RSStockFighter Framework

In this post I will talk about the StockFighter framework that I crated for the Swift language named RSStockFighter.  If you have not heard about StockFighter, it is a really nice (and free) online coding challenge where you need to solve a series of problems by using your coding skills.   You can read about StockFigher here.  The GitHub repository for RSStockFighter is here.  I am currently finishing up my third book on the Swift programming language so I have not had a lot of time to play the StockFighter game (yet) but I have used RSStockFighter to make it though level 4. 

I first read about StockFighter on DZone.  The idea behind StockFighter is you are a developer that has been tasked to create an automated trading system designed to automate stock trading for your firm.  The game has several levels that progressively get harder.  The first few levels are designed to get you familiar with the API while later levels require you to incorporate logic to determine when to buy and when to sell your stocks.  Don’t worry if you do not have any stock market experience, the game teaches you everything you need to know to complete each level. 

All of the APIs are REST based which means you can use any language you want to complete the levels.  I obviously chose to use the Swift language.  Why?  Because I really like programming with the Swift language.  RSStockFighter is designed to make it very easy to interact with the StockFighter API so you can focus on solving the levels rather than the backend network code. 

When I started looking at the StockFighter API, my first thought was I needed to create a framework for the API so I would not need to keep creating or cutting/pasting the code to perform the network requests in each challenge.  In this post I will explain how I designed RSStockFighter and how you can use it in your code.

At the time I am writing this document, StockFighter has ten documented API calls (I have currently implemented seven of the calls in RSStockFighter). One of my biggest concerns is the StockFighter challenge has not been around very long therefore I suspect they will be changing and/or adding to the API.  This means that it needs to be easy to add new API calls to RSStockFighter.  When I have a design problem like this I usually start off by pulling out one of my design pattern books and look through it to see if any of the patterns may solve my problem.  In this case the command pattern was perfectly suited for my needs.

The command pattern falls under the behavioral patter group.  In this pattern we encapsulate the logic of our actions into types that conforms to a command protocol.  We can then provide instances of the command types for use by an invoker.  The invoker will use the interface provided by the protocol to invoke the actions on the types that conform to that protocol.  What this means is we will create a separate command type for each one of our API calls.   We will then submit instances of these types to our invoker which will make the call to the StockFighter API.

The class structure to make our API calls will look like this:

The StockFighter request types will encapsulate the information needed to make the individual requests.  The RSStockFighterTransaction type will contain an instance of a StockFighter request type and also the information to make the network request.  Finally the RSStockFighterTransactionRequest will contain methods to make the actual network request to the StockFighter API.

I created a type named StockFighterHelper which implements the Façade pattern to hide the underlying RSStockFighterTransaction and RSStockFighterTransactionRequest types.  I did this for two reasons.  The first is to hide the complexity of the types and the second is to make it easy to change the interfaces of these types if needed.

In addition to the seven StockFighter request types we also have seven StockFighter response types.  These response types will encapsulate the logic needed to parse the response.  This will make it easy to pull the information from the response, as we need it. 

Enough about the design of RSStockFighter, lets see how we would use it.  We will start off by seeing if the StockFighter API is up.  The following code will make a request to check the status of the stockFighter API.

func apiStatus() {

       

    let request = StockFighterAPIStatus()

       

    StockFighterHelper.sendStockFighterTransaction(request, success: {

       (dict:NSDictionary!) -> Void in

        self.checkStatusResponse(dict)

        }, failure: {

            (str: String) -> Void in

            print("Error:  \(str)")

        })
}

In this function we begin by creating an instance of the StockFighterAPIStatus type.  We then call the static sendStockFighterTransaction() method from the StockFighterHelper protocol.  The sendStockFighterTransaction() method takes three parameters which are:

• transType:  An instance of a type that conforms to the StockFigherRequestType type which contains the information for our request. 
• success:   A completion handler that will be called if the StockFighter API call was successful
• failure:  A completion handler that will be called if the StockFigher API call failed.

In this example we set the transType parameter to an instance of the StockFighterAPIStatus type.  In the success completion handler we call a method named checkStatusResponse() and in the failure completion handler we print the error message to the console.

The checkStatusResponse() method called if the API call was successful looks like this:

func checkStatusResponse(dict: NSDictionary) {

    let response = StockFighterAPIStatusResponse(dict: dict)

    print("StockFighter is up:  \(response.ok)")
}

In this method we use the StockfighterAPIStatusResponse type to parse the response.  We then print a message to the console letting us know if the Stock Fighter API is up or not.

That was pretty easy but lets look at a more difficult API call like placing a new order.  The following code demonstrates how to do this:

let order = StockFighterOrder(symbol:SF_STOCK_SYMBOL, price:95.00,qty: 100,direction:SF_OrderDirection.Buy,orderType:SF_OrderTypes.Limit.rawValue)

       

StockFighterHelper.sendStockFighterTransaction(order, success: {

    (dict:NSDictionary!) -> Void in

        self.checkOrderResponse(dict)

}, failure: {(str: String) -> Void in

        print(str)
})

 

In this example we create an instance of the StockFighterOrder type.  The initializer for this type takes five parameters which are:

• symbol:  The stock symbol for the stock to place the order for
• price:  The maximum price to buy or minimum price to sell the stock at
• qty:  The quantity to buy or sell
• direction:  Specifies if this is a buy or a sell order
• orderType:  The order type

We then pass the instance of the StockFighterOrder type to the static sendStockFighterTransaction() method, just like we did with the API status example, to make the request to the StockFighter API.

If the API call was successful we would parse the response using the StockFighterOrderResponse type like this:

let response = StockFighterOrderResponse(dict: order)

Pretty easy, don’t you think?  The seven StockFighter request types that I have implemented in RSStockFighter are:

• StockFighterAPIStatus
• StockFighterVenueStatus
• StockFighterStocksOnVenue
• StockFighterOrder
• StockFighterOrderBook
• StockFighterQueryOrder
• StockFighterCancelOrder

For each of these request types I also implemented a response parser type which are:

• StockFighterAPIStatusResponse
• StockFighterVenueStatusResponse
• StockFighterStocksOnVenueResponse
• StockFighterOrderResponse
• StockFighterOrderBookResponse
• StockFighterQueryOrderResponse
• StockFighterCancelOrderResponse

There are two enumerations defined in this framework that you will be working with.  These are the SF_OrderTypes and SF_OrderDirection enumerations.  The SF_OrderTypes enumeration defines the type of order we are placing (You can read more about order types in the StockFighter help pages) and has the following values:

• Limit - Immediately matches any order on the books that has an offer price as good or better than the one listed on the order.  This order type is good until cancelled
• Market - Immediately matches any order.  Do not use this type.
• FOK - Fill or Kill:  Immediately fills the whole order.  If it cannot fill the whole order immediately then the whole order is killed
• IOC - Immediate or Cancel:  Immediately fills or partially fills the whole order and then cancels any shares that remains.

The SF_OrderDirection is used to define if the order is a buy or sell order and have the following values:

• Buy - The order is a purchase shares
• Sell - The order is to sell shares

Before you can use the API, you do need to make changes to a couple global constants.  These constants are in the StockFighterInfo.swift file.

• SF_API_KEY – Your API ley
• SF_VENUE – The venue you are using for this challenge
• SF_ACCOUNT – The account you are using for this challenge
• SF_STOCK_SYMBOL – The symbol of the stock you are using for this challenge

Overall the framework consists of six files that you will need to download and include in your project.  These files are:

• StockFighterInfo
• StockFighterKeys
• StockFighterRequestTypes
• StockFighterResponseTypes
• RSStockFighterTransactionRequest
• RSStockFighterTransaction

The GitHub repository for RSStockFighter is located here: 

So why am I releasing this framework?  I would like to see how many Swift Developers can complete this challenge.  As you go though the levels using RSStockFighter, post a comment to this blog posting telling us how far you have advanced (please no hints on how to complete a level).

Is using RSStockFighter cheating at the challenge?  Personally I do not think so.  I see using RSStockFighter like using any other third party framework (like the Alamofire framework) in our applications.  We use third party frameworks in our day to day development work to make our lives easier so we can focus on the business logic, why not use them to make challenges like this easier so we can focus on the business logic.

Good luck and happy coding.

Create a Data Access Layer with SQLite.swift and Swift 2

In an earlier post I showed how to crate a data access layer with SQLite.swift and Swift 1.2.  In this post I will be updating the previous post to use Swift 2 and SQLite.swift for Swift 2.  In both of these posts I describe the data access layer.  In my new book Protocol-Oriented programming I take this example a step further by not only showing the data access layer but also showing how to integrate it with your application using the bridge design pattern.

SQLite is an open source, lightweight and cross platform relational database however it does require good knowledge of SQL to use.  For me that is not much of a problem however it is always better if we can avoid embedding SQL statements in our source code.  This is where frameworks like SQLite.swift come in.  SQLite.swift is a type-safe, Swift language layer over SQLite3 that allows us to access an SQLite database in a pure Swift type interface. 

It is important, when designing an application, to design a good data access layer between our application and its backend data storage.  While some may argue that is what Core Data is for, I am just not a big fan of it especially if we are planning on porting our applications to other platforms.  In this blog post, I will show how we can use the SQLite.swift framework to design a good data access layer for our applications written in Swift 2.

We can download sqlite.swift for Swift 2 from it’s github repository.  There are good instructions on how to install it either though CocoaPods or manually in our applications.  Once we have sqlite.swift setup within our application we will be ready to design our data access layer.

In this blog post, we will mainly be discussing how to design the data access layer.  We will back up our design with a working example but the focus of the post will be on the design and the code is there to reinforce that design.  You can find the code for this post in this github repository.

Data Access Layer Introduction

Our data access layer will consist of three layers.  The bottom most, connection, layer will consist of one class named SQLiteDataStore which will contain the connection handle for our SQLite database.  This class will implement the singleton pattern so all access to the database will go through one connection handle.

The next layer will be a data helper layer that will contain one class for each of our tables.  These data helper classes will contain methods to create, insert, delete and query a specific table.  We will want to create a protocol for the data helper classes to ensure that they contain a minimum set of functionality like creating the table.

The final layer is the model layer.  This layer will contain tuples that model our database tables and will be used to write or retrieve data from the data store.

In our example, we will have two tables.  These are the Teams and the Players table.  From our description above, we can probably guess that this means we will need two data helper classes (TeamDataHelper and PlayerDataHelper) and two tuples (Team and Player).  The data access layer design will look like this:

Connection Layer

Lets begin by defining our errors that we can receive from our data access layer.  We will be using Swift 2’s new error handling feature so our errors are defined in an enum like this:

enum DataAccessError: ErrorType {

    case Datastore_Connection_Error

    case Insert_Error

    case Delete_Error

    case Search_Error

    case Nil_In_Data

}

We will see where we use these errors as we go though our code.

Now lets look at the code.  Next we will look at the SQLiteDataStore class that will contain the connection handle for our database.  The SQLiteDataStore class will contain the following code:

import Foundation

import SQLite

class SQLiteDataStore {

    static let sharedInstance = SQLiteDataStore()

    let BBDB: Connection?

   

    private init() {

       

        var path = "BaseballDB.sqlite"

       

        if let dirs: [NSString] =          NSSearchPathForDirectoriesInDomains(NSSearchPathDirectory.DocumentDirectory,

            NSSearchPathDomainMask.AllDomainsMask, true) as [NSString] {

               

             let dir = dirs[0]

             path = dir.stringByAppendingPathComponent("BaseballDB.sqlite");

        }

       

        do {

            BBDB = try Connection(path)

        } catch _ {

            BBDB = nil

        }

    }

   

    func createTables() throws{

        do {

            try TeamDataHelper.createTable()

            try PlayerDataHelper.createTable()

        } catch {

            throw DataAccessError.Datastore_Connection_Error

        }

    }

}

The SQLiteDataStore class implements the singleton pattern so there is only one instance of the class for the lifecycle of our application.  We implement this pattern by creating a private initializer which is accessed from a static constant named sharedInstance.  Inside the initializer, we set the path to the file that will contain our database and then create the BBDB Database connection with that path.

The SQLiteDataStore class contains a second method named createTables.  The createTables method calls the createTable methods from our data helper classes.  This method will let us create all of the tables from a single method.

Model Layer

Now lets look at the model layer.  The first thing we need to do in the model layer is to create an enumeration that will contain the position that the players could play.  This Positions enum looks like this.

enum Positions: String {

    case Pitcher = "Pitcher"

    case Catcher = "Catcher"

    case FirstBase = "First Base"

    case SecondBase = "Second Base"

    case ThirdBase = "Third Base"

    case Shortstop = "Shortstop"

    case LeftField = "Left Field"

    case CenterField = "Center Field"

    case RightField = "Right field"

    case DesignatedHitter = "Designated Hitter"

}

Next we need to define the types that we use to model our data.  These types will be used to transfer the data from our data access layer to the rest of our code.  Since these types should be used exclusively to transfer the data, using value types are definitely preferred.  The reason why we should avoid using these types in our business logic layer is they may change if we change the table structure of our database and we do not want these changes to effect the code in our business logic layer.  With this in mind we will use tuples to model our data as shown in the following code:

typealias Team = (

    teamId: Int64?,

    city: String?,

    nickName: String?,

    abbreviation: String?

)

typealias Player = (

    playerId: Int64?,

    firstName: String?,

    lastName: String?,

    number: Int?,

    teamId: Int64?,

    position: Positions?

)

Keep in mind that we should avoid using these types in our business logic layer so we have the ability to change them without having to change our business logic code.  We should have an adapter that will convert the data from these types into data structures used by the business logic layer.

Data Helper Layer

Now lets look at the data helper layer.  This layer will be the bridge to our connection layer and will contain one data helper class per table in our database.  We will begin by creating a data helper protocol that will define the set of methods that each data helper class needs to implement.  The DataHelperProtocol protocol looks like this:

protocol DataHelperProtocol {

    typealias T

    static func createTable() throws -> Void

    static func insert(item: T) throws -> Int64

    static func delete(item: T) throws -> Void

    static func findAll() throws -> [T]?

}

Within this protocol the four methods that we are defining are:    

createTable:  Creates the table

insert:  insets a row into the table

delete:  deletes a row from the table

findAll:  returns all rows in the table

Notice that we only define one method to query our data stores.  We do this because the methods to query each individual table could be different therefore the method(s) needed to query the tables could be different.  We would need to evaluate the query methods needed for each table on a table-by-table basis.

Now lets look at the TeamDataHelper class that will confirm to the DataHelperProtocol.  This class will be used to read and write information to the Teams table of the SQLite data store.

import Foundation

import SQLite

class TeamDataHelper: DataHelperProtocol {

    static let TABLE_NAME = "Teams"

   

    static let table = Table(TABLE_NAME)

    static let teamId = Expression<Int64>("teamid")

    static let city = Expression<String>("city")

    static let nickName = Expression<String>("nickname")

    static let abbreviation = Expression<String>("abbreviation")

   

   

    typealias T = Team

   

    static func createTable() throws {

        guard let DB = SQLiteDataStore.sharedInstance.BBDB else {

            throw DataAccessError.Datastore_Connection_Error

        }

        do {

            let _ = try DB.run( table.create(ifNotExists: true) {t in

                t.column(teamId, primaryKey: true)

                t.column(city)

                t.column(nickName)

                t.column(abbreviation)

                })

           

        } catch _ {

            // Error throw if table already exists

        }

       

    }

   

    static func insert(item: T) throws -> Int64 {

        guard let DB = SQLiteDataStore.sharedInstance.BBDB else {

            throw DataAccessError.Datastore_Connection_Error

        }

        if (item.city != nil && item.nickName != nil && item.abbreviation != nil) {

            let insert = table.insert(city <- item.city!, nickName <- item.nickName!, abbreviation <- item.abbreviation!)

            do {

                let rowId = try DB.run(insert)

                guard rowId > 0 else {

                    throw DataAccessError.Insert_Error

                }

                return rowId

            } catch _ {

                throw DataAccessError.Insert_Error

            }

        }

        throw DataAccessError.Nil_In_Data

       

    }

   

    static func delete (item: T) throws -> Void {

        guard let DB = SQLiteDataStore.sharedInstance.BBDB else {

            throw DataAccessError.Datastore_Connection_Error

        }

        if let id = item.teamId {

            let query = table.filter(teamId == id)

            do {

                let tmp = try DB.run(query.delete())

                guard tmp == 1 else {

                    throw DataAccessError.Delete_Error

                }

            } catch _ {

                throw DataAccessError.Delete_Error

            }

        }

    }

   

    static func find(id: Int64) throws -> T? {

        guard let DB = SQLiteDataStore.sharedInstance.BBDB else {

            throw DataAccessError.Datastore_Connection_Error

        }

        let query = table.filter(teamId == id)

        let items = DB.prepare(query)

        for item in  items {

            return Team(teamId: item[teamId] , city: item[city], nickName: item[nickName], abbreviation: item[abbreviation])

        }

       

        return nil

       

    }

   

    static func findAll() throws -> [T]? {

        guard let DB = SQLiteDataStore.sharedInstance.BBDB else {

            throw DataAccessError.Datastore_Connection_Error

        }

        var retArray = [T]()

        let items = DB.prepare(table)

        for item in items {

            retArray.append(Team(teamId: item[teamId], city: item[city], nickName: item[nickName], abbreviation: item[abbreviation]))

        }

       

        return retArray

       

    }

}

We start the TeamDataHelper class by setting the TABLE_NAME property.  This property defines the name for the table within our database.  Next we define the table property which is the SQLite table that contains our data.  We use the table property to access the Teams table with in our database.  The next four properties (teamId, city, nickname and abbreviation) define the name and type of each row in the table. 

Finally we implement each of the four methods defined in the DataHelperProtocol protocol plus one extra method which will search the table by its unique identifier.

The PlayerDataHelper class is similar to the TeamDataHelper class except it is used to read and write to the Players table of the SQLite data store.  The PlayerDataHelper class looks like this:

import Foundation

import SQLite

class PlayerDataHelper: DataHelperProtocol {

    static let TABLE_NAME = "Players"

   

    static let playerId = Expression<Int64>("playerid")

    static let firstName = Expression<String>("firstName")

    static let lastName = Expression<String>("lastName")

    static let number = Expression<Int>("number")

    static let teamId = Expression<Int64>("teamid")

    static let position = Expression<String>("position")

   

   

    static let table = Table(TABLE_NAME)

   

    typealias T = Player

   

    static func createTable() throws {

        guard let DB = SQLiteDataStore.sharedInstance.BBDB else {

            throw DataAccessError.Datastore_Connection_Error

        }

        do {

            _ = try DB.run( table.create(ifNotExists: true) {t in

               

                t.column(playerId, primaryKey: true)

                t.column(firstName)

                t.column(lastName)

                t.column(number)

                t.column(teamId)

                t.column(position)

               

                })

        } catch _ {

            // Error thrown when table exists

        }

    }

   

    static func insert(item: T) throws -> Int64 {

        guard let DB = SQLiteDataStore.sharedInstance.BBDB else {

            throw DataAccessError.Datastore_Connection_Error

        }

        if (item.firstName != nil && item.lastName != nil && item.teamId != nil && item.position != nil) {

            let insert = table.insert(firstName <- item.firstName!, number <- item.number!, lastName <- item.lastName!, teamId <- item.teamId!, position <- item.position!.rawValue)

            do {

                let rowId = try DB.run(insert)

                guard rowId >= 0 else {

                    throw DataAccessError.Insert_Error

                }

                return rowId

            } catch _ {

                throw DataAccessError.Insert_Error

            }

        }

        throw DataAccessError.Nil_In_Data

    }

   

    static func delete (item: T) throws -> Void {

        guard let DB = SQLiteDataStore.sharedInstance.BBDB else {

            throw DataAccessError.Datastore_Connection_Error

        }

        if let id = item.playerId {

            let query = table.filter(playerId == id)

            do {

                let tmp = try DB.run(query.delete())

                guard tmp == 1 else {

                    throw DataAccessError.Delete_Error

                }

            } catch _ {

                throw DataAccessError.Delete_Error

            }

        }

       

    }

   

    static func find(id: Int64) throws -> T? {

        guard let DB = SQLiteDataStore.sharedInstance.BBDB else {

            throw DataAccessError.Datastore_Connection_Error

        }

        let query = table.filter(playerId == id)

        let items = DB.prepare(query)

        for item in  items {

            return Player(playerId: item[playerId], firstName: item[firstName], lastName: item[lastName], number: item[number], teamId: item[teamId], position: Positions(rawValue: item[position]))

        }

       

        return nil

       

    }

   

    static func findAll() throws -> [T]? {

        guard let DB = SQLiteDataStore.sharedInstance.BBDB else {

            throw DataAccessError.Datastore_Connection_Error

        }

        var retArray = [T]()

        let items = DB.prepare(table)

        for item in items {

            retArray.append(Player(playerId: item[playerId], firstName: item[firstName], lastName: item[lastName], number: item[number], teamId: item[teamId], position: Positions(rawValue: item[position])))

        }

       

        return retArray

    }

}

Now lets look at how we would use this data access layer to read and write information from our data store. 

Using the Data Access Layer

To insert a row into a table, all we need to do is to create an instance of the appropriate model type (Player or Team) and then pass it to the appropriate data helper class (PlayerDataHelper or TeamDataHelper).  As an example, here is how we would create four rows in the Teams table:

        do {

       let bosId = try TeamDataHelper.insert(

           Team(

               teamId: 0,

               city: "Boston",

               nickName: "Red Sox",

              abbreviation: "BOS"))

       print(bosId)

   } catch _{}

     

       

   do {

       let torId = try TeamDataHelper.insert(

           Team(

               teamId: 0,

               city: "Toronto",

               nickName: "Blue Jays",

               abbreviation: "TOR"))

       print(torId)

   } catch _ {}

Here is how we would create three rows in the Players table:

let ortizId = try PlayerDataHelper.insert(

    Player(

        playerId: 0,

        firstName: "David",

        lastName: "Ortiz",

        number: 34,

        teamId: bosId,

        position: Positions.DesignatedHitter

 ))

 print(ortizId)

      

 let bogeyId = try PlayerDataHelper.insert(

     Player(

         playerId: 0,

         firstName: "Xander",

         lastName: "Bogarts",

         number: 2,

         teamId: bosId,

         position: Positions.Shortstop

  ))

  print(bogeyId)

To query the database we would call the appropriate method with the data helper class.  The following code would return a list of all teams in the Teams table and prints out the city and nickname for each team.

do {

   if let teams = try TeamDataHelper.findAll() {

      for team in teams {

          print("\(team.city!) \(team.nickName!)")

       }

    }

} catch _ {}

Conclusion

In this blog post we showed how we could use sqlite.swift and good development practices to create a data access layer that simplifies access to the data and hides the complexity of the underlying data store.  The example shown gives us a good abstraction layer that separates the backend data store from our main application. 

The one thing that our example is missing is a way to query the tables by different elements of the table.  For example, we do not have a way to query all players by team or position.  There are a number of ways that we could implement these queries depending on the complexity needed.  We could create a method for each type of query needed if there are not too many or we could create a single method that will query by each element set in the model object.  How you implement these queries is up to you and really depends on the complexity needed in your application.