Hello! Thank you for review! New patch below. This time I wasn't able to insert diff between two last versions, but I will do it in next patches. On 10/10/2018 02:18 PM, Vladislav Shpilevoy wrote: > > > On 06/10/2018 12:04, imeevma@tarantool.org wrote: >> At the moment all collations that don't have their "strength" >> option set works the same way as ones with this option set to >> "identical". It is not efficient, according to ICU Comparison >> Levels. This patch updates this option of old collations and set >> its default value to "primary" for new ones. >> >> Closes #3573 >> --- >> Branch: >> https://github.com/tarantool/tarantool/tree/imeevma/gh-3573-fix-collation-strength >> Issue: https://github.com/tarantool/tarantool/issues/3573 >> >>   src/box/bootstrap.snap       | Bin 1888 -> 1886 bytes >>   src/box/coll_id_def.c        |   4 +- >>   src/box/lua/upgrade.lua      |  29 ++++++++++++++- >>   src/coll.c                   |  33 ++++++++-------- >>   src/coll_def.c               |   1 - >>   src/coll_def.h               |   3 +- >>   test/box-py/bootstrap.result |   2 +- >>   test/box/ddl.result          |   6 +-- >>   test/box/tree_pk.result      |   2 +- >>   test/box/tree_pk.test.lua    |   2 +- >>   test/sql/collation.result    |  87 >> +++++++++++++++++++++++++++++++++++++++++++ >>   test/sql/collation.test.lua  |  32 ++++++++++++++++ >>   test/unit/coll.cpp           |   2 + >>   13 files changed, 176 insertions(+), 27 deletions(-) >> >> diff --git a/src/box/bootstrap.snap b/src/box/bootstrap.snap > > 1. Please, remove binary diff before sending an email. Ok, understood. > >> index >> 65739384a66d6ba4a538553ccf4677536ba15280..84b27c0d8b9d6d683692e2ef55034ffcc68d11d0 >> 100644 >> GIT binary patch >> delta 1884 >> zcmV-i2c!7l4&Dxs7k@P^HZUz`Ff(N|W@R=CNp5p=VQyn(Iv_J> zG-feoEjVOmHZ3+eGhs9}W;HQnWj6{|Lu_wjYdRo%eF_TIx(m9^2Ic_Hx7lmzr2qf` >> z001bpFZ}>e?KS}FOGFS!z*dm~D2k#jilV6AqMxq(Vp{_Az<(~`=k4N^J9QNuj=Z`N >> z=~K!63-eO4{Xa?goO8~~I%_^CBZ1> zy%3s9> z)wQCd>0t|o-+y-Dh~($*C8WEVmHYnrJ?s76-+IqxnZ_~OWX>t2h{+{DrBamb3&R_{ >> z;m#^i0F_EH-wt>E9?reGOaJa1M|9-I2^O^Z-q&Fr-e!`irB*7%IjR0I-ygdXsZ@$r >> zrW6m`p$oIkN~NgMQRS>Um8$bR)M@pqdwLxd>fHSbwLRQvC`i)w)n+f9GNDZZi|h >> zlyVA75fPvxtQad)dZGZJvM8eKGmha6=5D?nx8-c_J_zeu`v1DV+jWcK4cc+893buI >> zVX8r;QhfFBBJJm)9VdzbkxqxLn6=FsmP(}<_UA-89k%Q^X%-~K>}{G9>> z#@04x5`P$pS*0n1pkL1_nK1_{m7>0%zmJtW*}uBmjeXzF$Dyk+jJDtPUE>@0o!5-$ >> diff --git a/src/box/lua/upgrade.lua b/src/box/lua/upgrade.lua >> index d9c2ae4..07cace0 100644 >> --- a/src/box/lua/upgrade.lua >> +++ b/src/box/lua/upgrade.lua >> @@ -578,6 +578,32 @@ local function upgrade_to_2_1_0() >>       box.space._schema:format(format) >>   end >> +-------------------------------------------------------------------------------- >> +-- Tarantool 2.1.1 >> +-------------------------------------------------------------------------------- >> >> + >> +function update_collation_strength_field() >> +    local _collation = box.space[box.schema.COLLATION_ID] >> +    local _format = _collation:format() >> +    for _, collation in ipairs(_collation:select()) do >> +        if (collation.opts.strength == nil) then >> +            local old_collation = _collation:get{collation.id} >> +            local new_collation = {} >> +            for _,field in ipairs(_format) do >> +                new_collation[field.name] = old_collation[field.name] >> +            end >> +            new_collation.opts.strength = 'identical' >> +            _collation:delete{old_collation.id} >> + _collation:insert(_collation:frommap(new_collation)) > > 2. This looks much simpler: > > @@ -584,17 +584,11 @@ end > >  function update_collation_strength_field() >      local _collation = box.space[box.schema.COLLATION_ID] > -    local _format = _collation:format() >      for _, collation in ipairs(_collation:select()) do > -        if (collation.opts.strength == nil) then > -            local old_collation = _collation:get{collation.id} > -            local new_collation = {} > -            for _,field in ipairs(_format) do > -                new_collation[field.name] = old_collation[field.name] > -            end > -            new_collation.opts.strength = 'identical' > -            _collation:delete{old_collation.id} > -            _collation:insert(_collation:frommap(new_collation)) > +        if collation.opts.strength == nil then > +            local collation = _collation:get{collation.id}:totable() > +            collation[6].strength = 'identical' > +            _collation:replace(collation) >          end >      end >  end > > > I did not test it though. Please, apply and debug if > necessary. Done. >> +        end >> +    end >> +end >> diff --git a/test/sql/collation.result b/test/sql/collation.result >> index 79ba9ab..3c53be1 100644 >> --- a/test/sql/collation.result >> +++ b/test/sql/collation.result >> @@ -110,3 +110,90 @@ cn:close() >>   box.schema.user.revoke('guest', 'read,write,execute', 'universe') >>   --- >>   ... >> +-- >> +-- gh-3573: Strength in the _collation space >> +-- Collation without 'strength' option set now works as one with >> +-- 'strength' set to 'primary'. >> +-- >> +box.internal.collation.create('c0', 'ICU', 'unicode') >> +--- >> +... >> +box.internal.collation.create('c1', 'ICU', 'unicode', >> {strength='primary'}) >> +--- >> +... >> +box.internal.collation.create('c2', 'ICU', 'unicode', >> {strength='secondary'}) >> +--- >> +... >> +box.internal.collation.create('c5', 'ICU', 'unicode', >> {strength='identical'}) >> +--- >> +... >> +box.sql.execute([[create table tc (id int primary key autoincrement, >> s0 string collate "c0", s1 string collate "c1", s2 string collate >> "c2", s5 string collate "c5")]]) >> +--- >> +... >> +box.sql.execute([[insert into tc values (null, 'a', 'a', 'a', 'a')]]) >> +--- >> +... >> +box.sql.execute([[insert into tc values (null, 'A', 'A', 'A', 'A')]]) >> +--- >> +... >> +box.sql.execute([[insert into tc values (null, 'á', 'á', 'á', 'á')]]) >> +--- >> +... >> +box.sql.execute([[insert into tc values (null, 'â', 'â', 'â', 'â')]]) >> +--- >> +... >> +box.sql.execute([[select * from tc where s0 = 'a']]) >> +--- >> +- - [1, 'a', 'a', 'a', 'a'] >> +  - [2, 'A', 'A', 'A', 'A'] >> +  - [3, 'á', 'á', 'á', 'á'] >> +  - [4, 'â', 'â', 'â', 'â'] >> +... >> +box.sql.execute([[select * from tc where s1 = 'a']]) >> +--- >> +- - [1, 'a', 'a', 'a', 'a'] >> +  - [2, 'A', 'A', 'A', 'A'] >> +  - [3, 'á', 'á', 'á', 'á'] >> +  - [4, 'â', 'â', 'â', 'â'] >> +... >> +box.sql.execute([[select * from tc where s2 = 'a']]) >> +--- >> +- - [1, 'a', 'a', 'a', 'a'] >> +  - [2, 'A', 'A', 'A', 'A'] >> +... >> +box.sql.execute([[select * from tc where s5 = 'a']]) >> +--- >> +- - [1, 'a', 'a', 'a', 'a'] >> +... >> +a = box.sql.execute([[select id from tc where s0 = 'a']]) >> +--- >> +... >> +b = box.sql.execute([[select id from tc where s1 = 'a']]) >> +--- >> +... >> +count = 0 >> +--- >> +... >> +for k,v in pairs(a) do if (a[k][1] ~= b[k][1]) then count = count + >> 1 end end > > 3. You printed both a and b whole above and they are the same. Why do you > need this additional check in a cycle? Fixed, removed unnecessary checks and added comments. *New patch:* commit e55064f796d9320035db5fb2e3d5030601ee7333 Author: Mergen Imeev Date:   Tue Sep 25 21:30:36 2018 +0300     box: update collation strength option.     At the moment all collations that don't have their "strength"     option set works the same way as ones with this option set to     "identical". It is not efficient, according to ICU Comparison     Levels. This patch updates this option of old collations and set     its default value to "primary" for new ones.     Closes #3573 diff --git a/src/box/bootstrap.snap b/src/box/bootstrap.snap index 6573938..3d3232f 100644 Binary files a/src/box/bootstrap.snap and b/src/box/bootstrap.snap differ diff --git a/src/box/coll_id_def.c b/src/box/coll_id_def.c index 9fe0cda..43f3b63 100644 --- a/src/box/coll_id_def.c +++ b/src/box/coll_id_def.c @@ -55,8 +55,8 @@ icu_case_first_from_str(const char *str, uint32_t len)  static int64_t  icu_strength_from_str(const char *str, uint32_t len)  { -    return strnindex(coll_icu_strength_strs + 1, str, len, -             coll_icu_strength_MAX - 1) + 1; +    return strnindex(coll_icu_strength_strs, str, len, +             coll_icu_strength_MAX);  }  const struct opt_def coll_icu_opts_reg[] = { diff --git a/src/box/lua/upgrade.lua b/src/box/lua/upgrade.lua index d9c2ae4..66dcef4 100644 --- a/src/box/lua/upgrade.lua +++ b/src/box/lua/upgrade.lua @@ -578,6 +578,27 @@ local function upgrade_to_2_1_0()      box.space._schema:format(format)  end +-------------------------------------------------------------------------------- +-- Tarantool 2.1.1 +-------------------------------------------------------------------------------- + +function update_collation_strength_field() +    local _collation = box.space[box.schema.COLLATION_ID] +    for _, collation in ipairs(_collation:select()) do +        if (collation.opts.strength == nil) then +            local new_collation = _collation:get{collation.id}:totable() +            new_collation[6].strength = 'identical' +            _collation:delete{collation.id} +            _collation:insert(new_collation) +        end +    end +end + +local function upgrade_to_2_1_1() +    update_collation_strength_field() +end + +  local function get_version()      local version = box.space._schema:get{'version'}      if version == nil then @@ -605,7 +626,8 @@ local function upgrade(options)          {version = mkversion(1, 7, 7), func = upgrade_to_1_7_7, auto = true},          {version = mkversion(1, 10, 0), func = upgrade_to_1_10_0, auto = true},          {version = mkversion(1, 10, 2), func = upgrade_to_1_10_2, auto = true}, -        {version = mkversion(2, 1, 0), func = upgrade_to_2_1_0, auto = true} +        {version = mkversion(2, 1, 0), func = upgrade_to_2_1_0, auto = true}, +        {version = mkversion(2, 1, 1), func = upgrade_to_2_1_1, auto = true}      }      for _, handler in ipairs(handlers) do diff --git a/src/coll.c b/src/coll.c index 6a76f1f..9b719f2 100644 --- a/src/coll.c +++ b/src/coll.c @@ -193,22 +193,25 @@ coll_icu_init_cmp(struct coll *coll, const struct coll_def *def)              return -1;          }      } -    if (def->icu.strength != COLL_ICU_STRENGTH_DEFAULT) { -        enum coll_icu_strength w = def->icu.strength; -        UColAttributeValue v = -            w == COLL_ICU_STRENGTH_PRIMARY ? UCOL_PRIMARY : -            w == COLL_ICU_STRENGTH_SECONDARY ? UCOL_SECONDARY : -            w == COLL_ICU_STRENGTH_TERTIARY ? UCOL_TERTIARY : -            w == COLL_ICU_STRENGTH_QUATERNARY ? UCOL_QUATERNARY : -            w == COLL_ICU_STRENGTH_IDENTICAL ? UCOL_IDENTICAL : -            UCOL_DEFAULT; -        ucol_setAttribute(collator, UCOL_STRENGTH, v, &status); -        if (U_FAILURE(status)) { -            diag_set(CollationError, "failed to set strength: %s", -                 u_errorName(status)); -            return -1; -        } +    /* +     * Set "strength" option of collation. Default values is +     * "primary". +     */ +    enum coll_icu_strength w = def->icu.strength; +    UColAttributeValue v = +        w == COLL_ICU_STRENGTH_PRIMARY ? UCOL_PRIMARY : +        w == COLL_ICU_STRENGTH_SECONDARY ? UCOL_SECONDARY : +        w == COLL_ICU_STRENGTH_TERTIARY ? UCOL_TERTIARY : +        w == COLL_ICU_STRENGTH_QUATERNARY ? UCOL_QUATERNARY : +        w == COLL_ICU_STRENGTH_IDENTICAL ? UCOL_IDENTICAL : +        UCOL_DEFAULT; +    ucol_setAttribute(collator, UCOL_STRENGTH, v, &status); +    if (U_FAILURE(status)) { +        diag_set(CollationError, "failed to set strength: %s", +             u_errorName(status)); +        return -1;      } +      if (def->icu.numeric_collation != COLL_ICU_DEFAULT) {          enum coll_icu_on_off w = def->icu.numeric_collation;          UColAttributeValue v = w == COLL_ICU_ON ? UCOL_ON : diff --git a/src/coll_def.c b/src/coll_def.c index df58cac..1323643 100644 --- a/src/coll_def.c +++ b/src/coll_def.c @@ -54,7 +54,6 @@ const char *coll_icu_case_first_strs[] = {  };  const char *coll_icu_strength_strs[] = { -    "DEFAULT",      "PRIMARY",      "SECONDARY",      "TERTIARY", diff --git a/src/coll_def.h b/src/coll_def.h index 7c20abf..2a1cb68 100644 --- a/src/coll_def.h +++ b/src/coll_def.h @@ -84,8 +84,7 @@ extern const char *coll_icu_case_first_strs[];  /** Strength ICU settings */  enum coll_icu_strength { -    COLL_ICU_STRENGTH_DEFAULT = 0, -    COLL_ICU_STRENGTH_PRIMARY, +    COLL_ICU_STRENGTH_PRIMARY = 0,      COLL_ICU_STRENGTH_SECONDARY,      COLL_ICU_STRENGTH_TERTIARY,      COLL_ICU_STRENGTH_QUATERNARY, diff --git a/test/box-py/bootstrap.result b/test/box-py/bootstrap.result index 506aca3..2532b70 100644 --- a/test/box-py/bootstrap.result +++ b/test/box-py/bootstrap.result @@ -5,7 +5,7 @@ box.space._schema:select{}  ---  - - ['cluster', '']    - ['max_id', 511] -  - ['version', 2, 1, 0] +  - ['version', 2, 1, 1]  ...  box.space._cluster:select{}  --- diff --git a/test/box/ddl.result b/test/box/ddl.result index c9a8e96..396318f 100644 --- a/test/box/ddl.result +++ b/test/box/ddl.result @@ -484,14 +484,14 @@ box.space._collation:auto_increment{'test', 0, 'ICU', 'ru_RU', setmap{}}  ...  box.space._collation:select{}  --- -- - [1, 'unicode', 1, 'ICU', '', {}] +- - [1, 'unicode', 1, 'ICU', '', {'strength': 'identical'}]    - [2, 'unicode_ci', 1, 'ICU', '', {'strength': 'primary'}]    - [3, 'test', 0, 'ICU', 'ru_RU', {}]  ...  test_run:cmd('restart server default')  box.space._collation:select{}  --- -- - [1, 'unicode', 1, 'ICU', '', {}] +- - [1, 'unicode', 1, 'ICU', '', {'strength': 'identical'}]    - [2, 'unicode_ci', 1, 'ICU', '', {'strength': 'primary'}]    - [3, 'test', 0, 'ICU', 'ru_RU', {}]  ... @@ -512,7 +512,7 @@ utf8.cmp('abc', 'def')  ...  box.space._collation:replace(t)  --- -- [1, 'unicode', 1, 'ICU', '', {}] +- [1, 'unicode', 1, 'ICU', '', {'strength': 'identical'}]  ...  --  -- gh-2839: allow to store custom fields in field definition. diff --git a/test/box/tree_pk.result b/test/box/tree_pk.result index df3c78b..7366ed8 100644 --- a/test/box/tree_pk.result +++ b/test/box/tree_pk.result @@ -690,7 +690,7 @@ s:drop()  ...  --https://github.com/tarantool/tarantool/issues/2649  -- create standart index and alter it to collation index -box.internal.collation.create('test', 'ICU', 'ru-RU') +box.internal.collation.create('test', 'ICU', 'ru-RU', {strength = 'identical'})  ---  ...  box.internal.collation.create('test-ci', 'ICU', 'ru-RU', {strength = 'secondary'}) diff --git a/test/box/tree_pk.test.lua b/test/box/tree_pk.test.lua index 1190ab4..b4ee65c 100644 --- a/test/box/tree_pk.test.lua +++ b/test/box/tree_pk.test.lua @@ -238,7 +238,7 @@ s:drop()  --https://github.com/tarantool/tarantool/issues/2649  -- create standart index and alter it to collation index -box.internal.collation.create('test', 'ICU', 'ru-RU') +box.internal.collation.create('test', 'ICU', 'ru-RU', {strength = 'identical'})  box.internal.collation.create('test-ci', 'ICU', 'ru-RU', {strength = 'secondary'})  s = box.schema.space.create('test')  i1 = s:create_index('i1', { type = 'tree', parts = {{1, 'str'}}, unique = true }) diff --git a/test/sql/collation.result b/test/sql/collation.result index 79ba9ab..5fdb8ee 100644 --- a/test/sql/collation.result +++ b/test/sql/collation.result @@ -110,3 +110,77 @@ cn:close()  box.schema.user.revoke('guest', 'read,write,execute', 'universe')  ---  ... +-- +-- gh-3573: Strength in the _collation space +-- Collation without 'strength' option set now works as one with +-- 'strength' set to 'primary'. +-- +box.internal.collation.create('c0', 'ICU', 'unicode') +--- +... +box.internal.collation.create('c1', 'ICU', 'unicode', {strength='primary'}) +--- +... +box.internal.collation.create('c2', 'ICU', 'unicode', {strength='secondary'}) +--- +... +box.internal.collation.create('c5', 'ICU', 'unicode', {strength='identical'}) +--- +... +box.sql.execute([[create table tc (id int primary key autoincrement, s0 string collate "c0", s1 string collate "c1", s2 string collate "c2", s5 string collate "c5")]]) +--- +... +box.sql.execute([[insert into tc values (null, 'a', 'a', 'a', 'a')]]) +--- +... +box.sql.execute([[insert into tc values (null, 'A', 'A', 'A', 'A')]]) +--- +... +box.sql.execute([[insert into tc values (null, 'á', 'á', 'á', 'á')]]) +--- +... +box.sql.execute([[insert into tc values (null, 'â', 'â', 'â', 'â')]]) +--- +... +-- Should be same as the next one. +box.sql.execute([[select * from tc where s0 = 'a']]) +--- +- - [1, 'a', 'a', 'a', 'a'] +  - [2, 'A', 'A', 'A', 'A'] +  - [3, 'á', 'á', 'á', 'á'] +  - [4, 'â', 'â', 'â', 'â'] +... +-- Should return all records. +box.sql.execute([[select * from tc where s1 = 'a']]) +--- +- - [1, 'a', 'a', 'a', 'a'] +  - [2, 'A', 'A', 'A', 'A'] +  - [3, 'á', 'á', 'á', 'á'] +  - [4, 'â', 'â', 'â', 'â'] +... +-- Should return two records. +box.sql.execute([[select * from tc where s2 = 'a']]) +--- +- - [1, 'a', 'a', 'a', 'a'] +  - [2, 'A', 'A', 'A', 'A'] +... +-- Should return one record. +box.sql.execute([[select * from tc where s5 = 'a']]) +--- +- - [1, 'a', 'a', 'a', 'a'] +... +box.sql.execute([[drop table tc]]) +--- +... +box.internal.collation.drop('c0') +--- +... +box.internal.collation.drop('c1') +--- +... +box.internal.collation.drop('c2') +--- +... +box.internal.collation.drop('c5') +--- +... diff --git a/test/sql/collation.test.lua b/test/sql/collation.test.lua index 935dea8..130245b 100644 --- a/test/sql/collation.test.lua +++ b/test/sql/collation.test.lua @@ -43,3 +43,32 @@ cn:execute('select 1 limit ? collate not_exist', {1})  cn:close()  box.schema.user.revoke('guest', 'read,write,execute', 'universe') + +-- +-- gh-3573: Strength in the _collation space +-- Collation without 'strength' option set now works as one with +-- 'strength' set to 'primary'. +-- +box.internal.collation.create('c0', 'ICU', 'unicode') +box.internal.collation.create('c1', 'ICU', 'unicode', {strength='primary'}) +box.internal.collation.create('c2', 'ICU', 'unicode', {strength='secondary'}) +box.internal.collation.create('c5', 'ICU', 'unicode', {strength='identical'}) +box.sql.execute([[create table tc (id int primary key autoincrement, s0 string collate "c0", s1 string collate "c1", s2 string collate "c2", s5 string collate "c5")]]) +box.sql.execute([[insert into tc values (null, 'a', 'a', 'a', 'a')]]) +box.sql.execute([[insert into tc values (null, 'A', 'A', 'A', 'A')]]) +box.sql.execute([[insert into tc values (null, 'á', 'á', 'á', 'á')]]) +box.sql.execute([[insert into tc values (null, 'â', 'â', 'â', 'â')]]) +-- Should be same as the next one. +box.sql.execute([[select * from tc where s0 = 'a']]) +-- Should return all records. +box.sql.execute([[select * from tc where s1 = 'a']]) +-- Should return two records. +box.sql.execute([[select * from tc where s2 = 'a']]) +-- Should return one record. +box.sql.execute([[select * from tc where s5 = 'a']]) + +box.sql.execute([[drop table tc]]) +box.internal.collation.drop('c0') +box.internal.collation.drop('c1') +box.internal.collation.drop('c2') +box.internal.collation.drop('c5') diff --git a/test/unit/coll.cpp b/test/unit/coll.cpp index 94374a7..acbc7f4 100644 --- a/test/unit/coll.cpp +++ b/test/unit/coll.cpp @@ -51,6 +51,7 @@ manual_test()      memset(&def, 0, sizeof(def));      snprintf(def.locale, sizeof(def.locale), "%s", "ru_RU");      def.type = COLL_TYPE_ICU; +    def.icu.strength = COLL_ICU_STRENGTH_IDENTICAL;      struct coll *coll;      cout << " -- default ru_RU -- " << endl; @@ -134,6 +135,7 @@ hash_test()      struct coll *coll;      /* Case sensitive */ +    def.icu.strength = COLL_ICU_STRENGTH_IDENTICAL;      coll = coll_new(&def);      assert(coll != NULL);      cout << "Case sensitive" << endl;