Skip to content

Box points hover & select #2094

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 19 commits into from
Oct 19, 2017
Merged

Box points hover & select #2094

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
19 commits
Select commit Hold shift + click to select a range
cfc8725
lint (mostly var\n , -> var\n var blocks)
etpinard Oct 16, 2017
53c446b
revamp Box.calc and Box.plot handling of boxpoints
etpinard Oct 16, 2017
581f3fa
add lasso/select dragmode support to box traces
etpinard Oct 17, 2017
e59f283
flatten box test suite
etpinard Oct 17, 2017
5c58049
add some box hover tests
etpinard Oct 17, 2017
aef61ae
introduce box 'hoveron'
etpinard Oct 17, 2017
d352fa5
add 'text' attribute to box traces
etpinard Oct 17, 2017
6a878b0
more linting in Box.plot
etpinard Oct 17, 2017
0027b84
Merge branch 'master' into boxpoints-hover-select
etpinard Oct 17, 2017
a1ea1e8
fixup box hoveron description
etpinard Oct 18, 2017
6bfcbe0
make 'boxes+points' the box hoveron dflt
etpinard Oct 18, 2017
9ca410a
improve `boxpoints: false` default logic
etpinard Oct 18, 2017
7e44c9c
add `boxmode: 'overlay'` hover test case
etpinard Oct 18, 2017
f228a5e
aj-proof box hover logic
etpinard Oct 18, 2017
b4d8044
hide modebar select2d & lasso2d buttons for boxes w/o 'all' boxpoints
etpinard Oct 18, 2017
9e69900
rm useless 'c' value in scattercarper point event data
etpinard Oct 19, 2017
8901528
report coords in input-data space in pt event data
etpinard Oct 19, 2017
b66628e
introduce Axes.hoverLabelText
etpinard Oct 19, 2017
a3ec75a
add prefix box stats hover labels
etpinard Oct 19, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
238 changes: 136 additions & 102 deletions src/traces/box/calc.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,17 +13,17 @@ var isNumeric = require('fast-isnumeric');
var Lib = require('../../lib');
var Axes = require('../../plots/cartesian/axes');


// outlier definition based on http://www.physics.csbsju.edu/stats/box2.html
module.exports = function calc(gd, trace) {
var xa = Axes.getFromId(gd, trace.xaxis || 'x'),
ya = Axes.getFromId(gd, trace.yaxis || 'y'),
orientation = trace.orientation,
cd = [],
valAxis, valLetter, val, valBinned,
posAxis, posLetter, pos, posDistinct, dPos;

// Set value (val) and position (pos) keys via orientation
var xa = Axes.getFromId(gd, trace.xaxis || 'x');
var ya = Axes.getFromId(gd, trace.yaxis || 'y');
var orientation = trace.orientation;
var cd = [];

var i;
var valAxis, valLetter;
var posAxis, posLetter;

if(orientation === 'h') {
valAxis = xa;
valLetter = 'x';
Expand All @@ -36,112 +36,146 @@ module.exports = function calc(gd, trace) {
posLetter = 'x';
}

val = valAxis.makeCalcdata(trace, valLetter); // get val

// size autorange based on all source points
// position happens afterward when we know all the pos
Axes.expand(valAxis, val, {padded: true});

// In vertical (horizontal) box plots:
// if no x (y) data, use x0 (y0), or name
// so if you want one box
// per trace, set x0 (y0) to the x (y) value or category for this trace
// (or set x (y) to a constant array matching y (x))
function getPos(gd, trace, posLetter, posAxis, val) {
var pos0;
if(posLetter in trace) pos = posAxis.makeCalcdata(trace, posLetter);
else {
if(posLetter + '0' in trace) pos0 = trace[posLetter + '0'];
else if('name' in trace && (
posAxis.type === 'category' ||
(isNumeric(trace.name) &&
['linear', 'log'].indexOf(posAxis.type) !== -1) ||
(Lib.isDateTime(trace.name) &&
posAxis.type === 'date')
)) {
pos0 = trace.name;
}
else pos0 = gd.numboxes;
pos0 = posAxis.d2c(pos0, 0, trace[posLetter + 'calendar']);
pos = val.map(function() { return pos0; });
}
return pos;
}

pos = getPos(gd, trace, posLetter, posAxis, val);
var val = valAxis.makeCalcdata(trace, valLetter);
var pos = getPos(trace, posLetter, posAxis, val, gd.numboxes);

// get distinct positions and min difference
var dv = Lib.distinctVals(pos);
posDistinct = dv.vals;
dPos = dv.minDiff / 2;

function binVal(cd, val, pos, posDistinct, dPos) {
var posDistinctLength = posDistinct.length,
valLength = val.length,
valBinned = [],
bins = [],
i, p, n, v;

// store distinct pos in cd, find bins, init. valBinned
for(i = 0; i < posDistinctLength; ++i) {
p = posDistinct[i];
cd[i] = {pos: p};
bins[i] = p - dPos;
valBinned[i] = [];
}
bins.push(posDistinct[posDistinctLength - 1] + dPos);

// bin the values
for(i = 0; i < valLength; ++i) {
v = val[i];
if(!isNumeric(v)) continue;
n = Lib.findBin(pos[i], bins);
if(n >= 0 && n < valLength) valBinned[n].push(v);
var posDistinct = dv.vals;
var dPos = dv.minDiff / 2;
var posBins = makeBins(posDistinct, dPos);

var vLen = val.length;
var pLen = posDistinct.length;
var ptsPerBin = initNestedArray(pLen);

// bin pts info per position bins
for(i = 0; i < vLen; i++) {
var v = val[i];
if(!isNumeric(v)) continue;

var n = Lib.findBin(pos[i], posBins);
if(n >= 0 && n < pLen) {
var pt = {v: v, i: i};
Copy link
Contributor Author

@etpinard etpinard Oct 17, 2017 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This here was the key to make box point hover and selection somewhat clean.

Box calcdata traces have one item per box to be displayed. Box points data-esque array of objects used to be created during Box.plot, now the structure is setup here. Note that it is important to track the original sample pt index (that i above) to convert hovered are selected calcdata items to event data pointNumber corresponding to indices in the input x/y sample arrays

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

... probably best to review this file as a whole:

https://github.com/plotly/plotly.js/blob/6a878b0eef3e52332bd84eca40d024eec1219264/src/traces/box/calc.js

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Definitely nice and clean - I was curious whether this degraded performance for large data sets, but it looks like very little if at all

function r(n) { var v = 0; for(var i = 0; i < n; i++) v += Math.random(); return v; }
function a(n, m) { var out = new Array(n); for(var i = 0; i < n; i++) out[i] = r(m); return out; }
var y = a(1000000, 10)
// I had to patch timeit to support n = 1
timeit(function() { Plotly.newPlot(gd,[{type: 'box', y: y, jitter: 0.5, hoveron: 'points'}]) }, 1)

On my computer 1 million points (of which ~5k are outliers) takes ~3 sec, and difference between this branch and master is within the noise (~10% or less)

Copy link
Contributor Author

@etpinard etpinard Oct 18, 2017 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

(~10% or less)

Thanks for taking a look at this 🐎

ptsPerBin[n].push(pt);
}

return valBinned;
}

valBinned = binVal(cd, val, pos, posDistinct, dPos);

// sort the bins and calculate the stats
function calculateStats(cd, valBinned) {
var v, l, cdi, i;

for(i = 0; i < valBinned.length; ++i) {
v = valBinned[i].sort(Lib.sorterAsc);
l = v.length;
cdi = cd[i];

cdi.val = v; // put all values into calcdata
cdi.min = v[0];
cdi.max = v[l - 1];
cdi.mean = Lib.mean(v, l);
cdi.sd = Lib.stdev(v, l, cdi.mean);
cdi.q1 = Lib.interp(v, 0.25); // first quartile
cdi.med = Lib.interp(v, 0.5); // median
cdi.q3 = Lib.interp(v, 0.75); // third quartile
// build calcdata trace items, one item per distinct position
for(i = 0; i < pLen; i++) {
if(ptsPerBin[i].length > 0) {
var pts = ptsPerBin[i].sort(sortByVal);
var boxVals = pts.map(extractVal);
var bvLen = boxVals.length;

var cdi = {
pos: posDistinct[i],
pts: pts
};

cdi.min = boxVals[0];
cdi.max = boxVals[bvLen - 1];
cdi.mean = Lib.mean(boxVals, bvLen);
cdi.sd = Lib.stdev(boxVals, bvLen, cdi.mean);

// first quartile
cdi.q1 = Lib.interp(boxVals, 0.25);
// median
cdi.med = Lib.interp(boxVals, 0.5);
// third quartile
cdi.q3 = Lib.interp(boxVals, 0.75);

// lower and upper fences - last point inside
// 1.5 interquartile ranges from quartiles
cdi.lf = Math.min(cdi.q1, v[
Math.min(Lib.findBin(2.5 * cdi.q1 - 1.5 * cdi.q3, v, true) + 1, l - 1)]);
cdi.uf = Math.max(cdi.q3, v[
Math.max(Lib.findBin(2.5 * cdi.q3 - 1.5 * cdi.q1, v), 0)]);
cdi.lf = Math.min(
cdi.q1,
boxVals[Math.min(
Lib.findBin(2.5 * cdi.q1 - 1.5 * cdi.q3, boxVals, true) + 1,
bvLen - 1
)]
);
cdi.uf = Math.max(
cdi.q3,
boxVals[Math.max(
Lib.findBin(2.5 * cdi.q3 - 1.5 * cdi.q1, boxVals),
0
)]
);

// lower and upper outliers - 3 IQR out (don't clip to max/min,
// this is only for discriminating suspected & far outliers)
cdi.lo = 4 * cdi.q1 - 3 * cdi.q3;
cdi.uo = 4 * cdi.q3 - 3 * cdi.q1;

cd.push(cdi);
}
}

calculateStats(cd, valBinned);

// remove empty bins
cd = cd.filter(function(cdi) { return cdi.val && cdi.val.length; });
if(!cd.length) return [{t: {emptybox: true}}];
Axes.expand(valAxis, val, {padded: true});

// add numboxes and dPos to cd
cd[0].t = {boxnum: gd.numboxes, dPos: dPos};
gd.numboxes++;
return cd;
if(cd.length > 0) {
cd[0].t = {
boxnum: gd.numboxes,
dPos: dPos
};
gd.numboxes++;
return cd;
} else {
return [{t: {emptybox: true}}];
}
};

// In vertical (horizontal) box plots:
// if no x (y) data, use x0 (y0), or name
// so if you want one box
// per trace, set x0 (y0) to the x (y) value or category for this trace
// (or set x (y) to a constant array matching y (x))
function getPos(trace, posLetter, posAxis, val, numboxes) {
if(posLetter in trace) {
return posAxis.makeCalcdata(trace, posLetter);
}

var pos0;

if(posLetter + '0' in trace) {
pos0 = trace[posLetter + '0'];
} else if('name' in trace && (
posAxis.type === 'category' || (
isNumeric(trace.name) &&
['linear', 'log'].indexOf(posAxis.type) !== -1
) || (
Lib.isDateTime(trace.name) &&
posAxis.type === 'date'
)
)) {
pos0 = trace.name;
} else {
pos0 = numboxes;
}

var pos0c = posAxis.d2c(pos0, 0, trace[posLetter + 'calendar']);
return val.map(function() { return pos0c; });
}

function makeBins(x, dx) {
var len = x.length;
var bins = new Array(len + 1);

for(var i = 0; i < len; i++) {
bins[i] = x[i] - dx;
}
bins[len] = x[len - 1] + dx;

return bins;
}

function initNestedArray(len) {
var arr = new Array(len);
for(var i = 0; i < len; i++) {
arr[i] = [];
}
return arr;
}

function sortByVal(a, b) { return a.v - b.v; }

function extractVal(o) { return o.v; }
41 changes: 22 additions & 19 deletions src/traces/box/plot.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,8 +132,6 @@ module.exports = function plot(gd, plotinfo, cdbox) {
.attr('class', 'points')
.selectAll('path')
.data(function(d) {
var pts = (trace.boxpoints === 'all') ? d.val :
d.val.filter(function(v) { return (v < d.lf || v > d.uf); }),
// normally use IQR, but if this is 0 or too small, use max-min
typicalSpread = Math.max((d.max - d.min) / 10, d.q3 - d.q1),
minSpread = typicalSpread * 1e-9,
Expand All @@ -147,6 +145,10 @@ module.exports = function plot(gd, plotinfo, cdbox) {
jitterFactor,
newJitter;

var pts = trace.boxpoints === 'all' ?
d.pts :
d.pts.filter(function(pt) { return (pt.v < d.lf || pt.v > d.uf); });

// dynamic jitter
if(trace.jitter) {
if(typicalSpread === 0) {
Expand Down Expand Up @@ -179,31 +181,32 @@ module.exports = function plot(gd, plotinfo, cdbox) {
newJitter = trace.jitter * 2 / maxJitterFactor;
}

return pts.map(function(v, i) {
var posOffset = trace.pointpos,
p;
if(trace.jitter) {
posOffset += newJitter * jitterFactors[i] * (rand() - 0.5);
}
// fills in 'x' and 'y' in calcdata 'pts' item
for(i = 0; i < pts.length; i++) {
var pt = pts[i];
var v = pt.v;

var jitterOffset = trace.jitter ?
bdPos * (newJitter * jitterFactors[i] * (rand() - 0.5)) :
0;

var posPx = d.pos + bPos + bdPos * trace.pointpos + jitterOffset;

if(trace.orientation === 'h') {
p = {
y: d.pos + posOffset * bdPos + bPos,
x: v
};
pt.y = posPx;
pt.x = v;
} else {
p = {
x: d.pos + posOffset * bdPos + bPos,
y: v
};
pt.x = posPx;
pt.y = v;
}

// tag suspected outliers
if(trace.boxpoints === 'suspectedoutliers' && v < d.uo && v > d.lo) {
p.so = true;
pt.so = true;
}
return p;
});
}

return pts;
})
.enter().append('path')
.classed('point', true)
Expand Down