Download the PHP package prewk/snapper-php without Composer

Snapper 🐡

Converts rows from a relational database into serialization snapshots that can be deserialized into new rows at a later time with complex relationships preserved.

Example

From database into the serializer

Out from the serializer

Deserialize into database again

How to serialize?

How to deserialize?

Ingredient types

Value

Pass through the value of the field.

Raw

Force the value of the field.

Ref

The field is a foreign key that references another table. The optional method takes arguments that, when equal to the encountered field value, is considered non-references and are just passed through like Value.

Morph

Specify a polymorphic relation using two fields, one type (foo_type) and one id (foo_id), and map values in foo_type against other tables.

Requires both fields to be present to function properly, supports optional values that when matched ignores the foreign relation and just passes though the value.

Match

Look at the given field (type in the example above) and become different ingredients depending on its value.

• on: Exact match
• pattern: Regexp match
• default: Fallback

If no match is successful and no default is provided, the field will not be included.

Json

Define references nested in JSON.

On string and integer refs inside the JSON

The library is built for the normal scenario of integer keys (AUTO_INCREMENT), but when serializing references they will be converted to UUIDS (v4). They are strings that look like this: a0ff60f5-87fe-4d4e-855b-8993f1c3b065.

This poses a problem in JSON when..

..gets serialized into..

..and back into a integer key:

"456" is not strictly equal to 456. Therefore, the deserialization logic is as follows:

1. If the id returned from the inserter is numeric then all "UUID" will be replaced with INSERT_ID
2. All UUID will be replaced with INSERT_ID

Circular

If two of your tables contain circular references to each other, wrap one of the references with a Circular ingredient. Specify a valid fallback as optional value and specify that value as a fallback Raw ingredient.

The resulting serialization will start with an INSERT op containing the fallback value, and end with an UPDATE op with the real reference.

Events

onDeps

Override recipes/updaters/inserters

Recipe JSON

Recipes can be converted to/from JSON which can be useful.

Validation

Validate a serialization

Check a serialization for unresolvable references

Validate a JSON recipe

Recipes can be validated with a JSON schema validator.

Batched inserts/updates

The insert/update closures will be called with batches of rows that each can be executed in one SQL operation if you want to optimize:

The following rules apply for the return value from inserters:

• Returning void is acceptable, but the serialization will fail if later rows depend on the skipped primary keys
• To return the primary keys, return an array of the same length as $rows, everything else is invalid

The batch grouping logic considers one of the following conditions as "start a new batch of operations":

• A new row type (table) is encountered
• The row is dependent on the primary key of another row earlier in the same batch
• The exact number of, or names of, fields has changed from one row to the next

Id manifest

The result of a compilation ($serializer->compile()) is a Serialization. It has two methods:

• getOps() - Get an array representing the sequence of operations (the actual "serialization" if you will)
• getIdManifest() - Get a dictionary mapping the internal uuids in the serialization to your given db ids, grouped by type (table name)

Why?

The library is useful for providing snapshot functionality to multi-tenant services where a user owns one or more complex sets of data with complex internal relationships.

Howtos

I don't want to create table X, it already exists

The deserializer doesn't care about the inserter closure's internal logic, it only requires you to return the primary key:

I need row metadata to be serialized

Just use $recipe->value() in the recipe and feed whatever fields you want to the Serializer and it'll end up in the snapshot.

When deserializing, pick and choose which fields to insert into the database in your inserter.

I have composite keys

Unsupported at the moment, a workaround is adding a unique id and pretending that's the primary key.

License

MIT