Laravel Batch v3.2.0

New bulk operations for large tables and files, JSON column updates, and two JSON bug fixes.
Everything is additive, so upgrading from 3.1 needs no code changes:

composer update mavinoo/laravel-batch

Added

Update keys inside JSON columns (#64, #83, #111)

Batch::update(new User, [
    ['id' => 1, 'settings->theme' => 'dark', 'settings->notify->email' => false],
]);

• Works in every update method, on MySQL, MariaDB, PostgreSQL and SQLite.
• Other keys are kept, missing objects are created, and values keep their JSON type.
• updated_at only changes when the document changes.
• Arrays and collections for columns the model casts to JSON (array, json, collection,
  AsCollection, ...) are encoded the way the model stores them, in update, insert and upsert.

insertGetIds(): insert rows and get their ids (#48)

$ids = Batch::insertGetIds(new User, $rows); // [101, 102, ...] in row order

• Checked with 8 concurrent writers and 78,720 rows on MySQL 8.4, MariaDB, PostgreSQL and SQLite,
  with 0 wrong ids.

import(): import files and iterables with constant memory

Batch::import(new User, storage_path('users.csv'), ['mode' => 'upsert', 'uniqueBy' => ['email']]);

• Reads CSV, TSV and JSON Lines files (also .gz), open streams, or any iterable such as
  Model::cursor().
• Supports column mapping, null values, transforms, and progress and error callbacks.

500,000-row CSV	import()	array + insertRows()
Memory	7 MB	765 MB
Time	30–45% faster

splitFile(): split large files

Batch::splitFile($path, lines: 50000); // users-001.csv, users-002.csv, ...

• Splits by number of records and / or size.
• Never cuts a record or a multi-line CSV field in half, and repeats the header in every part.

deleteInChunks(), updateInChunks(), archive()

Log::where('created_at', '<', now()->subYear())->deleteInChunks(10000);
User::where('active', false)->updateInChunks(['status' => 'archived']);
Order::where('created_at', '<', '2020-01-01')->archiveTo('orders_archive');

• Each runs a chunk at a time, in primary key order, so no lock is held for long.
• Deleting 200k rows on MySQL 8.4: the longest chunk took 0.10 s, against 0.8 s for one DELETE.
• archive() copies and deletes each chunk in one transaction.

sync(): make a table match a list

Product::where('supplier_id', 5)->syncRows($feedRows, ['sku']);
// ['inserted' => 120, 'updated' => 4800, 'deleted' => 35]

• Inserts, updates and deletes in one transaction.
• The database compares the keys itself, so its collation decides which are equal. On MySQL,
  "ABC" and "abc" are the same key.
• Soft-deleted rows that are in the list are restored.
• Syncing 100,000 rows takes 1.4–4.8 s with about 15 MB of memory.

Fixed

• Updating a PostgreSQL json column failed with operator does not exist: json <> unknown.
• MySQL json columns touched updated_at even when their value didn't change.

Also

• 219 tests on SQLite, MySQL, MariaDB and PostgreSQL, PHP 8.1 – 8.5 and Laravel 10 – 13.
• PHPStan level 9 with Larastan.

Full changelog: https://github.com/mavinoo/laravelBatch/blob/master/CHANGELOG.md
Pull request: #126