auto">The cache do not affect the API.
Where to implement:

Inside sqlite3Prepare and sqlite3_finalize. Such way allows to use cache both for iproto and lua requests (sql_prepare_and_execute and lua_sql_execute);
Inside tarantool as a special subsystem.

How to implement:

enum { PREPARED_STATEMENTS_MAX = 100 };
struct prepared_stmts {
    hash_t table = new_hash(); /* To fast search by SQL. */
    heap_t heap = new_heap(); /* To fast sort by id. */
    uint32_t id_sequence = 0;
};
hash_key_t = const char *; /* sql request. */
hash_value_t = struct {
    uint32_t schema_version;
    uint32_t id;
    sqlite3_stmt *stmt;
};

heap_key_t = uint32_t; /* id. */
heap_value_t = hash_elem_t;

If the statement to prepare is new, then prepare it, save the prepared statement, the current schema version and the current id_sequence in the hash and in the heap. If the hash is full, then pop the statement with minimal id from the heap and from the hash. If the statement is found in the hash, then check schema version and update its id to the ++id_sequence.

schema version from the hash == current schema version -> use the prepared statement;
schema version from the hash != current schema version -> reprepare the statement.

sql: implement prepared statements cache #2592

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions