Prg2/arcade4 flappy dot

จาก Theory Wiki
ไปยังการนำทาง ไปยังการค้นหา
This is part of the course Programming 2, the material is originally from 01219245/cocos2d-js/Sprites2 from 01219245, 2nd semester 2557.

In this tutorial, we will recreate a clone of a wonderful Flappy Bird. Let's call it Flappy Dot (as our player would look like a dot). We will develop basic game mechanics in this tutorial. We will try to add special effects to the game in the next tutorial.

Task breakdown

Before we start, make sure you know how this game works. You may want to try it for a bit. I guess many of your friends have it on their phones. This is how our game would look like:

219245-dotscr.png

As usual, let's start by thinking about the possible list of increments we would add to an empty project to get this game.

When you get your list, please see the steps that we plan to take here.

  • Show the player on the screen.
  • The player can jump and fall. (Implement player physics)
  • Show a single pillar pair.
  • Move the pillar pair across the screen.
  • Let the pillar pair reappear.
  • Check for player-pillar collision.
  • Make the game with one pillar pair.
  • Show more than one pillar pairs.

The player and its movement

Create a new project and set up a Git repository

We will start with an empty game template. Put the following code in our main program flappy.py.

File: flappy.py
import arcade
 
SCREEN_WIDTH = 800
SCREEN_HEIGHT = 600
 
class FlappyDotWindow(arcade.Window):
    def __init__(self, width, height):
        super().__init__(width, height)
 
        arcade.set_background_color(arcade.color.WHITE)

    def on_draw(self):
        arcade.start_render()
        
 
def main():
    window = FlappyDotWindow(SCREEN_WIDTH, SCREEN_HEIGHT)
    arcade.set_window(window)
    arcade.run()
 
if __name__ == '__main__':
    main()

Try to run the game to see if an empty white window appears. Then, create a git repository at the project directory and commit the code.

Gitmark.png Create your git repository.

Creating the dot model and the sprite

In this step, we shall create a sprite for the player, and show it in the middle of the screen.

Use a graphic editor to create an image for our player. The image should be of size 40 pixels x 40 pixels. Save the image as images/dot.png and try to make it look cute.

We will continue our model/window code structure. So let's create a dot model (called Player) and World in models.py as in our previous projects. Note that currently the Player do nothing in update

File: models.py
class Player:
    def __init__(self, world, x, y):
        self.world = world
        self.x = x
        self.y = y

    def update(self, delta):
        pass

class World:
    def __init__(self, width, height):
        self.width = width
        self.height = height
 
        self.player = Player(self, width // 2, height // 2)
 
     def update(self, delta):
        self.player.update(delta)

As in the previous projects, we then use ModelSprite to display the sprite. Add the class in flappy.py and update FlappyDotWindow accordingly.

File: flappy.py
class ModelSprite(arcade.Sprite):
    def __init__(self, *args, **kwargs):
        self.model = kwargs.pop('model', None)
 
        super().__init__(*args, **kwargs)
 
    def sync_with_model(self):
        if self.model:
            self.set_position(self.model.x, self.model.y)
 
    def draw(self):
        self.sync_with_model()
        super().draw()


class FlappyDotWindow(arcade.Window):
    def __init__(self, width, height):
        super().__init__(width, height)

        arcade.set_background_color(arcade.color.WHITE)

        self.world = World(SCREEN_WIDTH, SCREEN_HEIGHT)
        
        self.dot_sprite = ModelSprite('images/dot.png', model=self.world.player)

    def update(self, delta):
        self.world.update(delta)

    def on_draw(self):
        arcade.start_render()

        self.dot_sprite.draw()

Try to run the game. You should see your sprite in the middle of the screen.

Gitmark.png Commit your work.

Review of physics

You might forget all these, but if you want objects in your game to look and act a bit like real objects, you might have to recall stuffs you learned from mechanics.

Let's look at the basics. An object has a position, its position changes if it has non-zero velocity.

How can you change the player's position? We can set its x and y attribute on the model.

If you want to apply the velocity, you can change the player position based on the velocity.

If there is an acceleration, the object's velocity also changes. The Player currently does not have velocity as its attribute, so we will add it. Now, you can update the velocity based on the acceleration.

These attributes (the position, the velocity, and the acceleration) all have directions. Sometimes, you see negative velocity; this means the object is moving in an opposite direction as the positive direction. We shall follow the standard co-ordinate system for arcade, i.e., for the y-axis, we think of the direction as going upwards.

While in Physics, everything is continuous, but when writing games, we don't really need exact physics, so we can move objects in discrete steps. (In fact, method update is also called with parameter delta, the time period between this call and the last call, and you can use this to make your simulation more smooth.)

So the usual pseudo code for physics is as follows.

pos = pos + velocity;
velocity = velocity + acceleration

Falling dot

To simulate the player falls, we should maintain the player's current velocity, so that we can make it falls as close as the real object.

Let's add this line that initialize property vy in Player's initialization code:

File: models.py
class Player:
    def __init__(self, world, x, y):
        # ... [old code hidden]

        self.vy = 15

You may wonder why we put 15 here. It is just pure guess at this point. However, when you write games, you might want to try various possible values and pick the best one (i.e., the one that make the game fun).

The update method changes the player's position

File: models.py
class Player:
    # ...
    def update(self, delta):
        self.y += self.vy
        self.vy -= 1

Note that we update self.vy at the end of update. The constant -1 is the acceleration. The parameter delta represents delta time; we do not use it for now.

Try to run the program. You should see the player falling.

While our program works, don't just rush to commit right away. Let's try to get rid of the magic numbers first, by defining them explicitly.

Add these contants at the beginning of Player in models.py. These are class variables.

File: models.py
class Player:
    GRAVITY = 1
    STARTING_VELOCITY = 15

    # ...

Then replace 15 and 1 in the code with the appropriate constants as follows. Note that we can refer to these variables as self.GRAVITY or Player.GRAVITY.

File: models.py
class Player:
    # ...
    def __init__(self, world, x, y):
        # ... 

        self.vy = Player.STARTING_VELOCITY

    def update(self, delta):
        # ...
        self.vy -= Player.GRAVITY
Gitmark.png When your program looks good, commit it

Jumping dot

Now, let's make the dot jumps. Let's add method jump that set the velocity to some positive amount. Let's create a constant JUMPING_VELOCITY to represent this magic number as well.

File: models.py
class Player:
    # ...
    JUMPING_VELOCITY = 15

    # ...

    def jump(self):
        self.vy = Player.JUMPING_VELOCITY

To jump, we have to call player.jump() in an appropriate time. We will response to keyboard inputs. We shall follow the style we did in the last tutorial.

First, add method on_key_press in FlappyDotWindow to forward the call to the world.

File: flappy.py
class FlappyDotWindow(arcade.Window):
    # ...

    def on_key_press(self, key, key_modifiers):
         self.world.on_key_press(key, key_modifiers)

Then, in World for any key pressed, call jump.

File: models.py
class World:
    # ...

    def on_key_press(self, key, key_modifiers):
        self.player.jump()

To test this increment, you will have to click on the game canvas, and then quickly hit on any key to get the dot jumping. Try a few times to see how the dot moves. You can adjust the jumping velocity to make the movement nice.

Gitmark.png After a few trials to make sure your code works, please commit.
*********************************************************************************
********************************* CHECK POINT 4.1 *******************************
*********************************************************************************

Game states

It won't be nice to have the user start clicking right away after the game loads. Therefore we shouldn't let the player move before the game actually starts. To do so, we will add a state to the game. How various objects interact depends on the game state. In this tutorial, we shall simply add a property state to World, and perform various events and update methods according to this state. We will learn a cleaner way later on.

For now, the game has 2 states: FROZEN and STARTED. After the game loads, its state is FROZEN. In this state, nothing moves, and after the user hit any keyboard, it changes its state to STARTED with the dot start to jump. The dot falls and jumps as usual in the STARTED state.

Let's add the constants for states in class World:

File: models.py
class World:
    STATE_FROZEN = 1
    STATE_STARTED = 2
    # ...

With this, we can refer to states as World.STATE_FRONT and World.STATE_STARTED.

Initial state

We let attribute state of World keeps the current game state. We initialize the state in __init__:

File: models.py
class World:
    # ...
    
    def __init__(self, width, height):
        # ...

        self.state = World.STATE_FROZEN

State transition

We add methods start and freeze to World so that FlappyDotWindow can tell the world to start working or to get frozen. We also include method is_started for querying the world's state.

File: models.py
class World:
    # ...
    def start(self):
        self.state = World.STATE_STARTED

    def freeze(self):
        self.state = World.STATE_FROZEN     
   
    def is_started(self):
        return self.state == World.STATE_STARTED

Our game changes its state when the user hits the keyboard. We let FlappyDotWindow inform the world to get started.

Rewrite method on_key_press in FlappyDotWindow to do that.

File: flappy.py
    def on_key_press(self, key, key_modifiers):
        if not self.world.is_started():
            self.world.start()
            
        self.world.on_key_press(key, key_modifiers)

The world's state

We have to change how the world behave according to its state. Basically, it should not update anything when it is frozen. Add the following if at the beginning of World.update

class World:
    # ...

    def update(self, delta):
        if self.state == World.STATE_FROZEN:
            return
        
        # ...
Gitmark.png Try to see if the game works as expected. Commit your work after you are done.
*********************************************************************************
********************************* CHECK POINT 4.2 *******************************
*********************************************************************************

The pillar pair

In this section, we will implement a moving pair of opposing pillars. Let's call them a pillar pair. Clearly, we will use one or more sprites to represent it. Before we think about how to implement it, let's think about what we want from this thing: (1) we want to move the pillar pair and (2) we want to check if the dot hits the pillars.

There are basically two approaches for this.

1. Use 1 sprite with transparent background in the middle. 2. Use 2 sprites (one for the top pillar, another for the bottom pillar).

Both approaches are shown below.

Sprites-pillarpair1.png

Question: Can you think of the advantages and disadvantages for using each approach? Expand to see some of the possible advantages and disadvantages.

Using 1 sprite is easier to deal with. However, the space between two opposing pillars are fixed. Using 2 sprites gives flexibility but it might be hard to to deal with two objects. It might be even worse if we have to deal with many pillar pairs.

Here, we will try to get the best out of both approaches. We will use 2 sprites, but we shall combine them into one object.

Sprites-pillarpair2.png

The blue object in the figure above contains two red sprite objects. Specifically, we shall create a PillarPairSprite object that contains two sprites.

Sprite images

Draw two images for the top and bottom pillars. Since we can place each pillar at different height, we should make the images large enough so that the other end of the pillar is still outside the screen. The figure below shows the sprites and the screen; the red border shows the screen boundary.

Clipped-pillars.png

So let's create two images pillar-top.png and pillar-bottom.png, each of width 80 pixels and height 600 pixels. Save them in directory images.

Model class

Let's create class PillarPair in models.py. Put the following code before class World.

File: models.py
class PillarPair:
    def __init__(self, world, x, y):
        self.world = world
        self.x = x
        self.y = y

    def update(self, delta):
        pass

Currently we will have a single pillar pair. So create it in the world's initialization and make sure that we also call PillarPair's update in world's update:

File: models.py
class World:
    # ...

    def __init__(self, width, height):
        # ...

        self.pillar_pair = PillarPair(self, width - 100, height // 2)

    def update(self, delta):
        # ...

        self.pillar_pair.update(delta)

Pillar pair's sprite

We add the PillarPairSprite to draw PillarPair. While we call it a sprite, for simplicity, it is actually not a subclass of arcade.Sprite. This class basically deals with pillar pair drawing based on its model position (similar to ModelSprite). Inside the pillar pair sprite object, we create two sprites for two pillars.

File: flappy.py
class PillarPairSprite():
    def __init__(self, model):
        self.model = model
        
        self.top_pillar_sprite = arcade.Sprite('images/pillar-top.png')
        self.bottom_pillar_sprite = arcade.Sprite('images/pillar-bottom.png')

To draw the pillar pair, we add the following method.

File: flappy.py
class PillarPairSprite():
    # ...

    def draw(self):
        self.top_pillar_sprite.set_position(self.model.x, self.model.y + 400)
        self.top_pillar_sprite.draw()
        
        self.bottom_pillar_sprite.set_position(self.model.x, self.model.y - 400)
        self.bottom_pillar_sprite.draw()

Please pay attention to how we draw the top pillar and the bottom pillar. Their positions are calculated in relative to the model's x and y.

Finally, let's add the sprite to the game and see if it appears. Make sure you draw pillar pair sprite before drawing the player. (Why?)

File: flappy.py
class FlappyDotWindow(arcade.Window):
    def __init__(self, width, height):
        # ...
        self.pillar_pair_sprite = PillarPairSprite(model=self.world.pillar_pair)

    # ...

    def on_draw(self):
        # ...

        self.pillar_pair_sprite.draw()
        self.dot_sprite.draw()              # ... this is old code
Gitmark.png If the pillar pair appears, commit your change.

Question: What is the width of the space between the top pillar and the bottom pillar?

200 pixels.

Moving the pillars

Let's add the code that move the pillars, PillarPair.update. Note that we use another constant PILLAR_SPEED to make our code readable.

File: models.py
class PillarPair:
    PILLAR_SPEED = 5
    
    # ...

    def update(self, delta):
        self.x -= PillarPair.PILLAR_SPEED

Try to see if it moves.

Gitmark.png Don't forget to commit.
*********************************************************************************
********************************* CHECK POINT 4.3 *******************************
*********************************************************************************

Exercise: Reusing the pillars

In this game, our player will have to fly passing a lot of pillar pairs. However, we will not always create a new pillar pair. Instead, we shall reuse the old pillar pair that recently disappear on the left side of the screen.

EXERCISE: modify method PillarPair.update so that right after the pillar pair move outside the screen, it re-enter at the right side of the screen.

Gitmark.png Test and commit the changes.
*********************************************************************************
********************************* CHECK POINT 4.4 *******************************
*********************************************************************************

Exercise: Collision detection

We will check if our dot hits the pillar pair. Our goal is to write method hit in class PillarPair. But WAIT!! Don't start write it for now. There will be more steps before we will do that.

    def hit(self, player):
        # ....

Unit testing

Question: When we write method hit, how can we test if the method is correct?

I'll have to test it. Right now, we have to run the game and try to hit and miss the pillars in as many cases as possible. This is rather hard to do because we need to control the dot as well. This make conventional testing for this logic hard to do.

We will try to use automated testing for this part of the game. The idea is simple: we will try to explicitly call hit in various ways to check if it returns the correct decision.

For simple functions, this is not very hard to do. However, in this case, to call method hit, we have to create the world, the pillar pair and the player. Note that since we have already decouple our model's code from arcade we can actually do that in our unit testing code.

But since this is probably the first time you learn about unit testing, we will try to make the set-up very minimal by extracting the decision function as a stand-alone function called check_player_pillar_collision and test it independently of our other codes. Let's finish method hit first.

File: models.py
class PillarPair:
    # ...

    def hit(self, player):
        return check_player_pillar_collision(player.x, player.y,
                                             self.x, self.y)

We will write function check_player_pillar_collision in a separated module called coldetect, so that we can test it more independently. Before we forget, let's add a code to import this function at the beginning of models.py.

File: models.py
from coldetect import check_player_pillar_collision

Since this part is quite a detour, I'll give more detailed instructions.

Let's create a (broken) function check_player_pillar_collision in file coldetect.py.

File: coldetect.py
def check_player_pillar_collision(player_x, player_y, pillar_x, pillar_y):
    return False

We will use Python's doctest to implement our unit test. Each test and its expected output will be put as a document on the function.

So let's start adding a few test cases. Change the function check_player_pillar_collision to be

File: coldetect.py
def check_player_pillar_collision(player_x, player_y, pillar_x, pillar_y):
    """
    When the dot is very far left of the pillar pair

    >>> check_player_pillar_collision(100, 100, 300, 200)
    False

    When the dot hit the middle of the top pillar

    >>> check_player_pillar_collision( 300, 300, 300, 200 )
    True
    """

    return False

We put each test case as a function call in the docstring of the function. Each line that started with ">>>" (that looks like you are typing in Python's console) is a test case. Its expected output is on the next line. In the above code, we have two test cases.

We can call the doctest framework to test our code by calling:

python -m doctest coldetect.py -v

You will get the following output:

Trying:
    check_player_pillar_collision(100, 100, 300, 200)
Expecting:
    False
ok
Trying:
    check_player_pillar_collision( 300, 300, 300, 200 )
Expecting:
    True
**********************************************************************
File "/xxxxxxxx/flappy/coldetect.py", line 10, in coldetect.check_player_pillar_collision
Failed example:
    check_player_pillar_collision( 300, 300, 300, 200 )
Expected:
    True
Got:
    False
1 items had no tests:
    coldetect
**********************************************************************
1 items had failures:
   1 of   2 in coldetect.check_player_pillar_collision
2 tests in 2 items.
1 passed and 1 failed.
***Test Failed*** 1 failures.

It says that our function is broken in the second test case. You can start fixing the function in coldetect.py to make sure that you get both test passed. That would look like the following.

Trying:
    check_player_pillar_collision(100, 100, 300, 200)
Expecting:
    False
ok
Trying:
    check_player_pillar_collision( 300, 300, 300, 200 )
Expecting:
    True
ok
1 items had no tests:
    coldetect
1 items passed all tests:
   2 tests in coldetect.check_player_pillar_collision
2 tests in 2 items.
2 passed and 0 failed.
Test passed.

NOTES 1: You can remove the option -v when calling doctest to let it show only broken test cases. For example, if we run it with

python -m doctest coldetect.py

You will see

**********************************************************************
File "/home/jittat/prog/arcade/flappy/coldetect.py", line 10, in coldetect.check_player_pillar_collision
Failed example:
    check_player_pillar_collision( 300, 300, 300, 200 )
Expected:
    True
Got:
    False
**********************************************************************
1 items had failures:
   1 of   2 in coldetect.check_player_pillar_collision
***Test Failed*** 1 failures.

You will get an empty response if everything works.

NOTES 2: You don't have to make the collision detection perfect to make the game fun. For example, you don't have to think of the dot as a perfect shape that you have drawn or even a circle, but you can think of it as a simple rectangle which is large enough. Our dot is of size 40 x 40, but we have some curve. Therefore we can view it as a rectangle of size, say, 30 x 30 centered at the position of the player.

Clearly these two test cases are not enough to ensure that your check_player_pillar_collision works correctly. Therefore, after your code for check_player_pillar_collision passes these two test cases, you might want to add more test cases. The figure below shows a few other test cases that you should try adding to the test script.

Test-coldetect-moretests.png

Gitmark.png Don't forget to commit your work at this point.

Combining with the game mechanics

With method PillarPair.hit, we can regularly checks if the player hits the pillars inside our world's code.

File: models.py
class World:
    # ...

    def update(self, delta):
        # ...

        if self.pillar_pair.hit(self.player):
            # ... do something here

What should we do when the player hits a pillar?

First, the game state should change. Let's add another state constant and its related methods:

File: models.py
class World:
    STATE_FROZEN = 1      # .. old code
    STATE_STARTED = 2     # .. old code
    STATE_DEAD = 3
    
    # ...
    
    def die(self):
        self.state = World.STATE_DEAD

    def is_dead(self):
        return self.state == World.STATE_DEAD

With that state, we can change our update function to

File: models.py
class World:
    # ...

    def update(self, delta):
        if self.state in [World.STATE_FROZEN, World.STATE_DEAD]:
            return
        
        self.player.update(delta)
        self.pillar_pair.update(delta)

        if self.pillar_pair.hit(self.player):
            self.die()

Try the game. You might want to try a few hit/miss at the pillars to manually test the collision detection function as well.

Gitmark.png The game should be somewhat playable. Don't forget to commit your work.
*********************************************************************************
********************************* CHECK POINT 4.5 *******************************
*********************************************************************************

Exercise: Random the pillar heights

Let's create the pillar pairs with different heights. Our screen's height is 600 pixels, and we probably do not want the pillar to be too high or too low.

Exerciese: Add this method to PillarPair.

File: models.py
class PillarPair:
    # ...

    def random_position_y(self):
        # ... change the y position of the pillar pair.

Don't forget to call this method when (1) you create a pillar pair, and (2) when the piilar pair wraps around the screen.

Notes: When you test this method, you may want to switch off the collision detection. This can be easily done by adding return False at the top of check_player_pillar_collision function. When you are done, don't forget to remove this hack.

Gitmark.png Commit your work. Don't forget to remove the hack to avoid check_player_pillar_collision.

Exercise: The series of pillar pairs

As we have most objects written nicely, adding more pillars can be done by creating more PillarPair's and spread them out nicely across the screen.

We only outline the changes that you need to make. It is your exercises to get all these steps done.

  • PillarPair creation: Right now we only create one pillar pair. We should create more and keep them in a list. Make sure you spread them out nicely over the x co-ordinate. It would be good if you extract the pillar creation function from the world's initialization code.
  • Collision detection: We can't just call self.pillar_pair.hit( ... ). It would be nice to write a method for checking collision on all pillar pairs.
Gitmark.png For each of these step, don't forget to commit when you get a unit of work done.
*********************************************************************************
********************************* CHECK POINT 4.6 *******************************
*********************************************************************************

Additional Exercises

There are many ways to make the game more fun like a real game. Here we list a few:

  • After the player dies, the game stops. Implement the game restart feature, i.e., when the player hit any key after the player dies, the game should restart.
  • The game shows no score. Add the scoring to the game.
  • Flappy dot is extremely hard because you can hit only once. Add more lives to the player.
  • In our game, the pillar pair can have different spacing. Implement this feature to make the game extremely hard.
  • You can allow the dot to fire a few rockets to destroy the pillars.

Good luck. Have fun writing games!

Gitmark.png Don't forget to regularly commit your work.