D3 og Canvas i 3 trinn

Bindingen, trekningen og interaktiviteten

La oss si at du bygger en datavisualisering ved hjelp av D3 og SVG. Du kan treffe et tak når du prøver å vise flere tusen elementer samtidig. Nettleseren din kan begynne å puste under vekten av alle disse DOM-elementene.

Vel her kommer HTML5 Canvas til unnsetning! Det er mye raskere, så det kan løse oppblåsingsproblemene i nettleseren din.

Men du kan fort bli skremt. Fordi D3 og Canvas fungerer litt annerledes enn D3 og SVG - spesielt når det gjelder å tegne og legge til interaktivitet.

Men frykt ikke - det er ikke så komplisert. Enhver erfaring du har hatt med å bygge grafikk med D3 og SVG - eller nærme deg D3 med en annen gjengivelse - vil hjelpe deg enormt.

Denne opplæringen er bygd på skuldrene til giganter som allerede har dekket Canvas godt. Jeg lærte disse tre opplæringene utenat, og jeg anbefaler at du også gjør det:

  • Arbeide med D3.js og Canvas: Når og hvordan fra Irene Ros
  • Nåler, høstakker og Canvas API fra Yannick Assogba
  • Lærdom fra en D3.js-narkoman når man starter med Canvas fra Nadieh Bremer

Så hvorfor fortsette å lese dette, da? Vel, når jeg vil lære noe nytt, hjelper det meg mye å se på det samme emnet fra litt forskjellige vinkler. Og denne opplæringen er en litt annen vinkel.

Denne opplæringen dekker også de tre viktige trinnene: binder data , tegner elementer og legger til interaktivitet - og det gjør alt dette på en gang, med en ekstra trinnvis manual for å sette deg opp.

Hva bygger vi?

Et rutenett med (mange) firkanter. Fargene deres har ingen dyp mening, men ser de ikke pene ut? Den viktige biten er at du kan oppdatere den (for å dekke innbinding og oppdatering av data), at den har mange elementer (opptil 10 000 for at lerret skal betale seg), og at du kan sveve over hver firkant for å vise kvadrat-spesifikk informasjon (interaktivitet). Du kan leke med det her på fullskjerm eller her med all koden

Den mentale modellen

Før vi faktisk dykker inn, la oss raskt gå tilbake og forstå konseptuelt hva vi gjør når vi lager elementer med D3 for å trekke dem til skjermen. Hopp over dette hvis du bare vil lage ting.

Det første trinnet når du bruker D3 innebærer vanligvis ikke tegning - det innebærer å forberede alle elementene du vil tegne. Det er litt som å bygge litt LEGO. Du kan rive opp esken og begynne å bygge noe, eller du kan se på manualen først og bygge den i henhold til tegningen. Manualen er din mentale modell, en tegning eller oppskrift på hva du vil bygge.

Hva er D3s modell? Bortsett fra det store antallet nyttige funksjoner og metoder som beregner posisjoner, omformer datasett (layoutene) og genererer funksjoner som trekker for eksempel baner for oss, har D3 en modell for hvordan elementenes liv skal utvikle seg på skjermen. . Det har en bestemt måte å tenke på livssyklusen til hvert element.

Mindre eterisk injiserer du data i en DOM som ikke eksisterer, og D3 skaper nye elementer etter eget valg i henhold til dataene du injiserer. Vanligvis ett element per datapunkt. Hvis du vil injisere nye data i DOM, kan du gjøre det, og D3 identifiserer hvilke elementer som må opprettes, hvilke elementer som får være og hvilke elementer som skal pakkes sammen og forlate skjermen.

D3 brukes vanligvis i forbindelse med SVG eller noen ganger med HTML-elementer. I dette ortodokse tilfellet kan du se dataene i DOM når du for eksempel velger å se på det gjennom konsollen. Du kan ta tak i den, du kan flytte den opp eller ned i DOM, og du kan - enda viktigere - legge til interaktivitet til hvert element du vil vise, for eksempel en verktøytips.

Men - på baksiden - du kan ikke vise mange elementer. Hvorfor? Fordi jo flere elementer du skyver inn i DOM, jo vanskeligere må nettleseren arbeide for å vise dem alle. La dem også bevege seg, og nettleseren må beregne dem kontinuerlig på nytt. Jo mer knackered nettleseren blir, jo lavere blir bildefrekvensen eller FPS (Frames per Second), som måler hvor mange bilder nettleseren kan male hvert sekund. En bildefrekvens på 60 er god og muliggjør en flytende opplevelse så lenge ingen bilder blir savnet - en bildefrekvens på noe under 30 kan tilsvare en hakkete tur. Så når du vil vise flere elementer, kan du gå tilbake til lerret.

Hvorfor lerret? Canvas er et HTML5-element som kommer med sin egen API for å male på den. Alle elementene som er tegnet på lerretelementet, manifesterer seg ikke i DOM og sparer mye arbeid for nettleseren. De tegnes i umiddelbar modus. Dette betyr at de gjengitte elementene ikke blir lagret i DOM, men instruksjonene dine trekker dem direkte til en bestemt ramme. DOM kjenner bare det ene lerretelementet; alt på det er bare i minnet. Hvis du vil endre lerretelementene dine, må du tegne scenen på nytt for neste ramme.

Problemet med dette er selvfølgelig at du ikke kan kommunisere direkte med disse ikke-materielle elementene som lever i minnet. Du må finne en måte å snakke med dem indirekte. Det er her D3-modellen kommer i tillegg til tilpassede eller 'virtuelle' DOM-elementer. Det du vil gjøre i prinsippet er:

  1. Bind dataene dine til tilpassede DOM-elementer. De bor ikke i DOM, men bare i hukommelsen (i en 'virtuell' DOM) og beskriver livssyklusen til disse elementene på en kjent D3-måte.
  2. Bruk lerret til å tegne disse elementene.
  3. Legg til interaktivitet med en teknikk som kalles 'plukking'.

La oss gjøre det.

Dataen

La oss produsere noen data før vi begynner å kode. La oss si at du vil ha 5000 datapunkter. Så la oss lage en matrise med 5000 elementer, som hver er et objekt med bare en enkelt eiendomsverdi som bærer elementets indeks. Slik lager du det med d3.range(). d3.range()er en D3-funksjon, som lager en matrise basert på argumentet:

var data = [];
d3.range(5000).forEach(function(el) {
 data.push({ value: el }); 
});

Slik ser dataene ut i konsollen

Spenning!

Lerretsbeholderen og dens verktøy

Lerretelementet er et HTML-element. Det er konseptuelt veldig mye som ethvert SVG-foreldre-element, som jeg i det minste vanligvis legger til i en enkel container div som i:

Så la oss legge den til din container med D3 som i ...

var width = 750, height = 400;
var canvas = d3.select('#container') .append('canvas') .attr('width', width) .attr('height', height);
var context = canvas.node().getContext('2d');

Du må også legge til konteksten, som er lerretets verktøykasse. Kontekstvariabelen er fra nå av objektet som bærer alle egenskapene og metodene, penslene og fargene du trenger for å tegne på lerretet. Uten konteksten ville lerretelementet forbli tomt og hvitt. Det er alt du trenger å sette opp - et lerret og dets verktøy ...

HTML

... er enkelt. Den viktigste HTML-strukturen på nettstedet ditt vil være:

Coloured grids

 ...takes numbers between 1 and 10k 

Javascript-strukturen

På et toppnivå trenger du bare to funksjoner:

databind(data) {
 // Bind data to custom elements.
}
draw() {
 // Draw the elements on the canvas.
}

Ganske rett frem så langt.

Bind elementene

For å binde data til elementene oppretter du først et basiselement for alle dine tilpassede elementer du skal produsere og tegne. Hvis du kjenner D3 godt, kan du tenke på det som en erstatning for SVG-elementet:

var customBase = document.createElement('custom');
var custom = d3.select(customBase); // This is your SVG replacement and the parent of all other elements

Deretter legger du til noen innstillinger for rutenettet ditt. Kort sagt, disse innstillingene lar deg tegne et rutenett av firkanter. 100 firkanter bygger en "pakke" og det er en linjeskift etter 10 pakker (eller etter 1000 firkanter). Du kan justere dette for forskjellig "parsellering" av rutene eller annen linjebryting. Eller bare ikke bekymre deg for det. Jeg foreslår sistnevnte ...

// Settings for a grid with 10 cells in a row, // 100 cells in a block and 1000 cells in a row.
var groupSpacing = 4; var cellSpacing = 2; var offsetTop = height / 5; var cellSize = Math.floor((width - 11 * groupSpacing) / 100) - cellSpacing;

La oss nå starte det databindende oppdraget. La oss først få nødvendighetene ut av veien og lage en fargeskala du vil bruke på rutene dine litt senere.

function databind(data) {
// Get a scale for the colours - not essential but nice.
colourScale = d3.scaleSequential(d3.interpolateSpectral) .domain(d3.extent(data, function(d) { return d; }));

La oss nå koble dataene dine til 'erstatnings-SVG' du kalte customovenfor og legge til ikke-eksisterende tilpassede elementer med klassen.rect

var join = custom.selectAll('custom.rect') .data(data);

You enter the custom elements (remember nothing enters the DOM, this is all in memory).

var enterSel = join.enter() .append('custom') .attr('class', 'rect') .attr("x", function(d, i) { var x0 = Math.floor(i / 100) % 10, x1 = Math.floor(i % 10); return groupSpacing * x0 + (cellSpacing + cellSize) * (x1 + x0 * 10); }) .attr("y", function(d, i) { var y0 = Math.floor(i / 1000), y1 = Math.floor(i % 100 / 10); return groupSpacing * y0 + (cellSpacing + cellSize) * (y1 + y0 * 10); }) .attr('width', 0) .attr('height', 0);

When an element enters your model, you just give it an x and a y position as well as a width and a height of 0, which you’ll change in the upcoming update selection…

You merge the enter selection into the update selection and define all attributes for the update and enter selection. This includes a width and a height value as well as a colour from the colour scale you built earlier:

join .merge(enterSel) .transition() .attr('width', cellSize) .attr('height', cellSize) .attr('fillStyle', function(d) { return colourScale(d); });

Two things of note about this last line. When you work with SVG this line would be

.style('color', function(d) { return colourScale(d); })

But with canvas you use .attr(). Why? Your main interest here is to find a pain-free way to transfer some element-specific information. Here you want to transfer a colour-string from the databind() to the draw() function. You use the element simply as a vessel to transport your data over to where it is being rendered to the canvas.

That's a very important distinction: when working with SVG or HTML you can bind data to elements and draw or apply styles to the elements in one step. In canvas you need two steps. First you bind the data then you draw the data. You can't style the elements while binding. They only exist in memory and canvas can't be styled via CSS style properties, which is exactly what you access when using .style().

At first, this might seem limiting as you can do less in one step, but it’s conceptually almost cleaner and also gives you some freedom. .attr() allows us to send any key-value pairs on the journey. You could use other methods like the HTML .dataset property for example, but .attr() will do just fine.

Notice we don't say color but fillStyle. To be honest, you could use color or you could use chooChooTrain here. You would only need to remember this when you fetch the information later during drawing. However, as canvas uses a property called fillStyle to style elements, it seems more appropriate in this case.

Finally, you also define the exit selection, deciding what should happen to exiting elements.

var exitSel = join.exit() .transition() .attr('width', 0) .attr('height', 0) .remove();

That’s it! You can close your databind() function and move on...

} // databind()

This is not really scary coming from D3 as it’s pretty much exactly the same. You have now successfully created your data model, the way the application will think about data. Each element will get the properties it needs to be drawn via the .attr() functions and each element will be assigned a life-cycle state depending on the injected data. Our standard D3 model.

Drawing the elements

Now you need to write the draw function to get the elements on screen. Let’s just note here that nothing has happened yet. You haven’t called databind() yet because you need to find a way to draw it to the canvas first. So here we go... The draw() function doesn't need to take any arguments in this case:

function draw() {

As mentioned fleetingly above, you need to take care of cleaning the canvas every time you draw afresh. The DOM is material, in that when you draw a rect-element on it and you change its x value, it will move in the x-direction and the DOM will take care of this move (or the re-paint) automatically.

If you move a rect from x = 0 to x = 1 at a certain point in time (after a button press for example) the browser will move the rect from 0 to 1 within one tick or frame-paint (which is roughly 16ms long). If you move it from 0 to 10, it will do so in a time depending on the duration you asked this transition to happen, maybe 1 pixel per tick maybe 8 pixel per tick (for more read this blog post).

But it will tell the pixel at 0 that the rect has disappeared and the pixel at 1 that there is a rect now. Canvas doesn’t do this. You need to tell canvas what to paint, and if you paint something new, you need to tell it to remove the previous paint.

So let’s start with cleaning up anything that might be on the canvas before you draw. Here’s how:

context.clearRect(0, 0, width, height); // Clear the canvas.

Simple.

Now you…

  1. …get hold of all elements in order to
  2. loop through all elements and
  3. take the information you have stored in the databind() function to draw the element:
// Draw each individual custom element with their properties.
var elements = custom.selectAll('custom.rect');// Grab all elements you bound data to in the databind() function.
elements.each(function(d,i) { // For each virtual/custom element...
 var node = d3.select(this); // This is each individual element in the loop. context.fillStyle = node.attr('fillStyle'); // Here you retrieve the colour from the individual in-memory node and set the fillStyle for the canvas paint
 context.fillRect(node.attr('x'), node.attr('y'), node.attr('width'), node.attr('height')); // Here you retrieve the position of the node and apply it to the fillRect context function which will fill and paint the square.
}); // Loop through each element.

And that’s it! You can close the draw() function

} // draw()

When I started with canvas after a while of wanting to dive into it, this simplicity really upped my spirits.

However, nothing has happened in the browser yet. We have the tools in the databind() and the draw() function, but nothing has been drawn yet. How do you do this? If you just wanted to draw a static visual or image, you just call:

databind(data);
draw();

This would bind the data to the custom elements, which would live in memory and then draw it — once!

But you have transitions. Remember above: when you wrote the databind() function you transitioned the cell width and height from 0 to their size as well as the colour from black (the default) to the respective element’s colour. A default D3 transition lasts 250 milliseconds, so you need to redraw the squares many times in these 250 ms in order to get a smooth transition. How do you do this?

It’s again simple. You just call databind(data) to create our custom elements before you repeatedly call draw() for as long as it takes the transition to run. So in our case at least 250 ms. You could use setInterval() for this but we really should use requestAnimationFrame() in order to be as performant as possible (for more read this). There are a few ways to use it, but keeping within the D3 spirit, I suggest using d3.timer() which implements requestAnimationFrame() as well as being straight forward to use. So here we go:

// === First call === //
databind(d3.range(value)); // Build the custom elements in memory.
var t = d3.timer(function(elapsed) {
 draw();
 if (elapsed > 300) t.stop();
}); // Timer running the draw function repeatedly for 300 ms.

d3.timer() calls the callback repeatedly until elapsed (which is the passed time in milliseconds from instantiation) is past 300 and then the timer is stopped. In these 300 milliseconds, it runs the draw() at each tick (roughly each 16ms). draw() then looks at each element's attributes and draws them accordingly.

This is how a transition works in canvas. You call the drawing function right after the binding function many times. Whatever your D3-model is set up to transition (positions, colours, sizes) will be re-drawn many times with small incremental changes for each draw

Note that draw() needs to come right after the databind() function. You couldn't ask the machine to run databind(), then do something else for a second and then call draw(). Because after 1 second the transitioned states calculated by your databind() function have all transitioned already. Done, dusted and forgotten.

That’s it! You’ve bound data to custom elements and you’ve drawn it to the canvas.

Let the user update the number of squares

To give the user the chance to repeat this feat with a custom number of elements (ok, semi-custom with a maximum of 10,000) you add the following listener and handler to your text-input box:

// === Listeners/handlers === //
d3.select('#text-input').on('keydown', function() {
if (d3.event.keyCode === 13) { // Only do something if the user hits return (keycode 13).
 if (+this.value  10000) { // If the user goes lower than 1 or higher than 10k... d3.select('#text-explain').classed('alert', true); // ... highlight the note about the range and return.
 return;
 } else { // If the user types in a sensible number...
 d3.select('#text-explain').classed('alert', false); // ...remove potential alert colours from the note...
 value = +this.value; // ...set the value...
 databind(d3.range(value)); // ...and bind the data.
 var t = d3.timer(function(elapsed) {
 draw(); if (elapsed > 300) t.stop();
 }); // Timer running the draw function repeatedly for 300 ms. } // If user hits return.
}); // Text input listener/handler

Here it is again, our colourful grid of canvas squares, ready to be updated and redrawn:

Interactivity

The biggest ‘pain’ with canvas in comparison to SVG or HTML is that there are no material elements living in the DOM. If there were you could just register listeners to the elements and add handlers to the listeners. For example you can trigger a mouse-over on an SVG rect element and whenever the listener triggers, you could do something to the rect. Like showing data values stored with the rect in a tooltip.

With canvas you have to find another way to make an event heard on our canvas elements. Luckily there are a number of clever people who thought of an indirect but logical way.

So what interactivity do we want? As said above let’s go for a tooltip and let’s assume you want to show the index of the square in a tooltip as soon as you hover over the element. Not very thrilling, but the key is that you can access the data bound to the element by hovering over it.

Picking

There are a few steps involved (all logical though). But in short you will build two canvases to achieve this. One main canvas that produces our visual and one hidden canvas (as in we can’t see it) that produces the same visual. The key here is that all elements on the second canvas will be at the exact same position in relation to the canvas origin compared to the first canvas. So square 1 starts on 0,0 on the main canvas as well as on the hidden canvas. Square 2 starts on 8,0 on the main canvas as well as on the hidden canvas and so on.

There is only one important difference. Each element on the hidden canvas will get a unique colour. We will create an object (or rather an associative array or map for brevity) that links each unique colour to each element’s data.

Why? Because next we attach a mouse-move listener to the main-canvas to retrieve a stream of mouse-positions. At each mouse-position we can use a canvas-own method to “pick” the colour at this exact position. Then we just look up the colour in our associative array and we have the data ! And we’re flying…

You could say “well, my squares have already got a unique colour, I can use those?” And indeed, you could use them. However, your interactivity would go out of the window as soon as you decide to bereft your squares from the colours. So you should make sure to always have one canvas — the hidden canvas — that has a guaranteed set of unique colours for the squares.

Let’s apply this technique step by step. The code you’ve built so far can stay as it is — you just add to it as you go along.

1. Prepare the hidden canvas

First let’s create the hidden canvas that will harbour our visual with a unique colour per square.

1.1 Create hidden canvas element and set its CSS to { display: none; }.

// Rename the main canvas and add a 'mainCanvas' class to it.
var mainCanvas = d3.select('#container') .append('canvas') .classed('mainCanvas', true) .attr('width', width) .attr('height', height); // new ----------------------------------- 
// Add the hidden canvas and give it the 'hiddenCanvas' class.
var hiddenCanvas = d3.select('#container') .append('canvas') .classed('hiddenCanvas', true) .attr('width', width) .attr('height', height);

In fact, I won’t set the canvas to hidden in this example to show what is going on. But to do so, just add .hiddenCanvas { display: none; } to your CSS and the deed is done.

1.2 Build the context variable in the draw() function and pass two arguments to the function: the canvas as well as a boolean called 'hidden' determining which canvas we build (hidden = true || false) as in:

function draw(canvas, hidden) {

1.3 You now need to adapt all draw functions to include the two new draw() arguments. So from now on, you don't just call draw() you call either draw(mainCanvas, false) or draw(hiddenCanvas, true)

2. Apply unique colours to the hidden elements and map them

Here, dear reader, comes the key part of our operation, the engine of our truck, the spice in our soup.

2.1 Include a function to generate a new unique colour every time it gets called (via Stack Overflow)

// Function to create new colours for the picking.
var nextCol = 1;
function genColor(){ var ret = [];
 if(nextCol > 8); // G ret.push((nextCol & 0xff0000) >;> 16); // B
 nextCol += 1; }
var col = "rgb(" + ret.join(',') + ")";
return col;
}

genColour() produces a colour defining string in the form rgb(0,0,0). Every time it's called it increments the R value by one. Once it reaches 255, it increments the G value by 1 and resets the R value to 0. Once it reaches r(255,255,0) it increments the B value by 1 resetting the R and the G to 0 and so on.

So in total you can have 256*256*256 = 16.777.216 elements to retain a unique colour. However, I can assure you your browser will die beforehand. Even with canvas (webGL tutorial to follow).

2.2 Create the map-object that will keep track of which custom element has which unique colour:

var colourToNode = {}; // Map to track the colour of nodes.

You can add the genColour() function wherever you want in your script, as long as it's outside the databind() and draw() function scope. But note that your map variable needs to be created before and beyond the scope of the databind() function.

2.3 Add a a unique colour to each custom element as for example .attr('fillStyleHidden') and

2.4 build the map-object during element creation

Here you’ll use your ‘colour-canon’ genColour() in our databind() function when assigning the fillStyle to our elements. As you also have access to each datapoint while it's being bound to each element, you can bring colour and data together in your colourToNode map.

join .merge(enterSel) .transition() .attr('width', cellSize) .attr('height', cellSize) .attr('fillStyle', function(d) { return colorScale(d.value); });
 // new ----------------------------------------------------- .attr('fillStyleHidden', function(d) { 
 if (!d.hiddenCol) {
 d.hiddenCol = genColor(); colourToNode[d.hiddenCol] = d;
 }
 // Here you (1) add a unique colour as property to each element // and(2) map the colour to the node in the colourToNode-map.
 return d.hiddenCol;
});

2.5 You can now colour the elements according to the canvas the draw() function is rendering. You add a conditional on the fillStyle in the draw() function applying the colours for our visual to the main canvas and the unique colours to the hidden canvas. It's a simple one-liner:

context.fillStyle = hidden ? node.attr('fillStyleHidden') : node.attr('fillStyle');
// The node colour depends on the canvas you draw.

The main canvas still looks the same of course:

Let’s finally add some interactivity and start with drawing the hidden canvas whenever we move the mouse onto our main canvas.

3. Pick up the colours with the mouse

3.1 First, simply register a listener to the main canvas, listening to mouse-move events.

d3.select('.mainCanvas').on('mousemove', function() {
});

Why mousemove? As you can’t register listeners with individual squares but have to use the entire canvas you won’t be able to work with mouseover or -out events as they will only trigger when entering the canvas not the elements. In order to get the mouse position on your canvas you can do mousemove or click/mousedown.

d3.select('.mainCanvas').on('mousemove', function() {
 draw(hiddenCanvas, true); // Draw the hidden canvas.
});

This way, the first thing our user triggers when mousing over the main canvas is to unknowingly create the hidden canvas. As said, in production this canvas would be hidden, but for our educational purposes we want to see it and indeed, trigger the hidden canvas to be drawn when the mouse moves over the main canvas like so:

The colours on the main canvas range from black to red, from rgb(0,0,0) to rgb(255,0,0) and then it looks as if the same range from black to red is repeated. However, now the colour ranges from a slightly greener black, precisely from rgb(0,1,0) to rgb(255,1,0):

Zooming into the first couple of hundred squares, here are the colours of the first, the 256th and the 257th square:

3.3 As our hidden canvas is structurally a carbon copy of our main canvas, all the hidden canvas elements will be at the same position as the elements on our main canvas. So, you can now use the mouse’s x and y positions you are collecting from the listener on the main canvas to establish the same location on the hidden canvas. Back in the listener, you add:

d3.select('.mainCanvas').on('mousemove', function()  d3.event.offsetY; );

Note here we take the event.layerX and event.layerY properties which return the mouse position including scrolling. This can break so use offsetX as a fallback (or just use offsetX).

3.4 The picking: Canvas greatly allows access to the pixel-data the mouse is hovering over with the getImageData() function and its .data property. In full bloom this will look like:

getImageData(posX, posY, 1, 1).data .

It will return an array with four numbers: the R, the G, the B and the alpha value. As you diligently built the colourToNode map assigning the element data to each of its hidden colours, you can now access this element's data simply by looking up the colour in the map!

d3.select('.mainCanvas').on('mousemove', function()  d3.event.offsetX; var mouseY = d3.event.layerY );

And indeed, logging the nodeData to the console returns an object every time you hover over a square:

The data per node now shows the value which constitutes the original data as well as the key hiddenCol showing this node's colour for the hidden canvas:

3.5 Finally — and that’s a formality — you add the tooltip

d3.select('.mainCanvas').on('mousemove', function() {
 // Draw the hidden canvas. draw(hiddenCanvas, true);
 // Get mouse positions from the main canvas. var mouseX = d3.event.layerX || d3.event.offsetX; var mouseY = d3.event.layerY || d3.event.offsetY;
 // Get the toolbox for the hidden canvas. var hiddenCtx = hiddenCanvas.node().getContext('2d');
 // Pick the colour from the mouse position. var col = hiddenCtx.getImageData(mouseX, mouseY, 1, 1).data;
 // Then stringify the values in a way our map-object can read it. var colKey = 'rgb(' + col[0] + ',' + col[1] + ',' + col[2] + ')';
 // Get the data from our map! var nodeData = colourToNode[colKey]; console.log(nodeData);
 // new -----------------------------------------------
 if (nodeData) { // Show the tooltip only when there is nodeData found by the mouse
 d3.select('#tooltip') .style('opacity', 0.8) .style('top', d3.event.pageY + 5 + 'px') .style('left', d3.event.pageX + 5 + 'px') .html(nodeData.value); 
 } else { // Hide the tooltip when the mouse doesn't find nodeData. d3.select('#tooltip').style('opacity', 0); }
}); // canvas listener/handler

Det er det! Du har visualisert et stort antall elementer på lerret - mer enn du ville ha vært i stand til å nyte problemfri med SVG. Du brukte fortsatt D3s livssyklusmodell, og du la til litt interaktivitet for å få tilgang til dataene som er knyttet til hvert element. Disse tre trinnene skal gjøre det mulig for deg å gjøre stort sett hva som helst eller i det minste mer enn det du er vant til når du jobber med D3 og SVG.

Det er en trinnvis håndbok fra bunnen av til interaktiv D3 / lerret på bloggen min som tillater interne sidekoblinger. På denne måten kan du se hele prosessen i en visning og enkelt klikke deg gjennom den:

... og her er hele koden igjen.

Jeg håper du likte å lese dette og vennligst si hei og / eller ...

lars verspohl www.datamake.io @lars_vers //www.linkedin.com/in/larsverspohl

... er alltid takknemlig for en lignende? eller en følge han kan komme tilbake.