TinyBase logoTinyBase

Inferring TablesSchemas

The TinyBase tools module includes a way to analyze existing data in a Store and attempt to generate a TablesSchema from it.

This is useful, for example, if you have imported external data into a Store and want to understand the shape and structure of the data - and then to constrain that shape through the use of the derived TablesSchema from then on.

The Tools Object

Firstly, to run any developer tools against a Store, you need to create a Tools object with the createTools function. This is not included in the main build of the tinybase module and must be imported explicitly from the tinybase/tools submodule.

Rather than this code living in your production application, it is more likely (and expected) to be used in a build script or similar.

Usage of the createTools function should be familiar by now, taking a reference to a Store:

import {createStore} from 'tinybase';
import {createTools} from 'tinybase/tools';

const store = createStore()
  .setValues({employees: 3, open: true})
  .setTable('pets', {
    fido: {species: 'dog'},
    felix: {species: 'cat'},
    cujo: {species: 'dog'},
  });

const tools = createTools(store);

(In reality it is more likely you would be inferring schemas from dynamically imported data, but for simplicity here, we are setting the data deterministically inline.)

The resulting Tools object is now associated with the Store, and, if they are not already explicitly set, the schemas can be inferred. The getStoreValuesSchema method is used to infer the ValuesSchema, and the getStoreTablesSchema method the TablesSchema. Each method returns an object:

console.log(tools.getStoreValuesSchema());
// -> {employees: {type: 'number'}, open: {type: 'boolean'}}

console.log(tools.getStoreTablesSchema());
// -> {pets: {species: {type: 'string', default: 'dog'}}}

Since every Row in the pets Table has a string species Cell, and dog appears the most often, the Schema reflects that type and default. Remember that a Cell can be a string, number, or boolean:

store.setTable('pets', {
  fido: {price: 4},
  felix: {price: 5},
  cujo: {friendly: false},
});

console.log(tools.getStoreTablesSchema());
// -> {pets: {price: {type: 'number'}, friendly: {type: 'boolean'}}}

In this case, not every Row has the same set of Cell Ids, and so no default is inferred in the TablesSchema.

Applying An Inferred Schema

Of course, you can programmatically apply the resulting schemas back to the Store so that future data insertions adhere to the shape of the existing data:

store.setTable('pets', {
  fido: {price: 4},
  felix: {price: 5},
  cujo: {price: 4},
});

store.setTablesSchema(tools.getStoreTablesSchema());

store.setRow('pets', 'rex', {barks: true});
console.log(store.getTable('pets'));
// -> {fido: {price: 4}, felix: {price: 5}, cujo: {price: 4}, rex: {price: 4}}

store.delTablesSchema();

Failure Conditions

There are cases in which a TablesSchema cannot be inferred. Most commonly, this will happen if data types are not consistent across each Row, or if there is no data at all:

store.setTable('pets', {
  fido: {price: 4},
  felix: {price: 5},
  cujo: {price: 'not for sale'},
});

console.log(tools.getStoreTablesSchema());
// -> {}

store.delTable('pets');

console.log(tools.getStoreTablesSchema());
// -> {}

Note that inferring a schema is an all-or-nothing operation. Even if only one Table amongst many has inconsistent data, the TablesSchema as a whole will be empty. One of the things this ensures is that the workflow is idempotent: if you take an inferred schema and reapply it to the Store, there will be no loss of data.

We can take this technique one step further and actually derive type definitions and ORM-like implementations from our data or schemas. Move on to the Generating APIs guide for more details.