path: root/API.txt

Flot Reference
--------------

Consider a call to the plot function:

   var plot = $.plot(placeholder, data, options)

The placeholder is a jQuery object or DOM element or jQuery expression
that the plot will be put into. This placeholder needs to have its
width and height set as explained in the README (go read that now if
you haven't, it's short). The plot will modify some properties of the
placeholder so it's recommended you simply pass in a div that you
don't use for anything else. Make sure you check any fancy styling
you apply to the div, e.g. background images have been reported to be a
problem on IE 7.

The format of the data is documented below, as is the available
options. The "plot" object returned has some methods you can call.
These are documented separately below.

Note that in general Flot gives no guarantees if you change any of the
objects you pass in to the plot function or get out of it since
they're not necessarily deep-copied.


Data Format
-----------

The data is an array of data series:

  [ series1, series2, ... ]

A series can either be raw data or an object with properties. The raw
data format is an array of points:

  [ [x1, y1], [x2, y2], ... ]

E.g.

  [ [1, 3], [2, 14.01], [3.5, 3.14] ]

Note that to simplify the internal logic in Flot both the x and y
values must be numbers, even if specifying time series (see below for
how to do this). This is a common problem because you might retrieve
data from the database and serialize them directly to JSON without
noticing the wrong type. If you're getting mysterious errors, double
check that you're inputting numbers and not strings.

If a null is specified as a point or if one of the coordinates is null
or couldn't be converted to a number, the point is ignored when
drawing. As a special case, a null value for lines is interpreted as a
line segment end, i.e. the point before and after the null value are
not connected.

Lines and points take two coordinates. For bars, you can specify a
third coordinate which is the bottom of the bar (defaults to 0).

The format of a single series object is as follows:

  {
    color: color or number
    data: rawdata
    label: string
    lines: specific lines options
    bars: specific bars options
    points: specific points options
    threshold: specific threshold options
    xaxis: 1 or 2
    yaxis: 1 or 2
    clickable: boolean
    hoverable: boolean
    shadowSize: number
  }

You don't have to specify any of them except the data, the rest are
options that will get default values. Typically you'd only specify
label and data, like this:

  {
    label: "y = 3",
    data: [[0, 3], [10, 3]]
  }

The label is used for the legend, if you don't specify one, the series
will not show up in the legend.

If you don't specify color, the series will get a color from the
auto-generated colors. The color is either a CSS color specification
(like "rgb(255, 100, 123)") or an integer that specifies which of
auto-generated colors to select, e.g. 0 will get color no. 0, etc.

The latter is mostly useful if you let the user add and remove series,
in which case you can hard-code the color index to prevent the colors
from jumping around between the series.

The "xaxis" and "yaxis" options specify which axis to use, specify 2
to get the secondary axis (x axis at top or y axis to the right).
E.g., you can use this to make a dual axis plot by specifying
{ yaxis: 2 } for one data series.

"clickable" and "hoverable" can be set to false to disable
interactivity for specific series if interactivity is turned on in
plot, see below.

The rest of the options are all documented below as they are the same
as the default options passed in via the options parameter in the plot
commmand. When you specify them for a specific data series, they will
override the default options for the plot for that data series.

Here's a complete example of a simple data specification:

  [ { label: "Foo", data: [ [10, 1], [17, -14], [30, 5] ] },
    { label: "Bar", data: [ [11, 13], [19, 11], [30, -7] ] } ]


Plot Options
------------

All options are completely optional. They are documented individually
below, to change them you just specify them in an object, e.g.

  var options = {
    lines: { show: true },
    points: { show: true }
  };

  $.plot(placeholder, data, options);


Customizing the legend
======================

  legend: {
    show: boolean
    labelFormatter: null or (fn: string, series object -> string)
    labelBoxBorderColor: color
    noColumns: number
    position: "ne" or "nw" or "se" or "sw"
    margin: number of pixels or [x margin, y margin]
    backgroundColor: null or color
    backgroundOpacity: number between 0 and 1
    container: null or jQuery object/DOM element/jQuery expression
  }

The legend is generated as a table with the data series labels and
small label boxes with the color of the series. If you want to format
the labels in some way, e.g. make them to links, you can pass in a
function for "labelFormatter". Here's an example that makes them
clickable:

  labelFormatter: function(label, series) {
    // series is the series object for the label
    return '<a href="#' + label + '">' + label + '</a>';
  }

"noColumns" is the number of columns to divide the legend table into.
"position" specifies the overall placement of the legend within the
plot (top-right, top-left, etc.) and margin the distance to the plot
edge (this can be either a number or an array of two numbers like [x,
y]). "backgroundColor" and "backgroundOpacity" specifies the
background. The default is a partly transparent auto-detected
background.

If you want the legend to appear somewhere else in the DOM, you can
specify "container" as a jQuery object/expression to put the legend
table into. The "position" and "margin" etc. options will then be
ignored. Note that Flot will overwrite the contents of the container.
Most of the above settings do not apply



Customizing the axes
====================

  xaxis, yaxis, x2axis, y2axis: {
    mode: null or "time"
    min: null or number
    max: null or number
    autoscaleMargin: null or number
    labelWidth: null or number
    labelHeight: null or number

    ticks: null or number or ticks array or (fn: range -> ticks array)
    tickSize: number or array
    minTickSize: number or array
    tickFormatter: (fn: number, object -> string) or string
    tickDecimals: null or number
  }

All axes have the same kind of options. The "mode" option
determines how the data is interpreted, the default of null means as
decimal numbers. Use "time" for time series data, see the next section.

The options "min"/"max" are the precise minimum/maximum value on the
scale. If you don't specify either of them, a value will automatically
be chosen based on the minimum/maximum data values.

The "autoscaleMargin" is a bit esoteric: it's the fraction of margin
that the scaling algorithm will add to avoid that the outermost points
ends up on the grid border. Note that this margin is only applied
when a min or max value is not explicitly set. If a margin is
specified, the plot will furthermore extend the axis end-point to the
nearest whole tick. The default value is "null" for the x axis and
0.02 for the y axis which seems appropriate for most cases.

"labelWidth" and "labelHeight" specifies the maximum size of the tick
labels in pixels. They're useful in case you need to align several
plots.

The rest of the options deal with the ticks.

If you don't specify any ticks, a tick generator algorithm will make
some for you. The algorithm has two passes. It first estimates how
many ticks would be reasonable and uses this number to compute a nice
round tick interval size. Then it generates the ticks.

You can specify how many ticks the algorithm aims for by setting
"ticks" to a number. The algorithm always tries to generate reasonably
round tick values so even if you ask for three ticks, you might get
five if that fits better with the rounding. If you don't want any
ticks at all, set "ticks" to 0 or an empty array.

Another option is to skip the rounding part and directly set the tick
interval size with "tickSize". If you set it to 2, you'll get ticks at
2, 4, 6, etc. Alternatively, you can specify that you just don't want
ticks at a size less than a specific tick size with "minTickSize".
Note that for time series, the format is an array like [2, "month"],
see the next section.

If you want to completely override the tick algorithm, you can specify
an array for "ticks", either like this:

  ticks: [0, 1.2, 2.4]

Or like this (you can mix the two if you like):

  ticks: [[0, "zero"], [1.2, "one mark"], [2.4, "two marks"]]

For extra flexibility you can specify a function as the "ticks"
parameter. The function will be called with an object with the axis
min and max and should return a ticks array. Here's a simplistic tick
generator that spits out intervals of pi, suitable for use on the x
axis for trigonometric functions:

  function piTickGenerator(axis) {
    var res = [], i = Math.floor(axis.min / Math.PI);
    do {
      var v = i * Math.PI;
      res.push([v, i + "\u03c0"]);
      ++i;
    } while (v < axis.max);
    
    return res;
  }


You can control how the ticks look like with "tickDecimals", the
number of decimals to display (default is auto-detected).

Alternatively, for ultimate control over how ticks look like you can
provide a function to "tickFormatter". The function is passed two
parameters, the tick value and an "axis" object with information, and
should return a string. The default formatter looks like this:

  function formatter(val, axis) {
    return val.toFixed(axis.tickDecimals);
  }

The axis object has "min" and "max" with the range of the axis,
"tickDecimals" with the number of decimals to round the value to and
"tickSize" with the size of the interval between ticks as calculated
by the automatic axis scaling algorithm (or specified by you). Here's
an example of a custom formatter:

  function suffixFormatter(val, axis) {
    if (val > 1000000)
      return (val / 1000000).toFixed(axis.tickDecimals) + " MB";
    else if (val > 1000)
      return (val / 1000).toFixed(axis.tickDecimals) + " kB";
    else
      return val.toFixed(axis.tickDecimals) + " B";
  }


Time series data
================

Time series are a bit more difficult than scalar data because
calendars don't follow a simple base 10 system. For many cases, Flot
abstracts most of this away, but it can still be a bit difficult to
get the data into Flot. So we'll first discuss the data format.

The time series support in Flot is based on Javascript timestamps,
i.e. everywhere a time value is expected or handed over, a Javascript
timestamp number is used. This is a number, not a Date object. A
Javascript timestamp is the number of milliseconds since January 1,
1970 00:00:00 UTC. This is almost the same as Unix timestamps, except it's
in milliseconds, so remember to multiply by 1000!

You can see a timestamp like this

  alert((new Date()).getTime())

Normally you want the timestamps to be displayed according to a
certain time zone, usually the time zone in which the data has been
produced. However, Flot always displays timestamps according to UTC.
It has to as the only alternative with core Javascript is to interpret
the timestamps according to the time zone that the visitor is in,
which means that the ticks will shift unpredictably with the time zone
and daylight savings of each visitor.

So given that there's no good support for custom time zones in
Javascript, you'll have to take care of this server-side.

The easiest way to think about it is to pretend that the data
production time zone is UTC, even if it isn't. So if you have a
datapoint at 2002-02-20 08:00, you can generate a timestamp for eight
o'clock UTC even if it really happened eight o'clock UTC+0200.

In PHP you can get an appropriate timestamp with
'strtotime("2002-02-20 UTC") * 1000', in Python with
'calendar.timegm(datetime_object.timetuple()) * 1000', in .NET with
something like:

  public static int GetJavascriptTimestamp(System.DateTime input)
  {
    System.TimeSpan span = new System.TimeSpan(System.DateTime.Parse("1/1/1970").Ticks);
    System.DateTime time = input.Subtract(span);
    return (long)(time.Ticks / 10000);
  }

Javascript also has some support for parsing date strings, so it is
possible to generate the timestamps manually client-side.

If you've already got the real UTC timestamp, it's too late to use the
pretend trick described above. But you can fix up the timestamps by
adding the time zone offset, e.g. for UTC+0200 you would add 2 hours
to the UTC timestamp you got. Then it'll look right on the plot. Most
programming environments have some means of getting the timezone
offset for a specific date (note that you need to get the offset for
each individual timestamp to account for daylight savings).

Once you've gotten the timestamps into the data and specified "time"
as the axis mode, Flot will automatically generate relevant ticks and
format them. As always, you can tweak the ticks via the "ticks" option
- just remember that the values should be timestamps (numbers), not
Date objects.

Tick generation and formatting can also be controlled separately
through the following axis options:

  minTickSize: array
  timeformat: null or format string
  monthNames: null or array of size 12 of strings

Here "timeformat" is a format string to use. You might use it like
this:

  xaxis: {
    mode: "time"
    timeformat: "%y/%m/%d"
  }
  
This will result in tick labels like "2000/12/24". The following
specifiers are supported

  %h': hours
  %H': hours (left-padded with a zero)
  %M': minutes (left-padded with a zero)
  %S': seconds (left-padded with a zero)
  %d': day of month (1-31)
  %m': month (1-12)
  %y': year (four digits)
  %b': month name (customizable)

You can customize the month names with the "monthNames" option. For
instance, for Danish you might specify:

  monthNames: ["jan", "feb", "mar", "apr", "maj", "jun", "jul", "aug", "sep", "okt", "nov", "dec"]

The format string and month names are used by a very simple built-in
format function that takes a date object, a format string (and
optionally an array of month names) and returns the formatted string.
If needed, you can access it as $.plot.formatDate(date, formatstring,
monthNames) or even replace it with another more advanced function
from a date library if you're feeling adventurous.

If everything else fails, you can control the formatting by specifying
a custom tick formatter function as usual. Here's a simple example
which will format December 24 as 24/12:

  tickFormatter: function (val, axis) {
    var d = new Date(val);
    return d.getUTCDate() + "/" + (d.getUTCMonth() + 1);
  }

Note that for the time mode "tickSize" and "minTickSize" are a bit
special in that they are arrays on the form "[value, unit]" where unit
is one of "second", "minute", "hour", "day", "month" and "year". So
you can specify

  minTickSize: [1, "month"]

to get a tick interval size of at least 1 month and correspondingly,
if axis.tickSize is [2, "day"] in the tick formatter, the ticks have
been produced with two days in-between.



Customizing the data series
===========================

  lines, points, bars: {
    show: boolean
    lineWidth: number
    fill: boolean or number
    fillColor: null or color/gradient
  }

  points: {
    radius: number
  }

  bars: {
    barWidth: number
    align: "left" or "center"
    horizontal: boolean
  }

  lines: {
    steps: boolean
  }

  colors: [ color1, color2, ... ]

  shadowSize: number

  threshold: {
    below: number
    color: color
  }

The most important options are "lines", "points" and "bars" that
specifies whether and how lines, points and bars should be shown for
each data series. In case you don't specify anything at all, Flot will
default to showing lines (you can turn this off with
lines: { show: false}). You can specify the various types
independently of each other, and Flot will happily draw each of them
in turn, e.g.

  var options = {
    lines: { show: true, fill: true, fillColor: "rgba(255, 255, 255, 0.8)" },
    points: { show: true, fill: false }
  };

"lineWidth" is the thickness of the line or outline in pixels. You can
set it to 0 to prevent a line or outline from being drawn; this will
also hide the shadow.

"fill" is whether the shape should be filled. For lines, this produces
area graphs. You can use "fillColor" to specify the color of the fill.
If "fillColor" evaluates to false (default for everything except
points), the fill color is auto-set to the color of the data series.
You can adjust the opacity of the fill by setting fill to a number
between 0 (fully transparent) and 1 (fully opaque).

For bars, fillColor can be a gradient, see the gradient documentation
below. "barWidth" is the width of the bars in units of the x axis,
contrary to most other measures that are specified in pixels. For
instance, for time series the unit is milliseconds so 24 * 60 * 60 *
1000 produces bars with the width of a day. "align" specifies whether
a bar should be left-aligned (default) or centered on top of the value
it represents. When "horizontal" is on, the bars are drawn
horizontally, i.e. from the y axis instead of the x axis; note that
the bar end points are still defined in the same way so you'll
probably want to swap the coordinates if you've been plotting vertical
bars first.

For lines, "steps" specifies whether two adjacent data points are
connected with a straight (possibly diagonal) line or with first a
horizontal and then a vertical line.

The "colors" array specifies a default color theme to get colors for
the data series from. You can specify as many colors as you like, like
this:

  colors: ["#d18b2c", "#dba255", "#919733"]

If there are more data series than colors, Flot will try to generate
extra colors by lightening and darkening colors in the theme.

"shadowSize" is the default size of shadows in pixels. Set it to 0 to
remove shadows.

"threshold" specifies that the data points below "below" should be
drawn with the specified color. This makes it easy to mark points
below 0, e.g. for budget data.


Customizing the grid
====================

  grid: {
    color: color
    backgroundColor: color/gradient or null
    tickColor: color
    labelMargin: number
    markings: array of markings or (fn: axes -> array of markings)
    borderWidth: number
    borderColor: color or null
    clickable: boolean
    hoverable: boolean
    autoHighlight: boolean
    mouseActiveRadius: number
  }

The grid is the thing with the axes and a number of ticks. "color" is
the color of the grid itself whereas "backgroundColor" specifies the
background color inside the grid area. The default value of null means
that the background is transparent. You can also set a gradient, see
the gradient documentation below.

"tickColor" is the color of the ticks and "labelMargin" is the spacing
between tick labels and the grid. Note that you can style the tick
labels with CSS, e.g. to change the color. They have class "tickLabel".
"borderWidth" is the width of the border around the plot. Set it to 0
to disable the border. You can also set "borderColor" if you want the
border to have a different color than the grid lines.

"markings" is used to draw simple lines and rectangular areas in the
background of the plot. You can either specify an array of ranges on
the form { xaxis: { from, to }, yaxis: { from, to } } (secondary axis
coordinates with x2axis/y2axis) or with a function that returns such
an array given the axes for the plot in an object as the first
parameter.

You can set the color of markings by specifying "color" in the ranges
object. Here's an example array:

  markings: [ { xaxis: { from: 0, to: 2 }, yaxis: { from: 10, to: 10 }, color: "#bb0000" }, ... ]

If you leave out one of the values, that value is assumed to go to the
border of the plot. So for example if you only specify { xaxis: {
from: 0, to: 2 } } it means an area that extends from the top to the
bottom of the plot in the x range 0-2.

A line is drawn if from and to are the same, e.g.

  markings: [ { yaxis: { from: 1, to: 1 } }, ... ]

would draw a line parallel to the x axis at y = 1. You can control the
line width with "lineWidth" in the ranges objects.

An example function might look like this:

  markings: function (axes) {
    var markings = [];
    for (var x = Math.floor(axes.xaxis.min); x < axes.xaxis.max; x += 2)
      markings.push({ xaxis: { from: x, to: x + 1 } });
    return markings;
  }


If you set "clickable" to true, the plot will listen for click events
on the plot area and fire a "plotclick" event on the placeholder with
a position and a nearby data item object as parameters. The coordinates
are available both in the unit of the axes (not in pixels) and in
global screen coordinates.

Likewise, if you set "hoverable" to true, the plot will listen for
mouse move events on the plot area and fire a "plothover" event with
the same parameters as the "plotclick" event. If "autoHighlight" is
true (the default), nearby data items are highlighted automatically.
If needed, you can disable highlighting and control it yourself with
the highlight/unhighlight plot methods described elsewhere.

You can use "plotclick" and "plothover" events like this:

    $.plot($("#placeholder"), [ d ], { grid: { clickable: true } });

    $("#placeholder").bind("plotclick", function (event, pos, item) {
        alert("You clicked at " + pos.x + ", " + pos.y);
        // secondary axis coordinates if present are in pos.x2, pos.y2,
        // if you need global screen coordinates, they are pos.pageX, pos.pageY

        if (item) {
          highlight(item.series, item.datapoint);
          alert("You clicked a point!");
        }
    });

The item object in this example is either null or a nearby object on the form:

  item: {
      datapoint: the point as you specified it in the data, e.g. [0, 2]
      dataIndex: the index of the point in the data array
      series: the series object
      seriesIndex: the index of the series
      pageX, pageY: the global screen coordinates of the point
  }

For instance, if you have specified the data like this 

    $.plot($("#placeholder"), [ { label: "Foo", data: [[0, 10], [7, 3]] } ], ...);

and the mouse is near the point (7, 3), "datapoint" is the [7, 3] we
specified, "dataIndex" will be 1, "series" is a normalized series
object with among other things the "Foo" label in series.label and the
color in series.color, and "seriesIndex" is 0.

If you use the above events to update some other information and want
to clear out that info in case the mouse goes away, you'll probably
also need to listen to "mouseout" events on the placeholder div.

"mouseActiveRadius" specifies how far the mouse can be from an item
and still activate it. If there are two or more points within this
radius, Flot chooses the closest item. For bars, the top-most bar
(from the latest specified data series) is chosen.

If you want to disable interactivity for a specific data series, you
can set "hoverable" and "clickable" to false in the options for that
series, like this { data: [...], label: "Foo", clickable: false }.


Customizing the selection
=========================

  selection: {
    mode: null or "x" or "y" or "xy",
    color: color
  }

You enable selection support by setting the mode to one of "x", "y" or
"xy". In "x" mode, the user will only be able to specify the x range,
similarly for "y" mode. For "xy", the selection becomes a rectangle
where both ranges can be specified. "color" is color of the selection.

When selection support is enabled, a "plotselected" event will be emitted
on the DOM element you passed into the plot function. The event
handler gets one extra parameter with the ranges selected on the axes,
like this:

  placeholder.bind("plotselected", function(event, ranges) {
    alert("You selected " + ranges.xaxis.from + " to " + ranges.xaxis.to)
    // similar for yaxis, secondary axes are in x2axis
    // and y2axis if present
  });

The "plotselected" event is only fired when the user has finished
making the selection. A "plotselecting" event is fired during the
process with the same parameters as the "plotselected" event, in case
you want to know what's happening while it's happening,

A "plotunselected" event with no arguments is emitted when the user
clicks the mouse to remove the selection.


Customizing the crosshair
=========================

  crosshair: {
    mode: null or "x" or "y" or "xy"
    color: color
  }

You can enable crosshairs, thin lines, that follow the mouse by
setting the mode to one of "x", "y" or "xy". The "x" mode enables a
vertical crosshair that lets you trace the values on the x axis, "y"
enables a horizontal crosshair and "xy" enables them both.


Specifying gradients
====================

A gradient is specified like this:

  { colors: [ color1, color2, ... ] }

For instance, you might specify a background on the grid going from
black to gray like this:

  grid: {
    backgroundColor: { colors: ["#000", "#999"] }
  }

For the series you can specify the gradient as an object that
specifies the scaling of the brightness and the opacity of the series
color, e.g.

  { colors: [{ opacity: 0.8 }, { brightness: 0.6, opacity: 0.8 } ]

where the first color simply has its alpha scaled, whereas the second
is also darkened.
  
Flot currently only supports vertical gradients drawn from top to
bottom because that's what works with IE.


Plot Methods
------------

The Plot object returned from the plot function has some methods you
can call:

  - setSelection(ranges, preventEvent)

    Set the selection rectangle. The passed in ranges is on the same
    form as returned in the "plotselected" event. If the selection
    mode is "x", you should put in either an xaxis (or x2axis) object,
    if the mode is "y" you need to put in an yaxis (or y2axis) object
    and both xaxis/x2axis and yaxis/y2axis if the selection mode is
    "xy", like this:

      setSelection({ xaxis: { from: 0, to: 10 }, yaxis: { from: 40, to: 60 } });

    setSelection will trigger the "plotselected" event when called. If
    you don't want that to happen, e.g. if you're inside a
    "plotselected" handler, pass true as the second parameter.

    
  - clearSelection(preventEvent)

    Clear the selection rectangle. Pass in true to avoid getting a
    "plotunselected" event.


  - setCrosshair(pos)

    Set the position of the crosshair. Note that this is cleared if
    the user moves the mouse. "pos" should be on the form { x: xpos,
    y: ypos } (or x2 and y2 if you're using the secondary axes), which
    is coincidentally the same format as what you get from a "plothover"
    event. If "pos" is null, the crosshair is cleared.


  - clearCrosshair()

    Clear the crosshair.


  - highlight(series, datapoint)

    Highlight a specific datapoint in the data series. You can either
    specify the actual objects, e.g. if you got them from a
    "plotclick" event, or you can specify the indices, e.g.
    highlight(1, 3) to highlight the fourth point in the second series.

  
  - unhighlight(series, datapoint)

    Remove the highlighting of the point, same parameters as highlight.


  - setData(data)

    You can use this to reset the data used. Note that axis scaling,
    ticks, legend etc. will not be recomputed (use setupGrid() to do
    that). You'll probably want to call draw() afterwards.

    You can use this function to speed up redrawing a plot if you know
    that the axes won't change. Put in the new data with
    setData(newdata) and call draw() afterwards, and you're good to
    go.

    
  - setupGrid()

    Recalculate and set axis scaling, ticks, legend etc.

    Note that because of the drawing model of the canvas, this
    function will immediately redraw (actually reinsert in the DOM)
    the labels and the legend, but not the actual tick lines because
    they're drawn on the canvas. You need to call draw() to get the
    canvas redrawn.
    
  - draw()

    Redraws the canvas.
    

There are also some members that let you peek inside the internal
workings of Flot which in some cases is useful. Note that if you change
something in the objects returned, you're changing the objects used by
Flot to keep track of its state, so be careful.

  - getData()

    Returns an array of the data series currently used on normalized
    form with missing settings filled in according to the global
    options. So for instance to find out what color Flot has assigned
    to the data series, you could do this:

      var series = plot.getData();
      for (var i = 0; i < series.length; ++i)
        alert(series[i].color);


  - getAxes()

    Gets an object with the axes settings as { xaxis, yaxis, x2axis,
    y2axis }. Various things are stuffed inside an axis object, e.g.
    you could use getAxes().xaxis.ticks to find out what the ticks are
    for the xaxis.
    

  - getCanvas()

    Returns the canvas used for drawing in case you need to hack on it
    yourself. You'll probably need to get the plot offset too.

  
  - getPlotOffset()

    Gets the offset that the grid has within the canvas as an object
    with distances from the canvas edges as "left", "right", "top",
    "bottom". I.e., if you draw a circle on the canvas with the center
    placed at (left, top), its center will be at the top-most, left
    corner of the grid.