Hare is a pure Go database management system that stores each table as a text file of line-delimited JSON. Each line of JSON represents a record. It is a good fit for applications that require a simple embedded DBMS.
To start using Hare, install Go and run go get
:
$ go get github.com/jameycribbs/hare
A directory of JSON files is represented by a hare.Database. Each JSON file needs a struct with it's members cooresponding to the JSON field names. Additionally, you need to implement 3 simple boilerplate methods on that struct that allow it to satisfy the hare.Record interface.
A good way to structure this is to put this boilerplate code in a "models" package in your project. You can find an example of this boilerplate code in the examples/crud/models/episodes.go file.
Now you are ready to go!
Let's say you have a "data" directory with a file in it called "contacts.json".
The top-level object in Hare is a Database
. It represents the directory on
your disk where the JSON files are located.
To open your database, you first need a new instance of a datastore. In this
example, we are using the Disk
datastore:
ds, err := disk.New("./data", ".json")
Hare also has the Ram
datastore for in-memory databases.
Now, you will pass the datastore to Hare's New function and it will return
a Database
instance:
db, err := hare.New(ds)
To add a record, you can use the Insert
method:
recID, err := db.Insert("contacts", &models.Contact{FirstName: "John", LastName: "Doe", Phone: "888-888-8888", Age: 21})
To find a record if you know the record ID, you can use the Find
method:
var c models.Contact
err = db.Find("contacts", 1, &c)
To update a record, you can use the Update
method:
c.Age = 22
err = db.Update("contacts", &c)
To delete a record, you can use the Delete
method:
err = db.Delete("contacts", 3)
To query the database, you can write your query expression in pure Go and pass it to your model's QueryContacts function as a closure. You would need to create the QueryContacts function for your model as part of setup. You can find an example of what this function should look like in examples/models/episodes.go.
results, err := models.QueryContacts(db, func(c models.Contact) bool {
return c.firstname == "John" && c.lastname == "Doe"
}, 0)
You can create associations (similar to "belongs_to" in Rails, but with less features). For example, you could create another table called "relationships" with the fields "id" and "type" (i.e. "Spouse", "Sister", "Co-worker", etc.). Next, you would add a "relationship_id" field to the contacts table and you would also add an embeded Relationship struct. Finally, in the Contact models "AfterFind" method, which is automatically called by Hare everytime the "Find" method is executed, you would add code to look-up the associated relationship and populate the embedded Relationship struct. Take a look at the crud.go file in the "examples" directory for an example of how this is done.
You can also mimic a "has_many" association, using a similar technique. Take a look at the files in the examples directory for how to do that.
There are also built-in methods you can run against the database to create a new table or delete an existing table. Take a look at the examples/dbadmin/dbadmin.go file for examples of how these can be used.
When Hare updates an existing record, if the changed record's length is less than the old record's length, Hare will overwrite the old data and pad the extra space on the line with all "X"s.
If the changed record's length is greater than the old record's length, Hare will write the changed record at the end of the file and overwrite the old record with all "X"s.
Similarly, when Hare deletes a record, it simply overwrites the record with all "X"s.
Eventually, you will want to remove these obsolete records. For an example of how to do this, take a look at the examples/dbadmin/compact.go file.
-
Records for each table are stored in a newline-delimited JSON file.
-
Mutexes are used for table locking. You can have multiple readers or one writer for that table at one time, as long as all processes share the same Database connection.
-
Querying is done using Go itself. No need to use a DSL.
-
An AfterFind callback is run automatically, everytime a record is read, allowing you to do creative things like auto-populate associations, etc.
-
When using the
Disk
datastore, the database is not read into memory, but is queried from disk, so no need to worry about a large dataset filling up memory. Of course, if your database is THAT big, you should probably be using a real DBMS, instead of Hare! -
Two different back-end datastores to choose from:
Disk
orRam
.