Home

Awesome

BadgerHold

Go GoDoc Coverage Status Go Report Card

BadgerHold is a simple querying and indexing layer on top of a Badger instance. The goal is to create a simple, higher level interface on top of Badger DB that simplifies dealing with Go Types and finding data, but exposes the underlying Badger DB for customizing as you wish. By default the encoding used is Gob, so feel free to use the GobEncoder/Decoder interface for faster serialization. Or, alternately, you can use any serialization you want by supplying encode / decode funcs to the Options struct on Open.

One Go Type will be prefixed with it's type name, so you can store multiple types in a single Badger database with conflicts.

This project is a rewrite of the BoltHold project on the Badger KV database instead of Bolt. For a performance comparison between bolt and badger, see https://blog.dgraph.io/post/badger-lmdb-boltdb/. I've written up my own comparison of the two focusing on characteristics other than performance here: https://tech.townsourced.com/post/boltdb-vs-badger/.

Indexes

Indexes allow you to skip checking any records that don't meet your index criteria. If you have 1000 records and only 10 of them are of the Division you want to deal with, then you don't need to check to see if the other 990 records match your query criteria if you create an index on the Division field. The downside of an index is added disk reads and writes on every write operation. For read heavy operations datasets, indexes can be very useful.

In every BadgerHold store, there will be a reserved bucket _indexes which will be used to hold indexes that point back to another bucket's Key system. Indexes will be defined by setting the badgerhold:"index" struct tag on a field in a type.

type Person struct {
	Name string
	Division string `badgerhold:"index"`
}

// alternate struct tag if you wish to specify the index name
type Person struct {
	Name string
	Division string `badgerholdIndex:"IdxDivision"`
}

This means that there will be an index created for Division that will contain the set of unique divisions, and the main record keys they refer to.

Optionally, you can implement the Storer interface, to specify your own indexes, rather than using the badgerHoldIndex struct tag.

Queries

Queries are chain-able constructs that filters out any data that doesn't match it's criteria. An index will be used if the .Index() chain is called, otherwise BadgerHold won't use any index.

Queries will look like this:

s.Find(&result, badgerhold.Where("FieldName").Eq(value).And("AnotherField").Lt(AnotherValue).Or(badgerhold.Where("FieldName").Eq(anotherValue)))

Fields must be exported, and thus always need to start with an upper-case letter. Available operators include:

An empty / zero value query matches against all records, because it has no critiera. You can then use Skip and Limit to page through all records in your dataset:

q := &badgerhold.Query{}

err := store.Find(&result, q.Skip(10).Limit(50))

If you want to run a query's criteria against the Key value, you can use the badgerhold.Key constant:

store.Find(&result, badgerhold.Where(badgerhold.Key).Ne(value))

You can access nested structure fields in queries like this:

type Repo struct {
  Name string
  Contact ContactPerson
}

type ContactPerson struct {
  Name string
}

store.Find(&repo, badgerhold.Where("Contact.Name").Eq("some-name")

Instead of passing in a specific value to compare against in a query, you can compare against another field in the same struct. Consider the following struct:

type Person struct {
	Name string
	Birth time.Time
	Death time.Time
}

If you wanted to find any invalid records where a Person's death was before their birth, you could do the following:

store.Find(&result, badgerhold.Where("Death").Lt(badgerhold.Field("Birth")))

Queries can be used in more than just selecting data. You can delete or update data that matches a query.

Using the example above, if you wanted to remove all of the invalid records where Death < Birth:

// you must pass in a sample type, so BadgerHold knows which bucket to use and what indexes to update
store.DeleteMatching(&Person{}, badgerhold.Where("Death").Lt(badgerhold.Field("Birth")))

Or if you wanted to update all the invalid records to flip/flop the Birth and Death dates:


store.UpdateMatching(&Person{}, badgerhold.Where("Death").Lt(badgerhold.Field("Birth")), func(record interface{}) error {
	update, ok := record.(*Person) // record will always be a pointer
	if !ok {
		return fmt.Errorf("Record isn't the correct type!  Wanted Person, got %T", record)
	}

	update.Birth, update.Death = update.Death, update.Birth

	return nil
})

If you simply want to count the number of records returned by a query use the Count method:

// need to pass in empty datatype so badgerhold knows what type to count
count, err := store.Count(&Person{}, badgerhold.Where("Death").Lt(badgerhold.Field("Birth")))

You can also use FindOne which is a shorthand for Find + Limit(1) which returns a single record instead of a slice of records, and will return an ErrNotFound if no record is found, unlike a normal Find query where an empty slice would be returned with no error.

	result := &ItemTest{}
	err := store.FindOne(result, query)

Keys in Structs

A common scenario is to store the badgerhold Key in the same struct that is stored in the badgerDB value. You can automatically populate a record's Key in a struct by using the badgerhold:"key" struct tag when running Find queries.

Another common scenario is to insert data with an auto-incrementing key assigned by the database. When performing an Insert, if the type of the key matches the type of the badgerhold:"key" tagged field, the data is passed in by reference, and the field's current value is the zero-value for that type, then it is set on the data before insertion.

type Employee struct {
	ID uint64 `badgerhold:"key"`
	FirstName string
	LastName string
	Division string
	Hired time.Time
}

// old struct tag, currenty still supported but may be deprecated in the future
type Employee struct {
	ID uint64 `badgerholdKey`
	FirstName string
	LastName string
	Division string
	Hired time.Time
}

Badgerhold assumes only one of such struct tags exists. If a value already exists in the key field, it will be overwritten.

If you want to insert an auto-incrementing Key you can pass the badgerhold.NextSequence() func as the Key value.

err := store.Insert(badgerhold.NextSequence(), data)

The key value will be a uint64.

If you want to know the value of the auto-incrementing Key that was generated using badgerhold.NextSequence(), then make sure to pass a pointer to your data and that the badgerholdKey tagged field is of type uint64.

err := store.Insert(badgerhold.NextSequence(), &data)

Slices in Structs and Queries

When querying slice fields in structs you can use the Contains, ContainsAll and ContainsAny criterion.

val := struct {
    Set []string
}{
    Set: []string{"1", "2", "3"},
}
bh.Where("Set").Contains("1") // true
bh.Where("Set").ContainsAll("1", "3") // true
bh.Where("Set").ContainsAll("1", "3", "4") // false
bh.Where("Set").ContainsAny("1", "7", "4") // true

The In, ContainsAll and ContainsAny critierion accept a slice of interface{} values. This means you can build your queries by passing in your values as arguments:

where := badgerhold.Where("Id").In("1", "2", "3")

However if you have an existing slice of values to test against, you can't pass in that slice because it is not of type []interface{}.

t := []string{"1", "2", "3", "4"}
where := badgerhold.Where("Id").In(t...) // compile error

Instead you need to copy your slice into another slice of empty interfaces:

t := []string{"1", "2", "3", "4"}
s := make([]interface{}, len(t))
for i, v := range t {
    s[i] = v
}
where := badgerhold.Where("Id").In(s...)

You can use the helper function badgerhold.Slice which does exactly that.

t := []string{"1", "2", "3", "4"}
where := badgerhold.Where("Id").In(badgerhold.Slice(t)...)

Unique Constraints

You can create a unique constraint on a given field by using the badgerhold:"unique" struct tag:

type User struct {
  Name string
  Email string `badgerhold:"unique"` // this field will be indexed with a unique constraint
}

The example above will only allow one record of type User to exist with a given Email field. Any insert, update or upsert that would violate that constraint will fail and return the badgerhold.ErrUniqueExists error.

ForEach

When working with large datasets, you may not want to have to store the entire dataset in memory. It's be much more efficient to work with a single record at a time rather than grab all the records and loop through them, which is what cursors are used for in databases. In BadgerHold you can accomplish the same thing by calling ForEach:

err := store.ForEach(badgerhold.Where("Id").Gt(4), func(record *Item) error {
	// do stuff with record

	// if you return an error, then the query will stop iterating through records

	return nil
})

Aggregate Queries

Aggregate queries are queries that group results by a field. For example, lets say you had a collection of employees:

type Employee struct {
	FirstName string
	LastName string
	Division string
	Hired time.Time
}

And you wanted to find the most senior (first hired) employee in each division:

result, err := store.FindAggregate(&Employee{}, nil, "Division") //nil query matches against all records

This will return a slice of Aggregate Result from which you can extract your groups and find Min, Max, Avg, Count, etc.

for i := range result {
	var division string
	employee := &Employee{}

	result[i].Group(&division)
	result[i].Min("Hired", employee)

	fmt.Printf("The most senior employee in the %s division is %s.\n",
		division, employee.FirstName + " " + employee.LastName)
}

Aggregate queries become especially powerful when combined with the sub-querying capability of MatchFunc.

Many more examples of queries can be found in the find_test.go file in this repository.

Comparing

Just like with Go, types must be the same in order to be compared with each other. You cannot compare an int to a int32. The built-in Go comparable types (ints, floats, strings, etc) will work as expected. Other types from the standard library can also be compared such as time.Time, big.Rat, big.Int, and big.Float. If there are other standard library types that I missed, let me know.

You can compare any custom type either by using the MatchFunc criteria, or by satisfying the Comparer interface with your type by adding the Compare method: Compare(other interface{}) (int, error).

If a type doesn't have a predefined comparer, and doesn't satisfy the Comparer interface, then the types value is converted to a string and compared lexicographically.

Behavior Changes

Since BadgerHold is a higher level interface than Badger DB, there are some added helpers. Instead of Put, you have the options of:

When getting data instead of returning nil if a value doesn't exist, BadgerHold returns badgerhold.ErrNotFound, and similarly when deleting data, instead of silently continuing if a value isn't found to delete, BadgerHold returns badgerhold.ErrNotFound. The exception to this is when using query based functions such as Find (returns an empty slice), DeleteMatching and UpdateMatching where no error is returned.

When should I use BadgerHold?

BadgerHold will be useful in the same scenarios where BadgerDB is useful, with the added benefit of being able to retire some of your data filtering code and possibly improved performance.

You can also use it instead of SQLite for many scenarios. BadgerHold's main benefit over SQLite is its simplicity when working with Go Types. There is no need for an ORM layer to translate records to types, simply put types in, and get types out. You also don't have to deal with database initialization. Usually with SQLite you'll need several scripts to create the database, create the tables you expect, and create any indexes. With BadgerHold you simply open a new file and put any type of data you want in it.

options := badgerhold.DefaultOptions
options.Dir = "data"
options.ValueDir = "data"

store, err := badgerhold.Open(options)
defer store.Close()
if err != nil {
	// handle error
	log.Fatal(err)
}


err = store.Insert("key", &Item{
	Name:    "Test Name",
	Created: time.Now(),
})

That's it!

Badgerhold currently has over 80% coverage in unit tests, and it's backed by BadgerDB which is a very solid and well built piece of software, so I encourage you to give it a try.

If you end up using BadgerHold, I'd love to hear about it.