UNINDEXED column constraint #1279

fnc12 · 2024-05-04T16:35:01Z

This PR adds UNINDEXED column constraint which is used in FTS virtual table.
To improve: add static assertion in table_t to ensure that its columns do not contain unindexed_t.

fnc12 · 2024-05-04T16:36:07Z

dev/statement_serializer.h

                    ss << " " << type_printer<field_type_t<column_type>>().print();
                    ss << streaming_column_constraints(
                        call_as_template_base<column_constraints>(polyfill::identity{})(column),
                        column.is_not_null(),
                        context);
+                } else {
+                    using constraints_tuple = typename column_type::constraints_type;


@trueqbit probably this algorithm can be implemented way more beautiful. Do you have any idea?

Yes, indeed ;)

template<class G, class S, class... Op> struct statement_serializer<column_t<G, S, Op...>, void> { using statement_type = column_t<G, S, Op...>; template<class Ctx> std::string operator()(const statement_type& column, const Ctx& context) const { std::stringstream ss; ss << streaming_identifier(column.name); if(!context.fts5_columns) { ss << " " << type_printer<field_type_t<column_field<G, S>>>().print(); } ss << streaming_column_constraints( call_as_template_base<column_constraints>(polyfill::identity{})(column), column.is_not_null(), context); return ss.str(); } };

And the accompanying streaming_column_constraints in serializing_util.h:

template<class... Op, class Ctx> std::ostream& operator<<(std::ostream& ss, std::tuple<const streaming<stream_as::column_constraints>&, const column_constraints<Op...>&, const bool&, Ctx> tpl) { const auto& column = std::get<1>(tpl); const bool& isNotNull = std::get<2>(tpl); auto& context = std::get<3>(tpl); using constraints_tuple = decltype(column.constraints); if(!context.fts5_columns) { iterate_tuple(column.constraints, [&ss, &context](auto& constraint) { ss << ' ' << serialize(constraint, context); }); constexpr bool hasExplicitNullableConstraint = mpl::invoke_t<mpl::disjunction<check_if_has_type<null_t>, check_if_has_type<not_null_t>>, constraints_tuple>::value; if SQLITE_ORM_CONSTEXPR_IF(!hasExplicitNullableConstraint) { if(isNotNull) { ss << " NOT NULL"; } else { ss << " NULL"; } } } else { constexpr bool hasUnindexedOption = mpl::invoke_t<check_if_has_type<unindexed_t>, constraints_tuple>::value; if SQLITE_ORM_CONSTEXPR_IF(hasUnindexedOption) { ss << " UNINDEXED"; } } return ss; }

fnc12 · 2024-05-04T16:36:21Z

tests/schema/virtual_table.cpp

@@ -30,8 +30,6 @@ TEST_CASE("virtual table") {
        Post{"SQLite Tutorial", "Help you learn SQLite quickly and effectively"},
    };

-    //    storage.remove_all<Post>();


removed old artifact

fnc12 · 2024-05-04T16:36:55Z

tests/statement_serializer_tests/schema/table.cpp

+
+using namespace sqlite_orm;
+
+TEST_CASE("statement_serializer table_t") {


this test case is not related to PRs topic but I wanted to add this last time in the next PR so here we are

fnc12 · 2024-05-04T16:37:25Z

tests/storage_tests.cpp

@@ -268,3 +268,184 @@ TEST_CASE("non-unique DBOs") {
                                      make_column("time", &Record::setTime, &Record::time)));
    db.sync_schema();
 }
+
+TEST_CASE("insert") {


I moved all remaining tests from tests2.cpp here cause they are related to storage_t

trueqbit · 2024-05-04T17:58:27Z

dev/statement_serializer.h

+            using statement_type = unindexed_t;
+
+            template<class Ctx>
+            std::string operator()(const statement_type& c, const Ctx& context) const {


Should return serialize_result_type.

trueqbit · 2024-05-04T18:00:33Z

tests/statement_serializer_tests/schema/table.cpp

+        using context_t = internal::serializer_context<db_objects_t>;
+        context_t context{dbObjects};
+        value = internal::serialize(table, context);
+        expected = "CREATE TABLE \"users\" (\"id\" INTEGER NOT NULL, \"name\" TEXT NOT NULL)";


Let's use raw string literals, please!

trueqbit · 2024-05-04T18:00:39Z

tests/statement_serializer_tests/schema/table.cpp

+        using context_t = internal::serializer_context<db_objects_t>;
+        context_t context{dbObjects};
+        value = internal::serialize(table, context);
+        expected = "CREATE TABLE \"users\" (\"id\" INTEGER NOT NULL, \"name\" TEXT NOT NULL) WITHOUT ROWID";


Let's use raw string literals, please!

trueqbit · 2024-05-04T18:00:51Z

tests/statement_serializer_tests/schema/using_fts5.cpp

+    SECTION("simple") {
+        auto node = using_fts5(make_column("title", &Post::title), make_column("body", &Post::body));
+        value = serialize(node, context);
+        expected = "USING FTS5(\"title\", \"body\")";


Let's use raw string literals, please!

trueqbit · 2024-05-04T18:00:57Z

tests/statement_serializer_tests/schema/using_fts5.cpp

+    SECTION("unindexed") {
+        auto node = using_fts5(make_column("title", &Post::title), make_column("body", &Post::body, unindexed()));
+        value = serialize(node, context);
+        expected = "USING FTS5(\"title\", \"body\" UNINDEXED)";


Let's use raw string literals, please!

trueqbit · 2024-05-04T18:07:47Z

tests/CMakeLists.txt

@@ -41,7 +41,6 @@ add_executable(unit_tests
    xdestroy_handling.cpp
    sync_schema_tests.cpp
    tests.cpp
-    tests2.cpp


yeah. I can tell you why all ungrouped tests were divided up to several files. Cause Windows CI couldn't handle such a huge .cpp file to compile 🤦

trueqbit · 2024-05-04T20:36:41Z

dev/serializer_context.h

@@ -8,7 +8,7 @@ namespace sqlite_orm {
            bool replace_bindable_with_question = false;
            bool skip_table_name = true;
            bool use_parentheses = true;
-            bool skip_types_and_constraints = false;
+            bool skip_types_and_constraints_except_unindexed = false;


Can we rename this flag into something like fts5_columns? Also see below when checking it.

trueqbit · 2024-05-04T20:36:49Z

dev/statement_serializer.h

                    ss << " " << type_printer<field_type_t<column_type>>().print();
                    ss << streaming_column_constraints(
                        call_as_template_base<column_constraints>(polyfill::identity{})(column),
                        column.is_not_null(),
                        context);
+                } else {
+                    using constraints_tuple = typename column_type::constraints_type;


Yes, indeed ;)

template<class G, class S, class... Op> struct statement_serializer<column_t<G, S, Op...>, void> { using statement_type = column_t<G, S, Op...>; template<class Ctx> std::string operator()(const statement_type& column, const Ctx& context) const { std::stringstream ss; ss << streaming_identifier(column.name); if(!context.fts5_columns) { ss << " " << type_printer<field_type_t<column_field<G, S>>>().print(); } ss << streaming_column_constraints( call_as_template_base<column_constraints>(polyfill::identity{})(column), column.is_not_null(), context); return ss.str(); } };

And the accompanying streaming_column_constraints in serializing_util.h:

template<class... Op, class Ctx> std::ostream& operator<<(std::ostream& ss, std::tuple<const streaming<stream_as::column_constraints>&, const column_constraints<Op...>&, const bool&, Ctx> tpl) { const auto& column = std::get<1>(tpl); const bool& isNotNull = std::get<2>(tpl); auto& context = std::get<3>(tpl); using constraints_tuple = decltype(column.constraints); if(!context.fts5_columns) { iterate_tuple(column.constraints, [&ss, &context](auto& constraint) { ss << ' ' << serialize(constraint, context); }); constexpr bool hasExplicitNullableConstraint = mpl::invoke_t<mpl::disjunction<check_if_has_type<null_t>, check_if_has_type<not_null_t>>, constraints_tuple>::value; if SQLITE_ORM_CONSTEXPR_IF(!hasExplicitNullableConstraint) { if(isNotNull) { ss << " NOT NULL"; } else { ss << " NULL"; } } } else { constexpr bool hasUnindexedOption = mpl::invoke_t<check_if_has_type<unindexed_t>, constraints_tuple>::value; if SQLITE_ORM_CONSTEXPR_IF(hasUnindexedOption) { ss << " UNINDEXED"; } } return ss; }

added unindexed

df91a2b

fnc12 requested a review from trueqbit May 4, 2024 16:35

fnc12 commented May 4, 2024

View reviewed changes

fnc12 mentioned this pull request May 4, 2024

FTS5 Options: Unindexed columns, prefix indexes, tokenizers & content #1277

Closed

trueqbit requested changes May 4, 2024

View reviewed changes

improved algo

3f6a287

fnc12 requested a review from trueqbit May 5, 2024 06:30

minor fixes

467037f

trueqbit approved these changes May 5, 2024

View reviewed changes

fnc12 merged commit b8b5bfb into dev May 5, 2024
3 checks passed

trueqbit deleted the feature/unindexed branch May 5, 2024 19:28

trueqbit linked an issue May 8, 2024 that may be closed by this pull request

FTS5 Options: Unindexed columns, prefix indexes, tokenizers & content #1277

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

UNINDEXED column constraint #1279

UNINDEXED column constraint #1279

fnc12 commented May 4, 2024 •

edited

Loading

fnc12 May 4, 2024

trueqbit May 4, 2024

fnc12 May 5, 2024

fnc12 May 4, 2024

fnc12 May 4, 2024

fnc12 May 4, 2024

trueqbit May 4, 2024

fnc12 May 5, 2024

trueqbit May 4, 2024

fnc12 May 5, 2024

trueqbit May 4, 2024

fnc12 May 5, 2024

trueqbit May 4, 2024

fnc12 May 5, 2024

trueqbit May 4, 2024

fnc12 May 5, 2024

trueqbit May 4, 2024

fnc12 May 5, 2024

trueqbit May 4, 2024

fnc12 May 5, 2024

trueqbit May 4, 2024


		using namespace sqlite_orm;

		TEST_CASE("statement_serializer table_t") {

UNINDEXED column constraint #1279

UNINDEXED column constraint #1279

Conversation

fnc12 commented May 4, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fnc12 commented May 4, 2024 •

edited

Loading