A thin SQLite wrapper for Zig

// good idea to pass EXResCode to get extended result codes (more detailed error codes)
const flags =  zqlite.OpenFlags.Create | zqlite.OpenFlags.EXResCode;
var conn = try zqlite.open("/tmp/test.sqlite", flags);
defer conn.close();

try conn.exec("create table if not exists test (name text)", .{});
try conn.exec("insert into test (name) values (?1), (?2)", .{"Leto", "Ghanima"});

{
    if (try conn.row("select * from test order by name limit 1", .{})) |row| {
      defer row.deinit();
      std.debug.print("name: {s}\n", .{row.text(0)});
    }
}

{
    var rows = try conn.rows("select * from test order by name", .{});
    defer rows.deinit();
    while (rows.next()) |row| {
        std.debug.print("name: {s}\n", .{row.text(0)});
    }
    if (rows.err) |err| return err;
}

Unless zqlite.OpenFlags.ReadOnly is set in the open flags, zqlite.OpenFlags.ReadWrite is assumed (in other words, the database opens in read-write by default, and the ReadOnly flag must be used to open it in readony mode.)

Install

This library is tested with SQLite3 3.46.1 .

1. Add zqlite as a dependency in your build.zig.zon:

zig fetch --save git+https://github.com/karlseguin/zqlite.zig#master

2. The library doesn't attempt to link/include SQLite. You're free to do this how you want.

If you have sqlite3 installed on your system you might get away with just adding this to your build.zig

const zqlite = b.dependency("zqlite", .{
    .target = target,
    .optimize = optimize,
});

exe.linkLibC();
exe.linkSystemLibrary("sqlite3");
exe.root_module.addImport("zqlite", zqlite.module("zqlite"));

Alternatively, If you download the SQLite amalgamation from the SQLite download page and place the sqlite.c and sqlite.h file in your project's lib/ folder, you can then:

2. Add this in build.zig:

const zqlite = b.dependency("zqlite", .{
    .target = target,
    .optimize = optimize,
});
exe.addCSourceFile(.{
    .file = b.path("lib/sqlite3.c"),
    .flags = &[_][]const u8{
        "-DSQLITE_DQS=0",
        "-DSQLITE_DEFAULT_WAL_SYNCHRONOUS=1",
        "-DSQLITE_USE_ALLOCA=1",
        "-DSQLITE_THREADSAFE=1",
        "-DSQLITE_TEMP_STORE=3",
        "-DSQLITE_ENABLE_API_ARMOR=1",
        "-DSQLITE_ENABLE_UNLOCK_NOTIFY",
        "-DSQLITE_ENABLE_UPDATE_DELETE_LIMIT=1",
        "-DSQLITE_DEFAULT_FILE_PERMISSIONS=0600",
        "-DSQLITE_OMIT_DECLTYPE=1",
        "-DSQLITE_OMIT_DEPRECATED=1",
        "-DSQLITE_OMIT_LOAD_EXTENSION=1",
        "-DSQLITE_OMIT_PROGRESS_CALLBACK=1",
        "-DSQLITE_OMIT_SHARED_CACHE",
        "-DSQLITE_OMIT_TRACE=1",
        "-DSQLITE_OMIT_UTF16=1",
        "-DHAVE_USLEEP=0",
    },
});
exe.linkLibC();
exe.root_module.addImport("zqlite", zqlite.module("zqlite"));

You can tweak the SQLite build flags for your own needs/platform.

Conn

The Conn type returned by open has the following functions:

• row(sql, args) !?zqlite.Row - returns an optional row
• rows(sql, args) !zqlite.Rows - returns an iterator that yields rows
• exec(sql, args) !void - executes the statement,
• execNoArgs(sql) !void - micro-optimization if there are no args, sql must be a null-terminated string
• changes() usize - the number of rows inserted/updated/deleted by the previous statement
• lastInsertedRowId() i64 - the row id of the last inserted row
• lastError() [*:0]const u8 - an error string describing the last error
• transaction() !void and exclusiveTransaction() !void - begins a transaction
• commit() !void and rollback() void - commits and rollback the current transaction
• prepare(sql, args) !zqlite.Stmt - returns a thin wrapper around a *c.sqlite3_stmt. row and rows wrap this type.
• close() void and tryClsoe() !void - closes the database. close() silently ignores any error, if you care about the error, use tryClose()
• busyTimeout(ms) - Sets the busyHandler for the connection. See https://www.sqlite.org/c3ref/busy_timeout.html

Row and Rows

Both row and rows wrap an zqlite.Stmt which itself is a thin wrapper around an *c.sqlite3_stmt.

While zqlite.Row exposes a deinit and deinitErr method, it should only be called when the row was fetched directly from conn.row(...):

if (try conn.row("select 1", .{})) |row| {
    defer row.deinit();  // must be called
    std.debug.print("{d}\n", .{row.int(0)});
}

When the row comes from iterating rows, deinit or deinitErr should not be called on the individual row:

var rows = try conn.rows("select 1 union all select 2", .{})
defer rows.deinit();  // must be called

while (rows.next()) |row| {
    // row.deinit() should not be called!
    ...
}

Note that zqlite.Rows has an err: ?anyerror field which can be checked at any point. Calls to next() when err != null will return null. Thus, err need only be checked at the end of the loop:

var rows = try conn.rows("select 1 union all select 2", .{})
defer rows.deinit();  // must be called

while (rows.next()) |row| {
    ...
}

if (rows.err) |err| {
    // something went wrong 
}

Row Getters

A row exposes the following functions to fetch data:

• boolean(index) bool
• nullableBoolean(index) ?bool
• int(index) i64
• nullableInt(index) ?i64
• float(index) i64
• nullableFloat(index) f64
• text(index) []const u8
• nullableText(index) ?[]const u8
• blob(index) []const u8
• nullableBlob(index) ?[]const u8

The nullableXYZ functions can safely be called on a not null column. The non-nullable versions avoid a call to sqlite3_column_type (which is needed in the nullable versions to determine if the value is null or not).

Transaction:

The transaction(), exclusiveTransaction(), commit() and rollback() functions are simply wrappers to conn.execNoArgs("begin"), conn.execNoArgs("begin exclusive"), conn.execNoArgs("commit") and conn.execNoArgs("rollback")

conn.transaction()
errdefer conn.rollback();

try conn.exec(...);
try conn.exec(...);
try conn.commit();

Blobs

When binding a []const u8, this library has no way to tell whether the value should be treated as an text or blob. It defaults to text. To have the value bound as a blob use zqlite.blob(value):

conn.insert("insert into records (image) values (?1)", .{zqlite.blob(image)})

However, this should only be necessary in specific cases where SQLite blob-specific operations are used on the data. Text and blob are practically the same, except they have a different type.

Pool

zqlite.Pool is a simple thread-safe connection pool. After being created, the acquire and release functions are used to get a connection from the pool and to release it.

var pool = try zqlite.Pool.init(allocator, .{
    // The number of connection in the pool. The pool will not grow or
    // shrink beyond this count
    .size = 5,   // default 5

    // The path  of the DB connection
    .path = "/tmp/zqlite.sqlite",  // no default, required

    // The zqlite.OpenFlags to use when opening each connection in the pool
    // Defaults are as shown here:
    .flags = zqlite.OpenFlags.Create | zqlite.OpenFlags.EXResCode  

    // Callback function to execute for each connection in the pool when opened
    .on_connection = null,

    // Callback function to execute only for the first connection in the pool
    .on_first_connection = null,
});

const c1 = pool.acquire();
defer c1.release();
c1.execNoArgs(...);

Callbacks

Both the on_connection and on_first_connection have the same signature. For the first connection to be opened by the pool, if both callbacks are provided then both callbacks will be executed, with on_first_connection executing first.

var pool = zqlite.Pool.init(allocator, .{
    .size = 5,
    .on_first_connection = &initializeDB,
    .on_connection = &initializeConnection,
    // other required & optional fields
});
...

// Our size is 5, but this will only be executed once, for the first
// connection in our pool
fn initializeDB(conn: Conn) !void {
    try conn.execNoArgs("create table if not exists testing(id int)");
}

// Our size is 5, so this will be executed 5 times, once for each
// connection. `initializeDB` is guaranteed to be called before this
// function is called.
fn initializeConnection(conn: Conn) !void {
    return conn.busyTimeout(1000);
}