1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499
//! Provides a trait [`ReadColumn`] for extracting a single column from a [`crate::table::RowRef`].
//! This is desirable as frequently, e.g. when evaluating filtered queries,
//! we are interested in only a single column (or a small set of columns),
//! and would like to avoid the allocation required by a `ProductValue`.
use crate::{
bflatn_from,
indexes::{PageOffset, Size},
layout::{AlgebraicTypeLayout, HasLayout, PrimitiveType, ProductTypeElementLayout, VarLenType},
table::RowRef,
util::slice_assume_init_ref,
};
use spacetimedb_sats::{
algebraic_value::ser::ValueSerializer, AlgebraicType, AlgebraicValue, ArrayValue, MapValue, ProductType,
ProductValue, SumValue,
};
use std::{cell::Cell, mem};
use thiserror::Error;
#[derive(Error, Debug)]
pub enum TypeError {
#[error(
"Attempt to read column {} of a product with only {} columns of type {:?}",
desired,
found.elements.len(),
found,
)]
IndexOutOfBounds { desired: usize, found: ProductType },
#[error("Attempt to read a column at type `{desired}`, but the column's type is {found:?}")]
WrongType {
desired: &'static str,
found: AlgebraicType,
},
}
/// Types which can be stored in a column of a row,
/// and can be extracted directly from a row.
///
/// # Safety
///
/// The implementor must define `is_compatible_type` to return `true` only for `AlgebraicTypeLayout`s
/// for which `unchecked_read_column` is safe.
/// The provided `read_column` method uses `is_compatible_type` to detect type errors,
/// and calls `unchecked_read_column` if `is_compatible_type` returns true.
pub unsafe trait ReadColumn: Sized {
/// Is `ty` compatible with `Self`?
///
/// The definition of "compatible" here is left to the implementor,
/// to be defined by `Self::is_compatible_type`.
///
/// For most types,"compatibility" will mean that each Rust type which implements `ReadColumn`
/// has exactly one corresponding [`AlgebraicTypeLayout`] which represents it,
/// and the column in `table.row_layout` must be of that type.
///
/// Notable exceptions are [`AlgebraicValue`], [`ProductValue`] and [`SumValue`].
/// Any `ProductTypeLayout` is compatible with `ProductValue`,
/// any `SumTypeLayout` is compatible with `SumValue`,
/// and any `AlgebraicTypeLayout` at all is compatible with `AlgebraicValue`.
fn is_compatible_type(ty: &AlgebraicTypeLayout) -> bool;
/// Extract a value of type `Self` from the row pointed to by `row_ref`
/// which is stored in the column defined by `layout`.
///
/// # Safety
///
/// `layout` must appear as a column in the `table.row_layout.product().elements`,
/// *not* to a nested field of a column which is a product or sum value.
/// That column must have the same layout as `layout`.
/// This restriction may be loosened in the future.
///
/// Assuming that the `row_ref` refers to a properly-aligned row,
/// adding the `layout.offset` must result in a properly-aligned value of that compatible type.
///
/// `layout.ty` must be compatible with `Self`.
/// The definition of "compatible" here is left to the implementor,
/// to be defined by `Self::is_compatible_type`.
///
/// For most types,"compatibility" will mean that each Rust type which implements `ReadColumn`
/// has exactly one corresponding [`AlgebraicTypeLayout`] which represents it,
/// and the column in `table.row_layout` must be of that type.
///
/// Notable exceptions are [`AlgebraicValue`], [`ProductValue`] and [`SumValue`].
/// Any `ProductTypeLayout` is compatible with `ProductValue`,
/// any `SumTypeLayout` is compatible with `SumValue`,
/// and any `AlgebraicTypeLayout` at all is compatible with `AlgebraicValue`.
///
/// # Notes for implementors
///
/// Implementors may depend on all of the above safety requirements,
/// and on the validity of the `row_ref`.
/// Assuming all of the above safety requirements are met and the `row_ref` refers to a valid row,
/// this method *must never* invoke Undefined Behavior.
///
/// Implementors should carefully study the BFLATN format.
/// Currently BFLATN lacks a normative specification,
/// so implementors should read the definitions in [`layout.rs`], [`bflatn_to.rs`] and [`bflatn_from.rs`].
/// A few highlights are included here:
///
/// - Variable-length columns, i.e. `AlgebraicType::String`, `AlgebraicType::Array` and `AlgebraicType::Map`
/// are stored within the row as [`crate::var_len::VarLenRef`s],
/// which refer to an intrusive linked list of 62-byte "granules",
/// allocated separately in a space starting from the end of the page.
/// Strings are stored as UTF-8 bytes; all other var-len types are stored as BSATN-encoded bytes.
///
/// - Fixed-length columns, i.e. all types not listed above as variable-length,
/// are stored inline at a known offset.
/// Their layout generally matches the C ABI on an x86_64 Linux machine,
/// with the notable exception of sum types, since the C ABI doesn't define a layout for sums.
///
/// - Fixed-length columns are stored in order, with padding between to ensure proper alignment.
///
/// - Primitive (non-compound) fixed-length types, i.e. integers, floats and booleans,
/// have alignment equal to their size.
///
/// - Integers are stored little-endian.
///
/// - Floats are stored by bitwise converting to integers as per IEEE-754,
/// then storing those integers little-endian.
///
/// - Booleans are stored as `u8`, i.e. bytes, restricted to the values `0` and `1`.
///
/// - Products store their elements in order, with padding between to ensure proper alignment.
///
/// - The first element of a product has offset 0.
///
/// - The alignment of a product is the maximum alignment of its elements,
/// or 1 for the empty product.
///
/// - The size of a product is the number of bytes required to store its elements, including padding,
/// plus trailing padding bytes so that the size is a multiple of the alignment.
///
/// - Sums store their payload at offset 0, followed by a 1-byte tag.
///
/// - The alignment of a sum is the maximum alignment of its variants' payloads.
///
/// - The size of a sum is the maximum size of its variants' payloads, plus 1 (the tag),
/// plus trailing padding bytes so that the size is a multiple of the alignment.
///
/// - The offset of a sum's tag bit is the maximum size of its variants' payloads.
unsafe fn unchecked_read_column(row_ref: RowRef<'_>, layout: &ProductTypeElementLayout) -> Self;
/// Check that the `idx`th column of the row type stored by `row_ref` is compatible with `Self`,
/// and read the value of that column from `row_ref`.
fn read_column(row_ref: RowRef<'_>, idx: usize) -> Result<Self, TypeError> {
let layout = row_ref.row_layout().product();
// Look up the `ProductTypeElementLayout` of the requested column,
// or return an error on an out-of-bounds index.
let col = layout.elements.get(idx).ok_or_else(|| TypeError::IndexOutOfBounds {
desired: idx,
found: layout.product_type(),
})?;
// Check that the requested column is of the expected type.
if !Self::is_compatible_type(&col.ty) {
return Err(TypeError::WrongType {
desired: std::any::type_name::<Self>(),
found: col.ty.algebraic_type(),
});
}
Ok(unsafe {
// SAFETY:
// - We trust that the `row_ref.table` knows its own layout,
// and we've derived our type and layout info from it,
// so they are correct.
// - We trust `Self::is_compatible_type`, and it returned `true`,
// so the column must be of appropriate type.
Self::unchecked_read_column(row_ref, col)
})
}
}
unsafe impl ReadColumn for bool {
fn is_compatible_type(ty: &AlgebraicTypeLayout) -> bool {
matches!(ty, AlgebraicTypeLayout::Primitive(PrimitiveType::Bool))
}
unsafe fn unchecked_read_column(row_ref: RowRef<'_>, layout: &ProductTypeElementLayout) -> Self {
debug_assert!(Self::is_compatible_type(&layout.ty));
let (page, offset) = row_ref.page_and_offset();
let col_offset = offset + PageOffset(layout.offset);
let data = page.get_row_data(col_offset, Size(mem::size_of::<Self>() as u16));
let data: *const bool = data.as_ptr().cast();
// SAFETY: We trust that the `row_ref` refers to a valid, initialized row,
// and that the `offset_in_bytes` refers to a column of type `Bool` within that row.
// A valid row can never have an uninitialized column or a column of an invalid value,
// so `data` must be initialized as either 0 or 1.
unsafe { *data }
}
}
macro_rules! impl_read_column_number {
($primitive_type:ident => $native_type:ty) => {
unsafe impl ReadColumn for $native_type {
fn is_compatible_type(ty: &AlgebraicTypeLayout) -> bool {
matches!(ty, AlgebraicTypeLayout::Primitive(PrimitiveType::$primitive_type))
}
unsafe fn unchecked_read_column(
row_ref: RowRef<'_>,
layout: &ProductTypeElementLayout,
) -> Self {
debug_assert!(Self::is_compatible_type(&layout.ty));
let (page, offset) = row_ref.page_and_offset();
let col_offset = offset + PageOffset(layout.offset);
let data = page.get_row_data(col_offset, Size(mem::size_of::<Self>() as u16));
// SAFETY: We trust that the `row_ref` refers to a valid, initialized row,
// and that the `offset_in_bytes` refers to a column of type `Self` within that row.
// A valid row can never have an uninitialized column,
// so `data` must be initialized.
let data = unsafe { slice_assume_init_ref(data) };
let data: Result<[u8; mem::size_of::<Self>()], _> = data.try_into();
// SAFETY: `<[u8; N] as TryFrom<&[u8]>` succeeds if and only if the slice's length is `N`.
// We used `mem::size_of::<Self>()` as both the length of the slice and the array,
// so we know them to be equal.
let data = unsafe { data.unwrap_unchecked() };
Self::from_le_bytes(data)
}
}
};
($($primitive_type:ident => $native_type:ty);* $(;)*) => {
$(impl_read_column_number!($primitive_type => $native_type);)*
};
}
impl_read_column_number! {
I8 => i8;
U8 => u8;
I16 => i16;
U16 => u16;
I32 => i32;
U32 => u32;
I64 => i64;
U64 => u64;
I128 => i128;
U128 => u128;
F32 => f32;
F64 => f64;
}
unsafe impl ReadColumn for AlgebraicValue {
fn is_compatible_type(_ty: &AlgebraicTypeLayout) -> bool {
true
}
unsafe fn unchecked_read_column(row_ref: RowRef<'_>, layout: &ProductTypeElementLayout) -> Self {
let curr_offset = Cell::new(layout.offset as usize);
let blob_store = row_ref.blob_store();
let (page, page_offset) = row_ref.page_and_offset();
let fixed_bytes = page.get_row_data(page_offset, row_ref.row_layout().size());
// SAFETY:
// 1. Our requirements on `row_ref` and `layout` mean that the column is valid at `layout`.
// 2. As a result of the above, all `VarLenRef`s in the column are valid.
// 3. Our requirements on `offset_in_bytes` mean that our `curr_offset` is valid.
let res = unsafe {
bflatn_from::serialize_value(ValueSerializer, fixed_bytes, page, blob_store, &curr_offset, &layout.ty)
};
debug_assert_eq!(curr_offset.get(), layout.offset as usize + layout.ty.size());
debug_assert!(res.is_ok());
// SAFETY: `ValueSerializer` is infallible.
unsafe { res.unwrap_unchecked() }
}
}
macro_rules! impl_read_column_via_av {
($av_pattern:pat => $into_method:ident => $native_type:ty) => {
unsafe impl ReadColumn for $native_type {
fn is_compatible_type(ty: &AlgebraicTypeLayout) -> bool {
matches!(ty, $av_pattern)
}
unsafe fn unchecked_read_column(
row_ref: RowRef<'_>,
layout: &ProductTypeElementLayout,
) -> Self {
debug_assert!(Self::is_compatible_type(&layout.ty));
// SAFETY:
// - Any layout is valid for `AlgebraicValue`, including our `layout`.
// - Forward requirements on `offset_in_bytes`.
let av = unsafe { AlgebraicValue::unchecked_read_column(row_ref, layout) };
let res = av.$into_method();
debug_assert!(res.is_ok());
// SAFETY: We trust that the value `row_ref + offset_in_bytes` is of type `layout`,
// and that `layout` is the layout for `Self`,
// so the `av` above must be a `Self`.
unsafe { res.unwrap_unchecked() }
}
}
};
($($av_pattern:pat => $into_method:ident => $native_type:ty);* $(;)*) => {
$(impl_read_column_via_av!($av_pattern => $into_method => $native_type);)*
};
}
impl_read_column_via_av! {
AlgebraicTypeLayout::VarLen(VarLenType::String) => into_string => String;
AlgebraicTypeLayout::VarLen(VarLenType::Array(_)) => into_array => ArrayValue;
AlgebraicTypeLayout::VarLen(VarLenType::Map(_)) => into_map => MapValue;
AlgebraicTypeLayout::Sum(_) => into_sum => SumValue;
AlgebraicTypeLayout::Product(_) => into_product => ProductValue;
}
#[cfg(test)]
mod test {
use super::*;
use crate::{
blob_store::HashMapBlobStore, indexes::SquashedOffset, proptest_sats::generate_typed_row, table::Table,
};
use proptest::{prelude::*, prop_assert_eq, proptest, test_runner::TestCaseResult};
use spacetimedb_sats::{
db::def::{TableDef, TableSchema},
product,
};
fn table(ty: ProductType) -> Table {
let def = TableDef::from_product("", ty);
let schema = TableSchema::from_def(0.into(), def);
Table::new(schema, SquashedOffset::COMMITTED_STATE)
}
proptest! {
#![proptest_config(ProptestConfig::with_cases(2048))]
#[test]
/// Test that `AlgebraicValue::read_column` returns expected values.
///
/// That is, test that, for any row type and any row value,
/// inserting the row, then doing `AlgebraicValue::read_column` on each column of the row
/// returns the expected value.
fn read_column_same_value((ty, val) in generate_typed_row()) {
let mut blob_store = HashMapBlobStore::default();
let mut table = table(ty);
let (_, ptr) = table.insert(&mut blob_store, &val).unwrap();
let row_ref = table.get_row_ref(&blob_store, ptr).unwrap();
for (idx, orig_col_value) in val.elements.into_iter().enumerate() {
let read_col_value = AlgebraicValue::read_column(row_ref, idx).unwrap();
prop_assert_eq!(orig_col_value, read_col_value);
}
}
#[test]
/// Test that trying to read a column at a type more specific than `AlgebraicValue`
/// which does not match the actual column type
/// returns an appropriate error.
fn read_column_wrong_type((ty, val) in generate_typed_row()) {
let mut blob_store = HashMapBlobStore::default();
let mut table = table(ty.clone());
let (_, ptr) = table.insert(&mut blob_store, &val).unwrap();
let row_ref = table.get_row_ref(&blob_store, ptr).unwrap();
for (idx, col_ty) in ty.elements.into_iter().enumerate() {
assert_wrong_type_error::<u8>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::U8)?;
assert_wrong_type_error::<i8>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::I8)?;
assert_wrong_type_error::<u16>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::U16)?;
assert_wrong_type_error::<i16>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::I16)?;
assert_wrong_type_error::<u32>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::U32)?;
assert_wrong_type_error::<i32>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::I32)?;
assert_wrong_type_error::<u64>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::U64)?;
assert_wrong_type_error::<i64>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::I64)?;
assert_wrong_type_error::<u128>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::U128)?;
assert_wrong_type_error::<i128>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::I128)?;
assert_wrong_type_error::<f32>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::F32)?;
assert_wrong_type_error::<f64>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::F64)?;
assert_wrong_type_error::<bool>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::Bool)?;
assert_wrong_type_error::<String>(row_ref, idx, &col_ty.algebraic_type, AlgebraicType::String)?;
}
}
#[test]
/// Test that trying to read a column which does not exist,
/// i.e. with an out-of-bounds index,
/// returns an appropriate error.
fn read_column_out_of_bounds((ty, val) in generate_typed_row()) {
let mut blob_store = HashMapBlobStore::default();
let mut table = table(ty.clone());
let (_, ptr) = table.insert(&mut blob_store, &val).unwrap();
let row_ref = table.get_row_ref(&blob_store, ptr).unwrap();
let oob = ty.elements.len();
match AlgebraicValue::read_column(row_ref, oob) {
Err(TypeError::IndexOutOfBounds { desired, found }) => {
prop_assert_eq!(desired, oob);
// Constructing a table changes the `ProductType` by adding column names
// if the type has `None` for its element names,
// so we can't blindly `prop_assert_eq!(found, ty)`.
// Instead, check that they have the same number of elements
// and that each element has the same type.
prop_assert_eq!(found.elements.len(), ty.elements.len());
for (found_col, ty_col) in found.elements.iter().zip(ty.elements.iter()) {
prop_assert_eq!(&found_col.algebraic_type, &ty_col.algebraic_type);
}
}
Err(e) => panic!("Expected TypeError::IndexOutOfBounds but found {:?}", e),
Ok(val) => panic!("Expected error but found Ok({:?})", val),
}
}
}
/// Assert, if and only if `col_ty` is not `correct_col_ty`,
/// that `Col::read_column(row_ref, col_idx)` returns a `TypeError::WrongType`.
///
/// If `col_ty == correct_col_ty`, do nothing.
fn assert_wrong_type_error<Col: ReadColumn + PartialEq + std::fmt::Debug>(
row_ref: RowRef<'_>,
col_idx: usize,
col_ty: &AlgebraicType,
correct_col_ty: AlgebraicType,
) -> TestCaseResult {
if col_ty != &correct_col_ty {
match Col::read_column(row_ref, col_idx) {
Err(TypeError::WrongType { desired, found }) => {
prop_assert_eq!(desired, std::any::type_name::<Col>());
prop_assert_eq!(&found, col_ty);
}
Err(e) => panic!("Expected TypeError::WrongType but found {:?}", e),
Ok(val) => panic!("Expected error but found Ok({:?})", val),
}
}
Ok(())
}
/// Define a test or tests which construct a row containing a known value of a known type,
/// then uses `ReadColumn::read_column` to extract that type as a native type,
/// e.g. a Rust integer,
/// and asserts that the extracted value is as expected.
macro_rules! test_read_column_primitive {
($name:ident { $algebraic_type:expr => $rust_type:ty = $val:expr }) => {
#[test]
fn $name() {
let mut blob_store = HashMapBlobStore::default();
let mut table = table(ProductType::from_iter([$algebraic_type]));
let val: $rust_type = $val;
let (_, ptr) = table.insert(&mut blob_store, &product![val.clone()]).unwrap();
let row_ref = table.get_row_ref(&blob_store, ptr).unwrap();
assert_eq!(val, <$rust_type as ReadColumn>::read_column(row_ref, 0).unwrap());
}
};
($($name:ident { $algebraic_type:expr => $rust_type:ty = $val:expr };)*) => {
$(test_read_column_primitive! {
$name { $algebraic_type => $rust_type = $val }
})*
}
}
test_read_column_primitive! {
read_column_i8 { AlgebraicType::I8 => i8 = i8::MAX };
read_column_u8 { AlgebraicType::U8 => u8 = 0xa5 };
read_column_i16 { AlgebraicType::I16 => i16 = i16::MAX };
read_column_u16 { AlgebraicType::U16 => u16 = 0xa5a5 };
read_column_i32 { AlgebraicType::I32 => i32 = i32::MAX };
read_column_u32 { AlgebraicType::U32 => u32 = 0xa5a5a5a5 };
read_column_i64 { AlgebraicType::I64 => i64 = i64::MAX };
read_column_u64 { AlgebraicType::U64 => u64 = 0xa5a5a5a5_a5a5a5a5 };
read_column_i128 { AlgebraicType::I128 => i128 = i128::MAX };
read_column_u128 { AlgebraicType::U128 => u128 = 0xa5a5a5a5_a5a5a5a5_a5a5a5a5_a5a5a5a5 };
read_column_f32 { AlgebraicType::F32 => f32 = 1.0 };
read_column_f64 { AlgebraicType::F64 => f64 = 1.0 };
read_column_bool { AlgebraicType::Bool => bool = true };
read_column_empty_string { AlgebraicType::String => String = "".to_string() };
// Use a short string which fits in a single granule.
read_column_short_string { AlgebraicType::String => String = "short string".to_string() };
// Use a medium-sized string which takes multiple granules.
read_column_medium_string { AlgebraicType::String => String = "medium string.".repeat(16) };
// Use a long string which will hit the blob store.
read_column_long_string { AlgebraicType::String => String = "long string. ".repeat(2048) };
}
}