peerlibrary:peerdb-migrations

v1.1.1Published 2 years ago

PeerDB Migrations

Meteor smart package which provides migrations for PeerDB reactive database layer.

Adding this package to your Meteor application enables migrations on all documents defined using PeerDB.

Server side only.

Installation

meteor add peerlibrary:peerdb-migrations

Migrations

In any large project your documents' schemas will evolve through time. While MongoDB does not really enforce any schema, your program logic might require one. It is error prone to allow documents with various versions of schemas to exist at the same time and it makes your program code more complicated.

To address this, PeerDB migrations provide you with a way to define schema migrations. They are applied automatically on a startup of your application to migrate any documents as necessary. To know which documents are at which schema version, PeerDB migrations add a _schema field to all PeerDB documents with a semantic version number. This allows recovery of partially migrated collections and importing documents with a different schema version.

To define a migration, extend a class Document.PatchMigration, Document.MinorMigration, or Document.MajorMigration, depending on whether your schema change is a backwards-compatible bug fix, you are adding functionality in a backwards-compatible manner, or you are making an incompatible change, respectively. By default migrations just update the schema version:

class Migration extends Document.PatchMigration
  name: "Migration not really doing anything with data"

  forward: (document, collection, currentSchema, newSchema) =>
    migrated: 0
    all: collection.update {_schema: currentSchema}, {$set: _schema: newSchema}, {multi: true}

  backward: (document, collection, currentSchema, oldSchema) =>
    migrated: 0
    all: collection.update {_schema: currentSchema}, {$set: _schema: oldSchema}, {multi: true}

Post.addMigration new Migration()

The base migrations class uses the migration definitions above. You have to define both forward and backward methods, but you can leave them undefined to use inherited definitions. Methods should return an object containing migrated and all counts. migrated is the count of all documents which were migrated, while all includes documents which were not really migrated (maybe because they were not applicable for the given migration), but they still exist in the collection and only their schema version was updated. The method must update the schema version on all documents in a collection. (An easy pattern is to first migrate applicable documents and then call super for the rest, returning the combined counts.)

Inside a migration method you should use DirectCollection, provided to you as a collection argument. If you need access to other collections in your migrations, you should use DirectCollection as well. DirectCollection is a library which provides a Meteor-like API for accessing MongoDB, but bypasses Meteor completely, allowing you direct interaction with the database. This is necessary in migrations because collection names could be changing during migrations and Meteor cannot really handle that.

You should call addMigration in the order you want migrations to be applied. The easiest way to assure that is to create a directory server/migrations/ in your project and have multiple files with XXXX-description.coffee filenames, where XXXX is a consecutive number for the order you want, each file adding one migration.

There are some migration classes predefined:

  • AddReferenceFieldsMigration – you are adding fields to a reference to be synced and you want to trigger resyncing of fields
  • RemoveReferenceFieldsMigration – you are removing fields from a reference and you want to trigger resyncing of fields
  • AddGeneratedFieldsMigration – you are adding auto-generated fields and you want to trigger generation of fields - you should pass a list of field names
  • ModifyGeneratedFieldsMigration – you are modifying auto-generated fields and you want to trigger regeneration of fields - you should pass a list of field names
  • RemoveGeneratedFieldsMigration – you are removing auto-generated fields - you should pass a list of field names
  • AddOptionalFieldsMigration – you are adding optional fields - you should pass a list of field names
  • AddRequiredFieldsMigration – you are adding required fields - you should pass a map between new field names and their initial values
  • RemoveFieldsMigration – you are removing fields - you should pass a map between field names to be removed, and their values which should be set when reverting the migration
  • RenameFieldsMigration – you are renaming fields - you should pass a map between current field names and the new names

To rename a collection backing the document, use renameCollectionMigration:

Post.renameCollectionMigration 'oldPosts', 'Posts'

Settings

PEERDB_MIGRATIONS_DISABLED=

If you want migrations to not run, set PEERDB_MIGRATIONS_DISABLED to a true value. Recommended setting is that only one web-facing instance has migrations enabled and all other, including PeerDB instances, have them disabled. This prevents any possible conflicts which could happen because of running migrations in parallel (but you are writing all your migrations in a way that conflicts will never happen, using the Update If Current pattern, aren't you?).

Disabling migrations just disables running them; documents are still populated with the _schema field.

Examples

See migrations in: