sqlxx

C++11 headers only cursors for SQLite / MySQL(MariaDB) / PostreSQL

Motivation:

• extremely low-memory-footprint (it depends on result set)
• blazing fast speed (it depends on intra/inter net connection speed)
• interact with stl algorithms
• interact with range-based 'for' loops
• abstract layer over SQL native backends
• sql bindings
• streams
• easy to use
• thread safe (with limitations)

You should:

• use all database backends at same time or just only one
• use '?' for binding in all backends, in PostreSQL it's transform to '${\d}'
• use 'values' helper for bindings, it's std::tuple under the hood
• use BLOB as sql type in all backend, in PostreSQL it's transform to BYTEA
• use 'blob' helper for binding blobs
• use 'value' helper for named bindings (where available), it's std::pair
• use '{\d}' to build dynamic query only for column names
• use 'format' helper for '{\d}', it's std::initializer_list< std::string >
• use std::string for queries only with query contructor
• use operator<< C-string for queries
• use operator<< std::string for binding
• benefit auto escaped \ and '
• define USE_SHARED_CONNECTION in threaded environment

You should NOT:

• use operator<< std::string for queries (it will be bind)
• use '{\d}' to build dynamic queries except for column names, \ and ' are not escaped
• share cursor or iterator between threads
• cache iterators, cursor menage a shared statement, iterator has a weak reference to it
• allow your cursor/iterator to outlive your database connection i.e. auto cursor = sqlitexx::connection::create()->query()->execute();
• use 'using namespace', different backends shares same names

Test compilation on Linux:

• you need libmysqlclient/libmariadbclient, sqlite3, postresql-libs (abi+api packages)
• postgresql, mysql/mariadb server packages
• start service, create user, database
• g++ -Wall --std=c++11 -O3 -s test.cpp -o test -lmysqlclient -lsqlite3 -lpq -lpthread
• -lpthread may needed due to gcc bug

Contributions are welcome