Tuesday, September 8, 2015

Nimongo - pure Nimlang MongoDB driver

Foreword


Nim logo
Lately I've spent much time coding in Nim. If you're still not acquainted with this fascinating, one of the fastest growing programming languages, please visit its official site, and get your grips on it. To be short, it's a statically typed compiled programming language that translates its source code ANSI C, which is then compiled into native executables. Such flow makes it easy to employ Nimlang on a broad range of software and hardware platforms.

Other features like, possibly, the most progressive compile-time programming features, smart type inference, multi-methods, and many others makes coding in Nim much faster than in its counterparts.

Meet nimongo


We have been adopting Nim as an experiment for quite a time at one of our commercial projects here, at ZeoAlliance. Lack of powerful ecosystem and wide community made me to start working on some tools that are crucial for the project. One of such tools (located at Github), which is fully open source, is a nimongo - pure MongoDB driver for Nimlang. Although there's standard mongo client, it is still in an unusable state, and has one minus - it uses libbson.c, and libmongo.c, which makes setting it up  a bit more painful, than if the package was native to Nim language.

Note: Nimongo implements thread-safe synchronous client, AsyncMongo is still in an active development.

Installing nimongo


In order to install nimongo, you can either, clone it from Github, or use Nim native package manager called nimble:

$ nimble install https://github.com/SSPkrolik/nimongo

For now, you have to specify repository URL directly. Later, when some stable state of the project is achieved, I will add it to nimble repo, so you could just type nimble install nimongo. 

Importing prerequisities


In order to start using nimongo, you have to import two of its modules - nimongo.mongo and nimongo.bson which contains BSON serialization/deserialization routines, helper templates, and converters that help integrate Bson objects into Nim source code as seamlessly as possible. Also you have to import oids which is a Nim standard library module for working with MongoDB Object IDs. Other modules like times will also be helpful if you plan to use date and time data in your application.

import oids
import times

import nimongo.bson
import nimongo.mongo


nimongo.mongo module contains objects needed to work with MongoDB: Mongo, Database, Collection, Find, and other supportive types and constants.

Establishing connection with MongoDB server


The main object to start with is a Mongo, which is a reference type that manages communication between your application and MongoDB server. You can use newMongo(...) constructor by passing host and port connection parameters, or leaving those empty in order to connect to localhost:27017, - default MongoDB server parameters:

var m: Mongo = newMongo()
let connected = m.connect()

Also you can see procedure connect() call here, which actually establishes connection to server, and returns bool result that equals to true if connection was established successfully.

Performing commands on Mongo server


If you're familiar with how MongoDB operates, you are aware that there are different groups of operations that can be also split into different levels: server-, database-, collection-level. In nimongo those are distributed into procedures bound to Mongo, Database, and Collection types respectively. For now, there is only one operation available for server level - isMaster():

let ismaster = m.isMaster()

This procedure returns true if connected MongoDB instance is master in a replica set, or false if not. Other commands will also be added, but the work is still in progress. Also, beware using such kind of commands, having in mind that such commands perform actual network communication with server over network.

Accessing Databases and Collections


As Nim has nice operator overloading implemented, accessing to databases, and collections are made via [] (indexing operator) the same way as you take values by keys from hash tables. Also the syntax becomes quite friendly to Python developers ;) It looks like that:

let db = m["db"]
let collection = m["collection"]

Here we have db and collection objects. Applying these operator does not carry any real communication with server. Database object procedures perform database-level operations, and Collection object performs actual data insertion, updates, and queries.

Dealing with BSON


MongoDB operates documents encoded in BSON format, and that's why nimongo.bson module exists. In order to perform data manipulations with nimongo you have to create such objects. They are able to be serialized into a byte stream that MongoDB server understands.

BSON document can be created using initBsonDocument() constructor:

let empty = initBsonDocument()

This constructor creates empty BSON document, and then have to be filled with fields. You can also use B template, which simplifies the way you create BSON document objects, by squashing empty document creation, and offering ability to add fields right in place. Documents are filled with fields using procedure call operator ():

let doc = B("name", "John")("surname", "Smith")("salary", 100)
let doc2 = B(
    "name", "Sam")(
    "surname", "Uncle")(
    "salary", 1000)(
    "skills", @["power", "government", "army"]
    )
                 
As you can see from the example, two sample documents are created. Documents are created as "field-value" pairs, and value must be of Bson type. Fortunately, Nim offers such kind of procedures as converters, so all values like "Sam", "Uncle", and 1000 are implicitly converted to Bson type objects. Also mention the last field which is a sequence. This field is converted to BSON array. Also, () operator, as you could have mentioned, returns Bson type object, so you can append fields to objects any time in future.

Whether you are doing inserts, updates, or find queries, Bson objects is a strong foundation on which queries to MongoDB are made.

Inserting data into MonoDB using nimongo


When we have some documents created, we are ready to insert() those into database collection:

if collection.insert(doc):
    echo("Document was successfully inserted.")

Here we insert previously created BSON document into collection. insert() procedure returns bool value indicating success of the operation. This example shows how to insert single document, though it is also possible to insert serveral documents at a time with a single insert query by using overloaded insert() procedure that accepts sequence of Bson objects:

collection.insert(@[doc, doc, doc])


Updating documents using nimongo


Almost the same way as you do insertion, it is possible to perform document updates. Here how you can easily update single document using update() procedure:

let
    selector = B("integer", 100'i32)
    updater  = B("$set", B("integer", 200))

if c.update(selector, updater) == true:
    echo("Document updated succesfully.")

In this example we set new value 200 for a field named integer, where previous field's value was 100.

Removing document using nimongo


Document removal can be performed with the remove() procedure:

if c.remove(B("string", "hello")):
    echo("Document removed successfully.")

This procedure also return bool value, which indicates success of the operation.

Querying data from MongoDB


Querying data from MongoDB is a bit more complex than inserting, updating, or removal. Finding data in MongoDB usually requires some kind of configuration depending on the performed query. That's why find() procedure does not perfrom immediate querying, but creates Find object that is configurable, and the actual query is run with one(), all(), and items() procedures of the Find object. Let's fetch one document from database:

try:
    let fetched = collection.find(B("name", "John")).one()
    echo(fetched)
except NotFound:
    echo("Document was not found in database.")

Another way to fetch objects from database is to fetch a sequence of objects:

let documents = collection.find(B("name", "John")).all()
for item in document:
    echo item

And the last way to use query results is to use iterator in order to work with fetched documents:

for item in collection.find(B("name", "John")).items():
    echo item

Also you can tweak querying process by configuring Find object with such procedures as tailableCursor(), slaveOk(), noCursorTimeout(), and others, which you can find in MongoDB documentation.

Afterword


All this functionality, though does not fully implements MongoDB wire protocol, and does not expose whole power of MongoDB database, allows us to use it in small projects, or on early stages of large long-term projects.

I encourage all of those who plans writing Nim applications, and likes nimongo aproach, play with the driver, and support the project by contributing via pull requests, or filing issues, and wishlists.

In the near future, I plan to finish all BSON types support (now only widely used ones supported), add more service commands, and implement asynchronous driver based on asyncnet, and asyncdispatch Nim standard library modules.