schema_tables.cc

/*
 * Modified by ScyllaDB
 * Copyright (C) 2015-present ScyllaDB
 */

/*
 * SPDX-License-Identifier: (AGPL-3.0-or-later and Apache-2.0)
 */

#include "db/schema_tables.hh"

#include "service/migration_manager.hh"
#include "service/storage_proxy.hh"
#include "gms/feature_service.hh"
#include "partition_slice_builder.hh"
#include "dht/i_partitioner.hh"
#include "system_keyspace.hh"
#include "query_context.hh"
#include "query-result-set.hh"
#include "query-result-writer.hh"
#include "schema/schema_builder.hh"
#include "map_difference.hh"
#include "utils/UUID_gen.hh"
#include <seastar/coroutine/all.hh>
#include "log.hh"
#include "frozen_schema.hh"
#include "schema/schema_registry.hh"
#include "mutation_query.hh"
#include "system_keyspace.hh"
#include "system_distributed_keyspace.hh"
#include "cql3/cql3_type.hh"
#include "cql3/functions/functions.hh"
#include "cql3/functions/user_function.hh"
#include "cql3/functions/user_aggregate.hh"
#include "cql3/expr/evaluate.hh"
#include "cql3/expr/expr-utils.hh"
#include "cql3/util.hh"
#include "types/list.hh"
#include "types/set.hh"

#include "db/marshal/type_parser.hh"
#include "db/config.hh"
#include "db/extensions.hh"
#include "utils/hashers.hh"

#include <seastar/util/noncopyable_function.hh>
#include <seastar/rpc/rpc_types.hh>
#include <seastar/core/coroutine.hh>
#include <seastar/coroutine/parallel_for_each.hh>
#include <seastar/core/loop.hh>

#include <boost/algorithm/string/predicate.hpp>
#include <boost/range/algorithm/copy.hpp>
#include <boost/range/algorithm/transform.hpp>
#include <boost/range/adaptor/indirected.hpp>
#include <boost/range/adaptor/map.hpp>
#include <boost/range/join.hpp>

#include "compaction/compaction_strategy.hh"
#include "view_info.hh"
#include "cql_type_parser.hh"
#include "db/timeout_clock.hh"
#include "replica/database.hh"
#include "data_dictionary/user_types_metadata.hh"

#include "index/target_parser.hh"
#include "lang/lua.hh"

#include "db/query_context.hh"
#include "idl/mutation.dist.hh"
#include "idl/mutation.dist.impl.hh"
#include "db/system_keyspace.hh"
#include "cql3/untyped_result_set.hh"
#include "cql3/functions/user_aggregate.hh"

#include "cql3/CqlParser.hpp"
#include "cql3/expr/expression.hh"
#include "cql3/column_identifier.hh"
#include "cql3/column_specification.hh"
#include "types/types.hh"

using namespace db;
using namespace std::chrono_literals;


static logging::logger diff_logger("schema_diff");


/** system.schema_* tables used to store keyspace/table/type attributes prior to C* 3.0 */
namespace db {
namespace {
    const auto set_null_sharder = schema_builder::register_static_configurator([](const sstring& ks_name, const sstring& cf_name, schema_static_props& props) {
        if (ks_name == schema_tables::NAME) {
            props.use_null_sharder = true;
        }
    });
    const auto set_use_schema_commitlog = schema_builder::register_static_configurator([](const sstring& ks_name, const sstring& cf_name, schema_static_props& props) {
        if (ks_name == schema_tables::NAME) {
            props.use_schema_commitlog = true;
            props.load_phase = system_table_load_phase::phase2;
        }
    });
}

schema_ctxt::schema_ctxt(const db::config& cfg, std::shared_ptr<data_dictionary::user_types_storage> uts,
                         const gms::feature_service& features, replica::database* db)
    : _db(db)
    , _features(features)
    , _extensions(cfg.extensions())
    , _murmur3_partitioner_ignore_msb_bits(cfg.murmur3_partitioner_ignore_msb_bits())
    , _schema_registry_grace_period(cfg.schema_registry_grace_period())
    , _user_types(std::move(uts))
{}

schema_ctxt::schema_ctxt(replica::database& db)
    : schema_ctxt(db.get_config(), db.as_user_types_storage(), db.features(), &db)
{}

schema_ctxt::schema_ctxt(distributed<replica::database>& db)
    : schema_ctxt(db.local())
{}

schema_ctxt::schema_ctxt(distributed<service::storage_proxy>& proxy)
    : schema_ctxt(proxy.local().get_db())
{}

namespace schema_tables {

logging::logger slogger("schema_tables");

const sstring version = "3";

struct qualified_name {
    sstring keyspace_name;
    sstring table_name;

    qualified_name(sstring keyspace_name, sstring table_name)
            : keyspace_name(std::move(keyspace_name))
            , table_name(std::move(table_name))
    { }

    qualified_name(const schema_ptr& s)
            : keyspace_name(s->ks_name())
            , table_name(s->cf_name())
    { }

    auto operator<=>(const qualified_name&) const = default;
};

static future<schema_mutations> read_table_mutations(distributed<service::storage_proxy>& proxy, const qualified_name& table, schema_ptr s);

static future<> merge_tables_and_views(distributed<service::storage_proxy>& proxy,
    sharded<db::system_keyspace>& sys_ks,
    std::map<table_id, schema_mutations>&& tables_before,
    std::map<table_id, schema_mutations>&& tables_after,
    std::map<table_id, schema_mutations>&& views_before,
    std::map<table_id, schema_mutations>&& views_after,
    bool reload);

struct [[nodiscard]] user_types_to_drop final {
    seastar::noncopyable_function<future<> ()> drop;
};

static future<user_types_to_drop> merge_types(distributed<service::storage_proxy>& proxy,
    schema_result before,
    schema_result after);

static future<> merge_functions(distributed<service::storage_proxy>& proxy, schema_result before, schema_result after);
static future<> merge_aggregates(distributed<service::storage_proxy>& proxy, schema_result before, schema_result after, schema_result scylla_before, schema_result scylla_after);

static future<> do_merge_schema(distributed<service::storage_proxy>&, sharded<db::system_keyspace>& sys_ks, std::vector<mutation>, bool do_flush, bool reload);

using computed_columns_map = std::unordered_map<bytes, column_computation_ptr>;
static computed_columns_map get_computed_columns(const schema_mutations& sm);

static std::vector<column_definition> create_columns_from_column_rows(
                const schema_ctxt& ctxt,
                const query::result_set& rows, const sstring& keyspace,
                const sstring& table, bool is_super, column_view_virtual is_view_virtual, const computed_columns_map& computed_columns);


static std::vector<index_metadata> create_indices_from_index_rows(const query::result_set& rows,
                                const sstring& keyspace,
                                const sstring& table);

static index_metadata create_index_from_index_row(const query::result_set_row& row,
                     sstring keyspace,
                     sstring table);

static void add_column_to_schema_mutation(schema_ptr, const column_definition&,
                api::timestamp_type, mutation&);

static void add_computed_column_to_schema_mutation(schema_ptr, const column_definition&,
                api::timestamp_type, mutation&);

static void add_index_to_schema_mutation(schema_ptr table,
                const index_metadata& index, api::timestamp_type timestamp,
                mutation& mutation);

static void drop_column_from_schema_mutation(schema_ptr schema_table, schema_ptr table,
                const sstring& column_name, long timestamp,
                std::vector<mutation>&);

static void drop_index_from_schema_mutation(schema_ptr table,
                const index_metadata& column, long timestamp,
                std::vector<mutation>& mutations);

static future<schema_ptr> create_table_from_table_row(
                distributed<service::storage_proxy>&,
                const query::result_set_row&);

static void prepare_builder_from_table_row(const schema_ctxt&, schema_builder&, const query::result_set_row&);

using namespace v3;

using days = std::chrono::duration<int, std::ratio<24 * 3600>>;

future<> save_system_schema(cql3::query_processor& qp, const sstring & ksname) {
    auto ks = qp.db().find_keyspace(ksname);
    auto ksm = ks.metadata();

    // delete old, possibly obsolete entries in schema tables
    co_await coroutine::parallel_for_each(all_table_names(schema_features::full()), [ksm] (sstring cf) -> future<> {
        auto deletion_timestamp = system_keyspace::schema_creation_timestamp() - 1;
        co_await qctx->execute_cql(format("DELETE FROM {}.{} USING TIMESTAMP {} WHERE keyspace_name = ?", NAME, cf,
            deletion_timestamp), ksm->name()).discard_result();
    });
    {
        auto mvec  = make_create_keyspace_mutations(qp.db().features().cluster_schema_features(), ksm, system_keyspace::schema_creation_timestamp(), true);
        co_await qp.proxy().mutate_locally(std::move(mvec), tracing::trace_state_ptr());
    }
}

/** add entries to system_schema.* for the hardcoded system definitions */
future<> save_system_keyspace_schema(cql3::query_processor& qp) {
    return save_system_schema(qp, NAME);
}

namespace v3 {

static constexpr auto schema_gc_grace = std::chrono::duration_cast<std::chrono::seconds>(days(7)).count();

schema_ptr keyspaces() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, KEYSPACES), NAME, KEYSPACES,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {},
        // regular columns
        {
            {"durable_writes", boolean_type},
            {"replication", map_type_impl::get_instance(utf8_type, utf8_type, false)},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "keyspace definitions"
        );
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

schema_ptr scylla_keyspaces() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, SCYLLA_KEYSPACES), NAME, SCYLLA_KEYSPACES,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {},
        // regular columns
        {
            {"storage_type", utf8_type},
            {"storage_options", map_type_impl::get_instance(utf8_type, utf8_type, false)},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "scylla-specific information for keyspaces"
        );
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

schema_ptr tables() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, TABLES), NAME, TABLES,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {{"table_name", utf8_type}},
        // regular columns
        {
         {"bloom_filter_fp_chance", double_type},
         {"caching", map_type_impl::get_instance(utf8_type, utf8_type, false)},
         {"comment", utf8_type},
         {"compaction", map_type_impl::get_instance(utf8_type, utf8_type, false)},
         {"compression", map_type_impl::get_instance(utf8_type, utf8_type, false)},
         {"crc_check_chance", double_type},
         {"dclocal_read_repair_chance", double_type},
         {"default_time_to_live", int32_type},
         {"extensions", map_type_impl::get_instance(utf8_type, bytes_type, false)},
         {"flags", set_type_impl::get_instance(utf8_type, false)}, // SUPER, COUNTER, DENSE, COMPOUND
         {"gc_grace_seconds", int32_type},
         {"id", uuid_type},
         {"max_index_interval", int32_type},
         {"memtable_flush_period_in_ms", int32_type},
         {"min_index_interval", int32_type},
         {"read_repair_chance", double_type},
         {"speculative_retry", utf8_type},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "table definitions"
        );
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

// Holds Scylla-specific table metadata.
schema_ptr scylla_tables(schema_features features) {
    static auto make = [] (bool has_cdc_options, bool has_per_table_partitioners) -> schema_ptr {
        auto id = generate_legacy_id(NAME, SCYLLA_TABLES);
        auto sb = schema_builder(NAME, SCYLLA_TABLES, std::make_optional(id))
            .with_column("keyspace_name", utf8_type, column_kind::partition_key)
            .with_column("table_name", utf8_type, column_kind::clustering_key)
            .with_column("version", uuid_type)
            .set_gc_grace_seconds(schema_gc_grace);
        // 0 - false, false
        // 1 - true, false
        // 2 - false, true
        // 3 - true, true
        int offset = 0;
        if (has_cdc_options) {
            sb.with_column("cdc", map_type_impl::get_instance(utf8_type, utf8_type, false));
            ++offset;
        }
        if (has_per_table_partitioners) {
            sb.with_column("partitioner", utf8_type);
            offset += 2;
        }
        sb.with_version(system_keyspace::generate_schema_version(id, offset));
        return sb.build();
    };
    static thread_local schema_ptr schemas[2][2] = { {make(false, false), make(false, true)}, {make(true, false), make(true, true)} };
    return schemas[features.contains(schema_feature::CDC_OPTIONS)][features.contains(schema_feature::PER_TABLE_PARTITIONERS)];
}

// The "columns" table lists the definitions of all columns in all tables
// and views. Its schema needs to be identical to the one in Cassandra because
// it is the API through which drivers inspect the list of columns in a table
// (e.g., cqlsh's "DESCRIBE TABLE" and "DESCRIBE MATERIALIZED VIEW" get their
// information from the columns table).
// The "view_virtual_columns" table is an additional table with exactly the
// same schema (both are created by columns_schema()), but has a separate
// list of "virtual" columns. Those are used in materialized views for keeping
// rows without data alive (see issue #3362). These virtual columns cannot be
// listed in the regular "columns" table, otherwise the "DESCRIBE MATERIALIZED
// VIEW" would list them - while it should only list real, selected, columns.

static schema_ptr columns_schema(const char* columns_table_name) {
    schema_builder builder(generate_legacy_id(NAME, columns_table_name), NAME, columns_table_name,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {{"table_name", utf8_type},{"column_name", utf8_type}},
        // regular columns
        {
         {"clustering_order", utf8_type},
         {"column_name_bytes", bytes_type},
         {"kind", utf8_type},
         {"position", int32_type},
         {"type", utf8_type},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "column definitions"
        );
    builder.set_gc_grace_seconds(schema_gc_grace);
    builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
    return builder.build();
}
schema_ptr columns() {
    static thread_local auto schema = columns_schema(COLUMNS);
    return schema;
}
schema_ptr view_virtual_columns() {
    static thread_local auto schema = columns_schema(VIEW_VIRTUAL_COLUMNS);
    return schema;
}

// Returns table ids of all schema tables which contribute to schema_mutations,
// i.e. those which are used to define schema of a table or a view.
// All such tables have a clustering key whose first column is the table name.
static
const std::unordered_set<table_id>& schema_tables_holding_schema_mutations() {
    static const std::unordered_set<table_id> table_ids = [] {
        std::unordered_set<table_id> ids;
        for (auto&& s : {
                tables(),
                views(),
                columns(),
                view_virtual_columns(),
                computed_columns(),
                dropped_columns(),
                indexes(),
                scylla_tables(),
                db::system_keyspace::legacy::column_families(),
                db::system_keyspace::legacy::columns(),
                db::system_keyspace::legacy::triggers()}) {
            assert(s->clustering_key_size() > 0);
            auto&& first_column_name = s->clustering_column_at(0).name_as_text();
            assert(first_column_name == "table_name"
                || first_column_name == "view_name"
                || first_column_name == "columnfamily_name");
            ids.emplace(s->id());
        }
        return ids;
    }();
    return table_ids;
};

// Computed columns are a special kind of columns. Rather than having their value provided directly
// by the user, they are computed - possibly from other column values. This table stores which columns
// for a given table are computed, and a serialized computation itself. Full column information is stored
// in the `columns` table, this one stores only entries for computed columns, so it will be empty for tables
// without any computed columns defined in the schema. `computation` is a serialized blob and its format
// is defined in column_computation.hh and system_schema docs.
//
static schema_ptr computed_columns_schema(const char* columns_table_name) {
    schema_builder builder(generate_legacy_id(NAME, columns_table_name), NAME, columns_table_name,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {{"table_name", utf8_type}, {"column_name", utf8_type}},
        // regular columns
        {{"computation", bytes_type}},
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "computed columns"
        );
    builder.set_gc_grace_seconds(schema_gc_grace);
    builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
    return builder.build();
}

schema_ptr computed_columns() {
    static thread_local auto schema = computed_columns_schema(COMPUTED_COLUMNS);
    return schema;
}

schema_ptr dropped_columns() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, DROPPED_COLUMNS), NAME, DROPPED_COLUMNS,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {{"table_name", utf8_type},{"column_name", utf8_type}},
        // regular columns
        {
         {"dropped_time", timestamp_type},
         {"type", utf8_type},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "dropped column registry"
        );
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

schema_ptr triggers() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, TRIGGERS), NAME, TRIGGERS,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {{"table_name", utf8_type},{"trigger_name", utf8_type}},
        // regular columns
        {
         {"options", map_type_impl::get_instance(utf8_type, utf8_type, false)},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "trigger definitions"
        );
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

schema_ptr views() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, VIEWS), NAME, VIEWS,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {{"view_name", utf8_type}},
        // regular columns
        {
         {"base_table_id", uuid_type},
         {"base_table_name", utf8_type},
         {"where_clause", utf8_type},
         {"bloom_filter_fp_chance", double_type},
         {"caching", map_type_impl::get_instance(utf8_type, utf8_type, false)},
         {"comment", utf8_type},
         {"compaction", map_type_impl::get_instance(utf8_type, utf8_type, false)},
         {"compression", map_type_impl::get_instance(utf8_type, utf8_type, false)},
         {"crc_check_chance", double_type},
         {"dclocal_read_repair_chance", double_type},
         {"default_time_to_live", int32_type},
         {"extensions", map_type_impl::get_instance(utf8_type, bytes_type, false)},
         {"gc_grace_seconds", int32_type},
         {"id", uuid_type},
         {"include_all_columns", boolean_type},
         {"max_index_interval", int32_type},
         {"memtable_flush_period_in_ms", int32_type},
         {"min_index_interval", int32_type},
         {"read_repair_chance", double_type},
         {"speculative_retry", utf8_type},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "view definitions"
        );
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

schema_ptr indexes() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, INDEXES), NAME, INDEXES,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {{"table_name", utf8_type},{"index_name", utf8_type}},
        // regular columns
        {
         {"kind", utf8_type},
         {"options", map_type_impl::get_instance(utf8_type, utf8_type, false)},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "secondary index definitions"
        );
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

schema_ptr types() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, TYPES), NAME, TYPES,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {{"type_name", utf8_type}},
        // regular columns
        {
         {"field_names", list_type_impl::get_instance(utf8_type, false)},
         {"field_types", list_type_impl::get_instance(utf8_type, false)},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "user defined type definitions"
        );
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

schema_ptr functions() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, FUNCTIONS), NAME, FUNCTIONS,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {{"function_name", utf8_type}, {"argument_types", list_type_impl::get_instance(utf8_type, false)}},
        // regular columns
        {
         {"argument_names", list_type_impl::get_instance(utf8_type, false)},
         {"body", utf8_type},
         {"language", utf8_type},
         {"return_type", utf8_type},
         {"called_on_null_input", boolean_type},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "user defined function definitions"
        );
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

schema_ptr aggregates() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, AGGREGATES), NAME, AGGREGATES,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {{"aggregate_name", utf8_type}, {"argument_types", list_type_impl::get_instance(utf8_type, false)}},
        // regular columns
        {
         {"final_func", utf8_type},
         {"initcond", utf8_type},
         {"return_type", utf8_type},
         {"state_func", utf8_type},
         {"state_type", utf8_type},
        },
        // static columns
        {},
        // regular column name type
        utf8_type,
        // comment
        "user defined aggregate definitions"
        );
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

schema_ptr scylla_aggregates() {
    static thread_local auto schema = [] {
        schema_builder builder(generate_legacy_id(NAME, SCYLLA_AGGREGATES), NAME, SCYLLA_AGGREGATES,
        // partition key
        {{"keyspace_name", utf8_type}},
        // clustering key
        {
            {"aggregate_name", utf8_type}, 
            {"argument_types", list_type_impl::get_instance(utf8_type, false)}
        },
        //regular columns
        {
            {"reduce_func", utf8_type},
            {"state_type", utf8_type},
        },
        //static columns,
        {},
        // regular column name type
        utf8_type,
        // comment
        "scylla-specific information for user defined aggregates"
        );
        
        builder.set_gc_grace_seconds(schema_gc_grace);
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build();
    }();
    return schema;
}

schema_ptr scylla_table_schema_history() {
    static thread_local auto s = [] {
        schema_builder builder(db::system_keyspace::NAME, SCYLLA_TABLE_SCHEMA_HISTORY, generate_legacy_id(db::system_keyspace::NAME, SCYLLA_TABLE_SCHEMA_HISTORY));
        builder.with_column("cf_id", uuid_type, column_kind::partition_key);
        builder.with_column("schema_version", uuid_type, column_kind::clustering_key);
        builder.with_column("column_name", utf8_type, column_kind::clustering_key);
        builder.with_column("clustering_order", utf8_type);
        builder.with_column("column_name_bytes", bytes_type);
        builder.with_column("kind", utf8_type);
        builder.with_column("position", int32_type);
        builder.with_column("type", utf8_type);
        builder.set_comment("Scylla specific table to store a history of column mappings "
            "for each table schema version upon an CREATE TABLE/ALTER TABLE operations");
        builder.with_version(system_keyspace::generate_schema_version(builder.uuid()));
        return builder.build(schema_builder::compact_storage::no);
    }();
    return s;
}

}

#if 0
    public static void truncateSchemaTables()
    {
        for (String table : ALL)
            getSchemaCFS(table).truncateBlocking();
    }

    private static void flushSchemaTables()
    {
        for (String table : ALL)
            SystemKeyspace.forceBlockingFlush(table);
    }
#endif

static
mutation
redact_columns_for_missing_features(mutation m, schema_features features) {
    if (features.contains(schema_feature::CDC_OPTIONS) && features.contains(schema_feature::PER_TABLE_PARTITIONERS)) {
        return m;
    }
    if (m.schema()->cf_name() != SCYLLA_TABLES) {
        return m;
    }
    slogger.debug("adjusting schema_tables mutation due to possible in-progress cluster upgrade");
    // The global schema ptr make sure it will be registered in the schema registry.
    global_schema_ptr redacted_schema{scylla_tables(features)};
    m.upgrade(redacted_schema);
    return m;
}

/**
 * Read schema from system keyspace and calculate MD5 digest of every row, resulting digest
 * will be converted into UUID which would act as content-based version of the schema.
 */
future<table_schema_version> calculate_schema_digest(distributed<service::storage_proxy>& proxy, schema_features features, noncopyable_function<bool(std::string_view)> accept_keyspace)
{
    auto map = [&proxy, features, accept_keyspace = std::move(accept_keyspace)] (sstring table) mutable -> future<std::vector<mutation>> {
        auto& db = proxy.local().get_db();
        auto rs = co_await db::system_keyspace::query_mutations(db, NAME, table);
        auto s = db.local().find_schema(NAME, table);
        std::vector<mutation> mutations;
        for (auto&& p : rs->partitions()) {
            auto mut = p.mut().unfreeze(s);
            auto partition_key = value_cast<sstring>(utf8_type->deserialize(mut.key().get_component(*s, 0)));
            if (!accept_keyspace(partition_key)) {
                continue;
            }
            mut = redact_columns_for_missing_features(std::move(mut), features);
            mutations.emplace_back(std::move(mut));
        }
        co_return mutations;
    };
    auto reduce = [features] (auto& hash, auto&& mutations) {
        for (const mutation& m : mutations) {
            feed_hash_for_schema_digest(hash, m, features);
        }
    };
    auto hash = md5_hasher();
    auto tables = all_table_names(features);
    {
        for (auto& table: tables) {
            auto mutations = co_await map(table);
            if (diff_logger.is_enabled(logging::log_level::trace)) {
                for (const mutation& m : mutations) {
                    md5_hasher h;
                    feed_hash_for_schema_digest(h, m, features);
                    diff_logger.trace("Digest {} for {}, compacted={}", h.finalize(), m, compact_for_schema_digest(m));
                }
            }
            reduce(hash, mutations);
        }
        co_return utils::UUID_gen::get_name_UUID(hash.finalize());
    }
}

future<table_schema_version> calculate_schema_digest(distributed<service::storage_proxy>& proxy, schema_features features)
{
    return calculate_schema_digest(proxy, features, std::not_fn(&is_system_keyspace));
}

future<std::vector<canonical_mutation>> convert_schema_to_mutations(distributed<service::storage_proxy>& proxy, schema_features features)
{
    auto map = [&proxy, features] (sstring table) -> future<std::vector<canonical_mutation>> {
        auto& db = proxy.local().get_db();
        auto rs = co_await db::system_keyspace::query_mutations(db, NAME, table);
        auto s = db.local().find_schema(NAME, table);
        std::vector<canonical_mutation> results;
        for (auto&& p : rs->partitions()) {
            auto mut = p.mut().unfreeze(s);
            auto partition_key = value_cast<sstring>(utf8_type->deserialize(mut.key().get_component(*s, 0)));
            if (is_system_keyspace(partition_key)) {
                continue;
            }
            mut = redact_columns_for_missing_features(std::move(mut), features);
            results.emplace_back(mut);
        }
        co_return results;
    };
    auto reduce = [] (auto&& result, auto&& mutations) {
        std::move(mutations.begin(), mutations.end(), std::back_inserter(result));
        return std::move(result);
    };
    co_return co_await map_reduce(all_table_names(features), map, std::vector<canonical_mutation>{}, reduce);
}

std::vector<mutation>
adjust_schema_for_schema_features(std::vector<mutation> schema, schema_features features) {
    //Don't send the `computed_columns` table mutations to nodes that doesn't know it.
    if (!features.contains(schema_feature::COMPUTED_COLUMNS)) {
        schema.erase(std::remove_if(schema.begin(), schema.end(), [] (const mutation& m) {
            return m.schema()->cf_name() == COMPUTED_COLUMNS;
        }) , schema.end());
    }

    for (auto& m : schema) {
        m = redact_columns_for_missing_features(m, features);
    }
    return schema;
}

future<schema_result>
read_schema_for_keyspaces(distributed<service::storage_proxy>& proxy, const sstring& schema_table_name, const std::set<sstring>& keyspace_names)
{
    auto map = [&proxy, schema_table_name] (const sstring& keyspace_name) { return read_schema_partition_for_keyspace(proxy, schema_table_name, keyspace_name); };
    auto insert = [] (schema_result&& result, auto&& schema_entity) {
        if (!schema_entity.second->empty()) {
            result.insert(std::move(schema_entity));
        }
        return std::move(result);
    };
    co_return co_await map_reduce(keyspace_names.begin(), keyspace_names.end(), map, schema_result{}, insert);
}

static
future<mutation> query_partition_mutation(service::storage_proxy& proxy,
    schema_ptr s,
    lw_shared_ptr<query::read_command> cmd,
    partition_key pkey)
{
    auto dk = dht::decorate_key(*s, pkey);
    auto range = dht::partition_range::make_singular(dk);
    auto res_hit_rate = co_await proxy.query_mutations_locally(s, std::move(cmd), range, db::no_timeout, tracing::trace_state_ptr{});
    auto&& [res, hit_rate] = res_hit_rate;
    auto&& partitions = res->partitions();
    if (partitions.size() == 0) {
        co_return mutation(s, std::move(dk));
    } else if (partitions.size() == 1) {
        co_return partitions[0].mut().unfreeze(s);
    } else {
        auto&& ex = std::make_exception_ptr(std::invalid_argument("Results must have at most one partition"));
        co_return coroutine::exception(std::move(ex));
    }
}

future<schema_result_value_type>
read_schema_partition_for_keyspace(distributed<service::storage_proxy>& proxy, sstring schema_table_name, sstring keyspace_name)
{
    auto schema = proxy.local().get_db().local().find_schema(NAME, schema_table_name);
    auto keyspace_key = dht::decorate_key(*schema,
        partition_key::from_singular(*schema, keyspace_name));
    auto rs = co_await db::system_keyspace::query(proxy.local().get_db(), NAME, schema_table_name, keyspace_key);
    co_return schema_result_value_type{keyspace_name, std::move(rs)};
}

future<mutation>
read_schema_partition_for_table(distributed<service::storage_proxy>& proxy, schema_ptr schema, const sstring& keyspace_name, const sstring& table_name)
{
    assert(schema_tables_holding_schema_mutations().contains(schema->id()));
    auto keyspace_key = partition_key::from_singular(*schema, keyspace_name);
    auto clustering_range = query::clustering_range(clustering_key_prefix::from_clustering_prefix(
            *schema, exploded_clustering_prefix({utf8_type->decompose(table_name)})));
    auto slice = partition_slice_builder(*schema)
            .with_range(std::move(clustering_range))
            .build();
    auto cmd = make_lw_shared<query::read_command>(schema->id(), schema->version(), std::move(slice), proxy.local().get_max_result_size(slice),
            query::tombstone_limit::max, query::row_limit(query::max_rows));
    co_return co_await query_partition_mutation(proxy.local(), std::move(schema), std::move(cmd), std::move(keyspace_key));
}

future<mutation>
read_keyspace_mutation(distributed<service::storage_proxy>& proxy, const sstring& keyspace_name) {
    schema_ptr s = keyspaces();
    auto key = partition_key::from_singular(*s, keyspace_name);
    auto slice = s->full_slice();
    auto cmd = make_lw_shared<query::read_command>(s->id(), s->version(), std::move(slice), proxy.local().get_max_result_size(slice), query::tombstone_limit::max);
    co_return co_await query_partition_mutation(proxy.local(), std::move(s), std::move(cmd), std::move(key));
}

static thread_local semaphore the_merge_lock {1};

future<> merge_lock() {
    if (slogger.is_enabled(log_level::trace)) {
        slogger.trace("merge_lock at {}", current_backtrace());
    }
    return smp::submit_to(0, [] { return the_merge_lock.wait(); });
}

future<> merge_unlock() {
    if (slogger.is_enabled(log_level::trace)) {
        slogger.trace("merge_unlock at {}", current_backtrace());
    }
    return smp::submit_to(0, [] { the_merge_lock.signal(); });
}

future<semaphore_units<>> hold_merge_lock() noexcept {
    assert(this_shard_id() == 0);

    if (slogger.is_enabled(log_level::trace)) {
        slogger.trace("hold_merge_lock at {}", current_backtrace());
    }
    return get_units(the_merge_lock, 1);
}

static future<> with_merge_lock(noncopyable_function<future<> ()> func) {
    co_await merge_lock();
    std::exception_ptr ep;
    try {
        co_await func();
    } catch (...) {
        ep = std::current_exception();
    }
    co_await merge_unlock();
    if (ep) {
        std::rethrow_exception(std::move(ep));
    }
}

static
future<> update_schema_version_and_announce(sharded<db::system_keyspace>& sys_ks, distributed<service::storage_proxy>& proxy, schema_features features) {
    auto uuid = co_await calculate_schema_digest(proxy, features);
    co_await sys_ks.local().update_schema_version(uuid);
    co_await proxy.local().get_db().invoke_on_all([uuid] (replica::database& db) {
        db.update_version(uuid);
    });
    slogger.info("Schema version changed to {}", uuid);
}

/**
 * Merge remote schema in form of mutations with local and mutate ks/cf metadata objects
 * (which also involves fs operations on add/drop ks/cf)
 *
 * @param mutations the schema changes to apply
 *
 * @throws ConfigurationException If one of metadata attributes has invalid value
 * @throws IOException If data was corrupted during transportation or failed to apply fs operations
 */
future<> merge_schema(sharded<db::system_keyspace>& sys_ks, distributed<service::storage_proxy>& proxy, gms::feature_service& feat, std::vector<mutation> mutations, bool reload)
{
    if (this_shard_id() != 0) {
        // mutations must be applied on the owning shard (0).
        co_await smp::submit_to(0, [&, fmuts = freeze(mutations)] () mutable -> future<> {
            return merge_schema(sys_ks, proxy, feat, unfreeze(fmuts));
        });
        co_return;
    }
    co_await with_merge_lock([&] () mutable -> future<> {
        bool flush_schema = proxy.local().get_db().local().get_config().flush_schema_tables_after_modification();
        co_await do_merge_schema(proxy, sys_ks, std::move(mutations), flush_schema, reload);
        co_await update_schema_version_and_announce(sys_ks, proxy, feat.cluster_schema_features());
    });
}

future<> recalculate_schema_version(sharded<db::system_keyspace>& sys_ks, distributed<service::storage_proxy>& proxy, gms::feature_service& feat) {
    co_await with_merge_lock([&] () -> future<> {
        co_await update_schema_version_and_announce(sys_ks, proxy, feat.cluster_schema_features());
    });
}

// Returns names of live table definitions of given keyspace
future<std::vector<sstring>>
static read_table_names_of_keyspace(distributed<service::storage_proxy>& proxy, const sstring& keyspace_name, schema_ptr schema_table) {
    auto pkey = dht::decorate_key(*schema_table, partition_key::from_singular(*schema_table, keyspace_name));
    auto&& rs = co_await db::system_keyspace::query(proxy.local().get_db(), schema_table->ks_name(), schema_table->cf_name(), pkey);
    co_return boost::copy_range<std::vector<sstring>>(rs->rows() | boost::adaptors::transformed([schema_table] (const query::result_set_row& row) {
        const sstring name = schema_table->clustering_key_columns().begin()->name_as_text();
        return row.get_nonnull<sstring>(name);
    }));
}

static std::optional<table_id> table_id_from_mutations(const schema_mutations& sm) {
    auto table_rs = query::result_set(sm.columnfamilies_mutation());
    if (table_rs.empty()) {
        return std::nullopt;
    }
    query::result_set_row table_row = table_rs.row(0);
    return table_id(table_row.get_nonnull<utils::UUID>("id"));
}

using table_name = sstring;
using keyspace_name = sstring;

enum class table_kind { table, view };

static std::ostream& operator<<(std::ostream& os, table_kind k) {
    switch (k) {
        case table_kind::table: return os << "table";
        case table_kind::view: return os << "view";
    }
    abort();
}

static constexpr std::initializer_list<table_kind> all_table_kinds = {
    table_kind::table,
    table_kind::view
};

static schema_ptr get_table_holder(table_kind k) {
    switch (k) {
        case table_kind::table: return tables();
        case table_kind::view: return views();
    }
    abort();
}

struct table_selector {
    bool all_in_keyspace = false; // If true, selects all existing tables in a keyspace plus what's in "tables";
    std::unordered_map<table_kind, std::unordered_set<sstring>> tables;

    table_selector& operator+=(table_selector&& o) {
        all_in_keyspace |= o.all_in_keyspace;
        for (auto t : all_table_kinds) {
            tables[t].merge(std::move(o.tables[t]));
        }
        return *this;
    }

    void add(table_kind t, sstring name) {
        tables[t].emplace(std::move(name));
    }

    void add(sstring name) {
        for (auto t : all_table_kinds) {
            add(t, name);
        }
    }
};

static
future<std::map<table_id, schema_mutations>>
read_tables_for_keyspaces(distributed<service::storage_proxy>& proxy, const std::set<sstring>& keyspace_names, table_kind kind,
                          const std::unordered_map<sstring, table_selector>& tables_per_keyspace)
{
    std::map<table_id, schema_mutations> result;
    for (auto&& [keyspace_name, sel] : tables_per_keyspace) {
        if (!sel.tables.contains(kind)) {
            continue;
        }
        for (auto&& table_name : sel.tables.find(kind)->second) {
            auto qn = qualified_name(keyspace_name, table_name);
            auto muts = co_await read_table_mutations(proxy, qn, get_table_holder(kind));
            auto id = table_id_from_mutations(muts);
            if (id) {
                result.emplace(std::move(*id), std::move(muts));
            }
        }
    }
    co_return result;
}

mutation compact_for_schema_digest(const mutation& m) {
    // Cassandra is skipping tombstones from digest calculation
    // to avoid disagreements due to tombstone GC.
    // See https://issues.apache.org/jira/browse/CASSANDRA-6862.
    // We achieve similar effect with compact_for_compaction().
    mutation m_compacted(m);
    m_compacted.partition().compact_for_compaction_drop_tombstones_unconditionally(*m.schema(), m.decorated_key());
    return m_compacted;
}

void feed_hash_for_schema_digest(hasher& h, const mutation& m, schema_features features) {
    auto compacted = compact_for_schema_digest(m);
    if (!features.contains<schema_feature::DIGEST_INSENSITIVE_TO_EXPIRY>() || !compacted.partition().empty()) {
        feed_hash(h, compacted);
    }
}

// Applies deletion of the "version" column to a system_schema.scylla_tables mutation.
static void delete_schema_version(mutation& m) {
    if (m.column_family_id() != scylla_tables()->id()) {
        return;
    }
    const column_definition& version_col = *m.schema()->get_column_definition(to_bytes("version"));
    for (auto&& row : m.partition().clustered_rows()) {
        auto&& cells = row.row().cells();
        auto&& cell = cells.find_cell(version_col.id);
        api::timestamp_type t = api::new_timestamp();
        if (cell) {
            t = std::max(t, cell->as_atomic_cell(version_col).timestamp());
        }
        cells.apply(version_col, atomic_cell::make_dead(t, gc_clock::now()));
    }
}

/// Helper function which fills a given mutation with column information
/// provided the corresponding column_definition object.
static void fill_column_info(const schema& table,
                             const clustering_key& ckey,
                             const column_definition& column,
                             api::timestamp_type timestamp,
                             ttl_opt ttl,
                             mutation& m) {
    auto order = "NONE";
    if (column.is_clustering_key()) {
        order = "ASC";
    }
    auto type = column.type;
    if (type->is_reversed()) {
        type = type->underlying_type();
        if (column.is_clustering_key()) {
            order = "DESC";
        }
    }
    int32_t pos = -1;
    if (column.is_primary_key()) {
        pos = table.position(column);
    }

    m.set_clustered_cell(ckey, "column_name_bytes", data_value(column.name()), timestamp, ttl);
    m.set_clustered_cell(ckey, "kind", serialize_kind(column.kind), timestamp, ttl);
    m.set_clustered_cell(ckey, "position", pos, timestamp, ttl);
    m.set_clustered_cell(ckey, "clustering_order", sstring(order), timestamp, ttl);
    m.set_clustered_cell(ckey, "type", type->as_cql3_type().to_string(), timestamp, ttl);
}

future<> store_column_mapping(distributed<service::storage_proxy>& proxy, schema_ptr s, bool with_ttl) {
    // Skip "system*" tables -- only user-related tables are relevant
    if (static_cast<std::string_view>(s->ks_name()).starts_with(db::system_keyspace::NAME)) {
        co_return;
    }
    schema_ptr history_tbl = scylla_table_schema_history();

    // Insert the new column mapping for a given schema version (without TTL)
    std::vector<mutation> muts;
    partition_key pk = partition_key::from_exploded(*history_tbl, {uuid_type->decompose(s->id().uuid())});

    ttl_opt ttl;
    if (with_ttl) {
        ttl = gc_clock::duration(DEFAULT_GC_GRACE_SECONDS);
    }
    // Use one timestamp for all mutations for the ease of debugging
    const auto ts = api::new_timestamp();
    for (const auto& cdef : boost::range::join(s->static_columns(), s->regular_columns())) {
        mutation m(history_tbl, pk);
        auto ckey = clustering_key::from_exploded(*history_tbl, {uuid_type->decompose(s->version().uuid()),
                                                                 utf8_type->decompose(cdef.name_as_text())});
        fill_column_info(*s, ckey, cdef, ts, ttl, m);
        muts.emplace_back(std::move(m));
    }
    co_await proxy.local().mutate_locally(std::move(muts), tracing::trace_state_ptr());
}

// Limit concurrency of user tables to prevent stalls.
// See https://github.com/scylladb/scylladb/issues/11574
// Note: we aim at providing enough concurrency to utilize
// the cpu while operations are blocked on disk I/O
// and or filesystem calls, e.g. fsync.
constexpr size_t max_concurrent = 8;

// Extracts the names of tables affected by a schema mutation.
// The mutation must target one of the tables in schema_tables_holding_schema_mutations().
static
table_selector get_affected_tables(const sstring& keyspace_name, const mutation& m) {
    const schema& s = *m.schema();
    auto get_table_name = [&] (const clustering_key& ck) {
        // The first component of the clustering key in each table listed in
        // schema_tables_holding_schema_mutations contains the table name.
        return value_cast<sstring>(utf8_type->deserialize(ck.get_component(s, 0)));
    };
    table_selector result;
    if (m.partition().partition_tombstone()) {
        slogger.trace("Mutation of {}.{} for keyspace {} contains a partition tombstone",
                      m.schema()->ks_name(), m.schema()->cf_name(), keyspace_name);
        result.all_in_keyspace = true;
    }
    for (auto&& e : m.partition().row_tombstones()) {
        const range_tombstone& rt = e.tombstone();
        if (rt.start.size(s) == 0 || rt.end.size(s) == 0) {
            slogger.trace("Mutation of {}.{} for keyspace {} contains a multi-table range tombstone",
                          m.schema()->ks_name(), m.schema()->cf_name(), keyspace_name);
            result.all_in_keyspace = true;
            break;
        }
        auto table_name = get_table_name(rt.start);
        if (table_name != get_table_name(rt.end)) {
            slogger.trace("Mutation of {}.{} for keyspace {} contains a multi-table range tombstone",
                          m.schema()->ks_name(), m.schema()->cf_name(), keyspace_name);
            result.all_in_keyspace = true;
            break;
        }
        result.add(table_name);
    }
    for (auto&& row : m.partition().clustered_rows()) {
        result.add(get_table_name(row.key()));
    }
    slogger.trace("Mutation of {}.{} for keyspace {} affects tables: {}, all_in_keyspace: {}",
                  m.schema()->ks_name(), m.schema()->cf_name(), keyspace_name, result.tables, result.all_in_keyspace);
    return result;
}

static future<> do_merge_schema(distributed<service::storage_proxy>& proxy, sharded<db::system_keyspace>& sys_ks, std::vector<mutation> mutations, bool do_flush, bool reload)
{
    slogger.trace("do_merge_schema: {}", mutations);
    schema_ptr s = keyspaces();
    // compare before/after schemas of the affected keyspaces only
    std::set<sstring> keyspaces;
    std::set<table_id> column_families;
    std::unordered_map<keyspace_name, table_selector> affected_tables;
    bool has_tablet_mutations = false;
    for (auto&& mutation : mutations) {
        sstring keyspace_name = value_cast<sstring>(utf8_type->deserialize(mutation.key().get_component(*s, 0)));

        if (schema_tables_holding_schema_mutations().contains(mutation.schema()->id())) {
            affected_tables[keyspace_name] += get_affected_tables(keyspace_name, mutation);
        }

        if (mutation.schema()->id() == system_keyspace::tablets()->id()) {
            has_tablet_mutations = true;
        }

        keyspaces.emplace(std::move(keyspace_name));
        column_families.emplace(mutation.column_family_id());
        // We must force recalculation of schema version after the merge, since the resulting
        // schema may be a mix of the old and new schemas.
        delete_schema_version(mutation);
    }

    if (reload) {
        for (auto&& ks : proxy.local().get_db().local().get_non_system_keyspaces()) {
            keyspaces.emplace(ks);
            table_selector sel;
            sel.all_in_keyspace = true;
            affected_tables[ks] = sel;
        }
    }

    // Resolve sel.all_in_keyspace == true to the actual list of tables and views.
    for (auto&& [keyspace_name, sel] : affected_tables) {
        if (sel.all_in_keyspace) {
            // FIXME: Obtain from the database object
            slogger.trace("Reading table list for keyspace {}", keyspace_name);
            for (auto k : all_table_kinds) {
                for (auto&& n : co_await read_table_names_of_keyspace(proxy, keyspace_name, get_table_holder(k))) {
                    sel.add(k, std::move(n));
                }
            }
        }
        slogger.debug("Affected tables for keyspace {}: {}", keyspace_name, sel.tables);
    }

    // current state of the schema
    auto&& old_keyspaces = co_await read_schema_for_keyspaces(proxy, KEYSPACES, keyspaces);
    auto&& old_column_families = co_await read_tables_for_keyspaces(proxy, keyspaces, table_kind::table, affected_tables);
    auto&& old_types = co_await read_schema_for_keyspaces(proxy, TYPES, keyspaces);
    auto&& old_views = co_await read_tables_for_keyspaces(proxy, keyspaces, table_kind::view, affected_tables);
    auto old_functions = co_await read_schema_for_keyspaces(proxy, FUNCTIONS, keyspaces);
    auto old_aggregates = co_await read_schema_for_keyspaces(proxy, AGGREGATES, keyspaces);
    auto old_scylla_aggregates = co_await read_schema_for_keyspaces(proxy, SCYLLA_AGGREGATES, keyspaces);

    if (proxy.local().get_db().local().uses_schema_commitlog()) {
        co_await proxy.local().get_db().local().apply(freeze(mutations), db::no_timeout);
    } else {
        co_await proxy.local().mutate_locally(std::move(mutations), tracing::trace_state_ptr());

        if (do_flush) {
            auto& db = proxy.local().get_db();
            co_await max_concurrent_for_each(column_families, max_concurrent, [&db] (const table_id& id) -> future<> {
                return replica::database::flush_table_on_all_shards(db, id);
            });
        }
    }

    // with new data applied
    auto&& new_keyspaces = co_await read_schema_for_keyspaces(proxy, KEYSPACES, keyspaces);
    auto&& new_column_families = co_await read_tables_for_keyspaces(proxy, keyspaces, table_kind::table, affected_tables);
    auto&& new_types = co_await read_schema_for_keyspaces(proxy, TYPES, keyspaces);
    auto&& new_views = co_await read_tables_for_keyspaces(proxy, keyspaces, table_kind::view, affected_tables);
    auto new_functions = co_await read_schema_for_keyspaces(proxy, FUNCTIONS, keyspaces);
    auto new_aggregates = co_await read_schema_for_keyspaces(proxy, AGGREGATES, keyspaces);
    auto new_scylla_aggregates = co_await read_schema_for_keyspaces(proxy, SCYLLA_AGGREGATES, keyspaces);

    if (has_tablet_mutations) {
        slogger.info("Tablet metadata changed");
        co_await proxy.local().get_db().invoke_on_all([&] (replica::database& db) -> future<> {
            co_await db.get_notifier().update_tablet_metadata();
        });
    }

    std::set<sstring> keyspaces_to_drop = co_await merge_keyspaces(proxy, std::move(old_keyspaces), std::move(new_keyspaces));
    auto types_to_drop = co_await merge_types(proxy, std::move(old_types), std::move(new_types));
    co_await merge_tables_and_views(proxy, sys_ks,
        std::move(old_column_families), std::move(new_column_families),
        std::move(old_views), std::move(new_views), reload);
    co_await merge_functions(proxy, std::move(old_functions), std::move(new_functions));
    co_await merge_aggregates(proxy, std::move(old_aggregates), std::move(new_aggregates), std::move(old_scylla_aggregates), std::move(new_scylla_aggregates));
    co_await types_to_drop.drop();

    auto& sharded_db = proxy.local().get_db();
    // it is safe to drop a keyspace only when all nested ColumnFamilies where deleted
    for (auto keyspace_to_drop : keyspaces_to_drop) {
        co_await replica::database::drop_keyspace_on_all_shards(sharded_db, keyspace_to_drop);
    }
}

future<lw_shared_ptr<query::result_set>> extract_scylla_specific_keyspace_info(distributed<service::storage_proxy>& proxy, const schema_result_value_type& partition) {
    lw_shared_ptr<query::result_set> scylla_specific_rs;
    if (proxy.local().features().cluster_schema_features().contains<schema_feature::SCYLLA_KEYSPACES>()) {
        auto&& rs = partition.second;
        if (rs->empty()) {
            co_await coroutine::return_exception(std::runtime_error("query result has no rows"));
        }
        auto&& row = rs->row(0);
        auto keyspace_name = row.get_nonnull<sstring>("keyspace_name");
        auto keyspace_key = dht::decorate_key(*scylla_keyspaces(), partition_key::from_singular(*scylla_keyspaces(), keyspace_name));
        scylla_specific_rs = co_await db::system_keyspace::query(proxy.local().get_db(), NAME, SCYLLA_KEYSPACES, keyspace_key);
    }
    co_return scylla_specific_rs;
}

future<std::set<sstring>> merge_keyspaces(distributed<service::storage_proxy>& proxy, schema_result&& before, schema_result&& after)
{
    std::vector<schema_result_value_type> created;
    std::vector<sstring> altered;
    std::set<sstring> dropped;

    /*
     * - we don't care about entriesOnlyOnLeft() or entriesInCommon(), because only the changes are of interest to us
     * - of all entriesOnlyOnRight(), we only care about ones that have live columns; it's possible to have a ColumnFamily
     *   there that only has the top-level deletion, if:
     *      a) a pushed DROP KEYSPACE change for a keyspace hadn't ever made it to this node in the first place
     *      b) a pulled dropped keyspace that got dropped before it could find a way to this node
     * - of entriesDiffering(), we don't care about the scenario where both pre and post-values have zero live columns:
     *   that means that a keyspace had been recreated and dropped, and the recreated keyspace had never found a way
     *   to this node
     */
    auto diff = difference(before, after, indirect_equal_to<lw_shared_ptr<query::result_set>>());

    for (auto&& key : diff.entries_only_on_left) {
        slogger.info("Dropping keyspace {}", key);
        dropped.emplace(key);
    }
    for (auto&& key : diff.entries_only_on_right) {
        auto&& value = after[key];
        slogger.info("Creating keyspace {}", key);
        created.emplace_back(schema_result_value_type{key, std::move(value)});
    }
    for (auto&& key : diff.entries_differing) {
        slogger.info("Altering keyspace {}", key);
        altered.emplace_back(key);
    }
    auto& sharded_db = proxy.local().get_db();
    for (auto&& val : created) {
        auto scylla_specific_rs = co_await extract_scylla_specific_keyspace_info(proxy, val);
        auto ksm = create_keyspace_from_schema_partition(val, std::move(scylla_specific_rs));
        co_await replica::database::create_keyspace_on_all_shards(sharded_db, proxy, *ksm);
    }
    for (auto& name : altered) {
        auto v = co_await db::schema_tables::read_schema_partition_for_keyspace(proxy, db::schema_tables::KEYSPACES, name);
        auto scylla_specific_rs = co_await db::schema_tables::extract_scylla_specific_keyspace_info(proxy, v);
        auto tmp_ksm = db::schema_tables::create_keyspace_from_schema_partition(v, scylla_specific_rs);
        co_await replica::database::update_keyspace_on_all_shards(sharded_db, *tmp_ksm);
    }
    co_return dropped;
}

struct schema_diff {
    struct dropped_schema {
        global_schema_ptr schema;
    };

    struct altered_schema {
        global_schema_ptr old_schema;
        global_schema_ptr new_schema;
    };

    std::vector<global_schema_ptr> created;
    std::vector<altered_schema> altered;
    std::vector<dropped_schema> dropped;

    size_t size() const {
        return created.size() + altered.size() + dropped.size();
    }
};

// Which side of the diff this schema is on?
// Helps ensuring that when creating schema for altered views, we match "before"
// version of view to "before" version of base table and "after" to "after"
// respectively.
enum class schema_diff_side {
    left, // old, before
    right, // new, after
};

static schema_diff diff_table_or_view(distributed<service::storage_proxy>& proxy,
    std::map<table_id, schema_mutations>&& before,
    std::map<table_id, schema_mutations>&& after,
    bool reload,
    noncopyable_function<schema_ptr (schema_mutations sm, schema_diff_side)> create_schema)
{
    schema_diff d;
    auto diff = difference(before, after);
    for (auto&& key : diff.entries_only_on_left) {
        auto&& s = proxy.local().get_db().local().find_schema(key);
        slogger.info("Dropping {}.{} id={} version={}", s->ks_name(), s->cf_name(), s->id(), s->version());
        d.dropped.emplace_back(schema_diff::dropped_schema{s});
    }
    for (auto&& key : diff.entries_only_on_right) {
        auto s = create_schema(std::move(after.at(key)), schema_diff_side::right);
        slogger.info("Creating {}.{} id={} version={}", s->ks_name(), s->cf_name(), s->id(), s->version());
        d.created.emplace_back(s);
    }
    for (auto&& key : diff.entries_differing) {
        auto s_before = create_schema(std::move(before.at(key)), schema_diff_side::left);
        auto s = create_schema(std::move(after.at(key)), schema_diff_side::right);
        slogger.info("Altering {}.{} id={} version={}", s->ks_name(), s->cf_name(), s->id(), s->version());
        d.altered.emplace_back(schema_diff::altered_schema{s_before, s});
    }
    if (reload) {
        for (auto&& key: diff.entries_in_common) {
            auto s = create_schema(std::move(after.at(key)), schema_diff_side::right);
            slogger.info("Reloading {}.{} id={} version={}", s->ks_name(), s->cf_name(), s->id(), s->version());
            d.altered.emplace_back(schema_diff::altered_schema {s, s});
        }
    }
    return d;
}

// see the comments for merge_keyspaces()
// Atomically publishes schema changes. In particular, this function ensures
// that when a base schema and a subset of its views are modified together (i.e.,
// upon an alter table or alter type statement), then they are published together
// as well, without any deferring in-between.
static future<> merge_tables_and_views(distributed<service::storage_proxy>& proxy,
    sharded<db::system_keyspace>& sys_ks,
    std::map<table_id, schema_mutations>&& tables_before,
    std::map<table_id, schema_mutations>&& tables_after,
    std::map<table_id, schema_mutations>&& views_before,
    std::map<table_id, schema_mutations>&& views_after,
    bool reload)
{
    auto tables_diff = diff_table_or_view(proxy, std::move(tables_before), std::move(tables_after), reload, [&] (schema_mutations sm, schema_diff_side) {
        return create_table_from_mutations(proxy, std::move(sm));
    });
    auto views_diff = diff_table_or_view(proxy, std::move(views_before), std::move(views_after), reload, [&] (schema_mutations sm, schema_diff_side side) {
        // The view schema mutation should be created with reference to the base table schema because we definitely know it by now.
        // If we don't do it we are leaving a window where write commands to this schema are illegal.
        // There are 3 possibilities:
        // 1. The table was altered - in this case we want the view to correspond to this new table schema.
        // 2. The table was just created - the table is guarantied to be published with the view in that case.
        // 3. The view itself was altered - in that case we already know the base table so we can take it from
        //    the database object.
        view_ptr vp = create_view_from_mutations(proxy, std::move(sm));
        schema_ptr base_schema;
        for (auto&& altered : tables_diff.altered) {
            // Chose the appropriate version of the base table schema: old -> old, new -> new.
            schema_ptr s = side == schema_diff_side::left ? altered.old_schema : altered.new_schema;
            if (s->ks_name() == vp->ks_name() && s->cf_name() == vp->view_info()->base_name() ) {
                base_schema = s;
                break;
            }
        }
        if (!base_schema) {
            for (auto&& s : tables_diff.created) {
                if (s.get()->ks_name() == vp->ks_name() && s.get()->cf_name() == vp->view_info()->base_name() ) {
                    base_schema = s;
                    break;
                }
            }
        }

        if (!base_schema) {
            base_schema = proxy.local().local_db().find_schema(vp->ks_name(), vp->view_info()->base_name());
        }

        // Now when we have a referenced base - just in case we are registering an old view (this can happen in a mixed cluster)
        // lets make it write enabled by updating it's compute columns.
        view_ptr fixed_vp = maybe_fix_legacy_secondary_index_mv_schema(proxy.local().get_db().local(), vp, base_schema, preserve_version::yes);
        if(fixed_vp) {
            vp = fixed_vp;
        }
        vp->view_info()->set_base_info(vp->view_info()->make_base_dependent_view_info(*base_schema));
        return vp;
    });

    // First drop views and *only then* the tables, if interleaved it can lead
    // to a mv not finding its schema when snapshoting since the main table
    // was already dropped (see https://github.com/scylladb/scylla/issues/5614)
    auto& db = proxy.local().get_db();
    co_await max_concurrent_for_each(views_diff.dropped, max_concurrent, [&db, &sys_ks] (schema_diff::dropped_schema& dt) {
        auto& s = *dt.schema.get();
        return replica::database::drop_table_on_all_shards(db, sys_ks, s.ks_name(), s.cf_name());
    });
    co_await max_concurrent_for_each(tables_diff.dropped, max_concurrent, [&db, &sys_ks] (schema_diff::dropped_schema& dt) -> future<> {
        auto& s = *dt.schema.get();
        return replica::database::drop_table_on_all_shards(db, sys_ks, s.ks_name(), s.cf_name());
    });

    co_await db.invoke_on_all([&] (replica::database& db) -> future<> {
        // In order to avoid possible races we first create the tables and only then the views.
        // That way if a view seeks information about its base table it's guarantied to find it.
        co_await max_concurrent_for_each(tables_diff.created, max_concurrent, [&] (global_schema_ptr& gs) -> future<> {
            co_await db.add_column_family_and_make_directory(gs);
        });
        co_await max_concurrent_for_each(views_diff.created, max_concurrent, [&] (global_schema_ptr& gs) -> future<> {
            co_await db.add_column_family_and_make_directory(gs);
        });
        for (auto&& gs : boost::range::join(tables_diff.created, views_diff.created)) {
            db.find_column_family(gs).mark_ready_for_writes();
            co_await coroutine::maybe_yield();
        }
        std::vector<bool> columns_changed;
        columns_changed.reserve(tables_diff.altered.size() + views_diff.altered.size());
        for (auto&& altered : boost::range::join(tables_diff.altered, views_diff.altered)) {
            columns_changed.push_back(db.update_column_family(altered.new_schema));
            co_await coroutine::maybe_yield();
        }

        auto it = columns_changed.begin();
        auto notify = [&] (auto& r, auto&& f) -> future<> {
            co_await max_concurrent_for_each(r, max_concurrent, std::move(f));
        };
        // View drops are notified first, because a table can only be dropped if its views are already deleted
        co_await notify(views_diff.dropped, [&] (auto&& dt) { return db.get_notifier().drop_view(view_ptr(dt.schema)); });
        co_await notify(tables_diff.dropped, [&] (auto&& dt) { return db.get_notifier().drop_column_family(dt.schema); });
        // Table creations are notified first, in case a view is created right after the table
        co_await notify(tables_diff.created, [&] (auto&& gs) { return db.get_notifier().create_column_family(gs); });
        co_await notify(views_diff.created, [&] (auto&& gs) { return db.get_notifier().create_view(view_ptr(gs)); });
        // Table altering is notified first, in case new base columns appear
        co_await notify(tables_diff.altered, [&] (auto&& altered) { return db.get_notifier().update_column_family(altered.new_schema, *it++); });
        co_await notify(views_diff.altered, [&] (auto&& altered) { return db.get_notifier().update_view(view_ptr(altered.new_schema), *it++); });
    });

    // Insert column_mapping into history table for altered and created tables.
    //
    // Entries for new tables are inserted without TTL, which means that the most
    // recent schema version should always be available.
    //
    // For altered tables we both insert a new column mapping without TTL and
    // overwrite the previous version entries with TTL to expire them eventually.
    //
    // Drop column mapping entries for dropped tables since these will not be TTLed automatically
    // and will stay there forever if we don't clean them up manually
    co_await max_concurrent_for_each(tables_diff.created, max_concurrent, [&proxy] (global_schema_ptr& gs) -> future<> {
        co_await store_column_mapping(proxy, gs.get(), false);
    });
    co_await max_concurrent_for_each(tables_diff.altered, max_concurrent, [&proxy] (schema_diff::altered_schema& altered) -> future<> {
        co_await when_all_succeed(
            store_column_mapping(proxy, altered.old_schema.get(), true),
            store_column_mapping(proxy, altered.new_schema.get(), false));
    });
    co_await max_concurrent_for_each(tables_diff.dropped, max_concurrent, [] (schema_diff::dropped_schema& dropped) -> future<> {
        schema_ptr s = dropped.schema.get();
        co_await drop_column_mapping(s->id(), s->version());
    });
}

static std::vector<const query::result_set_row*> collect_rows(const std::set<sstring>& keys, const schema_result& result) {
    std::vector<const query::result_set_row*> ret;
    for (const auto& key : keys) {
        for (const auto& row : result.find(key)->second->rows()) {
            ret.push_back(&row);
        }
    }
    return ret;
}

static std::vector<column_definition> get_primary_key_definition(const schema_ptr& schema) {
    std::vector<column_definition> primary_key;
    for (const auto& column : schema->partition_key_columns()) {
        primary_key.push_back(column);
    }
    for (const auto& column : schema->clustering_key_columns()) {
        primary_key.push_back(column);
    }
    
    return primary_key;
}

static std::vector<bytes> get_primary_key(const std::vector<column_definition>& primary_key, const query::result_set_row* row) {
    std::vector<bytes> key;
    for (const auto& column : primary_key) {
        const data_value *val = row->get_data_value(column.name_as_text());
        key.push_back(val->serialize_nonnull());
    }
    return key;
}

// Build a map from primary keys to rows.
static std::map<std::vector<bytes>, const query::result_set_row*> build_row_map(const query::result_set& result) {
    const std::vector<query::result_set_row>& rows = result.rows();
    auto primary_key = get_primary_key_definition(result.schema());
    std::map<std::vector<bytes>, const query::result_set_row*> ret;
    for (const auto& row: rows) {
        auto key = get_primary_key(primary_key, &row);
        ret.insert(std::pair(std::move(key), &row));
    }
    return ret;
}

struct row_diff {
    std::vector<const query::result_set_row*> altered;
    std::vector<const query::result_set_row*> created;
    std::vector<const query::result_set_row*> dropped;
};

// Compute which rows have been created, dropped or altered.
// A row is identified by its primary key.
// In the output, all entries of a given keyspace are together.
static row_diff diff_rows(const schema_result& before, const schema_result& after) {
    auto diff = difference(before, after, indirect_equal_to<lw_shared_ptr<query::result_set>>());

    // For new or empty keyspaces, just record each row.
    auto dropped = collect_rows(diff.entries_only_on_left, before); // Keyspaces now without rows
    auto created = collect_rows(diff.entries_only_on_right, after); // New keyspaces with rows
    std::vector<const query::result_set_row*> altered;

    for (const auto& key : diff.entries_differing) {
        // For each keyspace that changed, compute the difference of the corresponding result_set to find which rows
        // have changed.
        auto before_rows = build_row_map(*before.find(key)->second);
        auto after_rows = build_row_map(*after.find(key)->second);
        auto diff_row = difference(before_rows, after_rows, indirect_equal_to<const query::result_set_row*>());
        for (const auto& key : diff_row.entries_only_on_left) {
            dropped.push_back(before_rows.find(key)->second);
        }
        for (const auto& key : diff_row.entries_only_on_right) {
            created.push_back(after_rows.find(key)->second);
        }
        for (const auto& key : diff_row.entries_differing) {
            altered.push_back(after_rows.find(key)->second);
        }
    }
    return {std::move(altered), std::move(created), std::move(dropped)};
}

// User-defined aggregate stores its information in two tables: aggregates and scylla_aggregates
// The difference has to be joined to properly create an UDA.
//
// FIXME: Since UDA cannot be altered now, set of differing rows should be empty and those rows are
// ignored in calculating the diff.
struct aggregate_diff {
    std::vector<std::pair<const query::result_set_row*, const query::result_set_row*>> created;
    std::vector<std::pair<const query::result_set_row*, const query::result_set_row*>> dropped;
};

static aggregate_diff diff_aggregates_rows(const schema_result& aggr_before, const schema_result& aggr_after, 
        const schema_result& scylla_aggr_before, const schema_result& scylla_aggr_after) {
    using map = std::map<std::vector<bytes>, const query::result_set_row*>;
    auto aggr_diff = difference(aggr_before, aggr_after, indirect_equal_to<lw_shared_ptr<query::result_set>>());

    std::vector<std::pair<const query::result_set_row*, const query::result_set_row*>> created;
    std::vector<std::pair<const query::result_set_row*, const query::result_set_row*>> dropped;

    // Primary key for `aggregates` and `scylla_aggregates` tables
    auto primary_key = get_primary_key_definition(aggregates());

    // DROPPED
    for (const auto& key : aggr_diff.entries_only_on_left) {
        auto scylla_entry = scylla_aggr_before.find(key);
        auto scylla_aggr_rows = (scylla_entry != scylla_aggr_before.end()) ? build_row_map(*scylla_entry->second) : map();

        for (const auto& row : aggr_before.find(key)->second->rows()) {
            auto pk = get_primary_key(primary_key, &row);
            auto entry = scylla_aggr_rows.find(pk);
            dropped.push_back({&row, (entry != scylla_aggr_rows.end()) ? entry->second : nullptr});
        }
    }
    // CREATED
    for (const auto& key : aggr_diff.entries_only_on_right) {
        auto scylla_entry = scylla_aggr_after.find(key);
        auto scylla_aggr_rows = (scylla_entry != scylla_aggr_after.end()) ? build_row_map(*scylla_entry->second) : map();

        for (const auto& row : aggr_after.find(key)->second->rows()) {
            auto pk = get_primary_key(primary_key, &row);
            auto entry = scylla_aggr_rows.find(pk);
            created.push_back({&row, (entry != scylla_aggr_rows.end()) ? entry->second : nullptr});
        }
    }
    for (const auto& key : aggr_diff.entries_differing) {
        auto aggr_before_rows = build_row_map(*aggr_before.find(key)->second);
        auto aggr_after_rows = build_row_map(*aggr_after.find(key)->second);
        auto diff = difference(aggr_before_rows, aggr_after_rows, indirect_equal_to<const query::result_set_row*>());
        
        auto scylla_entry_before = scylla_aggr_before.find(key);
        auto scylla_aggr_rows_before = (scylla_entry_before != scylla_aggr_before.end()) ? build_row_map(*scylla_entry_before->second) : map();
        auto scylla_entry_after = scylla_aggr_after.find(key);
        auto scylla_aggr_rows_after = (scylla_entry_after != scylla_aggr_after.end()) ? build_row_map(*scylla_entry_after->second) : map();

        for (const auto& k : diff.entries_only_on_left) {
            auto entry = scylla_aggr_rows_before.find(k);
            dropped.push_back({
                aggr_before_rows.find(k)->second, (entry != scylla_aggr_rows_before.end()) ? entry->second : nullptr
            });
        }
        for (const auto& k : diff.entries_only_on_right) {
            auto entry = scylla_aggr_rows_after.find(k);
            created.push_back({
                aggr_after_rows.find(k)->second, (entry != scylla_aggr_rows_after.end()) ? entry->second : nullptr
            });
        }
    }

    return {std::move(created), std::move(dropped)};
}

template<typename V>
static std::vector<V> get_list(const query::result_set_row& row, const sstring& name);

// Create types for a given keyspace. This takes care of topologically sorting user defined types.
template <typename T> static std::vector<user_type> create_types(keyspace_metadata& ks, T&& range) {
    cql_type_parser::raw_builder builder(ks);
    std::unordered_set<bytes> names;
    for (const query::result_set_row& row : range) {
        auto name = row.get_nonnull<sstring>("type_name");
        names.insert(to_bytes(name));
        builder.add(std::move(name), get_list<sstring>(row, "field_names"), get_list<sstring>(row, "field_types"));
    }
    // Add user types that use any of the above types. From the
    // database point of view they haven't changed since the content
    // of system.types is the same for them. The runtime objects in
    // the other hand now point to out of date types, so we need to
    // recreate them.
    for (const auto& p : ks.user_types().get_all_types()) {
        const user_type& t = p.second;
        if (names.contains(t->_name)) {
            continue;
        }
        for (const auto& name : names) {
            if (t->references_user_type(t->_keyspace, name)) {
                std::vector<sstring> field_types;
                for (const data_type& f : t->field_types()) {
                    field_types.push_back(f->as_cql3_type().to_string());
                }
                builder.add(t->get_name_as_string(), t->string_field_names(), std::move(field_types));
            }
        }
    }
    return builder.build();
}

// Given a set of rows that is sorted by keyspace, create types for each keyspace.
// The topological sort in each keyspace is necessary when creating types, since we can only create a type when the
// types it reference have already been created.
static std::vector<user_type> create_types(replica::database& db, const std::vector<const query::result_set_row*>& rows) {
    std::vector<user_type> ret;
    for (auto i = rows.begin(), e = rows.end(); i != e;) {
        const auto &row = *i;
        auto keyspace = row->get_nonnull<sstring>("keyspace_name");
        auto next = std::find_if(i, e, [&keyspace](const query::result_set_row* r) {
            return r->get_nonnull<sstring>("keyspace_name") != keyspace;
        });
        auto ks = db.find_keyspace(keyspace).metadata();
        auto v = create_types(*ks, boost::make_iterator_range(i, next) | boost::adaptors::indirected);
        ret.insert(ret.end(), std::make_move_iterator(v.begin()), std::make_move_iterator(v.end()));
        i = next;
    }
    return ret;
}

// see the comments for merge_keyspaces()
static future<user_types_to_drop> merge_types(distributed<service::storage_proxy>& proxy, schema_result before, schema_result after)
{
    auto diff = diff_rows(before, after);

    // Create and update user types before any tables/views are created that potentially
    // use those types. Similarly, defer dropping until after tables/views that may use
    // some of these user types are dropped.

    co_await proxy.local().get_db().invoke_on_all([&] (replica::database& db) -> future<> {
        for (auto&& user_type : create_types(db, diff.created)) {
            db.find_keyspace(user_type->_keyspace).add_user_type(user_type);
            co_await db.get_notifier().create_user_type(user_type);
        }
        for (auto&& user_type : create_types(db, diff.altered)) {
            db.find_keyspace(user_type->_keyspace).add_user_type(user_type);
            co_await db.get_notifier().update_user_type(user_type);
        }
    });

    co_return user_types_to_drop{[&proxy, before = std::move(before), rows = std::move(diff.dropped)] () mutable -> future<> {
        co_await proxy.local().get_db().invoke_on_all([&] (replica::database& db) -> future<> {
            auto dropped = create_types(db, rows);
            for (auto& user_type : dropped) {
                db.find_keyspace(user_type->_keyspace).remove_user_type(user_type);
                co_await db.get_notifier().drop_user_type(user_type);
            }
        });
    }};
}

static std::vector<data_type> read_arg_types(replica::database& db, const query::result_set_row& row, const sstring& keyspace) {
    std::vector<data_type> arg_types;
    for (const auto& arg : get_list<sstring>(row, "argument_types")) {
        arg_types.push_back(db::cql_type_parser::parse(keyspace, arg, db.user_types()));
    }
    return arg_types;
}

static std::vector<data_value> read_arg_values(const query::result_set_row& row) {
    auto args = get_list<sstring>(row, "argument_types");
    return std::vector<data_value>(args.begin(), args.end());
}

#if 0
    // see the comments for mergeKeyspaces()
    private static void mergeAggregates(Map<DecoratedKey, ColumnFamily> before, Map<DecoratedKey, ColumnFamily> after)
    {
        List<UDAggregate> created = new ArrayList<>();
        List<UDAggregate> altered = new ArrayList<>();
        List<UDAggregate> dropped = new ArrayList<>();

        MapDifference<DecoratedKey, ColumnFamily> diff = Maps.difference(before, after);

        // New keyspace with functions
        for (Map.Entry<DecoratedKey, ColumnFamily> entry : diff.entriesOnlyOnRight().entrySet())
            if (entry.getValue().hasColumns())
                created.addAll(createAggregatesFromAggregatesPartition(new Row(entry.getKey(), entry.getValue())).values());

        for (Map.Entry<DecoratedKey, MapDifference.ValueDifference<ColumnFamily>> entry : diff.entriesDiffering().entrySet())
        {
            ColumnFamily pre = entry.getValue().leftValue();
            ColumnFamily post = entry.getValue().rightValue();

            if (pre.hasColumns() && post.hasColumns())
            {
                MapDifference<ByteBuffer, UDAggregate> delta =
                    Maps.difference(createAggregatesFromAggregatesPartition(new Row(entry.getKey(), pre)),
                                    createAggregatesFromAggregatesPartition(new Row(entry.getKey(), post)));

                dropped.addAll(delta.entriesOnlyOnLeft().values());
                created.addAll(delta.entriesOnlyOnRight().values());
                Iterables.addAll(altered, Iterables.transform(delta.entriesDiffering().values(), new Function<MapDifference.ValueDifference<UDAggregate>, UDAggregate>()
                {
                    public UDAggregate apply(MapDifference.ValueDifference<UDAggregate> pair)
                    {
                        return pair.rightValue();
                    }
                }));
            }
            else if (pre.hasColumns())
            {
                dropped.addAll(createAggregatesFromAggregatesPartition(new Row(entry.getKey(), pre)).values());
            }
            else if (post.hasColumns())
            {
                created.addAll(createAggregatesFromAggregatesPartition(new Row(entry.getKey(), post)).values());
            }
        }

        for (UDAggregate udf : created)
            Schema.instance.addAggregate(udf);
        for (UDAggregate udf : altered)
            Schema.instance.updateAggregate(udf);
        for (UDAggregate udf : dropped)
            Schema.instance.dropAggregate(udf);
    }
#endif

static seastar::future<shared_ptr<cql3::functions::user_function>> create_func(replica::database& db, const query::result_set_row& row) {
    cql3::functions::function_name name{
            row.get_nonnull<sstring>("keyspace_name"), row.get_nonnull<sstring>("function_name")};
    auto arg_types = read_arg_types(db, row, name.keyspace);
    data_type return_type = db::cql_type_parser::parse(name.keyspace, row.get_nonnull<sstring>("return_type"), db.user_types());

    // FIXME: We already computed the bitcode in
    // create_function_statement, but it is not clear how to get it
    // here. In this point in the code we only get what was saved in
    // system_schema.functions, and we don't want to store the bitcode
    // If this was not the replica that the client connected to we do
    // have to produce bitcode in at least one shard. Right now this
    // gets run in each shard.

    auto arg_names = get_list<sstring>(row, "argument_names");
    auto body = row.get_nonnull<sstring>("body");
    auto language = row.get_nonnull<sstring>("language");
    if (language == "lua") {
        lua::runtime_config cfg = lua::make_runtime_config(db.get_config());
        cql3::functions::user_function::context ctx = cql3::functions::user_function::lua_context {
            .bitcode = lua::compile(cfg, arg_names, body),
            .cfg = cfg,
        };

        co_return ::make_shared<cql3::functions::user_function>(std::move(name), std::move(arg_types), std::move(arg_names),
                std::move(body), language, std::move(return_type),
                row.get_nonnull<bool>("called_on_null_input"), std::move(ctx));
    } else if (language == "wasm") {
        wasm::context ctx{qctx->qp().wasm_engine(), name.name, qctx->qp().wasm_instance_cache(), db.get_config().wasm_udf_yield_fuel(), db.get_config().wasm_udf_total_fuel()};
        co_await wasm::precompile(qctx->qp().alien_runner(), ctx, arg_names, body);
        co_return ::make_shared<cql3::functions::user_function>(std::move(name), std::move(arg_types), std::move(arg_names),
                std::move(body), language, std::move(return_type),
                row.get_nonnull<bool>("called_on_null_input"), std::move(ctx));
    } else {
        throw std::runtime_error(format("Unsupported language for UDF: {}", language));
    }
}

static shared_ptr<cql3::functions::user_aggregate> create_aggregate(replica::database& db, const query::result_set_row& row, const query::result_set_row* scylla_row) {
    cql3::functions::function_name name{
            row.get_nonnull<sstring>("keyspace_name"), row.get_nonnull<sstring>("aggregate_name")};
    auto arg_types = read_arg_types(db, row, name.keyspace);
    data_type state_type = db::cql_type_parser::parse(name.keyspace, row.get_nonnull<sstring>("state_type"), db.user_types());
    sstring sfunc = row.get_nonnull<sstring>("state_func");
    auto ffunc = row.get<sstring>("final_func");
    auto initcond_str = row.get<sstring>("initcond");

    std::vector<data_type> acc_types{state_type};
    acc_types.insert(acc_types.end(), arg_types.begin(), arg_types.end());
    auto state_func = dynamic_pointer_cast<cql3::functions::scalar_function>(
            cql3::functions::functions::find(cql3::functions::function_name{name.keyspace, sfunc}, acc_types));
    if (!state_func) {
        throw std::runtime_error(format("State function {} needed by aggregate {} not found", sfunc, name.name));
    }
    if (state_func->return_type() != state_type) {
        throw std::runtime_error(format("State function {} needed by aggregate {} doesn't return state", sfunc, name.name));
    }

    ::shared_ptr<cql3::functions::scalar_function> reduce_func = nullptr;
    if (scylla_row) {
        auto rfunc_name = scylla_row->get<sstring>("reduce_func");
        auto rfunc = cql3::functions::functions::find(cql3::functions::function_name{name.keyspace, rfunc_name.value()}, {state_type, state_type});
        if (!rfunc) {
            throw std::runtime_error(format("Reduce function {} needed by aggregate {} not found", rfunc_name.value(), name.name));
        }
        reduce_func = dynamic_pointer_cast<cql3::functions::scalar_function>(rfunc);
        if (!reduce_func) {
            throw std::runtime_error(format("Reduce function {} needed by aggregate {} is not a scalar function", rfunc_name.value(), name.name));
        }
    }
    
    ::shared_ptr<cql3::functions::scalar_function> final_func = nullptr;
    if (ffunc) {
        final_func = dynamic_pointer_cast<cql3::functions::scalar_function>(
            cql3::functions::functions::find(cql3::functions::function_name{name.keyspace, ffunc.value()}, {state_type}));
        if (!final_func) {
            throw std::runtime_error(format("Final function {} needed by aggregate {} not found", ffunc.value(), name.name));
        }
    }

    bytes_opt initcond = std::nullopt;
    if (initcond_str) {
        auto expr = cql3::util::do_with_parser(*initcond_str, std::mem_fn(&cql3_parser::CqlParser::term));
        auto dummy_ident = ::make_shared<cql3::column_identifier>("", true);
        auto column_spec = make_lw_shared<cql3::column_specification>("", "", dummy_ident, state_type);
        auto raw = cql3::expr::evaluate(prepare_expression(expr, db.as_data_dictionary(), "", nullptr, {column_spec}), cql3::query_options::DEFAULT);
        initcond = std::move(raw).to_bytes_opt();
    }
    return ::make_shared<cql3::functions::user_aggregate>(name, initcond, std::move(state_func), std::move(reduce_func), std::move(final_func));
}

static void drop_cached_func(replica::database& db, const query::result_set_row& row) {
    auto language = row.get_nonnull<sstring>("language");
    if (language == "wasm") {
        cql3::functions::function_name name{
            row.get_nonnull<sstring>("keyspace_name"), row.get_nonnull<sstring>("function_name")};
        auto arg_types = read_arg_types(db, row, name.keyspace);
        qctx->qp().wasm_instance_cache().remove(name, arg_types);
    }
}

static future<> merge_functions(distributed<service::storage_proxy>& proxy, schema_result before, schema_result after) {
    auto diff = diff_rows(before, after);

    co_await proxy.local().get_db().invoke_on_all(coroutine::lambda([&] (replica::database& db) -> future<> {
        for (const auto& val : diff.created) {
            cql3::functions::functions::add_function(co_await create_func(db, *val));
        }
        for (const auto& val : diff.dropped) {
            cql3::functions::function_name name{
                val->get_nonnull<sstring>("keyspace_name"), val->get_nonnull<sstring>("function_name")};
            auto arg_types = read_arg_types(db, *val, name.keyspace);
            drop_cached_func(db, *val);
            cql3::functions::functions::remove_function(name, arg_types);
            co_await db.get_notifier().drop_function(name, arg_types);
        }
        for (const auto& val : diff.altered) {
            drop_cached_func(db, *val);
            cql3::functions::functions::replace_function(co_await create_func(db, *val));
        }
    }));
}

static future<> merge_aggregates(distributed<service::storage_proxy>& proxy, schema_result before, schema_result after, 
        schema_result scylla_before, schema_result scylla_after) {
    auto diff = diff_aggregates_rows(before, after, scylla_before, scylla_after);

    co_await proxy.local().get_db().invoke_on_all([&] (replica::database& db)-> future<> {
        for (const auto& val : diff.created) {
            cql3::functions::functions::add_function(create_aggregate(db, *val.first, val.second));
        }
        for (const auto& val : diff.dropped) {
            cql3::functions::function_name name{
                val.first->get_nonnull<sstring>("keyspace_name"), val.first->get_nonnull<sstring>("aggregate_name")};
            auto arg_types = read_arg_types(db, *val.first, name.keyspace);
            cql3::functions::functions::remove_function(name, arg_types);
            co_await db.get_notifier().drop_aggregate(name, arg_types);
        }
    });
}

template<typename... Args>
void set_cell_or_clustered(mutation& m, const clustering_key & ckey, Args && ...args) {
    m.set_clustered_cell(ckey, std::forward<Args>(args)...);
}

template<typename... Args>
void set_cell_or_clustered(mutation& m, const exploded_clustering_prefix & ckey, Args && ...args) {
    m.set_cell(ckey, std::forward<Args>(args)...);
}

template<typename Map>
static atomic_cell_or_collection
make_map_mutation(const Map& map,
                  const column_definition& column,
                  api::timestamp_type timestamp,
                  noncopyable_function<map_type_impl::native_type::value_type (const typename Map::value_type&)> f)
{
    auto column_type = static_pointer_cast<const map_type_impl>(column.type);
    auto ktyp = column_type->get_keys_type();
    auto vtyp = column_type->get_values_type();

    if (column_type->is_multi_cell()) {
        collection_mutation_description mut;

        for (auto&& entry : map) {
            auto te = f(entry);
            mut.cells.emplace_back(ktyp->decompose(data_value(te.first)), atomic_cell::make_live(*vtyp, timestamp, vtyp->decompose(data_value(te.second)), atomic_cell::collection_member::yes));
        }

        return mut.serialize(*column_type);
    } else {
        map_type_impl::native_type tmp;
        tmp.reserve(map.size());
        std::transform(map.begin(), map.end(), std::inserter(tmp, tmp.end()), std::move(f));
        return atomic_cell::make_live(*column.type, timestamp, column_type->decompose(make_map_value(column_type, std::move(tmp))));
    }
}

template<typename Map>
static atomic_cell_or_collection
make_map_mutation(const Map& map,
                  const column_definition& column,
                  api::timestamp_type timestamp)
{
    return make_map_mutation(map, column, timestamp, [](auto&& p) {
        return std::make_pair(data_value(p.first), data_value(p.second));
    });
}

template<typename K, typename Map>
static void store_map(mutation& m, const K& ckey, const bytes& name, api::timestamp_type timestamp, const Map& map) {
    auto s = m.schema();
    auto column = s->get_column_definition(name);
    assert(column);
    set_cell_or_clustered(m, ckey, *column, make_map_mutation(map, *column, timestamp));
}

/*
 * Keyspace metadata serialization/deserialization.
 */

std::vector<mutation> make_create_keyspace_mutations(schema_features features, lw_shared_ptr<keyspace_metadata> keyspace, api::timestamp_type timestamp, bool with_tables_and_types_and_functions)
{
    std::vector<mutation> mutations;
    schema_ptr s = keyspaces();
    auto pkey = partition_key::from_singular(*s, keyspace->name());
    mutation m(s, pkey);
    auto ckey = clustering_key_prefix::make_empty();
    m.set_cell(ckey, "durable_writes", keyspace->durable_writes(), timestamp);

    auto map = keyspace->strategy_options();
    map["class"] = keyspace->strategy_name();
    store_map(m, ckey, "replication", timestamp, map);

    if (features.contains<schema_feature::SCYLLA_KEYSPACES>()) {
        schema_ptr scylla_keyspaces_s = scylla_keyspaces();
        mutation scylla_m(scylla_keyspaces_s, pkey); // pkey can be reused, it's identical in both tables
        auto& storage_options = keyspace->get_storage_options();
        sstring storage_type(storage_options.type_string());
        auto storage_map = storage_options.to_map();
        if (!storage_map.empty()) {
            scylla_m.set_cell(ckey, "storage_type", storage_type, timestamp);
            store_map(scylla_m, ckey, "storage_options", timestamp, storage_map);
        }
        mutations.emplace_back(std::move(scylla_m));
    }

    mutations.emplace_back(std::move(m));

    if (with_tables_and_types_and_functions) {
        for (const auto& kv : keyspace->user_types().get_all_types()) {
            add_type_to_schema_mutation(kv.second, timestamp, mutations);
        }
        for (auto&& s : keyspace->cf_meta_data() | boost::adaptors::map_values) {
            add_table_or_view_to_schema_mutation(s, timestamp, true, mutations);
        }
    }
    return mutations;
}

std::vector<mutation> make_drop_keyspace_mutations(schema_features features, lw_shared_ptr<keyspace_metadata> keyspace, api::timestamp_type timestamp)
{
    std::vector<mutation> mutations;
    for (auto&& schema_table : all_tables(schema_features::full())) {
        auto pkey = partition_key::from_exploded(*schema_table, {utf8_type->decompose(keyspace->name())});
        mutation m{schema_table, pkey};
        m.partition().apply(tombstone{timestamp, gc_clock::now()});
        mutations.emplace_back(std::move(m));
    }
    if (features.contains<schema_feature::SCYLLA_KEYSPACES>()) {
        auto pkey = partition_key::from_exploded(*scylla_keyspaces(), {utf8_type->decompose(keyspace->name())});
        mutation km{scylla_keyspaces(), pkey};
        km.partition().apply(tombstone{timestamp, gc_clock::now()});
        mutations.emplace_back(std::move(km));
    }
    return mutations;
}

/**
 * Deserialize only Keyspace attributes without nested tables or types
 *
 * @param partition Keyspace attributes in serialized form
 */
lw_shared_ptr<keyspace_metadata> create_keyspace_from_schema_partition(const schema_result_value_type& result, lw_shared_ptr<query::result_set> scylla_specific_rs)
{
    auto&& rs = result.second;
    if (rs->empty()) {
        throw std::runtime_error("query result has no rows");
    }
    auto&& row = rs->row(0);
    auto keyspace_name = row.get_nonnull<sstring>("keyspace_name");
    // We get called from multiple shards with result set originating on only one of them.
    // Cannot use copying accessors for "deep" types like map, because we will hit shared_ptr asserts
    // (or screw up shared pointers)
    const auto& replication = row.get_nonnull<map_type_impl::native_type>("replication");

    std::map<sstring, sstring> strategy_options;
    for (auto& p : replication) {
        strategy_options.emplace(value_cast<sstring>(p.first), value_cast<sstring>(p.second));
    }
    auto strategy_name = strategy_options["class"];
    strategy_options.erase("class");
    bool durable_writes = row.get_nonnull<bool>("durable_writes");

    data_dictionary::storage_options storage_opts;
    // Scylla-specific row will only be present if SCYLLA_KEYSPACES schema feature is available in the cluster
    if (scylla_specific_rs) {
        if (!scylla_specific_rs->empty()) {
            auto row = scylla_specific_rs->row(0);
            auto storage_type = row.get<sstring>("storage_type");
            auto options = row.get<map_type_impl::native_type>("storage_options");
            if (storage_type && options) {
                std::map<sstring, sstring> values;
                for (const auto& entry : *options) {
                    values.emplace(value_cast<sstring>(entry.first), value_cast<sstring>(entry.second));
                }
                storage_opts.value = data_dictionary::storage_options::from_map(std::string_view(*storage_type), values);
            }
        }
    }
    return make_lw_shared<keyspace_metadata>(keyspace_name, strategy_name, strategy_options, durable_writes,
            std::vector<schema_ptr>{}, data_dictionary::user_types_metadata{}, storage_opts);
}

template<typename V>
static std::vector<V> get_list(const query::result_set_row& row, const sstring& name) {
    std::vector<V> list;

    const auto& values = row.get_nonnull<const list_type_impl::native_type&>(name);
    for (auto&& v : values) {
        list.emplace_back(value_cast<V>(v));
    };

    return list;
}

std::vector<user_type> create_types_from_schema_partition(
        keyspace_metadata& ks, lw_shared_ptr<query::result_set> result) {
    return create_types(ks, result->rows());
}

seastar::future<std::vector<shared_ptr<cql3::functions::user_function>>> create_functions_from_schema_partition(
        replica::database& db, lw_shared_ptr<query::result_set> result) {
    std::vector<shared_ptr<cql3::functions::user_function>> ret;
    for (const auto& row : result->rows()) {
        ret.emplace_back(co_await create_func(db, row));
    }
    co_return ret;
}

std::vector<shared_ptr<cql3::functions::user_aggregate>> create_aggregates_from_schema_partition(
        replica::database& db, lw_shared_ptr<query::result_set> result, lw_shared_ptr<query::result_set> scylla_result) {
    std::unordered_multimap<sstring, const query::result_set_row*> scylla_aggs;
    if (scylla_result) {
        for (const auto& scylla_row : scylla_result->rows()) {
            auto scylla_agg_name = scylla_row.get_nonnull<sstring>("aggregate_name");
            scylla_aggs.emplace(scylla_agg_name, &scylla_row);
        }
    }

    std::vector<shared_ptr<cql3::functions::user_aggregate>> ret;
    for (const auto& row : result->rows()) {
        auto agg_name = row.get_nonnull<sstring>("aggregate_name");
        auto agg_args = read_arg_types(db, row, row.get_nonnull<sstring>("keyspace_name"));
        const query::result_set_row *scylla_row_ptr = nullptr;
        for (auto [it, end] = scylla_aggs.equal_range(agg_name); it != end; ++it) {
            auto scylla_agg_args = read_arg_types(db, *it->second, it->second->get_nonnull<sstring>("keyspace_name"));
            if (agg_args == scylla_agg_args) {
                scylla_row_ptr = it->second;
                break;
            }
        }
        ret.emplace_back(create_aggregate(db, row, scylla_row_ptr));
    }
    return ret;
}

/*
 * User type metadata serialization/deserialization
 */

template<typename Func, typename T, typename... Args>
static atomic_cell_or_collection
make_list_mutation(const std::vector<T, Args...>& values,
                const column_definition& column,
                api::timestamp_type timestamp,
                Func&& f)
{
    auto column_type = static_pointer_cast<const list_type_impl>(column.type);
    auto vtyp = column_type->get_elements_type();

    if (column_type->is_multi_cell()) {
        collection_mutation_description m;
        m.cells.reserve(values.size());
        m.tomb.timestamp = timestamp - 1;
        m.tomb.deletion_time = gc_clock::now();

        for (auto&& value : values) {
            auto dv = f(value);
            auto uuid = utils::UUID_gen::get_time_UUID_bytes();
            m.cells.emplace_back(
                bytes(reinterpret_cast<const int8_t*>(uuid.data()), uuid.size()),
                atomic_cell::make_live(*vtyp, timestamp, vtyp->decompose(std::move(dv)), atomic_cell::collection_member::yes));
        }

        return m.serialize(*column_type);
    } else {
        list_type_impl::native_type tmp;
        tmp.reserve(values.size());
        std::transform(values.begin(), values.end(), std::back_inserter(tmp), f);
        return atomic_cell::make_live(*column.type, timestamp, column_type->decompose(make_list_value(column_type, std::move(tmp))));
    }
}

void add_type_to_schema_mutation(user_type type, api::timestamp_type timestamp, std::vector<mutation>& mutations)
{
    schema_ptr s = types();
    auto pkey = partition_key::from_singular(*s, type->_keyspace);
    auto ckey = clustering_key::from_singular(*s, type->get_name_as_string());
    mutation m{s, pkey};

    auto field_names_column = s->get_column_definition("field_names");
    auto field_names = make_list_mutation(type->field_names(), *field_names_column, timestamp, [](auto&& name) {
        return utf8_type->deserialize(name);
    });
    m.set_clustered_cell(ckey, *field_names_column, std::move(field_names));

    auto field_types_column = s->get_column_definition("field_types");
    auto field_types = make_list_mutation(type->field_types(), *field_types_column, timestamp, [](auto&& type) {
        return data_value(type->as_cql3_type().to_string());
    });
    m.set_clustered_cell(ckey, *field_types_column, std::move(field_types));

    mutations.emplace_back(std::move(m));
}

std::vector<mutation> make_create_type_mutations(lw_shared_ptr<keyspace_metadata> keyspace, user_type type, api::timestamp_type timestamp)
{
    std::vector<mutation> mutations;
    add_type_to_schema_mutation(type, timestamp, mutations);
    return mutations;
}

std::vector<mutation> make_drop_type_mutations(lw_shared_ptr<keyspace_metadata> keyspace, user_type type, api::timestamp_type timestamp)
{
    std::vector<mutation> mutations;
    schema_ptr s = types();
    auto pkey = partition_key::from_singular(*s, type->_keyspace);
    auto ckey = clustering_key::from_singular(*s, type->get_name_as_string());
    mutation m{s, pkey};
    m.partition().apply_delete(*s, ckey, tombstone(timestamp, gc_clock::now()));
    mutations.emplace_back(std::move(m));

    return mutations;
}

/*
 * UDF metadata serialization/deserialization.
 */

static std::pair<mutation, clustering_key> get_mutation(schema_ptr s, const cql3::functions::function& func) {
    auto name = func.name();
    auto pkey = partition_key::from_singular(*s, name.keyspace);

    list_type_impl::native_type arg_types;
    for (const auto& arg_type : func.arg_types()) {
        arg_types.emplace_back(arg_type->as_cql3_type().to_string());
    }
    auto arg_list_type = list_type_impl::get_instance(utf8_type, false);
    data_value arg_types_val = make_list_value(arg_list_type, std::move(arg_types));
    auto ckey = clustering_key::from_exploded(
            *s, {utf8_type->decompose(name.name), arg_list_type->decompose(arg_types_val)});
    mutation m{s, pkey};
    return {std::move(m), std::move(ckey)};
}

std::vector<mutation> make_create_function_mutations(shared_ptr<cql3::functions::user_function> func,
        api::timestamp_type timestamp) {
    schema_ptr s = functions();
    auto p = get_mutation(s, *func);
    mutation& m = p.first;
    clustering_key& ckey = p.second;
    auto argument_names_column = s->get_column_definition("argument_names");
    auto argument_names = make_list_mutation(func->arg_names(), *argument_names_column, timestamp, [] (auto&& name) {
        return name;
    });
    m.set_clustered_cell(ckey, *argument_names_column, std::move(argument_names));
    m.set_clustered_cell(ckey, "body", func->body(), timestamp);
    m.set_clustered_cell(ckey, "language", func->language(), timestamp);
    m.set_clustered_cell(ckey, "return_type", func->return_type()->as_cql3_type().to_string(), timestamp);
    m.set_clustered_cell(ckey, "called_on_null_input", func->called_on_null_input(), timestamp);
    return {m};
}

std::vector<mutation> make_drop_function_mutations(schema_ptr s, const cql3::functions::function& func, api::timestamp_type timestamp) {
    auto p = get_mutation(s, func);
    mutation& m = p.first;
    clustering_key& ckey = p.second;
    m.partition().apply_delete(*s, ckey, tombstone(timestamp, gc_clock::now()));
    return {std::move(m)};
}

std::vector<mutation> make_drop_function_mutations(shared_ptr<cql3::functions::user_function> func, api::timestamp_type timestamp) {
    return make_drop_function_mutations(functions(), *func, timestamp);
}

/*
 * UDA metadata serialization/deserialization
 */

static std::pair<mutation, clustering_key> get_mutation(schema_ptr s, const cql3::functions::user_aggregate& aggregate) {
    auto name = aggregate.name();
    auto pkey = partition_key::from_singular(*s, name.keyspace);

    list_type_impl::native_type arg_types;
    for (const auto& arg_type : aggregate.arg_types()) {
        arg_types.emplace_back(arg_type->as_cql3_type().to_string());
    }
    auto arg_list_type = list_type_impl::get_instance(utf8_type, false);
    data_value arg_types_val = make_list_value(arg_list_type, std::move(arg_types));
    auto ckey = clustering_key::from_exploded(
            *s, {utf8_type->decompose(name.name), arg_list_type->decompose(arg_types_val)});
    mutation m{s, pkey};
    return {std::move(m), std::move(ckey)};
}

std::vector<mutation> make_create_aggregate_mutations(schema_features features, shared_ptr<cql3::functions::user_aggregate> aggregate, api::timestamp_type timestamp) {
    schema_ptr s = aggregates();
    auto p = get_mutation(s, *aggregate);
    mutation& m = p.first;
    clustering_key& ckey = p.second;

    data_type state_type = aggregate->sfunc()->arg_types()[0];
    if (aggregate->has_finalfunc()) {
        m.set_clustered_cell(ckey, "final_func", aggregate->finalfunc()->name().name, timestamp);
    }
    if (aggregate->initcond()) {
        m.set_clustered_cell(ckey, "initcond", state_type->deserialize(*aggregate->initcond()).to_parsable_string(), timestamp);
    }
    m.set_clustered_cell(ckey, "return_type", aggregate->return_type()->as_cql3_type().to_string(), timestamp);
    m.set_clustered_cell(ckey, "state_func", aggregate->sfunc()->name().name, timestamp);
    m.set_clustered_cell(ckey, "state_type", state_type->as_cql3_type().to_string(), timestamp);
    std::vector<mutation> muts = {m};

    if (features.contains<schema_feature::SCYLLA_AGGREGATES>() && aggregate->is_reducible()) {
        schema_ptr sa_schema = scylla_aggregates();
        auto sa_p = get_mutation(sa_schema, *aggregate);
        mutation& sa_mut = sa_p.first;
        clustering_key& sa_ckey = sa_p.second;
        sa_mut.set_clustered_cell(sa_ckey, "reduce_func", aggregate->reducefunc()->name().name, timestamp);
        sa_mut.set_clustered_cell(sa_ckey, "state_type", state_type->as_cql3_type().to_string(), timestamp);

        muts.emplace_back(sa_mut);
    }

    return muts;
}

std::vector<mutation> make_drop_aggregate_mutations(schema_features features, shared_ptr<cql3::functions::user_aggregate> aggregate, api::timestamp_type timestamp) {
    auto muts = make_drop_function_mutations(aggregates(), *aggregate, timestamp);
    if (features.contains<schema_feature::SCYLLA_AGGREGATES>() && aggregate->is_reducible()) {
        auto scylla_muts = make_drop_function_mutations(scylla_aggregates(), *aggregate, timestamp);
        muts.insert(muts.end(), scylla_muts.begin(), scylla_muts.end());
    }

    return muts;
}

/*
 * Table metadata serialization/deserialization.
 */

/// Returns mutations which when applied to the database will cause all schema changes
/// which create or alter the table with a given name, made with timestamps smaller than t,
/// have no effect. Used when overriding schema to shadow concurrent conflicting schema changes.
/// Shouldn't be needed if schema changes are serialized with RAFT.
static schema_mutations make_table_deleting_mutations(const sstring& ks, const sstring& table, bool is_view, api::timestamp_type t) {
    tombstone tomb;

    // Generate neutral mutations if t == api::min_timestamp
    if (t > api::min_timestamp) {
        tomb = tombstone(t - 1, gc_clock::now());
    }

    auto tables_m_s = is_view ? views() : tables();
    mutation tables_m{tables_m_s, partition_key::from_singular(*tables_m_s, ks)};
    {
        auto ckey = clustering_key::from_singular(*tables_m_s, table);
        tables_m.partition().apply_delete(*tables_m_s, ckey, tomb);
    }

    mutation scylla_tables_m{scylla_tables(), partition_key::from_singular(*scylla_tables(), ks)};
    {
        auto ckey = clustering_key::from_singular(*scylla_tables(), table);
        scylla_tables_m.partition().apply_delete(*scylla_tables(), ckey, tomb);
    }

    auto make_drop_columns = [&] (const schema_ptr& s) {
        mutation m{s, partition_key::from_singular(*s, ks)};
        auto ckey = clustering_key::from_exploded(*s, {utf8_type->decompose(table)});
        m.partition().apply_delete(*s, ckey, tomb);
        return m;
    };

    return schema_mutations(std::move(tables_m),
                            make_drop_columns(columns()),
                            make_drop_columns(view_virtual_columns()),
                            make_drop_columns(computed_columns()),
                            mutation(indexes(), partition_key::from_singular(*indexes(), ks)),
                            make_drop_columns(dropped_columns()),
                            std::move(scylla_tables_m)
    );
}

std::vector<mutation> make_create_table_mutations(schema_ptr table, api::timestamp_type timestamp)
{
    std::vector<mutation> mutations;
    add_table_or_view_to_schema_mutation(table, timestamp, true, mutations);
    make_table_deleting_mutations(table->ks_name(), table->cf_name(), table->is_view(), timestamp)
        .copy_to(mutations);
    return mutations;
}

static void add_table_params_to_mutations(mutation& m, const clustering_key& ckey, schema_ptr table, api::timestamp_type timestamp) {
    m.set_clustered_cell(ckey, "bloom_filter_fp_chance", table->bloom_filter_fp_chance(), timestamp);
    m.set_clustered_cell(ckey, "comment", table->comment(), timestamp);
    m.set_clustered_cell(ckey, "dclocal_read_repair_chance", table->dc_local_read_repair_chance(), timestamp);
    m.set_clustered_cell(ckey, "default_time_to_live", gc_clock::as_int32(table->default_time_to_live()), timestamp);
    m.set_clustered_cell(ckey, "gc_grace_seconds", gc_clock::as_int32(table->gc_grace_seconds()), timestamp);
    m.set_clustered_cell(ckey, "max_index_interval", table->max_index_interval(), timestamp);
    m.set_clustered_cell(ckey, "memtable_flush_period_in_ms", table->memtable_flush_period(), timestamp);
    m.set_clustered_cell(ckey, "min_index_interval", table->min_index_interval(), timestamp);
    m.set_clustered_cell(ckey, "read_repair_chance", table->read_repair_chance(), timestamp);
    m.set_clustered_cell(ckey, "speculative_retry", table->speculative_retry().to_sstring(), timestamp);
    m.set_clustered_cell(ckey, "crc_check_chance", table->crc_check_chance(), timestamp);

    store_map(m, ckey, "caching", timestamp, table->caching_options().to_map());

    {
        auto map = table->compaction_strategy_options();
        map["class"] = sstables::compaction_strategy::name(table->configured_compaction_strategy());
        store_map(m, ckey, "compaction", timestamp, map);
    }

    store_map(m, ckey, "compression", timestamp, table->get_compressor_params().get_options());

    std::map<sstring, bytes> map;

    if (!table->extensions().empty()) {
        for (auto& p : table->extensions()) {
            map.emplace(p.first, p.second->serialize());
        }
    }

    store_map(m, ckey, "extensions", timestamp, map);
}

static data_type expand_user_type(data_type);

static std::vector<data_type> expand_user_types(const std::vector<data_type>& types) {
    std::vector<data_type> result;
    result.reserve(types.size());
    std::transform(types.begin(), types.end(), std::back_inserter(result), &expand_user_type);
    return result;
}

static data_type expand_user_type(data_type original) {
    if (original->is_user_type()) {
        return tuple_type_impl::get_instance(
                        expand_user_types(
                                        static_pointer_cast<const user_type_impl>(
                                                        original)->field_types()));
    }
    if (original->is_tuple()) {
        return tuple_type_impl::get_instance(
                        expand_user_types(
                                        static_pointer_cast<
                                                        const tuple_type_impl>(
                                                        original)->all_types()));
    }
    if (original->is_reversed()) {
        return reversed_type_impl::get_instance(
                        expand_user_type(original->underlying_type()));
    }

    if (original->is_collection()) {

        auto ct = static_pointer_cast<const collection_type_impl>(original);

        if (ct->is_list()) {
            return list_type_impl::get_instance(
                            expand_user_type(ct->value_comparator()),
                            ct->is_multi_cell());
        }
        if (ct->is_map()) {
            return map_type_impl::get_instance(
                            expand_user_type(ct->name_comparator()),
                            expand_user_type(ct->value_comparator()),
                            ct->is_multi_cell());
        }
        if (ct->is_set()) {
            return set_type_impl::get_instance(
                            expand_user_type(ct->name_comparator()),
                            ct->is_multi_cell());
        }
    }

    return original;
}

static void add_dropped_column_to_schema_mutation(schema_ptr table, const sstring& name, const schema::dropped_column& column, api::timestamp_type timestamp, mutation& m) {
    auto ckey = clustering_key::from_exploded(*dropped_columns(), {utf8_type->decompose(table->cf_name()), utf8_type->decompose(name)});
    db_clock::time_point tp(db_clock::duration(column.timestamp));
    m.set_clustered_cell(ckey, "dropped_time", tp, timestamp);

    /*
     * From origin:
     * we never store actual UDT names in dropped column types (so that we can safely drop types if nothing refers to
     * them anymore), so before storing dropped columns in schema we expand UDTs to tuples. See expandUserTypes method.
     * Because of that, we can safely pass Types.none() to parse()
     */
    m.set_clustered_cell(ckey, "type", expand_user_type(column.type)->as_cql3_type().to_string(), timestamp);
}

mutation make_scylla_tables_mutation(schema_ptr table, api::timestamp_type timestamp) {
    schema_ptr s = tables();
    auto pkey = partition_key::from_singular(*s, table->ks_name());
    auto ckey = clustering_key::from_singular(*s, table->cf_name());
    mutation m(scylla_tables(), pkey);
    m.set_clustered_cell(ckey, "version", table->version().uuid(), timestamp);
    // Since 4.0, we stopped using cdc column in scylla tables. Extensions are
    // used instead. Since we stopped reading this column in commit 861c7b5, we
    // can now keep it always empty.
    auto& cdc_cdef = *scylla_tables()->get_column_definition("cdc");
    m.set_clustered_cell(ckey, cdc_cdef, atomic_cell::make_dead(timestamp, gc_clock::now()));
    if (table->has_custom_partitioner()) {
        m.set_clustered_cell(ckey, "partitioner", table->get_partitioner().name(), timestamp);
    } else {
        // Avoid storing anything for default partitioner, so we don't end up with
        // different digests on different nodes due to the other node redacting
        // the partitioner column when the per_table_partitioners cluster feature is disabled.
        //
        // Tombstones are not considered for schema digest, so this is okay (and
        // needed in order for disabling of per_table_partitioners to have effect).
        auto& cdef = *scylla_tables()->get_column_definition("partitioner");
        m.set_clustered_cell(ckey, cdef, atomic_cell::make_dead(timestamp, gc_clock::now()));
    }
    return m;
}

static schema_mutations make_table_mutations(schema_ptr table, api::timestamp_type timestamp, bool with_columns_and_triggers)
{
    // When adding new schema properties, don't set cells for default values so that
    // both old and new nodes will see the same version during rolling upgrades.

    // For property that can be null (and can be changed), we insert tombstones, to make sure
    // we don't keep a property the user has removed
    schema_ptr s = tables();
    auto pkey = partition_key::from_singular(*s, table->ks_name());
    mutation m{s, pkey};
    auto ckey = clustering_key::from_singular(*s, table->cf_name());
    m.set_clustered_cell(ckey, "id", table->id().uuid(), timestamp);

    auto scylla_tables_mutation = make_scylla_tables_mutation(table, timestamp);

    list_type_impl::native_type flags;
    if (table->is_super()) {
        flags.emplace_back("super");
    }
    if (table->is_dense()) {
        flags.emplace_back("dense");
    }
    if (table->is_compound()) {
        flags.emplace_back("compound");
    }
    if (table->is_counter()) {
        flags.emplace_back("counter");
    }

    m.set_clustered_cell(ckey, "flags", make_list_value(s->get_column_definition("flags")->type, flags), timestamp);

    add_table_params_to_mutations(m, ckey, table, timestamp);

    mutation columns_mutation(columns(), pkey);
    mutation computed_columns_mutation(computed_columns(), pkey);
    mutation dropped_columns_mutation(dropped_columns(), pkey);
    mutation indices_mutation(indexes(), pkey);

    if (with_columns_and_triggers) {
        for (auto&& column : table->v3().all_columns()) {
            if (column.is_view_virtual()) {
                throw std::logic_error("view_virtual column found in non-view table");
            }
            add_column_to_schema_mutation(table, column, timestamp, columns_mutation);
            if (column.is_computed()) {
                add_computed_column_to_schema_mutation(table, column, timestamp, computed_columns_mutation);
            }
        }
        for (auto&& index : table->indices()) {
            add_index_to_schema_mutation(table, index, timestamp, indices_mutation);
        }
        // TODO: triggers

        for (auto&& e : table->dropped_columns()) {
            add_dropped_column_to_schema_mutation(table, e.first, e.second, timestamp, dropped_columns_mutation);
        }
    }

    return schema_mutations{std::move(m),
                            std::move(columns_mutation),
                            std::nullopt,
                            std::move(computed_columns_mutation),
                            std::move(indices_mutation),
                            std::move(dropped_columns_mutation),
                            std::move(scylla_tables_mutation)};
}

void add_table_or_view_to_schema_mutation(schema_ptr s, api::timestamp_type timestamp, bool with_columns, std::vector<mutation>& mutations)
{
    make_schema_mutations(s, timestamp, with_columns).copy_to(mutations);
}

static schema_mutations make_view_mutations(view_ptr view, api::timestamp_type timestamp, bool with_columns);
static void make_drop_table_or_view_mutations(schema_ptr schema_table, schema_ptr table_or_view, api::timestamp_type timestamp, std::vector<mutation>& mutations);

static void make_update_indices_mutations(
        replica::database& db,
        schema_ptr old_table,
        schema_ptr new_table,
        api::timestamp_type timestamp,
        std::vector<mutation>& mutations)
{
    mutation indices_mutation(indexes(), partition_key::from_singular(*indexes(), old_table->ks_name()));

    auto diff = difference(old_table->all_indices(), new_table->all_indices());
    bool new_token_column_computation = db.features().correct_idx_token_in_secondary_index;

    // indices that are no longer needed
    for (auto&& name : diff.entries_only_on_left) {
        const index_metadata& index = old_table->all_indices().at(name);
        drop_index_from_schema_mutation(old_table, index, timestamp, mutations);
        auto& cf = db.find_column_family(old_table);
        auto view = cf.get_index_manager().create_view_for_index(index, new_token_column_computation);
        make_drop_table_or_view_mutations(views(), view, timestamp, mutations);
    }

    // newly added indices and old indices with updated attributes
    for (auto&& name : boost::range::join(diff.entries_differing, diff.entries_only_on_right)) {
        const index_metadata& index = new_table->all_indices().at(name);
        add_index_to_schema_mutation(new_table, index, timestamp, indices_mutation);
        auto& cf = db.find_column_family(new_table);
        auto view = cf.get_index_manager().create_view_for_index(index, new_token_column_computation);
        auto view_mutations = make_view_mutations(view, timestamp, true);
        view_mutations.copy_to(mutations);
    }

    mutations.emplace_back(std::move(indices_mutation));
}

static void add_drop_column_to_mutations(schema_ptr table, const sstring& name, const schema::dropped_column& dc, api::timestamp_type timestamp, std::vector<mutation>& mutations) {
    schema_ptr s = dropped_columns();
    auto pkey = partition_key::from_singular(*s, table->ks_name());
    auto ckey = clustering_key::from_exploded(*s, {utf8_type->decompose(table->cf_name()), utf8_type->decompose(name)});
    mutation m(s, pkey);
    add_dropped_column_to_schema_mutation(table, name, dc, timestamp, m);
    mutations.emplace_back(std::move(m));
}

static void make_update_columns_mutations(schema_ptr old_table,
        schema_ptr new_table,
        api::timestamp_type timestamp,
        bool from_thrift,
        std::vector<mutation>& mutations) {
    mutation columns_mutation(columns(), partition_key::from_singular(*columns(), old_table->ks_name()));
    mutation view_virtual_columns_mutation(view_virtual_columns(), partition_key::from_singular(*columns(), old_table->ks_name()));
    mutation computed_columns_mutation(computed_columns(), partition_key::from_singular(*columns(), old_table->ks_name()));

    auto diff = difference(old_table->v3().columns_by_name(), new_table->v3().columns_by_name());

    // columns that are no longer needed
    for (auto&& name : diff.entries_only_on_left) {
        // Thrift only knows about the REGULAR ColumnDefinition type, so don't consider other type
        // are being deleted just because they are not here.
        const column_definition& column = *old_table->v3().columns_by_name().at(name);
        if (from_thrift && !column.is_regular()) {
            continue;
        }
        if (column.is_view_virtual()) {
            drop_column_from_schema_mutation(view_virtual_columns(), old_table, column.name_as_text(), timestamp, mutations);
        } else {
            drop_column_from_schema_mutation(columns(), old_table, column.name_as_text(), timestamp, mutations);
        }
        if (column.is_computed()) {
            drop_column_from_schema_mutation(computed_columns(), old_table, column.name_as_text(), timestamp, mutations);
        }
    }

    // newly added columns and old columns with updated attributes
    for (auto&& name : boost::range::join(diff.entries_differing, diff.entries_only_on_right)) {
        const column_definition& column = *new_table->v3().columns_by_name().at(name);
        if (column.is_view_virtual()) {
            add_column_to_schema_mutation(new_table, column, timestamp, view_virtual_columns_mutation);
        } else {
            add_column_to_schema_mutation(new_table, column, timestamp, columns_mutation);
        }
        if (column.is_computed()) {
            add_computed_column_to_schema_mutation(new_table, column, timestamp, computed_columns_mutation);
        }
    }

    mutations.emplace_back(std::move(columns_mutation));
    mutations.emplace_back(std::move(view_virtual_columns_mutation));
    mutations.emplace_back(std::move(computed_columns_mutation));

    // dropped columns
    auto dc_diff = difference(old_table->dropped_columns(), new_table->dropped_columns());

    // newly dropped columns
    // columns added then dropped again
    for (auto& name : boost::range::join(dc_diff.entries_differing, dc_diff.entries_only_on_right)) {
        add_drop_column_to_mutations(new_table, name, new_table->dropped_columns().at(name), timestamp, mutations);
    }
}

std::vector<mutation> make_update_table_mutations(replica::database& db,
    lw_shared_ptr<keyspace_metadata> keyspace,
    schema_ptr old_table,
    schema_ptr new_table,
    api::timestamp_type timestamp,
    bool from_thrift)
{
    std::vector<mutation> mutations;
    add_table_or_view_to_schema_mutation(new_table, timestamp, false, mutations);
    make_update_indices_mutations(db, old_table, new_table, timestamp, mutations);
    make_update_columns_mutations(std::move(old_table), std::move(new_table), timestamp, from_thrift, mutations);

    warn(unimplemented::cause::TRIGGERS);
#if 0
        MapDifference<String, TriggerDefinition> triggerDiff = Maps.difference(oldTable.getTriggers(), newTable.getTriggers());

        // dropped triggers
        for (TriggerDefinition trigger : triggerDiff.entriesOnlyOnLeft().values())
            dropTriggerFromSchemaMutation(oldTable, trigger, timestamp, mutation);

        // newly created triggers
        for (TriggerDefinition trigger : triggerDiff.entriesOnlyOnRight().values())
            addTriggerToSchemaMutation(newTable, trigger, timestamp, mutation);

#endif
    return mutations;
}

static void make_drop_table_or_view_mutations(schema_ptr schema_table,
            schema_ptr table_or_view,
            api::timestamp_type timestamp,
            std::vector<mutation>& mutations) {
    auto pkey = partition_key::from_singular(*schema_table, table_or_view->ks_name());
    mutation m{schema_table, pkey};
    auto ckey = clustering_key::from_singular(*schema_table, table_or_view->cf_name());
    m.partition().apply_delete(*schema_table, ckey, tombstone(timestamp, gc_clock::now()));
    mutations.emplace_back(m);
    for (auto& column : table_or_view->v3().all_columns()) {
        if (column.is_view_virtual()) {
            drop_column_from_schema_mutation(view_virtual_columns(), table_or_view, column.name_as_text(), timestamp, mutations);
        } else {
            drop_column_from_schema_mutation(columns(), table_or_view, column.name_as_text(), timestamp, mutations);
        }
        if (column.is_computed()) {
            drop_column_from_schema_mutation(computed_columns(), table_or_view, column.name_as_text(), timestamp, mutations);
        }
    }
    for (auto& column : table_or_view->dropped_columns() | boost::adaptors::map_keys) {
        drop_column_from_schema_mutation(dropped_columns(), table_or_view, column, timestamp, mutations);
    }
    mutation m1{scylla_tables(), pkey};
    m1.partition().apply_delete(*scylla_tables(), ckey, tombstone(timestamp, gc_clock::now()));
    mutations.emplace_back(m1);
}

std::vector<mutation> make_drop_table_mutations(lw_shared_ptr<keyspace_metadata> keyspace, schema_ptr table, api::timestamp_type timestamp)
{
    std::vector<mutation> mutations;
    make_drop_table_or_view_mutations(tables(), std::move(table), timestamp, mutations);

#if 0
    for (TriggerDefinition trigger : table.getTriggers().values())
        dropTriggerFromSchemaMutation(table, trigger, timestamp, mutation);

    // TODO: get rid of in #6717
    ColumnFamily indexCells = mutation.addOrGet(SystemKeyspace.BuiltIndexes);
    for (String indexName : Keyspace.open(keyspace.name).getColumnFamilyStore(table.cfName).getBuiltIndexes())
        indexCells.addTombstone(indexCells.getComparator().makeCellName(indexName), ldt, timestamp);
#endif
    return mutations;
}

static future<schema_mutations> read_table_mutations(distributed<service::storage_proxy>& proxy, const qualified_name& table, schema_ptr s)
{
    auto&& [cf_m, col_m, vv_col_m, c_col_m, dropped_m, idx_m, st_m] = co_await coroutine::all(
        [&] { return read_schema_partition_for_table(proxy, s, table.keyspace_name, table.table_name); },
        [&] { return read_schema_partition_for_table(proxy, columns(), table.keyspace_name, table.table_name); },
        [&] { return read_schema_partition_for_table(proxy, view_virtual_columns(), table.keyspace_name, table.table_name); },
        [&] { return read_schema_partition_for_table(proxy, computed_columns(), table.keyspace_name, table.table_name); },
        [&] { return read_schema_partition_for_table(proxy, dropped_columns(), table.keyspace_name, table.table_name); },
        [&] { return read_schema_partition_for_table(proxy, indexes(), table.keyspace_name, table.table_name); },
        [&] { return read_schema_partition_for_table(proxy, scylla_tables(), table.keyspace_name, table.table_name); }
    );
    co_return schema_mutations{std::move(cf_m), std::move(col_m), std::move(vv_col_m), std::move(c_col_m), std::move(idx_m), std::move(dropped_m), std::move(st_m)};
#if 0
        // FIXME:
    Row serializedTriggers = readSchemaPartitionForTable(TRIGGERS, ksName, cfName);
    try
    {
        for (TriggerDefinition trigger : createTriggersFromTriggersPartition(serializedTriggers))
            cfm.addTriggerDefinition(trigger);
    }
    catch (InvalidRequestException e)
    {
        throw new RuntimeException(e);
    }
#endif
}

future<schema_ptr> create_table_from_name(distributed<service::storage_proxy>& proxy, const sstring& keyspace, const sstring& table)
{
    auto qn = qualified_name(keyspace, table);
    auto sm = co_await read_table_mutations(proxy, qn, tables());
    if (!sm.live()) {
        co_await coroutine::return_exception(std::runtime_error(format("{}:{} not found in the schema definitions keyspace.", qn.keyspace_name, qn.table_name)));
    }
    co_return create_table_from_mutations(proxy, std::move(sm));
}

/**
 * Deserialize tables from low-level schema representation, all of them belong to the same keyspace
 *
 * @return map containing name of the table and its metadata for faster lookup
 */
future<std::map<sstring, schema_ptr>> create_tables_from_tables_partition(distributed<service::storage_proxy>& proxy, const schema_result::mapped_type& result)
{
    auto tables = std::map<sstring, schema_ptr>();
    co_await max_concurrent_for_each(result->rows().begin(), result->rows().end(), max_concurrent, [&] (const query::result_set_row& row) -> future<> {
        schema_ptr cfm = co_await create_table_from_table_row(proxy, row);
        tables.emplace(cfm->cf_name(), std::move(cfm));
    });
    co_return std::move(tables);
}

#if 0
    public static CFMetaData createTableFromTablePartitionAndColumnsPartition(Row serializedTable, Row serializedColumns)
    {
        String query = String.format("SELECT * FROM %s.%s", SystemKeyspace.NAME, COLUMNFAMILIES);
        return createTableFromTableRowAndColumnsPartition(QueryProcessor.resultify(query, serializedTable).one(), serializedColumns);
    }
#endif

/**
 * Deserialize table metadata from low-level representation
 *
 * @return Metadata deserialized from schema
 */
static future<schema_ptr> create_table_from_table_row(distributed<service::storage_proxy>& proxy, const query::result_set_row& row)
{
    auto ks_name = row.get_nonnull<sstring>("keyspace_name");
    auto cf_name = row.get_nonnull<sstring>("table_name");
    return create_table_from_name(proxy, ks_name, cf_name);
}

static void prepare_builder_from_table_row(const schema_ctxt& ctxt, schema_builder& builder, const query::result_set_row& table_row)
{
    // These row reads have been purposefully reordered to match the origin counterpart. For easier matching.
    if (auto val = table_row.get<double>("bloom_filter_fp_chance")) {
        builder.set_bloom_filter_fp_chance(*val);
    } else {
        builder.set_bloom_filter_fp_chance(builder.get_bloom_filter_fp_chance());
    }

    if (auto map = get_map<sstring, sstring>(table_row, "caching")) {
        builder.set_caching_options(caching_options::from_map(*map));
    }

    if (auto val = table_row.get<sstring>("comment")) {
        builder.set_comment(*val);
    }

    if (auto opt_map = get_map<sstring, sstring>(table_row, "compaction")) {
        auto &map = *opt_map;
        auto i = map.find("class");
        if (i != map.end()) {
            try {
                builder.set_compaction_strategy(sstables::compaction_strategy::type(i->second));
                map.erase(i);
            } catch (const exceptions::configuration_exception& e) {
                // If compaction strategy class isn't supported, fallback to size tiered.
                slogger.warn("Falling back to size-tiered compaction strategy after the problem: {}", e.what());
                builder.set_compaction_strategy(sstables::compaction_strategy_type::size_tiered);
            }
        }
        if (map.contains("max_threshold")) {
            builder.set_max_compaction_threshold(std::stoi(map["max_threshold"]));
        }
        if (map.contains("min_threshold")) {
            builder.set_min_compaction_threshold(std::stoi(map["min_threshold"]));
        }
        if (map.contains("enabled")) {
            builder.set_compaction_enabled(boost::algorithm::iequals(map["enabled"], "true"));
        }

        builder.set_compaction_strategy_options(std::move(map));
    }

    if (auto map = get_map<sstring, sstring>(table_row, "compression")) {
        compression_parameters cp(*map);
        builder.set_compressor_params(cp);
    }

    if (auto val = table_row.get<double>("dclocal_read_repair_chance")) {
        builder.set_dc_local_read_repair_chance(*val);
    }

    if (auto val = table_row.get<int32_t>("default_time_to_live")) {
        builder.set_default_time_to_live(gc_clock::duration(*val));
    }

    if (auto val = get_map<sstring, bytes>(table_row, "extensions")) {
        auto &map = *val;
        schema::extensions_map result;
        auto& exts = ctxt.extensions().schema_extensions();
        for (auto&p : map) {
            auto i = exts.find(p.first);
            if (i != exts.end()) {
                try {
                    auto ep = i->second(p.second);
                    if (ep) {
                        result.emplace(p.first, std::move(ep));
                    }
                    continue;
                } catch (...) {
                    slogger.warn("Error parsing extension {}: {}", p.first, std::current_exception());
                }
            }

            // unknown. we should still preserve it.
            class placeholder : public schema_extension {
                bytes _bytes;
            public:
                placeholder(bytes bytes)
                                : _bytes(std::move(bytes)) {
                }
                bytes serialize() const override {
                    return _bytes;
                }
                virtual bool is_placeholder() const override {
                    return true;
                }
            };

            result.emplace(p.first, ::make_shared<placeholder>(p.second));
        }
        builder.set_extensions(std::move(result));
    }

    if (auto val = table_row.get<int32_t>("gc_grace_seconds")) {
        builder.set_gc_grace_seconds(*val);
    }

    if (auto val = table_row.get<int>("min_index_interval")) {
        builder.set_min_index_interval(*val);
    }

    if (auto val = table_row.get<int32_t>("memtable_flush_period_in_ms")) {
        builder.set_memtable_flush_period(*val);
    }

    if (auto val = table_row.get<int>("max_index_interval")) {
        builder.set_max_index_interval(*val);
    }

    if (auto val = table_row.get<double>("read_repair_chance")) {
        builder.set_read_repair_chance(*val);
    }

    if (auto val = table_row.get<double>("crc_check_chance")) {
        builder.set_crc_check_chance(*val);
    }

    if (auto val = table_row.get<sstring>("speculative_retry")) {
        builder.set_speculative_retry(*val);
    }
}

schema_ptr create_table_from_mutations(const schema_ctxt& ctxt, schema_mutations sm, std::optional<table_schema_version> version)
{
    slogger.trace("create_table_from_mutations: version={}, {}", version, sm);

    auto table_rs = query::result_set(sm.columnfamilies_mutation());
    query::result_set_row table_row = table_rs.row(0);

    auto ks_name = table_row.get_nonnull<sstring>("keyspace_name");
    auto cf_name = table_row.get_nonnull<sstring>("table_name");
    auto id = table_id(table_row.get_nonnull<utils::UUID>("id"));
    schema_builder builder{ks_name, cf_name, id};

    auto cf = cf_type::standard;
    auto is_dense = false;
    auto is_counter = false;
    auto is_compound = false;
    auto flags = table_row.get<set_type_impl::native_type>("flags");

    if (flags) {
        for (auto& s : *flags) {
            if (s == "super") {
                // cf = cf_type::super;
                fail(unimplemented::cause::SUPER);
            } else if (s == "dense") {
                is_dense = true;
            } else if (s == "compound") {
                is_compound = true;
            } else if (s == "counter") {
                is_counter = true;
            }
        }
    }

    auto computed_columns = get_computed_columns(sm);
    std::vector<column_definition> column_defs = create_columns_from_column_rows(
            ctxt,
            query::result_set(sm.columns_mutation()),
            ks_name,
            cf_name,/*,
            fullRawComparator, */
            cf == cf_type::super,
            column_view_virtual::no,
            computed_columns);


    builder.set_is_dense(is_dense);
    builder.set_is_compound(is_compound);
    builder.set_is_counter(is_counter);

    prepare_builder_from_table_row(ctxt, builder, table_row);

    v3_columns columns(std::move(column_defs), is_dense, is_compound);
    columns.apply_to(builder);

    std::vector<index_metadata> index_defs;
    if (sm.indices_mutation()) {
        index_defs = create_indices_from_index_rows(query::result_set(*sm.indices_mutation()), ks_name, cf_name);
    }
    for (auto&& index : index_defs) {
        builder.with_index(index);
    }

    if (sm.dropped_columns_mutation()) {
        query::result_set dcr(*sm.dropped_columns_mutation());
        for (auto& row : dcr.rows()) {
            auto name = row.get_nonnull<sstring>("column_name");
            auto type = cql_type_parser::parse(ks_name, row.get_nonnull<sstring>("type"), ctxt.user_types());
            auto time = row.get_nonnull<db_clock::time_point>("dropped_time");
            builder.without_column(name, type, time.time_since_epoch().count());
        }
    }

    if (version) {
        builder.with_version(*version);
    } else {
        builder.with_version(sm.digest(ctxt.features().cluster_schema_features()));
    }

    if (auto partitioner = sm.partitioner()) {
        builder.with_partitioner(*partitioner);
        builder.with_sharder(smp::count, ctxt.murmur3_partitioner_ignore_msb_bits());
    }

    return builder.build();
}

/*
 * Column metadata serialization/deserialization.
 */

static void add_column_to_schema_mutation(schema_ptr table,
                                   const column_definition& column,
                                   api::timestamp_type timestamp,
                                   mutation& m)
{
    auto ckey = clustering_key::from_exploded(*m.schema(), {utf8_type->decompose(table->cf_name()),
                                                            utf8_type->decompose(column.name_as_text())});
    fill_column_info(*table, ckey, column, timestamp, std::nullopt, m);
}

static void add_computed_column_to_schema_mutation(schema_ptr table,
        const column_definition& column,
        api::timestamp_type timestamp,
        mutation& m) {
    auto ckey = clustering_key::from_exploded(*m.schema(),
            {utf8_type->decompose(table->cf_name()), utf8_type->decompose(column.name_as_text())});

    m.set_clustered_cell(ckey, "computation", data_value(column.get_computation().serialize()), timestamp);
}

sstring serialize_kind(column_kind kind)
{
    switch (kind) {
    case column_kind::partition_key:  return "partition_key";
    case column_kind::clustering_key: return "clustering";
    case column_kind::static_column:  return "static";
    case column_kind::regular_column: return "regular";
    default:                          throw std::invalid_argument("unknown column kind");
    }
}

column_kind deserialize_kind(sstring kind) {
    if (kind == "partition_key") {
        return column_kind::partition_key;
    } else if (kind == "clustering_key" || kind == "clustering") {
        return column_kind::clustering_key;
    } else if (kind == "static") {
        return column_kind::static_column;
    } else if (kind == "regular") {
        return column_kind::regular_column;
    } else if (kind == "compact_value") { // backward compatibility
        return column_kind::regular_column;
    } else {
        throw std::invalid_argument("unknown column kind: " + kind);
    }
}

sstring serialize_index_kind(index_metadata_kind kind)
{
    switch (kind) {
    case index_metadata_kind::keys:       return "KEYS";
    case index_metadata_kind::composites: return "COMPOSITES";
    case index_metadata_kind::custom:     return "CUSTOM";
    }
    throw std::invalid_argument("unknown index kind");
}

index_metadata_kind deserialize_index_kind(sstring kind) {
    if (kind == "KEYS") {
        return index_metadata_kind::keys;
    } else if (kind == "COMPOSITES") {
        return index_metadata_kind::composites;
    } else if (kind == "CUSTOM") {
        return index_metadata_kind::custom;
    } else {
        throw std::invalid_argument("unknown column kind: " + kind);
    }
}

static void add_index_to_schema_mutation(schema_ptr table,
                                  const index_metadata& index,
                                  api::timestamp_type timestamp,
                                  mutation& m)
{
    auto ckey = clustering_key::from_exploded(*m.schema(), {utf8_type->decompose(table->cf_name()), utf8_type->decompose(index.name())});
    m.set_clustered_cell(ckey, "kind", serialize_index_kind(index.kind()), timestamp);
    store_map(m, ckey, "options", timestamp, index.options());
}

static void drop_index_from_schema_mutation(schema_ptr table, const index_metadata& index, long timestamp, std::vector<mutation>& mutations)
{
    schema_ptr s = indexes();
    auto pkey = partition_key::from_singular(*s, table->ks_name());
    auto ckey = clustering_key::from_exploded(*s, {utf8_type->decompose(table->cf_name()), utf8_type->decompose(index.name())});
    mutation m{s, pkey};
    m.partition().apply_delete(*s, ckey, tombstone(timestamp, gc_clock::now()));
    mutations.push_back(std::move(m));
}

static void drop_column_from_schema_mutation(
        schema_ptr schema_table,
        schema_ptr table,
        const sstring& column_name,
        long timestamp,
        std::vector<mutation>& mutations)
{
    auto pkey = partition_key::from_singular(*schema_table, table->ks_name());
    auto ckey = clustering_key::from_exploded(*schema_table, {utf8_type->decompose(table->cf_name()),
                                                              utf8_type->decompose(column_name)});

    mutation m{schema_table, pkey};
    m.partition().apply_delete(*schema_table, ckey, tombstone(timestamp, gc_clock::now()));
    mutations.emplace_back(m);
}

static computed_columns_map get_computed_columns(const schema_mutations& sm) {
    if (!sm.computed_columns_mutation()) {
        return {};
    }
    query::result_set computed_result(*sm.computed_columns_mutation());
    return boost::copy_range<computed_columns_map>(
            computed_result.rows() | boost::adaptors::transformed([] (const query::result_set_row& row) {
        return computed_columns_map::value_type{to_bytes(row.get_nonnull<sstring>("column_name")), column_computation::deserialize(row.get_nonnull<bytes>("computation"))};
    }));
}

static std::vector<column_definition> create_columns_from_column_rows(const schema_ctxt& ctxt,
                                                               const query::result_set& rows,
                                                               const sstring& keyspace,
                                                               const sstring& table, /*,
                                                               AbstractType<?> rawComparator, */
                                                               bool is_super,
                                                               column_view_virtual is_view_virtual,
                                                               const computed_columns_map& computed_columns)
{
    std::vector<column_definition> columns;
    for (auto&& row : rows.rows()) {
        auto kind = deserialize_kind(row.get_nonnull<sstring>("kind"));
        auto type = cql_type_parser::parse(keyspace, row.get_nonnull<sstring>("type"), ctxt.user_types());
        auto name_bytes = row.get_nonnull<bytes>("column_name_bytes");
        column_id position = row.get_nonnull<int32_t>("position");

        if (auto val = row.get<sstring>("clustering_order")) {
            auto order = *val;
            std::transform(order.begin(), order.end(), order.begin(), ::toupper);
            if (order == "DESC") {
                type = reversed_type_impl::get_instance(type);
            }
        }
        column_computation_ptr computation;
        auto computed_it = computed_columns.find(name_bytes);
        if (computed_it != computed_columns.end()) {
            computation = computed_it->second->clone();
        }

        columns.emplace_back(name_bytes, type, kind, position, is_view_virtual, std::move(computation));
    }
    return columns;
}

static std::vector<index_metadata> create_indices_from_index_rows(const query::result_set& rows,
                                                           const sstring& keyspace,
                                                           const sstring& table)
{
    return boost::copy_range<std::vector<index_metadata>>(rows.rows() | boost::adaptors::transformed([&keyspace, &table] (auto&& row) {
        return create_index_from_index_row(row, keyspace, table);
    }));
}

static index_metadata create_index_from_index_row(const query::result_set_row& row,
                                           sstring keyspace,
                                           sstring table)
{
    auto index_name = row.get_nonnull<sstring>("index_name");
    index_options_map options;
    auto map = row.get_nonnull<map_type_impl::native_type>("options");
    for (auto&& entry : map) {
        options.emplace(value_cast<sstring>(entry.first), value_cast<sstring>(entry.second));
    }
    index_metadata_kind kind = deserialize_index_kind(row.get_nonnull<sstring>("kind"));
    sstring target_string = options.at(cql3::statements::index_target::target_option_name);
    const index_metadata::is_local_index is_local(secondary_index::target_parser::is_local(target_string));
    return index_metadata{index_name, options, kind, is_local};
}

/*
 * View metadata serialization/deserialization.
 */

view_ptr create_view_from_mutations(const schema_ctxt& ctxt, schema_mutations sm, std::optional<table_schema_version> version)  {
    auto table_rs = query::result_set(sm.columnfamilies_mutation());
    query::result_set_row row = table_rs.row(0);

    auto ks_name = row.get_nonnull<sstring>("keyspace_name");
    auto cf_name = row.get_nonnull<sstring>("view_name");
    auto id = table_id(row.get_nonnull<utils::UUID>("id"));

    schema_builder builder{ks_name, cf_name, id};
    prepare_builder_from_table_row(ctxt, builder, row);

    auto computed_columns = get_computed_columns(sm);
    auto column_defs = create_columns_from_column_rows(ctxt, query::result_set(sm.columns_mutation()), ks_name, cf_name, false, column_view_virtual::no, computed_columns);
    for (auto&& cdef : column_defs) {
        builder.with_column_ordered(cdef);
    }
    if (sm.view_virtual_columns_mutation()) {
        column_defs = create_columns_from_column_rows(ctxt, query::result_set(*sm.view_virtual_columns_mutation()), ks_name, cf_name, false, column_view_virtual::yes, computed_columns);
        for (auto&& cdef : column_defs) {
            builder.with_column_ordered(cdef);
        }
    }

    if (version) {
        builder.with_version(*version);
    } else {
        builder.with_version(sm.digest(ctxt.features().cluster_schema_features()));
    }

    auto base_id = table_id(row.get_nonnull<utils::UUID>("base_table_id"));
    auto base_name = row.get_nonnull<sstring>("base_table_name");
    auto include_all_columns = row.get_nonnull<bool>("include_all_columns");
    auto where_clause = row.get_nonnull<sstring>("where_clause");

    builder.with_view_info(std::move(base_id), std::move(base_name), include_all_columns, std::move(where_clause));
    return view_ptr(builder.build());
}

static future<view_ptr> create_view_from_table_row(distributed<service::storage_proxy>& proxy, const query::result_set_row& row) {
    qualified_name qn(row.get_nonnull<sstring>("keyspace_name"), row.get_nonnull<sstring>("view_name"));
    schema_mutations sm = co_await read_table_mutations(proxy, qn, views());
    if (!sm.live()) {
        co_await coroutine::return_exception(std::runtime_error(format("{}:{} not found in the view definitions keyspace.", qn.keyspace_name, qn.table_name)));
    }
    co_return create_view_from_mutations(proxy, std::move(sm));
}

/**
 * Deserialize views from low-level schema representation, all of them belong to the same keyspace
 *
 * @return vector containing the view definitions
 */
future<std::vector<view_ptr>> create_views_from_schema_partition(distributed<service::storage_proxy>& proxy, const schema_result::mapped_type& result)
{
    std::vector<view_ptr> views;
    co_await max_concurrent_for_each(result->rows().begin(), result->rows().end(), max_concurrent, [&] (auto&& row) -> future<> {
        auto v = co_await create_view_from_table_row(proxy, row);
        views.push_back(std::move(v));
    });
    co_return std::move(views);
}

static schema_mutations make_view_mutations(view_ptr view, api::timestamp_type timestamp, bool with_columns)
{
    // When adding new schema properties, don't set cells for default values so that
    // both old and new nodes will see the same version during rolling upgrades.

    // For properties that can be null (and can be changed), we insert tombstones, to make sure
    // we don't keep a property the user has removed
    schema_ptr s = views();
    auto pkey = partition_key::from_singular(*s, view->ks_name());
    mutation m{s, pkey};
    auto ckey = clustering_key::from_singular(*s, view->cf_name());

    m.set_clustered_cell(ckey, "base_table_id", view->view_info()->base_id().uuid(), timestamp);
    m.set_clustered_cell(ckey, "base_table_name", view->view_info()->base_name(), timestamp);
    m.set_clustered_cell(ckey, "where_clause", view->view_info()->where_clause(), timestamp);
    m.set_clustered_cell(ckey, "bloom_filter_fp_chance", view->bloom_filter_fp_chance(), timestamp);
    m.set_clustered_cell(ckey, "include_all_columns", view->view_info()->include_all_columns(), timestamp);
    m.set_clustered_cell(ckey, "id", view->id().uuid(), timestamp);

    add_table_params_to_mutations(m, ckey, view, timestamp);

    mutation columns_mutation(columns(), pkey);
    mutation view_virtual_columns_mutation(view_virtual_columns(), pkey);
    mutation computed_columns_mutation(computed_columns(), pkey);
    mutation dropped_columns_mutation(dropped_columns(), pkey);
    mutation indices_mutation(indexes(), pkey);

    if (with_columns) {
        for (auto&& column : view->v3().all_columns()) {
            if (column.is_view_virtual()) {
                add_column_to_schema_mutation(view, column, timestamp, view_virtual_columns_mutation);
            } else {
                add_column_to_schema_mutation(view, column, timestamp, columns_mutation);
            }
            if (column.is_computed()) {
                add_computed_column_to_schema_mutation(view, column, timestamp, computed_columns_mutation);
            }
        }

        for (auto&& e : view->dropped_columns()) {
            add_dropped_column_to_schema_mutation(view, e.first, e.second, timestamp, dropped_columns_mutation);
        }
        for (auto&& index : view->indices()) {
            add_index_to_schema_mutation(view, index, timestamp, indices_mutation);
        }
    }

    auto scylla_tables_mutation = make_scylla_tables_mutation(view, timestamp);

    return schema_mutations{std::move(m),
                            std::move(columns_mutation),
                            std::move(view_virtual_columns_mutation),
                            std::move(computed_columns_mutation),
                            std::move(indices_mutation),
                            std::move(dropped_columns_mutation),
                            std::move(scylla_tables_mutation)};
}

schema_mutations make_schema_mutations(schema_ptr s, api::timestamp_type timestamp, bool with_columns)
{
    return s->is_view() ? make_view_mutations(view_ptr(s), timestamp, with_columns) : make_table_mutations(s, timestamp, with_columns);
}

std::vector<mutation> make_create_view_mutations(lw_shared_ptr<keyspace_metadata> keyspace, view_ptr view, api::timestamp_type timestamp)
{
    std::vector<mutation> mutations;
    // And also the serialized base table.
    auto base = keyspace->cf_meta_data().at(view->view_info()->base_name());
    add_table_or_view_to_schema_mutation(base, timestamp, true, mutations);
    add_table_or_view_to_schema_mutation(view, timestamp, true, mutations);
    make_table_deleting_mutations(view->ks_name(), view->cf_name(), view->is_view(), timestamp)
        .copy_to(mutations);
    return mutations;
}

/**
 * Note: new_view can be generated due to an ALTER on its base table; in that
 * case, the new base schema isn't yet loaded, thus can't be accessed from this
 * function.
 */
std::vector<mutation> make_update_view_mutations(lw_shared_ptr<keyspace_metadata> keyspace,
                                                 view_ptr old_view,
                                                 view_ptr new_view,
                                                 api::timestamp_type timestamp,
                                                 bool include_base)
{
    std::vector<mutation> mutations;
    if (include_base) {
        // Include the serialized base table mutations in case the target node is missing them.
        auto base = keyspace->cf_meta_data().at(new_view->view_info()->base_name());
        add_table_or_view_to_schema_mutation(base, timestamp, true, mutations);
    }
    add_table_or_view_to_schema_mutation(new_view, timestamp, false, mutations);
    make_update_columns_mutations(old_view, new_view, timestamp, false, mutations);
    return mutations;
}

std::vector<mutation> make_drop_view_mutations(lw_shared_ptr<keyspace_metadata> keyspace, view_ptr view, api::timestamp_type timestamp) {
    std::vector<mutation> mutations;
    make_drop_table_or_view_mutations(views(), view, timestamp, mutations);
    return mutations;
}

#if 0
    private static AbstractType<?> getComponentComparator(AbstractType<?> rawComparator, Integer componentIndex)
    {
        return (componentIndex == null || (componentIndex == 0 && !(rawComparator instanceof CompositeType)))
               ? rawComparator
               : ((CompositeType)rawComparator).types.get(componentIndex);
    }

    /*
     * Trigger metadata serialization/deserialization.
     */

    private static void addTriggerToSchemaMutation(CFMetaData table, TriggerDefinition trigger, long timestamp, Mutation mutation)
    {
        ColumnFamily cells = mutation.addOrGet(Triggers);
        Composite prefix = Triggers.comparator.make(table.cfName, trigger.name);
        CFRowAdder adder = new CFRowAdder(cells, prefix, timestamp);
        adder.addMapEntry("trigger_options", "class", trigger.classOption);
    }

    private static void dropTriggerFromSchemaMutation(CFMetaData table, TriggerDefinition trigger, long timestamp, Mutation mutation)
    {
        ColumnFamily cells = mutation.addOrGet(Triggers);
        int ldt = (int) (System.currentTimeMillis() / 1000);

        Composite prefix = Triggers.comparator.make(table.cfName, trigger.name);
        cells.addAtom(new RangeTombstone(prefix, prefix.end(), timestamp, ldt));
    }

    /**
     * Deserialize triggers from storage-level representation.
     *
     * @param partition storage-level partition containing the trigger definitions
     * @return the list of processed TriggerDefinitions
     */
    private static List<TriggerDefinition> createTriggersFromTriggersPartition(Row partition)
    {
        List<TriggerDefinition> triggers = new ArrayList<>();
        String query = String.format("SELECT * FROM %s.%s", SystemKeyspace.NAME, TRIGGERS);
        for (UntypedResultSet.Row row : QueryProcessor.resultify(query, partition))
        {
            String name = row.getString("trigger_name");
            String classOption = row.getMap("trigger_options", UTF8Type.instance, UTF8Type.instance).get("class");
            triggers.add(new TriggerDefinition(name, classOption));
        }
        return triggers;
    }

    /*
     * Aggregate UDF metadata serialization/deserialization.
     */

    public static Mutation makeCreateAggregateMutation(KSMetaData keyspace, UDAggregate aggregate, long timestamp)
    {
        // Include the serialized keyspace in case the target node missed a CREATE KEYSPACE migration (see CASSANDRA-5631).
        Mutation mutation = makeCreateKeyspaceMutation(keyspace, timestamp, false);
        addAggregateToSchemaMutation(aggregate, timestamp, mutation);
        return mutation;
    }

    private static void addAggregateToSchemaMutation(UDAggregate aggregate, long timestamp, Mutation mutation)
    {
        ColumnFamily cells = mutation.addOrGet(Aggregates);
        Composite prefix = Aggregates.comparator.make(aggregate.name().name, UDHelper.calculateSignature(aggregate));
        CFRowAdder adder = new CFRowAdder(cells, prefix, timestamp);

        adder.resetCollection("argument_types");
        adder.add("return_type", aggregate.returnType().toString());
        adder.add("state_func", aggregate.stateFunction().name().name);
        if (aggregate.stateType() != null)
            adder.add("state_type", aggregate.stateType().toString());
        if (aggregate.finalFunction() != null)
            adder.add("final_func", aggregate.finalFunction().name().name);
        if (aggregate.initialCondition() != null)
            adder.add("initcond", aggregate.initialCondition());

        for (AbstractType<?> argType : aggregate.argTypes())
            adder.addListEntry("argument_types", argType.toString());
    }

    private static Map<ByteBuffer, UDAggregate> createAggregatesFromAggregatesPartition(Row partition)
    {
        Map<ByteBuffer, UDAggregate> aggregates = new HashMap<>();
        String query = String.format("SELECT * FROM %s.%s", SystemKeyspace.NAME, AGGREGATES);
        for (UntypedResultSet.Row row : QueryProcessor.resultify(query, partition))
        {
            UDAggregate aggregate = createAggregateFromAggregateRow(row);
            aggregates.put(UDHelper.calculateSignature(aggregate), aggregate);
        }
        return aggregates;
    }

    private static UDAggregate createAggregateFromAggregateRow(UntypedResultSet.Row row)
    {
        String ksName = row.getString("keyspace_name");
        String functionName = row.getString("aggregate_name");
        FunctionName name = new FunctionName(ksName, functionName);

        List<String> types = row.getList("argument_types", UTF8Type.instance);

        List<AbstractType<?>> argTypes;
        if (types == null)
        {
            argTypes = Collections.emptyList();
        }
        else
        {
            argTypes = new ArrayList<>(types.size());
            for (String type : types)
                argTypes.add(parseType(type));
        }

        AbstractType<?> returnType = parseType(row.getString("return_type"));

        FunctionName stateFunc = new FunctionName(ksName, row.getString("state_func"));
        FunctionName finalFunc = row.has("final_func") ? new FunctionName(ksName, row.getString("final_func")) : null;
        AbstractType<?> stateType = row.has("state_type") ? parseType(row.getString("state_type")) : null;
        ByteBuffer initcond = row.has("initcond") ? row.getBytes("initcond") : null;

        try
        {
            return UDAggregate.create(name, argTypes, returnType, stateFunc, finalFunc, stateType, initcond);
        }
        catch (InvalidRequestException reason)
        {
            return UDAggregate.createBroken(name, argTypes, returnType, initcond, reason);
        }
    }

    public static Mutation makeDropAggregateMutation(KSMetaData keyspace, UDAggregate aggregate, long timestamp)
    {
        // Include the serialized keyspace in case the target node missed a CREATE KEYSPACE migration (see CASSANDRA-5631).
        Mutation mutation = makeCreateKeyspaceMutation(keyspace, timestamp, false);

        ColumnFamily cells = mutation.addOrGet(Aggregates);
        int ldt = (int) (System.currentTimeMillis() / 1000);

        Composite prefix = Aggregates.comparator.make(aggregate.name().name, UDHelper.calculateSignature(aggregate));
        cells.addAtom(new RangeTombstone(prefix, prefix.end(), timestamp, ldt));

        return mutation;
    }
#endif

data_type parse_type(sstring str)
{
    return db::marshal::type_parser::parse(str);
}

std::vector<schema_ptr> all_tables(schema_features features) {
    // Don't forget to update this list when new schema tables are added.
    // The listed schema tables are the ones synchronized between nodes,
    // and forgetting one of them in this list can cause bugs like #4339.
    //
    // This list must be kept backwards-compatible because it's used
    // for schema digest calculation. Refs #4457.
    std::vector<schema_ptr> result = {
        keyspaces(), tables(), scylla_tables(), columns(), dropped_columns(), triggers(),
        views(), types(), functions(), aggregates(), indexes()
    };
    if (features.contains<schema_feature::VIEW_VIRTUAL_COLUMNS>()) {
        result.emplace_back(view_virtual_columns());
    }
    if (features.contains<schema_feature::COMPUTED_COLUMNS>()) {
        result.emplace_back(computed_columns());
    }
    if (features.contains<schema_feature::SCYLLA_KEYSPACES>()) {
        result.emplace_back(scylla_keyspaces());
    }
    if (features.contains<schema_feature::SCYLLA_AGGREGATES>()) {
        result.emplace_back(scylla_aggregates());
    }
    return result;
}

std::vector<sstring> all_table_names(schema_features features) {
    return boost::copy_range<std::vector<sstring>>(all_tables(features) |
           boost::adaptors::transformed([] (auto schema) { return schema->cf_name(); }));
}

view_ptr maybe_fix_legacy_secondary_index_mv_schema(replica::database& db, const view_ptr& v, schema_ptr base_schema, preserve_version preserve_version) {
    // Legacy format for a secondary index used a hardcoded "token" column, which ensured a proper
    // order for indexed queries. This "token" column is now implemented as a computed column,
    // but for the sake of compatibility we assume that there might be indexes created in the legacy
    // format, where "token" is not marked as computed. Once we're sure that all indexes have their
    // columns marked as computed (because they were either created on a node that supports computed
    // columns or were fixed by this utility function), it's safe to remove this function altogether.
    if (v->clustering_key_size() == 0) {
        return view_ptr(nullptr);
    }
    const auto ck_cols = v->clustering_key_columns();
    const column_definition& first_view_ck = ck_cols.front();
    if (first_view_ck.is_computed()) {
        return view_ptr(nullptr);
    }

    if (!base_schema) {
        base_schema = db.find_schema(v->view_info()->base_id());
    }

    // If the first clustering key part of a view is a column with name not found in base schema,
    // it implies it might be backing an index created before computed columns were introduced,
    // and as such it must be recreated properly.
    if (!base_schema->columns_by_name().contains(first_view_ck.name())) {
        schema_builder builder{schema_ptr(v)};
        builder.mark_column_computed(first_view_ck.name(), std::make_unique<legacy_token_column_computation>());
        if (preserve_version) {
            builder.with_version(v->version());
        }
        return view_ptr(builder.build());
    }
    return view_ptr(nullptr);
}


namespace legacy {

table_schema_version schema_mutations::digest() const {
    md5_hasher h;
    const db::schema_features no_features;
    db::schema_tables::feed_hash_for_schema_digest(h, _columnfamilies, no_features);
    db::schema_tables::feed_hash_for_schema_digest(h, _columns, no_features);
    return table_schema_version(utils::UUID_gen::get_name_UUID(h.finalize()));
}

future<schema_mutations> read_table_mutations(distributed<service::storage_proxy>& proxy,
    sstring keyspace_name, sstring table_name, schema_ptr s)
{
    mutation cf_m = co_await read_schema_partition_for_table(proxy, s, keyspace_name, table_name);
    mutation col_m = co_await read_schema_partition_for_table(proxy, db::system_keyspace::legacy::columns(), keyspace_name, table_name);
    co_return schema_mutations{std::move(cf_m), std::move(col_m)};
}

} // namespace legacy

static auto GET_COLUMN_MAPPING_QUERY = format("SELECT column_name, clustering_order, column_name_bytes, kind, position, type FROM system.{} WHERE cf_id = ? AND schema_version = ?",
    db::schema_tables::SCYLLA_TABLE_SCHEMA_HISTORY);

future<column_mapping> get_column_mapping(::table_id table_id, table_schema_version version) {
    shared_ptr<cql3::untyped_result_set> results = co_await qctx->qp().execute_internal(
        GET_COLUMN_MAPPING_QUERY,
        db::consistency_level::LOCAL_ONE,
        {table_id.uuid(), version.uuid()},
        cql3::query_processor::cache_internal::no
    );
    if (results->empty()) {
        // If we don't have a stored column_mapping for an obsolete schema version
        // then it means it's way too old and been cleaned up already.
        // Fail the whole learn stage in this case.
        co_await coroutine::return_exception(std::runtime_error(
            format("Failed to look up column mapping for schema version {}",
                version)));
    }
    std::vector<column_definition>  static_columns, regular_columns;
    for (const auto& row : *results) {
        auto kind = deserialize_kind(row.get_as<sstring>("kind"));
        auto type = cql_type_parser::parse("" /*unused*/, row.get_as<sstring>("type"), data_dictionary::dummy_user_types_storage());
        auto name_bytes = row.get_blob("column_name_bytes");
        column_id position = row.get_as<int32_t>("position");

        auto order = row.get_as<sstring>("clustering_order");
        std::transform(order.begin(), order.end(), order.begin(), ::toupper);
        if (order == "DESC") {
            type = reversed_type_impl::get_instance(type);
        }
        if (kind == column_kind::static_column) {
            static_columns.emplace_back(name_bytes, type, kind, position);
        } else if (kind == column_kind::regular_column) {
            regular_columns.emplace_back(name_bytes, type, kind, position);
        }
    }
    std::vector<column_mapping_entry> cm_columns;
    for (const column_definition& def : boost::range::join(static_columns, regular_columns)) {
        cm_columns.emplace_back(column_mapping_entry{def.name(), def.type});
    }
    column_mapping cm(std::move(cm_columns), static_columns.size());
    co_return std::move(cm);
}

future<bool> column_mapping_exists(table_id table_id, table_schema_version version) {
    shared_ptr<cql3::untyped_result_set> results = co_await qctx->qp().execute_internal(
        GET_COLUMN_MAPPING_QUERY,
        db::consistency_level::LOCAL_ONE,
        {table_id.uuid(), version.uuid()},
        cql3::query_processor::cache_internal::yes
    );
    co_return !results->empty();
}

future<> drop_column_mapping(table_id table_id, table_schema_version version) {
    const static sstring DEL_COLUMN_MAPPING_QUERY =
        format("DELETE FROM system.{} WHERE cf_id = ? and schema_version = ?",
            db::schema_tables::SCYLLA_TABLE_SCHEMA_HISTORY);
    co_await qctx->qp().execute_internal(
        DEL_COLUMN_MAPPING_QUERY,
        db::consistency_level::LOCAL_ONE,
        {table_id.uuid(), version.uuid()},
        cql3::query_processor::cache_internal::no);
}

} // namespace schema_tables
} // namespace schema