NodeJS ORM based on ObjectionJS with some extra utilities
ObjectionJS documentation: http://vincit.github.io/objection.js/ ObjectionJS Repo: https://github.com/Vincit/objection.js/
Xorm adds some more functionalities in ObjectionJS.
You don't need to give tableName in your model. By default the tableName = name of the model.
// eg. instead of writing this
class Person extends Model {
static tableName = 'Person';
...
}
// you can write
class Person extends Model {
...
}
Every model has automatically managed timestamps (createdAt
and updatedAt
), which gets updated whenever the model is created or updated.
For disabling the timestamps you can do static timestamps = false;
in your model.
timestamps
can be true, false or an object of {createdAt, updatedAt}
. If true, createdAt
and updatedAt
columns will automatically be updated.
You can change column names using an object.
static timestamps = {createdAt: 'add_time', updatedAt: 'modify_time'};
If you omit one column in the object, that column won't be touched at all eg. if you don't want updatedAt:
static timestamps = {createdAt: 'createdAt'};
For dealing with timestamp few extra methods are added:
touch
: update updatedAt
of the model to current time.
dontTouch
: don't update the timestamps after executing the query
For turning on soft delete you can do static softDelete = true;
(make sure your schema has deletedAt
column as a nullable timestamp)
So whenever you delete a model, it won't actually be deleted from the database, instead the value of deletedAt
column will be set to current time.
When retriving, it will automatically take care of the deleted models (they won't come in results).
For dealing with softDelete few extra methods are added:
delete
: soft delete or delete a model (depending on whether softDelete is true or false)
trash
: soft delete a model (you won't usually need this method, just use delete
)
withTrashed
: get all models (whether they are soft deleted or not)
onlyTrashed
: only get soft deleted models.
restore
: restore (undelete) a soft deleted model.
forceDelete
: actually delete a model from the database
If you want to use some other column name instead of deletedAt
you can do static softDelete = 'deleted_at';
class Person extends Model {
// Turn on soft delete
static softDelete = true;
}
// soft delete
Person.query().where('id', 1).delete();
// restore a soft deleted model
Person.query().where('id', 1).restore();
// get all persons (except deleted ones)
Person.query();
// get all persons (including deleted ones)
Person.query().withTrashed();
// get only deleted persons
Person.query().onlyTrashed();
// delete a person from the database
Person.query().forceDelete();
You can set timestampColumns
property of the model class to convert some columns to Date
automatically while making objects from fromJson
or insert
.
createdAt
, updatedAt
and deletedAt
are automatically added to timestampColumns
class Person extends Model {
static timestampColumns = ['lastLoginTime', 'bannedAt'];
}
There are some graphql helper functions defined. You should use them because they take care of batching and optimizations.
returns a resolver for GraphQL query where you query the item by a single unique field
category(id: ID, name: String): Category
Query: { category: Category.getFindOneResolver() }
Returns a resolver for a relation that is defined in the model.
You can override the default related resolver by defining selfRelationResolver(relation, options)
in the related model.
type Store { id: ID!, name: String!, category: Category }
Store: { category: Store.getRelationResolver('category') }
Options can be:
// options can be
// default: default value to return if the relation returns null, can be a function
returns a resolver for GraphQL delete mutation
deleteCategory(id: ID) { DeletedItem }
Mutation: { deleteCategory: Category.getDeleteByIdResolver() }
returns a resolver for finding model by id. It optionally takes a propName argument which denotes the name of the field containing the id.
type Store { id: ID!, name: String!, category: Category }
Store: { category: Category.getFindByIdSubResolver() }
Store: { category: Category.getFindByIdSubResolver('categoryId') }
These functions are used to modify (or any other action like logging) the result of resolver before and after the item has been resolved.
afterResolve(options) {
const params = {options.args};
if (params) {
this.url = `${this.url}?${params}`;
}
return this;
}
// options will be
// args: graphql query arguments (it'll be automatically filled by xorm)
// isDefault: whether the query returned null and default value is used
returns a Data Loader for batching and caching purposes. You can optionally give it a context to return different loaders for different contexts.
// get store by id 1
Store.getIdLoader().load(1);
// get multiple stores
Store.getIdLoader().loadMany([1, 2, 3]);
Note: findById
is modified to use loader automatically. You should use findById
wherever
possible to get the benefits of batching.
Store.query().findById(1);
// optionally give it a context
Store.query().loaderContext(ctx).findById(1);
returns a Data Loader for batching and caching purposes. You can optionally give it a context to return different loaders for different contexts. It can return a loader for any arbitary column. Use this when the column returns only one result (unique)
// get user by email
User.getLoader('email').load('abc@xyz.com');
// get multiple users
User.getLoader('email').loadMany(['abc@xyz.com', 'def@xyz.com']);
returns a Data Loader for batching and caching purposes. You can optionally give it a context to return different loaders for different contexts. It can return a loader for any arbitary column. Use this when the column returns many results (non-unique)
// get user by email
User.getManyLoader('country').load('IN');
// get multiple users
User.getManyLoader('country').loadMany(['IN', 'US']);
Short for getIdLoader(options).load(id)
Short for getLoader(columnName, options).load(columnValue)
Both loadById
and loadByColumn
can accept options which can an object of
{
ctx: null, // context for the dataloader [optional / default null]
nonNull: false, // only return nonNull results [default false]
limit: null, // only return this many results [default null => return as many results as possible]
offset: 0, // in conjunction with limit [default 0]
}
Short for getManyLoader(columnName, options).load(columnValue)
You can use this to get a related model. This will automatically take care of batching. args
can be used to pass arguments to beforeResolve
and afterResolve
const product = await Product.query().findById(1);
const category = await product.loadByRelation('category');
You can use this function to make a loader if the query is not covered by existing loaders.
class Product extends Model {
async setIndexed() {
const loader = Product.makeLoader('setIndexed', async (keys) => {
return Product.query().update('indexed', 1).whereIn('id', keys);
}, {ignoreResults: true});
return loader.load(this.id);
}
}
// options can be
// ignoreResults: [default false] ignore the results returned by the loader function
// mapBy: [default 'id'] map results returned by the loaderFn using this key
// cache: [default false] cache the results returned by loaderFn indefinitely
// filterKeys: [default true] filter the falsy keys before calling loaderFn, filterKeys can also be a function
save
: inserts a model if the id column does not exist, otherwise updates it.
saveAndFetch
: saves the model and then fetches it.
updateById(id, fields)
: updates the model by id
patchById(id, fields)
: patches the model by id
deleteById(id)
: deletes the model by id
All three merge the id property into the query context
creates an and (or - or - or) condition
q.where('id', 1).whereByOr({votes: 100, user: 'smpx'})
// where `id` = 1 and (`votes` = 100 or `user` = 'smpx')
creates an or (and - and - and) condition
q.where('id', 1).whereByAnd({votes: 100, user: 'smpx'})
// where `id` = 1 or (`votes` = 100 and `user` = 'smpx')
order the items by position in the array given (of column). this is mostly useful in whereIn queries where you need ordered results.
const ids = [1, 4, 3, 5, 8];
q.whereIn('id', ids).orderByArrayPos('id', ids);
equivalent to this.where(column, values).orderByArrayPos(column, values)
.
use this for returning results ordered in the way you gave them
const ids = [1, 4, 3, 5, 8];
q.whereInOrdered('id', ids)
find
: find is like where except if only a single argument is given, it treats the argument as an id.
// all three are equivalent
Person.query().find(1);
Person.query().where('id', 1);
Person.query().findById(1);
// both are equivalent
Person.query().find('name', 'Hitesh');
Person.query().where('name', 'Hitesh');
Instead of doing Person.query().where()
you can do Person.where()
Instead of doing Person.query().find()
you can do Person.find()
You can use these properties to cache things.
$cache
is for instance and cache
is for the Model
redisCache
is cache backed by Redis for the Model
class Person extends Model {
static getAll() {
return this.redisCache.getOrSet('all', () => Person.query());
}
async getChildren() {
return this.$cache.getOrSet('children', () => this.$loadRelated('children'));
}
}
You can set cacheById
if you want to cache items by their id.
cache will be used when you use loadById
function.
Items are automatically removed from cache on $afterUpdate
and $afterDelete
cacheById
is an object of {ttl, columns, excludeColumns, maxLocalItems}
class Person extends Model {
static cacheById = {
ttl: '3h',
};
}
// items are now being cached by their id for 3 hours
// the following will result in only 2 DB queries
await Person.loadById(1);
await Person.loadById(1);
await Person.loadById([1, 2]);
await Person.loadById(2);
You can define all your relationships in the $relations
method.
belongsTo(model, options) hasOne(model, options) hasMany(model, options) hasManyThrough(model, options)
options are totally optional. they can be:
name: name of the relationship (by default name of the model in camelCase)
joinFrom: join field from (by default ModelTableName.camelCasedModelNameId)
joinTo: join field to (by default ModelTableName.id)
filter: apply this filter to the model
through: {
model: if there is a model for the through table
table: tableName of the table to join through
from: join through from
to: join through to
filter: filter to apply on the join table
extra: extra fields to select from the through table
}
class Pet extends Model {
static $relations() {
this.hasOne(Pet);
this.hasMany(Toy, {name: 'toys'});
this.belongsTo(Person);
this.hasManyThrough(Pet, {
name: 'siblings',
through: {
table: 'Pet_Siblings',
from: 'pet_id',
to: 'sibling_id',
},
});
}
}
Model can be a model object, or an absolute path to a model class. It can also be a relative path if you set the basePath of all models using Model.setBasePath(path)
If you throw any error using UserError class, it'll be sent to the API client (if you're using graphql & gqutils). Otherwise errors are simply sent as Server Error
. You can also throw an object using it.
import {UserError} from 'xorm';
async function saveStore(store) {
if (!store.name) {
throw new UserError({
name: 'Name is required',
});
}
}
// Alternatively instead of importing UserError you can simply throw Model.Error
async function saveStore(store) {
if (!store.name) {
throw new Store.Error({
name: 'Name is required',
});
}
}