I'm having an issue where I needed to use .nice() on one of my axis scales so that the ticks would make use of the end points and spread out properly (instead of leaving the end points unused).
Now that I fixed that, an unexpected problem has arisen. After I called .nice() like so:
var x = d3.time.scale().range([0, width])
x.domain(d3.extent(data, function(d) { return d.date; })).nice();
The axis now ranges up to the year 2018. This could spell trouble, because it looks very pretentious, as its only Feb. 2017. Obviously I don't have a crystal ball, there is no 2018 in my data anywhere, so my only guess is that the function went overboard when it was interpolating.
I love/literally need the tick spacing interpolating from .nice() or something equivalent to that, but at that same time I can't afford to have interpolations out of the domain of my data. Has anyone else had this problem? What can be brought to bear?
You are between a rock and a hard place.
The problem here is that D3 axis generator (specially when using a time scale) was not created having in mind such customizations.
Let's look at your problem: when using nice() you get the end ticks. However, as you said, the end value may exceed the domain, and that's a well known issue.
Here is a demo, the end date is today but the axis goes to 2020:
var svg = d3.select("svg");
var xScale = d3.scaleTime().domain([new Date("1980-01-01"), new Date()]).range([20,480]).nice();
var xAxis = d3.axisBottom(xScale);
var gX = svg.append("g").attr("transform","translate(0,50)").call(xAxis);
<script src="https://d3js.org/d3.v4.min.js"></script>
<svg width="500" height="80"></svg>
A possible solution is using concat with the scale's domain to guarantee that the first and last value in the domain will be the first and last tick, like this:
.tickValues(xScale.ticks(10).concat(xScale.domain()))
And setting the approximate number of ticks using ticks. Here is the demo:
var svg = d3.select("svg");
var xScale = d3.scaleTime().domain([new Date("1980-01-01"), new Date()]).range([20,480]);
var xAxis = d3.axisBottom(xScale).tickFormat(d3.timeFormat("%Y")).tickValues(xScale.ticks(10).concat(xScale.domain()));
var gX = svg.append("g").attr("transform","translate(0,50)").call(xAxis);
<script src="https://d3js.org/d3.v4.min.js"></script>
<svg width="500" height="80"></svg>
However, as you can see, the ticks are not evenly spaced, as you wish.
Thus, an alternative solution for your problem, still using this concat approach, is tweaking the ticks value until we make the ticks appear to be more or less evenly spaced:
var svg = d3.select("svg");
var xScale = d3.scaleTime().domain([new Date("1980-01-01"), new Date()]).range([20,480]);
var xAxis = d3.axisBottom(xScale).tickFormat(d3.timeFormat("%Y")).tickValues(xScale.ticks(5).concat(xScale.domain()));
var gX = svg.append("g").attr("transform","translate(0,50)").call(xAxis);
<script src="https://d3js.org/d3.v4.min.js"></script>
<svg width="500" height="80"></svg>
That being said, this is not a proper solution, because I doubt anyone can make (without a cumbersome hack) the axis starting and ending at the exact domain limits and having all the ticks evenly spaced .
This should be resolved with D3.js v6. The end tick out of domain issue is easily solved when using d3.scaleTime() together with .nice(). However, be aware of d3.scaleUtc(). When using d3.scaleUtc() together with .nice(), the same issue persists - so just switch to d3.scaleTime(). :)
Related
I am trying to create a chart that will have a custom tick range. Having an issue on how to set up the axis though, tried using logscale too but it didn't work properly. Any help is appreciated, attaching a pic for reference.
How I want the axis to be
var y = d3.scaleLinear()
.domain(d3.extent(props.dailyDataAll, function (d) { return d.confirmed }))
.range([height, 0])
I know this really does not answer your question, but I think it may be just a common XY problem.
As such I would suggest instead of struggling to plot ticks properly, you could simply replace Y values with Math.log10(Y) which would make ticks work exactly as you wanted (100 being 2 on the Y scale, 1000 being 3, 10000 being 4 and so on, effectively one tick per order of magnitude just like in your requirements)
I want the date labels to automatically be calculated, appear, and disappear when I change the focus range so that they don't overlap.
I am using a MultiBar graph with a focus chart with the default ordinal scale for nv.models.multiBar(). When I use .ticks(availableWidth / 100 ) on the xAxis, it seems to generate a tick label for EVERY date, or at least a very large number of them:
On nv.models.lineWithFocusChart(), the labels are automatically reduced to fit in a space. This could be because it uses the scale for nv.models.scatter() which is a d3.scale.linear(), but I'm not sure. I tried creating my own scale with the following:
x = d3.scale.ordinal() //as well as x = d3.scale.linear()
.domain(d3.extent(data.map(function(d) {
return d.values.map(function(d,i) {
var X = getX(d,i);
return X.getTime();
});
})))
.range([0, availableWidth]);
I get the following for an ordinal scale:
and no labels for a linear scale. Will this approach work? If so, what am I doing wrong?
On nv.models.multiBarChart(), there is a .reduceXTicks(BOOLEAN) option but this only applies to multiBarChart and there doesn't seem to be an easy way to add it to nv.models.multiBar(). Can I somehow use this?
If there is anything I haven't tried please let me know. I don't want to calculate the labels myself and specify them using .tickValues()
The solution was in fact to use d3.scale.linear() for the x axis. What I tried above didn't work because I was specifying the whole domain of the context chart, when I needed to specify the min and max of the current selection.
In chart(selection) {...}, I put
x = d3.scale.linear()
.range([0, availableWidth]);
and in onBrush(), I put x.domain([new Date(extent[0]), new Date(extent[1])]);, where extent contains the min and max dates of the selection in milliseconds.
I'm having trouble with my X axis width on a D3 scatterplot - what's happening is that my x-axis isn't filling the space allotted, that the Data happily fill. JSfiddle here: http://jsfiddle.net/u4cGJ/
I've defined my d3 scale's output range thusly:
.range([padding, svgWidth - padding]);
and the output range of points on the scatterplot perfectly lines up with this, but the x-axis doesn't - it stops short of the range of points displayed, - it's doing exactly what I tell it to do, and yet, as the svg extends farther out, the data fill in that space too, leaving a section of data that are visible, but not being placed in context of an axis.
Thanks for any help you can offer!
Likely the problem is that your defined domain doesn't include the full extent of your data. The scale domain/range are used to map ranges of data values so that if your domain is [0,10] and your range is [0,100], you will get values like range(0) = 0, range(1) = 10, range(2) = 20... range(10) = 100 (depending on the kind of scale used).
Since it's just mapping domain to range, if you give it a value that it can't map, you aren't likely to get a value inside of your range's extent. Eg. if you do range(11) you won't get a value between 0 and 100. Since the range value is what your axis and plot are likely using to build the actual svg, the points that fall outside of your domain are going to end up off your plot (or NaN or something).
Try deriving the domain from the data itself. D3 has min, max, and extent functions to make this easy for you. If you have an array of point objects, you can use the accessor as a function to provide the right value to use in the calculation.
Here's an example:
var data = [{x:1, y:1}, {x:2, y:4}, {x:3, y:2}, {x:4, y:1}, {x:5, y:2}];
var xExtent = d3.extent(data, function(d) { return d.x; });
var yExtent = d3.extent(data, function(d) { return d.y; });
See the jsFiddle: http://jsfiddle.net/reblace/c9qw8/
Looking at your fiddle, it looks like there's an error in the logic where you're using the wrong domain for what's actually being plotted. If you get rid of the default min and max stuff, it seems to correct the problem you're describing:
function makeScales(xAxisRepresents, domainMinOverride, domainMaxOverride) {
var domainMin = d3.min(dataset, function datumToValueTransformer(datum) { return datum[xAxisRepresents]; });
var domainMax = d3.max(dataset, function datumToValueTransformer(datum) { return datum[xAxisRepresents]; });
I asked a question before about d3, and they suggested me to use an ordinal scale, this would solve my problems. Indeed it solved my problems, but know I'm stuck with another issue...
It draws perfectly, but my X-axis is full of text.
As an example, I want:
1900 1904 1908 1912 ...
but I got:
190119021903190419051906. As you can see this is not clear. (this is just an example, if there were only dates I could use another scale).
Everywhere I looked they talk about axis.ticks(number). But this doesn't work. Nothing happens and I still get the same result.
I hacked a result to get less results on the x-axis:
var str = [];
var i = 0;
while(i < data.length) {
str.push(data[i].age);
i=i+8;
}
x.domain(str);
But if I do this it creates a random line and doesn't draw it perfectly anymore. Don't know how to solve this.. It's a simple line chart, nothing difficult, the only difficulty (for me) is the ordinal scale...
Hope someone can help me out.
this is how my x and x-axis is defined:
var x = d3.scale.ordinal()
.rangeRoundBands([0, width-150],1);
var xAxis = d3.svg.axis()
.scale(x)
.orient("bottom")
without the while-loop (the nasty hack), I just had the following line defining the x.domain:
x.domain(data.map(function(d) { return d.age; }));
Have a look at the documentation for axes, in particular the ticks() function. You can use it (or tickValues()) to control how many (and what) values you want to show.
If you're working with dates, you might want to use a time scale instead of the ordinal one. In particular it will allow you to control the ticks in a more meaningful way, e.g. specify that you want ticks every five years.
I really like this graph and its functionality and it is perfect for what I want/need. The only thing I need to change is I need it to allow ordinal data on the y-axis and I cannot seem to get that to work (I am a beginner).
When I change the y scale from linear to ordinal:
yscale[k] = d3.scale.linear()
.domain(d3.extent(data, function(d) { return +d[k]; }))
.range([h, 0]));
to
yscale[k] = d3.scale.ordinal().rangePoints([h, 0]),
yscale[k].domain(data.map(function(d) { return d[k]; })))
Brushing still shows up and works by itself but it does not filter leaving the selected lines. No lines show up unless I move it to the very top of the axis then, all or mostly all show up. When I stepped through the code with firebug it looked like it was just not getting the lines that were in the brush area but all(?)... and I can't seem to figure out. :(
If anyone could help out with this (especially all the places I have to change and how), I would love to get this working and learn what I am doing wrong :-\
Brushing an ordinal axis returns the pixels, while brushing a quantitative axis returns the domain.
https://github.com/mbostock/d3/wiki/SVG-Controls#wiki-brush_x
The scale is typically defined as a
quantitative scale, in which case the extent is in data space from the
scale's domain; however, it may instead be defined as an ordinal
scale, where the extent is in pixel space from the scale's range
extent.
My guess is that you need to work backwards and translate the pixels to the domain values. I found this question because I'm trying to do the same thing. If I figure it out, I'll let you know.
EDIT: Here's an awesome example to get you started.
http://philau.willbowman.com/2012/digitalInnovation/DevelopmentReferences/LIBS/d3JS/examples/brush/brush-ordinal.html
function brushmove() {
var s = d3.event.target.extent();
symbol.classed("selected", function(d) { return s[0] <= (d = x(d)) && d <= s[1]; });
}
He grabs the selection extent (in pixels), then selects all of the series elements and determines whether they lie within the extent. You can filter elements based on that, and return data keys or what have you to add to your filters.
There is an example of an ordinal scale with brushing here:
http://bl.ocks.org/chrisbrich/4173587
The basic idea is as #gumballhead suggests, you are responsible for projecting the pixel values back onto the input domain. The relevant snippet from the example is:
brushed = function(){var selected = yScale.domain().filter(function(d){return (brush.extent()[0] <= yScale(d)) && (yScale(d) <= brush.extent()[1])});
d3.select(".selected").text(selected.join(","));}