TableEngine and SqlHandler impl (#45)

* Impl TableEngine, bridge to storage

* Impl sql handler to process insert sql

* fix: minor changes and typo

* test: add datanode test

* test: add table-engine test

* fix: code style

* refactor: split out insert mod from sql and minor changes by CR

* refactor: replace with_context with context
This commit is contained in:
dennis zhuang
2022-06-17 11:36:49 +08:00
committed by GitHub
parent e03ac2fc2b
commit e78c015fc0
36 changed files with 1438 additions and 110 deletions

View File

@@ -0,0 +1,20 @@
[package]
name = "table-engine"
version = "0.1.0"
edition = "2021"
[dependencies]
async-trait = "0.1"
chrono = { version = "0.4", features = ["serde"] }
common-error = {path = "../common/error" }
common-query = {path = "../common/query" }
common-recordbatch = {path = "../common/recordbatch" }
common-telemetry = {path = "../common/telemetry" }
snafu = { version = "0.7", features = ["backtraces"] }
storage ={ path = "../storage" }
store-api ={ path = "../store-api" }
table = { path = "../table" }
[dev-dependencies]
datatypes = { path = "../datatypes" }
tokio = { version = "1.18", features = ["full"] }

View File

@@ -0,0 +1,251 @@
use std::collections::HashMap;
use std::sync::atomic::{AtomicU64, Ordering};
use std::sync::Arc;
use std::sync::RwLock;
use async_trait::async_trait;
use snafu::ResultExt;
use store_api::storage::ConcreteDataType;
use store_api::storage::{
self as store, ColumnDescriptorBuilder, ColumnFamilyDescriptorBuilder,
EngineContext as StorageContext, Region, RegionDescriptor, RegionId, RegionMeta,
RowKeyDescriptorBuilder, StorageEngine,
};
use table::engine::{EngineContext, TableEngine};
use table::requests::{AlterTableRequest, CreateTableRequest, DropTableRequest};
use table::{
metadata::{TableId, TableInfoBuilder, TableMetaBuilder, TableType},
table::TableRef,
};
use crate::error::{CreateTableSnafu, Error, Result};
use crate::table::MitoTable;
pub const DEFAULT_ENGINE: &str = "mito";
/// [TableEngine] implementation.
///
/// About mito <https://en.wikipedia.org/wiki/Alfa_Romeo_MiTo>.
/// "you can't be a true petrolhead until you've owned an Alfa Romeo" -- by Jeremy Clarkson
#[derive(Clone)]
pub struct MitoEngine<Store: StorageEngine> {
inner: Arc<MitoEngineInner<Store>>,
}
impl<Store: StorageEngine> MitoEngine<Store> {
pub fn new(storage_engine: Store) -> Self {
Self {
inner: Arc::new(MitoEngineInner::new(storage_engine)),
}
}
}
#[async_trait]
impl<Store: StorageEngine> TableEngine for MitoEngine<Store> {
type Error = Error;
async fn create_table(
&self,
ctx: &EngineContext,
request: CreateTableRequest,
) -> Result<TableRef> {
self.inner.create_table(ctx, request).await
}
async fn alter_table(
&self,
_ctx: &EngineContext,
_request: AlterTableRequest,
) -> Result<TableRef> {
unimplemented!();
}
fn get_table(&self, ctx: &EngineContext, name: &str) -> Result<Option<TableRef>> {
self.inner.get_table(ctx, name)
}
fn table_exists(&self, _ctx: &EngineContext, _name: &str) -> bool {
unimplemented!();
}
async fn drop_table(&self, _ctx: &EngineContext, _request: DropTableRequest) -> Result<()> {
unimplemented!();
}
}
/// FIXME(boyan) impl system catalog to keep table metadata.
struct MitoEngineInner<Store: StorageEngine> {
tables: RwLock<HashMap<String, TableRef>>,
storage_engine: Store,
next_table_id: AtomicU64,
}
impl<Store: StorageEngine> MitoEngineInner<Store> {
fn new(storage_engine: Store) -> Self {
Self {
tables: RwLock::new(HashMap::default()),
storage_engine,
next_table_id: AtomicU64::new(0),
}
}
fn next_table_id(&self) -> TableId {
self.next_table_id.fetch_add(1, Ordering::Relaxed)
}
}
impl<Store: StorageEngine> MitoEngineInner<Store> {
async fn create_table(
&self,
_ctx: &EngineContext,
request: CreateTableRequest,
) -> Result<TableRef> {
//FIXME(boyan): we only supports creating a demo table right now
//The create table sql is like:
// create table demo(host string,
// ts int64,
// cpu float64,
// memory float64,
// PRIMARY KEY(ts, host)) with regions=1;
//TODO(boyan): supports multi regions
let region_id: RegionId = 0;
let name = store::gen_region_name(region_id);
let host_column =
ColumnDescriptorBuilder::new(0, "host", ConcreteDataType::string_datatype())
.is_nullable(false)
.build();
let cpu_column =
ColumnDescriptorBuilder::new(1, "cpu", ConcreteDataType::float64_datatype())
.is_nullable(true)
.build();
let memory_column =
ColumnDescriptorBuilder::new(2, "memory", ConcreteDataType::float64_datatype())
.is_nullable(true)
.build();
let ts_column =
ColumnDescriptorBuilder::new(0, "ts", ConcreteDataType::int64_datatype()).build();
let row_key = RowKeyDescriptorBuilder::new(ts_column)
.push_column(host_column)
.enable_version_column(false)
.build();
let default_cf = ColumnFamilyDescriptorBuilder::default()
.push_column(cpu_column)
.push_column(memory_column)
.build();
let region = self
.storage_engine
.create_region(
&StorageContext::default(),
RegionDescriptor {
id: region_id,
name,
row_key,
default_cf,
extra_cfs: Vec::default(),
},
)
.await
.map_err(|e| Box::new(e) as _)
.context(CreateTableSnafu)?;
// Use region meta schema instead of request schema
let table_meta = TableMetaBuilder::new(region.in_memory_metadata().schema().clone())
.engine(DEFAULT_ENGINE)
.build();
let table_name = request.name;
let table_info = TableInfoBuilder::new(table_name.clone(), table_meta)
.table_id(self.next_table_id())
.table_version(0u64)
.table_type(TableType::Base)
.desc(request.desc)
.build();
let table = Arc::new(MitoTable::new(table_info, region));
self.tables
.write()
.unwrap()
.insert(table_name, table.clone());
Ok(table)
}
fn get_table(&self, _ctx: &EngineContext, name: &str) -> Result<Option<TableRef>> {
Ok(self.tables.read().unwrap().get(name).cloned())
}
}
#[cfg(test)]
mod tests {
use datatypes::schema::{ColumnSchema, Schema};
use datatypes::vectors::*;
use storage::EngineImpl;
use table::requests::InsertRequest;
use super::*;
#[tokio::test]
async fn test_creat_table_insert() {
let column_schemas = vec![
ColumnSchema::new("host", ConcreteDataType::string_datatype(), false),
ColumnSchema::new("ts", ConcreteDataType::int64_datatype(), true),
ColumnSchema::new("cpu", ConcreteDataType::float64_datatype(), true),
ColumnSchema::new("memory", ConcreteDataType::float64_datatype(), true),
];
let table_engine = MitoEngine::<EngineImpl>::new(EngineImpl::new());
let table_name = "demo";
let schema = Arc::new(Schema::new(column_schemas));
let table = table_engine
.create_table(
&EngineContext::default(),
CreateTableRequest {
name: table_name.to_string(),
desc: Some(" a test table".to_string()),
schema: schema.clone(),
},
)
.await
.unwrap();
assert_eq!(TableType::Base, table.table_type());
assert_eq!(schema, table.schema());
let insert_req = InsertRequest {
table_name: table_name.to_string(),
columns_values: HashMap::default(),
};
assert_eq!(0, table.insert(insert_req).await.unwrap());
let mut columns_values: HashMap<String, VectorRef> = HashMap::with_capacity(4);
columns_values.insert(
"host".to_string(),
Arc::new(StringVector::from(vec!["host1", "host2"])),
);
columns_values.insert(
"cpu".to_string(),
Arc::new(Float64Vector::from_vec(vec![55.5, 66.6])),
);
columns_values.insert(
"memory".to_string(),
Arc::new(Float64Vector::from_vec(vec![1024f64, 4096f64])),
);
columns_values.insert(
"ts".to_string(),
Arc::new(Int64Vector::from_vec(vec![1, 2])),
);
let insert_req = InsertRequest {
table_name: table_name.to_string(),
columns_values,
};
assert_eq!(2, table.insert(insert_req).await.unwrap());
}
}

View File

@@ -0,0 +1,35 @@
use std::any::Any;
use common_error::prelude::*;
// TODO(boyan): use ErrorExt instead.
pub type BoxedError = Box<dyn std::error::Error + Send + Sync>;
#[derive(Debug, Snafu)]
#[snafu(visibility(pub))]
pub enum Error {
#[snafu(display("Fail to create table, source: {}", source))]
CreateTable {
source: BoxedError,
backtrace: Backtrace,
},
}
pub type Result<T> = std::result::Result<T, Error>;
impl ErrorExt for Error {
fn status_code(&self) -> StatusCode {
match self {
//TODO: should return the source's status code after use ErrorExt in BoxedError.
Error::CreateTable { .. } => StatusCode::InvalidArguments,
}
}
fn backtrace_opt(&self) -> Option<&Backtrace> {
ErrorCompat::backtrace(self)
}
fn as_any(&self) -> &dyn Any {
self
}
}

View File

@@ -0,0 +1,3 @@
pub mod engine;
pub mod error;
pub mod table;

View File

@@ -0,0 +1,94 @@
use std::any::Any;
use async_trait::async_trait;
use common_query::logical_plan::Expr;
use common_recordbatch::SendableRecordBatchStream;
use snafu::OptionExt;
use store_api::storage::SchemaRef;
use store_api::storage::{PutOperation, Region, WriteContext, WriteRequest};
use table::error::{Error as TableError, MissingColumnSnafu, Result as TableResult};
use table::requests::InsertRequest;
use table::{
metadata::{TableInfo, TableType},
table::Table,
};
/// [Table] implementation.
pub struct MitoTable<R: Region> {
table_info: TableInfo,
//TODO(dennis): a table contains multi regions
region: R,
}
#[async_trait]
impl<R: Region> Table for MitoTable<R> {
fn as_any(&self) -> &dyn Any {
self
}
fn schema(&self) -> SchemaRef {
self.table_info.meta.schema.clone()
}
async fn insert(&self, request: InsertRequest) -> TableResult<usize> {
if request.columns_values.is_empty() {
return Ok(0);
}
let mut write_request = R::WriteRequest::new(self.schema());
//FIXME(boyan): we can only insert to demo table right now
let mut put_op = <<R as Region>::WriteRequest as WriteRequest>::PutOp::new();
let mut columns_values = request.columns_values;
let key_columns = vec!["ts", "host"];
let value_columns = vec!["cpu", "memory"];
//Add row key and columns
for name in key_columns {
put_op
.add_key_column(
name,
columns_values
.get(name)
.context(MissingColumnSnafu { name })?
.clone(),
)
.map_err(TableError::new)?;
}
// Add vaue columns
let mut rows_num = 0;
for name in value_columns {
if let Some(v) = columns_values.remove(name) {
rows_num = v.len();
put_op.add_value_column(name, v).map_err(TableError::new)?;
}
}
write_request.put(put_op).map_err(TableError::new)?;
let _resp = self
.region
.write(&WriteContext::default(), write_request)
.await
.map_err(TableError::new)?;
Ok(rows_num)
}
fn table_type(&self) -> TableType {
self.table_info.table_type
}
async fn scan(
&self,
_projection: &Option<Vec<usize>>,
_filters: &[Expr],
_limit: Option<usize>,
) -> TableResult<SendableRecordBatchStream> {
unimplemented!();
}
}
impl<R: Region> MitoTable<R> {
pub fn new(table_info: TableInfo, region: R) -> Self {
Self { table_info, region }
}
}