НИР

.gitignore

@@ -10,3 +10,11 @@ FlameGraph/
 /perf.data.old
 **/.auctex-auto/
 **/build/
+/.odbc/
+**/__pycache__
+/build-tsan/
+/benchmarks/datasets/ssb/generated/
+/benchmarks/datasets/ssb/raw/
+/.plans/
+/report/build-practice/
+/report/build-vkr/

.gitmodules

@@ -0,0 +1,9 @@
+[submodule "research/datasets"]
+	path = research/datasets
+	url = git@github.com:ronaldbradford/data.git
+[submodule "research/imdb-sqlserver"]
+	path = research/imdb-sqlserver
+	url = git@github.com:sqlsunday/imdb-to-sqlserver.git
+[submodule "research/ssb-dbgen"]
+	path = research/ssb-dbgen
+	url = git@github.com:stewkk/ssb-dbgen.git

CMakeLists.txt

@@ -30,6 +30,10 @@ if (SANITIZE_THREADS)
     add_link_options(-fsanitize=thread)
 endif()
 
+if(STEWKK_SQL_DEBUG)
+  add_definitions(-DSTEWKK_SQL_DEBUG)
+endif()
+
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${PROJECT_BINARY_DIR}/bin)
 set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY ${PROJECT_BINARY_DIR}/lib)
 set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${PROJECT_BINARY_DIR}/lib)

Makefile

@@ -5,7 +5,16 @@ PARSER_SOURCE_DIR := $(CURRENT_DIR)/src/stewkk/sql/logic/parser
 build:
 	cmake --build build -- -j 6
 
+sanitize:
+	cmake --build build-sanitizers -- -j 8
+
 codegen:
 	@antlr -Dlanguage=Cpp -visitor -o $(CODEGEN_DIR) -package stewkk::sql::codegen $(PARSER_SOURCE_DIR)/PostgreSQLParser.g4 $(PARSER_SOURCE_DIR)/PostgreSQLLexer.g4
 
-.PHONY: codegen build
+test-ssb-converter:
+	pytest benchmarks/datasets/ssb
+
+dot:
+	dot -T png -O .plans/*.dot
+
+.PHONY: codegen build sanitize test-ssb-converter

benchmarks/CMakeLists.txt

@@ -1,6 +1,6 @@
 include(FetchGBenchmark)
 
-add_executable(benchmarks main.cpp)
+add_executable(benchmarks main.cpp operator_cost.cpp)
 target_compile_features(benchmarks PRIVATE cxx_std_23)
 set_target_properties(benchmarks PROPERTIES
   CXX_STANDART 23

benchmarks/datasets/ssb/README.md

@@ -0,0 +1,59 @@
+# Star Schema Benchmark data
+
+This directory contains the schema and conversion tooling for Star Schema
+Benchmark data. It does not download or build an SSB data generator; generate
+the `.tbl` files with an external SSB dbgen-compatible tool first.
+
+## Convert `.tbl` files
+
+Expected input files:
+
+- `lineorder.tbl`
+- `customer.tbl`
+- `supplier.tbl`
+- `part.tbl`
+- `date.tbl`
+
+Convert them to the CSV format used by this project:
+
+```sh
+python3 benchmarks/datasets/ssb/convert_tbl_to_csv.py \
+  --input-dir /path/to/ssb/tbl \
+  --output-dir benchmarks/datasets/ssb/generated/sf1
+```
+
+The converter reads `schema.json` for table and column definitions. Output CSV
+headers use `column:type`, for example `lo_orderkey:int`. String columns are
+preserved as `string` values in the generated CSV.
+
+Generated full-scale data should stay under `benchmarks/datasets/ssb/generated/`
+or another ignored location.
+
+## Queries
+
+The standard 13 SSB query templates are in `queries/`. They are written in the
+subset supported by this engine: explicit joins, table aliases, string
+predicates, `BETWEEN`, list `IN`, `SUM`, `COUNT`, `GROUP BY`, and `ORDER BY`.
+
+Run one query through the CLI:
+
+```sh
+./build/bin/sql --data-dir benchmarks/datasets/ssb/generated/sf1 \
+  < benchmarks/datasets/ssb/queries/q1.1.sql
+```
+
+Run SSB benchmarks after generating data:
+
+```sh
+SSB_DATA_DIR=benchmarks/datasets/ssb/generated/sf1 \
+  ./build/bin/benchmarks --benchmark_filter='SSB/'
+```
+
+The JIT expression executor still rejects string and `IN` expressions, so the
+SSB benchmark registrations currently use the interpreted expression executor.
+
+## Test the converter
+
+```sh
+make test-ssb-converter
+```

benchmarks/datasets/ssb/convert_tbl_to_csv.py

@@ -0,0 +1,137 @@
+#!/usr/bin/env python3
+"""Convert Star Schema Benchmark .tbl files to this project's CSV format."""
+
+from __future__ import annotations
+
+import argparse
+import csv
+import json
+import sys
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Iterable
+
+
+SUPPORTED_TYPES = {"int", "string"}
+
+
+@dataclass(frozen=True)
+class Column:
+    name: str
+    type: str
+
+
+@dataclass(frozen=True)
+class Table:
+    name: str
+    columns: tuple[Column, ...]
+
+
+class ConversionError(RuntimeError):
+    pass
+
+
+def load_schema(path: Path) -> list[Table]:
+    with path.open(encoding="utf-8") as f:
+        raw = json.load(f)
+
+    tables = []
+    for table_raw in raw.get("tables", []):
+        columns = []
+        for column_raw in table_raw.get("columns", []):
+            column_type = column_raw["type"]
+            if column_type not in SUPPORTED_TYPES:
+                raise ConversionError(
+                    f"{path}: unsupported type {column_type!r} in table {table_raw['name']}"
+                )
+            columns.append(Column(column_raw["name"], column_type))
+        if not columns:
+            raise ConversionError(f"{path}: table {table_raw.get('name')!r} has no columns")
+        tables.append(Table(table_raw["name"], tuple(columns)))
+
+    if not tables:
+        raise ConversionError(f"{path}: schema does not define any tables")
+    return tables
+
+
+def parse_tbl_line(line: str) -> list[str]:
+    fields = line.rstrip("\r\n").split("|")
+    if fields and fields[-1] == "":
+        fields.pop()
+    return fields
+
+
+def validate_row(table: Table, row: list[str], input_path: Path, line_no: int) -> None:
+    if len(row) != len(table.columns):
+        raise ConversionError(
+            f"{input_path}:{line_no}: expected {len(table.columns)} fields for "
+            f"{table.name}, got {len(row)}"
+        )
+
+    for value, column in zip(row, table.columns, strict=True):
+        if column.type == "int":
+            try:
+                int(value)
+            except ValueError as exc:
+                raise ConversionError(
+                    f"{input_path}:{line_no}: {column.name} expects int, got {value!r}"
+                ) from exc
+
+
+def header(table: Table) -> list[str]:
+    return [f"{column.name}:{column.type}" for column in table.columns]
+
+
+def convert_table(table: Table, input_dir: Path, output_dir: Path) -> int:
+    input_path = input_dir / f"{table.name}.tbl"
+    output_path = output_dir / f"{table.name}.csv"
+    if not input_path.exists():
+        raise ConversionError(f"missing input file: {input_path}")
+
+    rows = 0
+    with input_path.open(encoding="utf-8", newline="") as src, output_path.open(
+        "w", encoding="utf-8", newline=""
+    ) as dst:
+        writer = csv.writer(dst, lineterminator="\n")
+        writer.writerow(header(table))
+        for line_no, line in enumerate(src, start=1):
+            row = parse_tbl_line(line)
+            validate_row(table, row, input_path, line_no)
+            writer.writerow(row)
+            rows += 1
+    return rows
+
+
+def convert_all(schema_path: Path, input_dir: Path, output_dir: Path) -> dict[str, int]:
+    tables = load_schema(schema_path)
+    output_dir.mkdir(parents=True, exist_ok=True)
+    return {
+        table.name: convert_table(table, input_dir, output_dir)
+        for table in tables
+    }
+
+
+def parse_args(argv: Iterable[str]) -> argparse.Namespace:
+    default_schema = Path(__file__).with_name("schema.json")
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--input-dir", required=True, type=Path, help="Directory containing SSB .tbl files")
+    parser.add_argument("--output-dir", required=True, type=Path, help="Directory for generated CSV files")
+    parser.add_argument("--schema", default=default_schema, type=Path, help="Schema JSON path")
+    return parser.parse_args(argv)
+
+
+def main(argv: Iterable[str] = sys.argv[1:]) -> int:
+    args = parse_args(argv)
+    try:
+        counts = convert_all(args.schema, args.input_dir, args.output_dir)
+    except ConversionError as exc:
+        print(f"error: {exc}", file=sys.stderr)
+        return 1
+
+    for table_name, rows in counts.items():
+        print(f"wrote {args.output_dir / (table_name + '.csv')} ({rows} rows)")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

benchmarks/datasets/ssb/fixtures/expected_csv/customer.csv

@@ -0,0 +1,3 @@
+c_custkey:int,c_name:string,c_address:string,c_city:string,c_nation:string,c_region:string,c_phone:string,c_mktsegment:string
+10,Customer#000000010,Addr 10,UNITED KI1,UNITED KINGDOM,EUROPE,10-111-111-1111,BUILDING
+20,Customer#000000020,"Addr, 20",BRAZIL   1,BRAZIL,AMERICA,20-222-222-2222,AUTOMOBILE

benchmarks/datasets/ssb/fixtures/expected_csv/date.csv

@@ -0,0 +1,3 @@
+d_datekey:int,d_date:string,d_dayofweek:string,d_month:string,d_year:int,d_yearmonthnum:int,d_yearmonth:string,d_daynuminweek:int,d_daynuminmonth:int,d_daynuminyear:int,d_monthnuminyear:int,d_weeknuminyear:int,d_sellingseason:string,d_lastdayinweekfl:int,d_lastdayinmonthfl:int,d_holidayfl:int,d_weekdayfl:int
+19940101,"January 1, 1994",Saturday,January,1994,199401,Jan1994,6,1,1,1,1,Winter,0,0,1,0
+19940102,"January 2, 1994",Sunday,January,1994,199401,Jan1994,7,2,2,1,1,Winter,1,0,0,0

benchmarks/datasets/ssb/fixtures/expected_csv/lineorder.csv

@@ -0,0 +1,3 @@
+lo_orderkey:int,lo_linenumber:int,lo_custkey:int,lo_partkey:int,lo_suppkey:int,lo_orderdate:int,lo_orderpriority:string,lo_shippriority:string,lo_quantity:int,lo_extendedprice:int,lo_ordtotalprice:int,lo_discount:int,lo_revenue:int,lo_supplycost:int,lo_tax:int,lo_commitdate:int,lo_shipmode:string
+1,1,10,100,1000,19940101,1-URGENT,0,5,10000,12000,3,9700,6000,1,19940105,AIR
+2,1,20,200,2000,19940102,2-HIGH,0,7,20000,23000,4,19200,12000,2,19940106,RAIL

benchmarks/datasets/ssb/fixtures/expected_csv/part.csv

@@ -0,0 +1,3 @@
+p_partkey:int,p_name:string,p_mfgr:string,p_category:string,p_brand:string,p_color:string,p_type:string,p_size:int,p_container:string
+100,almond antique,MFGR#1,MFGR#11,MFGR#111,almond,STANDARD ANODIZED TIN,3,SM CASE
+200,green burnished,MFGR#2,MFGR#22,MFGR#222,green,PROMO POLISHED STEEL,7,LG BOX

benchmarks/datasets/ssb/fixtures/expected_csv/supplier.csv

@@ -0,0 +1,3 @@
+s_suppkey:int,s_name:string,s_address:string,s_city:string,s_nation:string,s_region:string,s_phone:string
+1000,Supplier#000001000,Supplier Addr 1000,RUSSIA   1,RUSSIA,EUROPE,30-333-333-3333
+2000,Supplier#000002000,Supplier Addr 2000,ARGENTINA1,ARGENTINA,AMERICA,40-444-444-4444

benchmarks/datasets/ssb/fixtures/tbl/customer.tbl

@@ -0,0 +1,2 @@
+10|Customer#000000010|Addr 10|UNITED KI1|UNITED KINGDOM|EUROPE|10-111-111-1111|BUILDING|
+20|Customer#000000020|Addr, 20|BRAZIL   1|BRAZIL|AMERICA|20-222-222-2222|AUTOMOBILE|

benchmarks/datasets/ssb/fixtures/tbl/date.tbl

@@ -0,0 +1,2 @@
+19940101|January 1, 1994|Saturday|January|1994|199401|Jan1994|6|1|1|1|1|Winter|0|0|1|0|
+19940102|January 2, 1994|Sunday|January|1994|199401|Jan1994|7|2|2|1|1|Winter|1|0|0|0|

benchmarks/datasets/ssb/fixtures/tbl/lineorder.tbl

@@ -0,0 +1,2 @@
+1|1|10|100|1000|19940101|1-URGENT|0|5|10000|12000|3|9700|6000|1|19940105|AIR|
+2|1|20|200|2000|19940102|2-HIGH|0|7|20000|23000|4|19200|12000|2|19940106|RAIL|

benchmarks/datasets/ssb/fixtures/tbl/part.tbl

@@ -0,0 +1,2 @@
+100|almond antique|MFGR#1|MFGR#11|MFGR#111|almond|STANDARD ANODIZED TIN|3|SM CASE|
+200|green burnished|MFGR#2|MFGR#22|MFGR#222|green|PROMO POLISHED STEEL|7|LG BOX|

benchmarks/datasets/ssb/fixtures/tbl/supplier.tbl

@@ -0,0 +1,2 @@
+1000|Supplier#000001000|Supplier Addr 1000|RUSSIA   1|RUSSIA|EUROPE|30-333-333-3333|
+2000|Supplier#000002000|Supplier Addr 2000|ARGENTINA1|ARGENTINA|AMERICA|40-444-444-4444|

benchmarks/datasets/ssb/queries/q1.1.sql

@@ -0,0 +1,6 @@
+SELECT SUM(lo.lo_extendedprice * lo.lo_discount) AS revenue
+FROM lineorder AS lo
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+WHERE d.d_year = 1993
+  AND lo.lo_discount BETWEEN 1 AND 3
+  AND lo.lo_quantity < 25;

benchmarks/datasets/ssb/queries/q1.2.sql

@@ -0,0 +1,6 @@
+SELECT SUM(lo.lo_extendedprice * lo.lo_discount) AS revenue
+FROM lineorder AS lo
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+WHERE d.d_yearmonthnum = 199401
+  AND lo.lo_discount BETWEEN 4 AND 6
+  AND lo.lo_quantity BETWEEN 26 AND 35;

benchmarks/datasets/ssb/queries/q1.3.sql

@@ -0,0 +1,7 @@
+SELECT SUM(lo.lo_extendedprice * lo.lo_discount) AS revenue
+FROM lineorder AS lo
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+WHERE d.d_weeknuminyear = 6
+  AND d.d_year = 1994
+  AND lo.lo_discount BETWEEN 5 AND 7
+  AND lo.lo_quantity BETWEEN 26 AND 35;

benchmarks/datasets/ssb/queries/q2.1.sql

@@ -0,0 +1,9 @@
+SELECT SUM(lo.lo_revenue) AS lo_revenue, d.d_year, p.p_brand
+FROM lineorder AS lo
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+JOIN part AS p ON lo.lo_partkey = p.p_partkey
+JOIN supplier AS s ON lo.lo_suppkey = s.s_suppkey
+WHERE p.p_category = 'MFGR#12'
+  AND s.s_region = 'AMERICA'
+GROUP BY d.d_year, p.p_brand
+ORDER BY d.d_year, p.p_brand;

benchmarks/datasets/ssb/queries/q2.2.sql

@@ -0,0 +1,9 @@
+SELECT SUM(lo.lo_revenue) AS lo_revenue, d.d_year, p.p_brand
+FROM lineorder AS lo
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+JOIN part AS p ON lo.lo_partkey = p.p_partkey
+JOIN supplier AS s ON lo.lo_suppkey = s.s_suppkey
+WHERE p.p_brand BETWEEN 'MFGR#2221' AND 'MFGR#2228'
+  AND s.s_region = 'ASIA'
+GROUP BY d.d_year, p.p_brand
+ORDER BY d.d_year, p.p_brand;

benchmarks/datasets/ssb/queries/q2.3.sql

@@ -0,0 +1,9 @@
+SELECT SUM(lo.lo_revenue) AS lo_revenue, d.d_year, p.p_brand
+FROM lineorder AS lo
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+JOIN part AS p ON lo.lo_partkey = p.p_partkey
+JOIN supplier AS s ON lo.lo_suppkey = s.s_suppkey
+WHERE p.p_brand = 'MFGR#2221'
+  AND s.s_region = 'EUROPE'
+GROUP BY d.d_year, p.p_brand
+ORDER BY d.d_year, p.p_brand;

benchmarks/datasets/ssb/queries/q3.1.sql

@@ -0,0 +1,10 @@
+SELECT c.c_nation, s.s_nation, d.d_year, SUM(lo.lo_revenue) AS lo_revenue
+FROM lineorder AS lo
+JOIN customer AS c ON lo.lo_custkey = c.c_custkey
+JOIN supplier AS s ON lo.lo_suppkey = s.s_suppkey
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+WHERE c.c_region = 'ASIA'
+  AND s.s_region = 'ASIA'
+  AND d.d_year BETWEEN 1992 AND 1997
+GROUP BY c.c_nation, s.s_nation, d.d_year
+ORDER BY d.d_year, lo_revenue DESC;

benchmarks/datasets/ssb/queries/q3.2.sql

@@ -0,0 +1,10 @@
+SELECT c.c_city, s.s_city, d.d_year, SUM(lo.lo_revenue) AS lo_revenue
+FROM lineorder AS lo
+JOIN customer AS c ON lo.lo_custkey = c.c_custkey
+JOIN supplier AS s ON lo.lo_suppkey = s.s_suppkey
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+WHERE c.c_nation = 'UNITED STATES'
+  AND s.s_nation = 'UNITED STATES'
+  AND d.d_year BETWEEN 1992 AND 1997
+GROUP BY c.c_city, s.s_city, d.d_year
+ORDER BY d.d_year, lo_revenue DESC;

benchmarks/datasets/ssb/queries/q3.3.sql

@@ -0,0 +1,10 @@
+SELECT c.c_city, s.s_city, d.d_year, SUM(lo.lo_revenue) AS lo_revenue
+FROM lineorder AS lo
+JOIN customer AS c ON lo.lo_custkey = c.c_custkey
+JOIN supplier AS s ON lo.lo_suppkey = s.s_suppkey
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+WHERE c.c_city IN ('UNITED KI1', 'UNITED KI5')
+  AND s.s_city IN ('UNITED KI1', 'UNITED KI5')
+  AND d.d_year BETWEEN 1992 AND 1997
+GROUP BY c.c_city, s.s_city, d.d_year
+ORDER BY d.d_year, lo_revenue DESC;

benchmarks/datasets/ssb/queries/q3.4.sql

@@ -0,0 +1,10 @@
+SELECT c.c_city, s.s_city, d.d_year, SUM(lo.lo_revenue) AS lo_revenue
+FROM lineorder AS lo
+JOIN customer AS c ON lo.lo_custkey = c.c_custkey
+JOIN supplier AS s ON lo.lo_suppkey = s.s_suppkey
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+WHERE c.c_city IN ('UNITED KI1', 'UNITED KI5')
+  AND s.s_city IN ('UNITED KI1', 'UNITED KI5')
+  AND d.d_yearmonth = 'Dec1997'
+GROUP BY c.c_city, s.s_city, d.d_year
+ORDER BY d.d_year, lo_revenue DESC;

benchmarks/datasets/ssb/queries/q4.1.sql

@@ -0,0 +1,11 @@
+SELECT d.d_year, c.c_nation, SUM(lo.lo_revenue - lo.lo_supplycost) AS profit
+FROM lineorder AS lo
+JOIN date AS d ON lo.lo_orderdate = d.d_datekey
+JOIN customer AS c ON lo.lo_custkey = c.c_custkey
+JOIN supplier AS s ON lo.lo_suppkey = s.s_suppkey
+JOIN part AS p ON lo.lo_partkey = p.p_partkey
+WHERE c.c_region = 'AMERICA'
+  AND s.s_region = 'AMERICA'
+  AND p.p_mfgr IN ('MFGR#1', 'MFGR#2')
+GROUP BY d.d_year, c.c_nation
+ORDER BY d.d_year, c.c_nation;