AI Context: Dynamic Query & Low-Code Engine

1. Project Overview

This project explores two distinct approaches for building a "Low-Code" style backend that supports dynamic field selection, complex filtering, and deep graph traversal without writing boilerplate code.

2. Core Architectures

A. Hibernate/JPA Strategy (Runtime Dynamic)

Goal: Fully dynamic query construction at runtime.
Key Components:
- BaseRepository<T, ID>: Extends JpaRepository & JpaSpecificationExecutor.
- BaseController<T>: Provides generic /list endpoint.
- GenericSpecification: Converts Map params to JPA Predicates.
  - Supports exact match (default).
  - Supports LIKE match via wildcard * (e.g., name=John*).
- Squiggly: Used for dynamic JSON field projection (Graph-trimming) after DB fetch.
Pros: Zero code generation, extremely flexible, standard JPA ecosystem.
Cons: "N+1" issues if not careful (Entity Graph needed for optimization), limited by JPA's performance on complex joins.

B. MyBatis XML Generation Strategy (Compile/Runtime Generation)

Goal: Generate optimized, static-like SQL/XML artifacts for high performance and complex mapping.
Key Components:
- MyBatisGeneratorUtils: The core engine.
Mechanism:
1. ResultMap Generation:
  - Generates standard <resultMap> for each involved Entity.
  - Uses columnPrefix (e.g., author$, books$) to link nested associations/collections recursively.
  - This allows infinite recursion (Book -> Author -> Book) using the same set of ResultMaps.
2. SQL Generation:
  - Builds a Join Tree based on requested fields (e.g., ["title", "author.books.title"]).
  - Generates LEFT JOINs automatically.
  - Smart Alias Strategy:
    - Table Alias: Full path (e.g., book$author$books) to avoid collisions.
    - Column Alias: path$ColumnName (e.g., author$books$publish_month).
    - Crucial: The last part of the alias MUST be the DB Column Name (not Java Field Name) to match <result column="..."> during prefix stripping.
3. One-To-Many Handling:
  - Detects @OneToMany(mappedBy=...) to generate correct ON parent.id = child.parent_id join condition.
  - Forced ID Projection: Automatically selects id columns for all nodes to ensure correct MyBatis collection folding.
Pros: High performance (single SQL), solves N+1, precise column selection.
Cons: Requires generation step (or runtime string manipulation), more complex engine logic.

3. Key Files & Locations

com.example.demo.util.MyBatisGeneratorUtils: The "Brain" of the MyBatis strategy.
com.example.demo.repository.GenericSpecification: The "Brain" of the JPA strategy.
com.example.demo.controller.BaseController: The generic entry point.

4. Current Status

JPA Strategy: Fully functional. Read operations support wildcard search and dynamic projection. Write operations support cascading creation and deep-merge patching.
MyBatis Strategy: Engine is complete and tested (MyBatisGeneratorUtilsTest). Capable of generating full valid XML with recursive mappings and List support. Integration into Controller is the next logical step.

3.3 KiB Raw Blame History