| 1 |
# These tests are specifically written to test the regex-lite crate. While it |
| 2 |
# largely has the same semantics as the regex crate, there are some differences |
| 3 |
# around Unicode support and UTF-8. |
| 4 |
# |
| 5 |
# To be clear, regex-lite supports far fewer patterns because of its lack of |
| 6 |
# Unicode support, nested character classes and character class set operations. |
| 7 |
# What we're talking about here are the patterns that both crates support but |
| 8 |
# where the semantics might differ. |
| 9 |
|
| 10 |
# regex-lite uses ASCII definitions for Perl character classes. |
| 11 |
[[test]] |
| 12 |
name = "perl-class-decimal" |
| 13 |
regex = '\d' |
| 14 |
haystack = '᠕' |
| 15 |
matches = [] |
| 16 |
unicode = true |
| 17 |
|
| 18 |
# regex-lite uses ASCII definitions for Perl character classes. |
| 19 |
[[test]] |
| 20 |
name = "perl-class-space" |
| 21 |
regex = '\s' |
| 22 |
haystack = "\u2000" |
| 23 |
matches = [] |
| 24 |
unicode = true |
| 25 |
|
| 26 |
# regex-lite uses ASCII definitions for Perl character classes. |
| 27 |
[[test]] |
| 28 |
name = "perl-class-word" |
| 29 |
regex = '\w' |
| 30 |
haystack = 'δ' |
| 31 |
matches = [] |
| 32 |
unicode = true |
| 33 |
|
| 34 |
# regex-lite uses the ASCII definition of word for word boundary assertions. |
| 35 |
[[test]] |
| 36 |
name = "word-boundary" |
| 37 |
regex = '\b' |
| 38 |
haystack = 'δ' |
| 39 |
matches = [] |
| 40 |
unicode = true |
| 41 |
|
| 42 |
# regex-lite uses the ASCII definition of word for negated word boundary |
| 43 |
# assertions. But note that it should still not split codepoints! |
| 44 |
[[test]] |
| 45 |
name = "word-boundary-negated" |
| 46 |
regex = '\B' |
| 47 |
haystack = 'δ' |
| 48 |
matches = [[0, 0], [2, 2]] |
| 49 |
unicode = true |
| 50 |
|
| 51 |
# While we're here, the empty regex---which matches at every |
| 52 |
# position---shouldn't split a codepoint either. |
| 53 |
[[test]] |
| 54 |
name = "empty-no-split-codepoint" |
| 55 |
regex = '' |
| 56 |
haystack = '💩' |
| 57 |
matches = [[0, 0], [4, 4]] |
| 58 |
unicode = true |
| 59 |
|
| 60 |
# A dot always matches a full codepoint. |
| 61 |
[[test]] |
| 62 |
name = "dot-always-matches-codepoint" |
| 63 |
regex = '.' |
| 64 |
haystack = '💩' |
| 65 |
matches = [[0, 4]] |
| 66 |
unicode = false |
| 67 |
|
| 68 |
# A negated character class also always matches a full codepoint. |
| 69 |
[[test]] |
| 70 |
name = "negated-class-always-matches-codepoint" |
| 71 |
regex = '[^a]' |
| 72 |
haystack = '💩' |
| 73 |
matches = [[0, 4]] |
| 74 |
unicode = false |
| 75 |
|
| 76 |
# regex-lite only supports ASCII-aware case insensitive matching. |
| 77 |
[[test]] |
| 78 |
name = "case-insensitive-is-ascii-only" |
| 79 |
regex = 's' |
| 80 |
haystack = 'ſ' |
| 81 |
matches = [] |
| 82 |
unicode = true |
| 83 |
case-insensitive = true |
| 84 |
|
| 85 |
# Negated word boundaries shouldn't split a codepoint, but they will match |
| 86 |
# between invalid UTF-8. |
| 87 |
# |
| 88 |
# This test is only valid for a 'bytes' API, but that doesn't (yet) exist in |
| 89 |
# regex-lite. This can't happen in the main API because &str can't contain |
| 90 |
# invalid UTF-8. |
| 91 |
# [[test]] |
| 92 |
# name = "word-boundary-invalid-utf8" |
| 93 |
# regex = '\B' |
| 94 |
# haystack = '\xFF\xFF\xFF\xFF' |
| 95 |
# unescape = true |
| 96 |
# matches = [[0, 0], [1, 1], [2, 2], [3, 3], [4, 4]] |
| 97 |
# unicode = true |
| 98 |
# utf8 = false |