Skip to content

PIPELINE.md

Latest commit

 

History

History
100 lines (59 loc) · 2.74 KB

PIPELINE.md

File metadata and controls

100 lines (59 loc) · 2.74 KB

SQL pipeline

This document describes how a SQL string moves through compilation stages and becomes an executable plan.

Общая схема

flowchart LR
  SQL["SQL string"] --> LEX["LexerImpl\nTokens"]
  LEX --> PAR["ParserImpl\nAST"]
  PAR --> SEM["SemanticAnalyzerImpl\nQueryTree"]
  SEM --> PLN["PlannerImpl\nLogical plan"]
  PLN --> OPT["OptimizerImpl\nPhysical plan"]
  OPT --> EXF["ExecutorFactoryImpl\nExecutor tree"]
  EXF --> QEE["QueryExecutionEngineImpl\nRows / status"]
Loading

1) Lexer (lexical analysis)

File: src/main/java/ru/open/cu/student/lexer/LexerImpl.java

Responsibilities:

  • transform the input string into a token stream (keywords, identifiers, literals, operators)
  • provide predictable lexical-level errors

2) Parser (syntax analysis)

File: src/main/java/ru/open/cu/student/parser/ParserImpl.java

Responsibilities:

  • build an abstract syntax tree (AST)
  • enforce the supported grammar rules for the dialect

AST node types live under src/main/java/ru/open/cu/student/parser/nodes/.

3) Semantic analyzer

File: src/main/java/ru/open/cu/student/semantic/SemanticAnalyzerImpl.java

Responsibilities:

  • validate tables and columns exist
  • perform name resolution and basic expression typing
  • build QueryTree — a typed representation of the query

4) Planner (logical planning)

File: src/main/java/ru/open/cu/student/planner/PlannerImpl.java

Responsibilities:

  • build a logical plan (operators: scan/filter/project/insert/create)
  • separate “what to do” from “how to execute”

Logical plan nodes: src/main/java/ru/open/cu/student/planner/node/.

5) Optimizer (physical planning)

File: src/main/java/ru/open/cu/student/optimizer/OptimizerImpl.java

Current strategy: a rule-based scan selection:

  • SeqScan by default
  • IndexScan when the predicate can leverage an index

Main rule entry point: optimizer/strategy/RuleBasedScanStrategy.java.

Physical plan nodes: src/main/java/ru/open/cu/student/optimizer/node/.

6) Executor factory (building the executor tree)

File: src/main/java/ru/open/cu/student/execution/ExecutorFactoryImpl.java

Responsibilities:

  • map physical plan nodes to executor implementations
  • construct a Volcano-style executor tree

7) Execution engine

File: src/main/java/ru/open/cu/student/execution/QueryExecutionEngineImpl.java

Responsibilities:

  • call open() on the root executor
  • iteratively pull rows via next()
  • finalize execution and produce a QueryResult

Error handling

Stages are separated, which makes it possible to:

  • return errors annotated with the stage (LEXER, PARSER, SEMANTIC, PLANNER, OPTIMIZER, EXECUTION)
  • test negative scenarios more easily

The error format and network response are described in PROTOCOL.md.