Javascript Tetris

Mon, Oct 10, 2011

My previous game, snakes, took a while to ship. It was a good experience to learn what it means to finish a game, but in reality, these are just side projects for me to play with some basic game mechanics and learn a little game programming. If I spend that amount of time on every minor project, I will still be doing this in 2035…

… with that in mind, I needed to ensure my next game was a short, single weekend project, so I chose an easy one, and I didn’t spend any time polishing it beyond the primary game mechanic of …


Without the polish, it’s a little ugly, but its fully functional, and I can show you how to implement it yourself…

Implementation Details

Tetris is a pretty easy game to implement, just a simple html file with some inline css/javascript.

The only tricky part is dealing with rotation of the 7 tetrominoes.

You could try to do mathematical rotations at run time, but you will quickly discover that some of the pieces rotate around a central block (j, l, s, t and z), whilst others rotate around a point between blocks (i and o).

You could special case the 2 different behaviors, or you could accept the fact that Tetris is hard-coded to always have 7 pieces with 4 rotations and simply hard code all 28 patterns in advance.

You can avoid other special case code if you assume that all pieces are, conceptually, laid out on a 4x4 grid, where each cell is either occupied or not. Interestingly 4x4 = 16, and ‘occupied or not’ = 1 or 0.

That sounds like we can represent each pattern as a simple 16 bit integer to define exactly how we want each piece to rotate:

var i = { blocks: [0x0F00, 0x2222, 0x00F0, 0x4444], color: 'cyan'   };
var j = { blocks: [0x44C0, 0x8E00, 0x6440, 0x0E20], color: 'blue'   };
var l = { blocks: [0x4460, 0x0E80, 0xC440, 0x2E00], color: 'orange' };
var o = { blocks: [0xCC00, 0xCC00, 0xCC00, 0xCC00], color: 'yellow' };
var s = { blocks: [0x06C0, 0x8C40, 0x6C00, 0x4620], color: 'green'  };
var t = { blocks: [0x0E40, 0x4C40, 0x4E00, 0x4640], color: 'purple' };
var z = { blocks: [0x0C60, 0x4C80, 0xC600, 0x2640], color: 'red'    };

We can then provide a helper method that given:

… will iterate over all of the cells in the tetris grid that the piece will occupy:

function eachblock(type, x, y, dir, fn) {
  var bit, result, row = 0, col = 0, blocks = type.blocks[dir];
  for(bit = 0x8000 ; bit > 0 ; bit = bit >> 1) {
    if (blocks & bit) {
      fn(x + col, y + row);
    if (++col === 4) {
      col = 0;

Valid Piece Positioning

We need to be careful about our bounds checking when sliding a piece left and right, or dropping it down a row. We can build on our eachblock helper to provide an occupied method that returns true if any of the blocks required to place a piece at a position on the tetris grid, with a particular rotation direction, would be occupied or out of bounds:

function occupied(type, x, y, dir) {
  var result = false
  eachblock(type, x, y, dir, function(x, y) {
    if ((x < 0) || (x >= nx) || (y < 0) || (y >= ny) || getBlock(x,y))
      result = true;
  return result;

function unoccupied(type, x, y, dir) {
  return !occupied(type, x, y, dir);

NOTE: assume getBlock returns true or false to indicate if that position on the tetris grid is occupied.

Randomizing the Next Piece

Chosing the next random piece is an interesting puzzle. If we chose in a purely random fashion, e.g:

var pieces = [i,j,l,o,s,t,z];
var next = pieces[Math.round(Math.random(0, pieces.length-1))];

… we find that the game is quite frustrating since we often get long sequences of the same type, and sometimes dont get a piece we want for quite some time.

It seems the standard way for picking the next piece in Tetris is to imagine a bag with 4 instances of each piece, we randomly pull one item from the bag until it is empty, then rinse and repeat.

This ensures that each piece will show up at least 4 times in every 28 pieces, it also ensures that the same piece will only repeat in a chain, at most, 4 times… well, technically it can be up to 8 because we could get a chain of 4 at the end of one bag followed by a chain of 4 at the start of the next bag, but the chances of that are quite remote.

This makes for a much more playable game, so we implement our randomPiece method something like this:

var pieces = [];
function randomPiece() {
  if (pieces.length == 0)
    pieces = [i,i,i,i,j,j,j,j,l,l,l,l,o,o,o,o,s,s,s,s,t,t,t,t,z,z,z,z];
  var type = pieces.splice(random(0, pieces.length-1), 1)[0]; // remove a single piece
  return { type: type, dir: DIR.UP, x: 2, y: 0 };

Once we have our data structure and helper methods in place, the remainder of the game becomes fairly straight forward.

Game Constants and Variables

We declare some constants that never change:

var KEY     = { ESC: 27, SPACE: 32, LEFT: 37, UP: 38, RIGHT: 39, DOWN: 40 },
    DIR     = { UP: 0, RIGHT: 1, DOWN: 2, LEFT: 3, MIN: 0, MAX: 3 },
    stats   = new Stats(),
    canvas  = get('canvas'),
    ctx     = canvas.getContext('2d'),
    ucanvas = get('upcoming'),
    uctx    = ucanvas.getContext('2d'),
    speed   = { start: 0.6, decrement: 0.005, min: 0.1 }, // seconds until current piece drops 1 row
    nx      = 10, // width of tetris court (in blocks)
    ny      = 20, // height of tetris court (in blocks)
    nu      = 5;  // width/height of upcoming preview (in blocks)

and some variables that will change, possibly reset() for every game:

var dx, dy,        // pixel size of a single tetris block
    blocks,        // 2 dimensional array (nx*ny) representing tetris court - either empty block or occupied by a 'piece'
    actions,       // queue of user actions (inputs)
    playing,       // true|false - game is in progress
    dt,            // time since starting this game
    current,       // the current piece
    next,          // the next piece
    score,         // the current score
    rows,          // number of completed rows in the current game
    step;          // how long before current piece drops by 1 row

In a more complex game these should all be encapsulated within one or more classes, but to keep Tetris simple we are using simple global variables. But that doesn’t mean we want them modified all over the code, so it still makes sense to write getter and setter methods for most of the variable game state.

function setScore(n)            { score = n; invalidateScore(); };
function addScore(n)            { score = score + n; };
function setRows(n)             { rows = n; step = Math.max(speed.min, speed.start - (speed.decrement*rows)); invalidateRows(); };
function addRows(n)             { setRows(rows + n); };
function getBlock(x,y)          { return (blocks && blocks[x] ? blocks[x][y] : null); };
function setBlock(x,y,type)     { blocks[x] = blocks[x] || []; blocks[x][y] = type; invalidate(); };
function setCurrentPiece(piece) { current = piece || randomPiece(); invalidate();     };
function setNextPiece(piece)    { next    = piece || randomPiece(); invalidateNext(); };

This also allows us to have a controlled way to know when a value has changed, so that we can invalidate the UI and know that section needs re-rendering. This will allow us to optimize our rendering and only draw things when they change.

The Game Loop

The core game loop is a simplified version of the same loop from pong, breakout and snakes. Using requestAnimationFrame (or a polyfill) we simply need to update our game state based on the time interval and then draw the result:

var last = now = timestamp();
function frame() {
  now = timestamp();
  update((now - last) / 1000.0);
  last = now;
  requestAnimationFrame(frame, canvas);
frame(); // start the first frame

Handling Keyboard Input

Our keyboard handler is very simple, we dont actually do anything on a keypress (except for starting/abandoning the game). Instead we simply record the action in a queue to be handled during our update process.

function keydown(ev) {
  if (playing) {
    switch(ev.keyCode) {
      case KEY.LEFT:   actions.push(DIR.LEFT);  break;
      case KEY.RIGHT:  actions.push(DIR.RIGHT); break;
      case KEY.UP:     actions.push(DIR.UP);    break;
      case KEY.DOWN:   actions.push(DIR.DOWN);  break;
      case KEY.ESC:    lose();                  break;
  else if (ev.keyCode == KEY.SPACE) {

Playing the Game

Having defined our data structures, setup our constants and variables, provided our getters and setters, started a game loop and handled keyboard input, we can now look at the logic that implements the Tetris game mechanics:

The update loop consists of handling the next user action, and if the time accumulated is greater than some variable (based on the number of completed rows), then we drop the current piece by 1 row:

function update(idt) {
  if (playing) {
    dt = dt + idt;
    if (dt > step) {
      dt = dt - step;

Handling user input consists of either moving the current piece left or right, rotating it, or dropping it 1 more row:

function handle(action) {
  switch(action) {
    case DIR.LEFT:  move(DIR.LEFT);  break;
    case DIR.RIGHT: move(DIR.RIGHT); break;
    case DIR.UP:    rotate();        break;
    case DIR.DOWN:  drop();          break;

move and rotate simply change the current pieces x, y or dir variable, but they must check to ensure the new position/direction is unoccupied:

function move(dir) {
  var x = current.x, y = current.y;
  switch(dir) {
    case DIR.RIGHT: x = x + 1; break;
    case DIR.LEFT:  x = x - 1; break;
    case DIR.DOWN:  y = y + 1; break;
  if (unoccupied(current.type, x, y, current.dir)) {
    current.x = x;
    current.y = y;
    return true;
  else {
    return false;

function rotate(dir) {
  var newdir = (current.dir == DIR.MAX ? DIR.MIN : current.dir + 1);
  if (unoccupied(current.type, current.x, current.y, newdir)) {
    current.dir = newdir;

The drop method will move the current piece 1 row down, but if thats not possible then the current piece is as low as it can go and it will be broken into individual blocks. At this point we increase the score, check for any completed lines and setup a new piece. If the new piece will also be in an occupied position then the game is over:

function drop() {
  if (!move(DIR.DOWN)) {
    if (occupied(current.type, current.x, current.y, current.dir)) {

function dropPiece() {
  eachblock(current.type, current.x, current.y, current.dir, function(x, y) {
    setBlock(x, y, current.type);


Rendering becomes a fairly straight forward use of the <canvas> API, and a DOM helper method html:

  function html(id, html) { document.getElementById(id).innerHTML = html; }

Since tetris moves in a fairly ‘chunky’ manner, we can recognize that we dont need to redraw everything on every frame at 60fps. We can simply redraw elements only when they change. For this simple implementation we will break down the UI into 4 parts and only rerender those parts when we detect a change in:

This last part, the game court, is quite a broad category. Technically we could track each individual block in the grid and only redraw the ones that have changed, but that would be overkill. Redrawing the entire grid can be done in only a few ms, and ensuring we only do it when a change has occurred means we only take that small hit a couple of times a second.

var invalid = {};

function invalidate()         { invalid.court  = true; }
function invalidateNext()     {   = true; }
function invalidateScore()    { invalid.score  = true; }
function invalidateRows()     { invalid.rows   = true; }

function draw() {;
  ctx.lineWidth = 1;
  ctx.translate(0.5, 0.5); // for crisp 1px black lines

function drawCourt() {
  if (invalid.court) {
    ctx.clearRect(0, 0, canvas.width, canvas.height);
    if (playing)
      drawPiece(ctx, current.type, current.x, current.y, current.dir);
    var x, y, block;
    for(y = 0 ; y < ny ; y++) {
      for (x = 0 ; x < nx ; x++) {
        if (block = getBlock(x,y))
          drawBlock(ctx, x, y, block.color);
    ctx.strokeRect(0, 0, nx*dx - 1, ny*dy - 1); // court boundary
    invalid.court = false;

function drawNext() {
  if ( {
    var padding = (nu - next.type.size) / 2; // half-arsed attempt at centering next piece display;
    uctx.translate(0.5, 0.5);
    uctx.clearRect(0, 0, nu*dx, nu*dy);
    drawPiece(uctx, next.type, padding, padding, next.dir);
    uctx.strokeStyle = 'black';
    uctx.strokeRect(0, 0, nu*dx - 1, nu*dy - 1);
    uctx.restore(); = false;

function drawScore() {
  if (invalid.score) {
    html('score', ("00000" + Math.floor(score)).slice(-5));
    invalid.score = false;

function drawRows() {
  if (invalid.rows) {
    html('rows', rows);
    invalid.rows = false;

function drawPiece(ctx, type, x, y, dir) {
  eachblock(type, x, y, dir, function(x, y) {
    drawBlock(ctx, x, y, type.color);

function drawBlock(ctx, x, y, color) {
  ctx.fillStyle = color;
  ctx.fillRect(x*dx, y*dy, dx, dy);
  ctx.strokeRect(x*dx, y*dy, dx, dy)

Room for Improvement

Like I said, this is just a raw Tetris game mechanic. If you were to polish this game you would want to add things like:

You can play the game now or view the source code.

The game plays best on a desktop browser. Sorry, there is no mobile support.