fetchRows<T> method
Return data back and convert each row to a type T, using the
rowMapper function.
Implementation
List<T> fetchRows<T>(String query, T Function(List<Object?>) rowMapper) {
var q = query.toNativeUtf8().cast<Char>();
var resultPtr = calloc<duckdb_result>();
if (bindings.duckdb_query(ptrCon.value, q, resultPtr) ==
duckdb_state.DuckDBError) {
throw StateError(
bindings.duckdb_result_error(resultPtr).cast<Utf8>().toDartString());
}
var out = <T>[];
var chunkCount = bindings.duckdb_result_chunk_count(resultPtr.ref);
for (var chunk = 0; chunk < chunkCount; chunk++) {
var chunkPtr = bindings.duckdb_result_get_chunk(resultPtr.ref, chunk);
var rowCount = bindings.duckdb_data_chunk_get_size(chunkPtr);
var colCount = bindings.duckdb_data_chunk_get_column_count(chunkPtr);
var rs =
List.generate(rowCount, (i) => List<Object?>.filled(colCount, null));
for (var j = 0; j < colCount; j++) {
var vector = bindings.duckdb_data_chunk_get_vector(chunkPtr, j);
var logicalType = bindings.duckdb_vector_get_column_type(vector);
var typeId = bindings.duckdb_get_type_id(logicalType);
var values = bindings.duckdb_vector_get_data(vector);
var validity = bindings.duckdb_vector_get_validity(vector);
switch (typeId) {
case DUCKDB_TYPE.DUCKDB_TYPE_BOOLEAN:
var xs = values.cast<Bool>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_TINYINT:
var xs = values.cast<Int8>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_SMALLINT:
var xs = values.cast<Int16>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_INTEGER:
var xs = values.cast<Int32>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_BIGINT:
var xs = values.cast<Int64>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_UTINYINT:
var xs = values.cast<Uint8>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_USMALLINT:
var xs = values.cast<Uint16>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_UINTEGER:
var xs = values.cast<Uint32>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_UBIGINT:
var xs = values.cast<Uint64>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_FLOAT: // 4 bytes
var xs = values.cast<Float>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_DOUBLE: // 8 bytes
var xs = values.cast<Double>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE
.DUCKDB_TYPE_TIMESTAMP: // UTC DateTime, microsecond precision
var xs = values.cast<Int64>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_DATE: // number of days since 1970-01-01
var xs = values.cast<Int32>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE.DUCKDB_TYPE_VARCHAR:
// see test 'Test DataChunk varchar result fetch in C API'
// https://github.com/duckdb/duckdb/blob/main/test/api/capi/test_capi_data_chunk.cpp#L260
var xs = values.cast<duckdb_string_t>();
for (var i = 0; i < rowCount; i++) {
if (!bindings.duckdb_validity_row_is_valid(validity, i)) continue;
final tuple = xs[i];
if (bindings.duckdb_string_is_inlined(tuple)) {
// The data is small enough to fit in the string_t, it does not have a separate allocation
var array = tuple.value.inlined.inlined;
final charCodes = <int>[];
var k = 0;
while (array[k] != 0) {
charCodes.add(array[k]);
k++;
}
rs[i][j] = String.fromCharCodes(charCodes);
} else {
rs[i][j] = tuple.value.pointer.ptr.cast<Utf8>().toDartString();
}
}
case DUCKDB_TYPE.DUCKDB_TYPE_DECIMAL:
/// https://github.com/Giorgi/DuckDB.NET/blob/8520bf5005d9309f762ef61d71412d60d24ca32c/DuckDB.NET.Data/Internal/Reader/DecimalVectorDataReader.cs#L43
var type = bindings.duckdb_decimal_internal_type(logicalType);
var scale = bindings.duckdb_decimal_scale(logicalType);
for (var i = 0; i < rowCount; i++) {
switch (type) {
case DUCKDB_TYPE.DUCKDB_TYPE_SMALLINT ||
DUCKDB_TYPE.DUCKDB_TYPE_INTEGER:
var xs = values.cast<Int32>();
if (!bindings.duckdb_validity_row_is_valid(validity, i)) {
continue;
}
rs[i][j] = (Decimal.fromInt(xs[i]) /
Decimal.ten.pow(scale).toDecimal())
.toDecimal();
case DUCKDB_TYPE.DUCKDB_TYPE_BIGINT:
var xs = values.cast<Int64>();
if (!bindings.duckdb_validity_row_is_valid(validity, i)) {
continue;
}
rs[i][j] = (Decimal.fromInt(xs[i]) /
Decimal.ten.pow(scale).toDecimal())
.toDecimal();
case _:
throw StateError('Unsupported decimal type $type');
}
}
case DUCKDB_TYPE
.DUCKDB_TYPE_TIMESTAMP_S: // UTC DateTime, second precision
var xs = values.cast<Int64>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: DateTime.fromMillisecondsSinceEpoch(xs[i] * 1000,
isUtc: true);
}
case DUCKDB_TYPE
.DUCKDB_TYPE_TIMESTAMP_TZ: // return the number of microseconds
var xs = values.cast<Int64>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: xs[i];
}
case DUCKDB_TYPE
.DUCKDB_TYPE_TIMESTAMP_MS: // UTC DateTime, millisecond precision
var xs = values.cast<Int64>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: DateTime.fromMillisecondsSinceEpoch(xs[i], isUtc: true);
}
case DUCKDB_TYPE
.DUCKDB_TYPE_TIMESTAMP_NS: // UTC DateTime, nanosecond precision
var xs = values.cast<Int64>();
for (var i = 0; i < rowCount; i++) {
rs[i][j] = !bindings.duckdb_validity_row_is_valid(validity, i)
? null
: DateTime.fromMillisecondsSinceEpoch(xs[i] ~/ 1000,
isUtc: true);
}
case DUCKDB_TYPE.DUCKDB_TYPE_ENUM:
// there are several internal types for ENUMs based on the size
// of the dictionary (uint8_t, uint16_t, uint32_t)
// See https://github.com/duckdb/duckdb/blob/main/test/api/capi/test_capi_complex_types.cpp#L52
var enumInternalType =
bindings.duckdb_enum_internal_type(logicalType);
for (var i = 0; i < rowCount; i++) {
if (!bindings.duckdb_validity_row_is_valid(validity, i)) continue;
// get the index of the enum dictionary for this row
late int idx;
if (enumInternalType == DUCKDB_TYPE.DUCKDB_TYPE_UTINYINT) {
idx = (values as Pointer<Uint8>)[i];
} else if (enumInternalType ==
DUCKDB_TYPE.DUCKDB_TYPE_USMALLINT) {
idx = (values as Pointer<Uint16>)[i];
} else if (enumInternalType == DUCKDB_TYPE.DUCKDB_TYPE_UINTEGER) {
idx = (values as Pointer<Uint32>)[i];
}
rs[i][j] = bindings
.duckdb_enum_dictionary_value(logicalType, idx)
.cast<Utf8>()
.toDartString();
}
default:
throw StateError('TypeId $typeId has not been mapped yet');
}
// clean the column
var ptr = calloc<duckdb_logical_type>();
ptr.value = Pointer.fromAddress(logicalType.address);
bindings.duckdb_destroy_logical_type(ptr);
}
for (var i = 0; i < rowCount; i++) {
out.add(rowMapper(rs[i]));
}
// clean the chunk
var ptr = calloc<duckdb_data_chunk>();
ptr.value = Pointer.fromAddress(chunkPtr.address);
bindings.duckdb_destroy_data_chunk(ptr);
}
bindings.duckdb_destroy_result(resultPtr);
return out;
}