drift/sqlparser
Simon Binder 8661e0a28a
Refactoring on the model classes in moor_generator
2019-12-14 21:43:04 +01:00
..
example Enable more recent lints and enforce them 2019-12-04 21:45:09 +01:00
lib Refactoring on the model classes in moor_generator 2019-12-14 21:43:04 +01:00
test Support detailed analysis for fts5 functions 2019-12-13 22:30:36 +01:00
.gitignore Use builtin coverage collection from test package 2019-11-08 22:51:53 +01:00
CHANGELOG.md Parse CREATE VIRTUAL TABLE statements 2019-12-04 15:36:23 +01:00
LICENSE Extract sql parser to yet another subpackage 2019-06-15 23:01:10 +02:00
README.md RELEASE 2019-11-16 11:23:11 +01:00
analysis_options.yaml Extract sql parser to yet another subpackage 2019-06-15 23:01:10 +02:00
mono_pkg.yaml Provide coverage upload token variable 2019-07-28 11:08:39 +02:00
pubspec.yaml Refactoring on the model classes in moor_generator 2019-12-14 21:43:04 +01:00

README.md

sqlparser

Sql parser and static analyzer written in Dart. At the moment, this library targets the sqlite dialect only.

Features

This library can parse most sql statements and perform static analysis. We can resolve what type a column in a SELECT statement has, infer types for variables, find semantic errors and more.

This library supports most features, including joins, group by, nested and compound sql statements, window functions and foreign keys.

Using the parser

You can parse the abstract syntax tree of sqlite statements with SqlEngine.parse.

import 'package:sqlparser/sqlparser.dart';

final engine = SqlEngine();
final result = engine.parse('''
SELECT f.* FROM frameworks f
  INNER JOIN uses_language ul ON ul.framework = f.id
  INNER JOIN languages l ON l.id = ul.language
WHERE l.name = 'Dart'
ORDER BY f.name ASC, f.popularity DESC
LIMIT 5 OFFSET 5 * 3
  ''');
// result.rootNode contains the select statement in tree form

Analysis

Given information about all tables and a sql statement, this library can:

  1. Determine which result columns a query is going to have, including types and nullability
  2. Make an educated guess about what type the variables in the query should have (it's not really possible to be 100% accurate about this because sqlite is very flexible at types, but this library gets it mostly right)
  3. Issue basic warnings about queries that are syntactically valid but won't run (references unknown tables / columns, uses undefined functions, etc.)

To use the analyzer, first register all known tables via SqlEngine.registerTable. Then, SqlEngine.analyze(sql) gives you an AnalysisContext which contains an annotated AST and information about errors. The type of result columns and expressions can be inferred by using AnalysisContext.typeOf(). Here's an example:

final id = TableColumn('id', const ResolvedType(type: BasicType.int));
final content = TableColumn('content', const ResolvedType(type: BasicType.text));
final demoTable = Table(
  name: 'demo',
  resolvedColumns: [id, content],
);
final engine = SqlEngine()..registerTable(demoTable);

final context =
    engine.analyze('SELECT id, d.content, *, 3 + 4 FROM demo AS d');

final select = context.root as SelectStatement;
final resolvedColumns = select.resolvedColumns;

resolvedColumns.map((c) => c.name)); // id, content, id, content, 3 + 4
resolvedColumns.map((c) => context.typeOf(c).type.type) // int, text, int, text, int, int

But why?

Moor, a persistence library for Dart apps, uses this package to generate type-safe methods from sql.

Limitations

Most on this list is just not supported yet because I didn't found a use case for them yet. If you need them, just leave an issue and I'll try to implement them soon.

  • Some advanced expressions, like CASTs aren't supported yet.
  • An UPSERT clause is not yet supported on insert statements

If you run into parsing errors with what you think is valid sql, please create an issue.

Thanks