A single setting to disable all schema changes (DDL) at run time, Lookup contents included

オープン
#1,649 コメント 3 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 6 日以内に返信

まだ誰も着手していません。

評価

難易度
5/5
見積もり時間
1週間以上
初心者へのやさしさ
15/100
issue の種類
機能追加
明瞭さ
おおむね明確
活発さ
活発
技術スタック
mysql, postgresql, python
領域
backend, databases

調査の方向性

Start with schemas.py around lines 329-340 (Lookup contents insertion) and Schema.call / _decorate_table, then autopopulate.py lines 186-187 and jobs.py lines 383-384, where jobs tables are declared. The open design question is where a single allow_ddl check should sit, and whether it must cover alter() and drop(). Done means that with the setting false, each of these paths raises a clear error naming the setting, and an explicit declare_all() path exists for deployments.

索引モデルが issue の本文から書いたものです。

説明

What the library does today

Schema changes happen implicitly whenever a process activates a schema with a privileged login:

  • Activating a schema declares every table that doesn't exist yet (Schema.__call__ → _decorate_table, when create_tables is true, the default).
  • Declaring a Lookup table inserts any contents rows that are missing (schemas.py:329–340).
  • populate() and jobs.refresh() declare a table's ~~jobs table on first use (autopopulate.py:186–187, jobs.py:383–384), even when create_tables is false.
  • alter() and drop() run whenever they're called.

So setting create_tables=False, as the production guide advises, still leaves schema changes at run time, through the jobs tables and Lookup contents. Nothing turns them all off at once.

Proposal

A single setting, for example database.allow_ddl (env DJ_ALLOW_DDL), default true. With it set to false, nothing changes the schema:

  • no table, part or jobs-table declaration;
  • no Lookup contents inserted, updated or deleted;
  • no alter(), drop() or schema creation.

An attempt raises a clear error naming the setting, rather than silently doing nothing. A missing table is reported as missing, not declared, and populate() and refresh() fail clearly if a jobs table doesn't exist.

Why

In a controlled deployment, schema changes are a deployment step: planned, reviewed and applied together, before the new code runs. Run-time processes (workers, notebooks, dashboards) then read and write rows but never change the schema. This setting lets an operator enforce that in the library, not only through database privileges.

Also needed

An explicit way for the deployment step to apply schema changes (declare missing tables, their jobs tables and Lookup contents) with the setting on, e.g. a schema.declare_all(), or the setting being overridable for one call.

Expected behavior

With allow_ddl=false, a pipeline whose tables all exist runs as before; one with a missing table, a missing jobs table or changed Lookup contents fails at activation or populate, with a message that says what's missing.

Related

#1552 proposes declaring draft tables on the server beside the pipeline's tables. With schema changes treated as a deployment step, should drafting stay entirely in memory instead?

主要言語
Python
スター
198
フォーク
103
平均マージ
6日 1時間
マージ済み PR(30日)
10

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

datajoint/datajoint-python のほかの issue

datajoint/datajoint-python の issue をすべて見る

似ている issue

Python の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。