[tarantool-patches] Re: [PATCH v1 1/1] sql: UPPER and LOWER support COLLATE

  • From: Vladislav Shpilevoy <v.shpilevoy@xxxxxxxxxxxxx>
  • To: tarantool-patches@xxxxxxxxxxxxx
  • Date: Mon, 30 Jul 2018 13:35:35 +0300

Hi! Thanks for the patch!

Please, write me in 'Copy' when sending a patch on review.

See 3 comments below.

On 27/07/2018 18:40, imeevma@xxxxxxxxxxxxx wrote:

SQL functions UPPER and LOWER now works
with COLLATE as they should according to
ANSI Standart.

Closes #3052.
---
Branch: 
https://github.com/tarantool/tarantool/tree/imeevma/gh-3052-collate-for-upper-lower
Issue: https://github.com/tarantool/tarantool/issues/3052

  src/box/sql/func.c          | 18 ++++++++++-----
  test/sql/collation.result   | 53 +++++++++++++++++++++++++++++++++++++++++++++
  test/sql/collation.test.lua | 20 +++++++++++++++++
  3 files changed, 86 insertions(+), 5 deletions(-)

diff --git a/src/box/sql/func.c b/src/box/sql/func.c
index e211de1..637121c 100644
--- a/src/box/sql/func.c
+++ b/src/box/sql/func.c
@@ -503,6 +504,14 @@ case_type##ICUFunc(sqlite3_context *context, int argc, 
sqlite3_value **argv)   \
                return;                                                        \
        }                                                                      \
        UErrorCode status = U_ZERO_ERROR;                                      \
+       struct coll *pColl = sqlite3GetFuncCollSeq(context);                   \

1. Please, use Tarantool style for new code.

+       const char *locale = NULL;                                             \
+       if (pColl != NULL) {                                                   \
+               locale = ucol_getLocaleByType(pColl->collator,                 \
+                                             ULOC_VALID_LOCALE, &status);     \
+       }                                                                      \
+       UCaseMap *pUCaseMap = ucasemap_open(locale, 0, &status);               \
+       assert(pUCaseMap);                                                     \
        int len = ucasemap_utf8To##case_type(pUCaseMap, z1, n, z2, n, &status);\
        if (len > n) {                                                         \
                status = U_ZERO_ERROR;                                         \
diff --git a/test/sql/collation.result b/test/sql/collation.result
index 7fec96d..e057bb6 100644
--- a/test/sql/collation.result
+++ b/test/sql/collation.result
@@ -32,6 +32,59 @@ box.sql.execute("SELECT 1 LIMIT 1 COLLATE BINARY, 1;")
  ---
  - error: 'near "COLLATE": syntax error'
  ...
+-- gh-3052: sql: upper/lower support only default locale

2. I do not think that in /test/sql directory it is necessary to
specify 'sql:' subsystem.

+-- For tr-TR result depends on collation
+box.internal.collation.create('TURKISH', 'ICU', 'tr-TR', {strength='primary'});
+---
+...
+box.sql.execute([[CREATE TABLE tu (descriptor CHAR(50) PRIMARY KEY, letter 
CHAR)]]);
+---
+...
+box.sql.execute([[INSERT INTO tu VALUES ('Latin Capital Letter I 
U+0049','I');]])
+---
+...
+box.sql.execute([[INSERT INTO tu VALUES ('Latin Small Letter I U+0069','i');]])
+---
+...
+box.sql.execute([[INSERT INTO tu VALUES ('Latin Capital Letter I With Dot 
Above U+0130','İ');]])
+---
+...
+box.sql.execute([[INSERT INTO tu VALUES ('Latin Small Letter Dotless I 
U+0131','ı');]])
+---
+...
+-- Without collation
+box.sql.execute([[SELECT descriptor, upper(letter) AS upper,lower(letter) AS 
lower FROM tu;]])
+---
+- - ['Latin Capital Letter I U+0049', 'I', 'i']
+  - ['Latin Capital Letter I With Dot Above U+0130', 'İ', 'i̇']
+  - ['Latin Small Letter Dotless I U+0131', 'I', 'ı']
+  - ['Latin Small Letter I U+0069', 'I', 'i']
+...
+-- With collation
+box.sql.execute([[SELECT descriptor, upper(letter COLLATE "TURKISH") AS 
upper,lower(letter COLLATE "TURKISH") AS lower FROM tu;]])
+---
+- - ['Latin Capital Letter I U+0049', 'I', 'ı']
+  - ['Latin Capital Letter I With Dot Above U+0130', 'İ', 'i']
+  - ['Latin Small Letter Dotless I U+0131', 'I', 'ı']
+  - ['Latin Small Letter I U+0069', 'İ', 'i']
+...
+-- For de-DE result is actually the same
+box.internal.collation.create('GERMAN', 'ICU', 'de-DE', {strength='primary'});
+---
+...
+box.sql.execute([[INSERT INTO tu VALUES ('German Small Letter Sharp S 
U+00DF','ß');]])
+---
+...
+-- Without collation
+box.sql.execute([[SELECT descriptor, upper(letter), letter FROM tu where 
UPPER(letter) = 'SS';]])
+---
+- - ['German Small Letter Sharp S U+00DF', 'SS', 'ß']
+...
+-- With collation
+box.sql.execute([[SELECT descriptor, upper(letter COLLATE "GERMAN"), letter FROM tu 
where UPPER(letter COLLATE "GERMAN") = 'SS';]])
+---
+- - ['German Small Letter Sharp S U+00DF', 'SS', 'ß']
+...
  box.schema.user.grant('guest','read,write,execute', 'universe')
  ---
  ...

3. Cleanup the test: drop tables, collations.

Other related posts: