n

[xiongjiwei]

What problem does this PR solve?

charset: support collation utf8mb4_unicode_ci and utf8_unicode_ci

add utf8_unicode_ci and utf8mb4_unicode_ci collation interface, implement as utf8_bin and utf8mb4_bin now.

Unit Test Change

• TestForbidUnsupportedCollations tests utf8mb4_roman_ci and utf8_roman_ci cause utf8mb4_unicode_ci and utf8mb4_unicode_ci are support now.
• show collation will show utf8mb4_unicode_ci and utf8_unicode_ci
• many other tests for utf8mb4_unicode_ci and utf8_unicode_ci

Note

• Because of utf8mb4_unicode_ci and utf8_unicode_ci implement as utf8mb4_bin and utf8_bin, the test is also same as utf8mb4_bin and utf8_bin except when using ordering rule. These tests should change when actual implement.

• should add utf8mb4_unicode_ci and utf8_unicode_ci to collationPriority and CollationStrictness
  

  tidb/expression/collation.go

  Lines 115 to 136 in f31298f

  	// collationPriority is the priority when infer the result collation, the priority of collation a > b iff collationPriority[a] > collationPriority[b]
  	collationPriority = map[string]int{
  	charset.CollationASCII: 0,
  	charset.CollationLatin1: 1,
  	"utf8_general_ci": 2,
  	charset.CollationUTF8: 3,
  	"utf8mb4_general_ci": 4,
  	charset.CollationUTF8MB4: 5,
  	charset.CollationBin: 6,
  	}
  	// CollationStrictness indicates the strictness of comparison of the collation. The unequal order in a weak collation also holds in a strict collation.
  	// For example, if a < b in a weak collation(e.g. general_ci), then there must be a < b in a strict collation(e.g. _bin).
  	CollationStrictness = map[string]int{
  	"utf8_general_ci": 0,
  	"utf8mb4_general_ci": 0,
  	charset.CollationASCII: 1,
  	charset.CollationLatin1: 1,
  	charset.CollationUTF8: 1,
  	charset.CollationUTF8MB4: 1,
  	charset.CollationBin: 2,
  	}

see also 表达式中排序规则的 Coercibility 值

PS

I did not replace all the utf8_bin to utf8_unicode_ci and utf8mb4_bin to utf8mb4_unicode_ci in test, change the default collation to utf8mb4_unicode_ci.
It's no need to change them all cause it only used when new_collation_enabled is true.

Also, I write tests for utf8mb4_unicode_ci and utf8_unicode_ci. For pass the test now, I write the tests same as tests about utf8mb4_bin, utf8_bin. (In TDD, these tests should failed)