | 52.17K i/s | 78.28K i/s |
65
+ | Simple Selectors | 7.28M i/s | 506.61K i/s | 2.45M i/s |
66
+ | Compound Selectors | 6.4M i/s | 186.62K i/s | 405.15K i/s |
67
+ | Combinators | 5.04M i/s | 147.07K i/s | 255.92K i/s |
68
+ | Pseudo-classes & Pseudo-elements | 5.17M i/s | 117.16K i/s | 195.12K i/s |
69
+ | :not() Pseudo-class (CSS3) | 3.48M i/s | 103.47K i/s | 161.34K i/s |
70
+ | Complex Real-world Selectors | 4.01M i/s | 52.41K i/s | 77.48K i/s |
71
71
 
72
72
  ### Speedups
73
73
 
74
74
  | Comparison | Speedup |
75
75
  |------------|---------|
76
- | Native vs Pure (no YJIT) | 40.75x faster (avg) |
77
- | Native vs Pure (YJIT) | 23.38x faster (avg) |
78
- | YJIT impact on Pure Ruby | 2.29x faster (avg) |
76
+ | Native vs Pure (no YJIT) | 39.54x faster (avg) |
77
+ | Native vs Pure (YJIT) | 23.05x faster (avg) |
78
+ | YJIT impact on Pure Ruby | 2.24x faster (avg) |
79
79
 
80
80
  ---
81
81
 
@@ -85,20 +85,20 @@ Time to flatten multiple CSS rule sets with same selector
85
85
 
86
86
  | Test Case | Native | Pure (no YJIT) | Pure (YJIT) |
87
87
  |-----------|--------|----------------|-------------|
88
- | No shorthand properties (large) | 20.74K i/s | 3.15K i/s | 5.73K i/s |
89
- | Simple properties | 155.01K i/s | 73.1K i/s | 99.95K i/s |
90
- | Cascade with specificity | 194.63K i/s | 74.22K i/s | 104.8K i/s |
91
- | Important declarations | 195.45K i/s | 72.82K i/s | 105.65K i/s |
92
- | Shorthand expansion | 20.74K i/s | 3.15K i/s | 5.73K i/s |
93
- | Complex flattening | 31.19K i/s | 15.81K i/s | 21.52K i/s |
88
+ | No shorthand properties (large) | 13.91K i/s | 3.23K i/s | 6.39K i/s |
89
+ | Simple properties | 144.19K i/s | 71.11K i/s | 101.38K i/s |
90
+ | Cascade with specificity | 172.96K i/s | 69.16K i/s | 99.23K i/s |
91
+ | Important declarations | 173.35K i/s | 69.12K i/s | 100.16K i/s |
92
+ | Shorthand expansion | 13.91K i/s | 3.23K i/s | 6.39K i/s |
93
+ | Complex flattening | 29.86K i/s | 15.77K i/s | 22.09K i/s |
94
94
 
95
95
  ### Speedups
96
96
 
97
97
  | Comparison | Speedup |
98
98
  |------------|---------|
99
- | Native vs Pure (no YJIT) | 3.01x faster (avg) |
100
- | Native vs Pure (YJIT) | 1.97x faster (avg) |
101
- | YJIT impact on Pure Ruby | 1.47x faster (avg) |
99
+ | Native vs Pure (no YJIT) | 2.54x faster (avg) |
100
+ | Native vs Pure (YJIT) | 1.63x faster (avg) |
101
+ | YJIT impact on Pure Ruby | 1.53x faster (avg) |
102
102
 
103
103
  ---
104
104
 
data/CHANGELOG.md CHANGED
@@ -1,5 +1,15 @@
1
1
  ## [ Unreleased ]
2
2
 
3
+ ## [0.2.4 - 2025-11-23]
4
+ - MediaQuery first-class objects: Refactored media queries from simple symbols to proper structs with id, type, and conditions, enabling accurate
5
+ serialization and proper handling of complex queries like @media screen and (min-width: 768px)
6
+ - Fixed import resolution: Import statements now properly merge selector lists and media query lists from imported stylesheets with correct ID offsetting,
7
+ preventing data loss
8
+ - Sequential rule ID invariant: Parser now ensures rules[i].id == i via placeholder strategy, enabling O(1) array access instead of O(N) lookups during
9
+ serialization
10
+ - Improved nested media handling: Nested media queries in imports now combine correctly (e.g., @import "file.css" screen where file contains @media
11
+ (min-width: 768px))
12
+
3
13
  ## [0.2.3 - 2025-11-18]
4
14
  - Pure Parser: Bugs with url()
5
15
 
data/Gemfile CHANGED
@@ -10,13 +10,16 @@ gem 'rake', '~> 13.0'
10
10
  gem 'rake-compiler', '~> 1.0'
11
11
 
12
12
  # Development/benchmarking dependencies (not needed by gem users)
13
+ gem 'addressable', '~> 2.8' # for custom URI resolver testing
13
14
  gem 'benchmark-ips', '~> 2.0'
14
15
  gem 'css_parser', '~> 1.0' # for benchmarking against
15
16
  gem 'minitest'
16
17
  gem 'minitest-spec'
17
18
  gem 'nokogiri' # for docs
19
+ gem 'ruby-prof', require: false # for profiling
18
20
  gem 'simplecov', require: false
19
21
  gem 'simplecov-cobertura', require: false
22
+ gem 'stackprof', require: false # for profiling
20
23
  gem 'webmock', '~> 3.0' # for testing URL loading
21
24
 
22
25
  gem 'overcommit', '~> 0.64', group: :development