Description

violin traces can fail or hang when y values contain high-precision floats that differ in precision.
This causes either:

• Underrun: the internal KDE density array is not fully populated - "Cannot read properties of undefined (reading 'v')"
• Overrun: the density loop runs one extra iteration - "TimeoutError".

Screenshots/Video

Steps to reproduce

minimal reproduction:

Plotly.newPlot("myDiv", {"data":[      {x: [0,0],y: [0.5006312999999999,0.5006313],type:"violin"      }   ]})

This code pen shows the in correct calculations happening for both cases (over run and underrun):

https://codepen.io/Koookadooo/pen/gbPxEVy?editors=0012

I have also attached a file of a python script showing both the "TimeoutError" and the "Cannot read properties of undefined (reading 'v')" using a larger set of high precision variable floating point numbers.

debug_html.py

To run:

• Python debug_html.py
• input 'y' to open files in browser to observe errors

Root Cause

src/traces/violin/calc.js currently generates KDE samples like this:

for(vark=0,t=span[0];t<(span[1]+step/2);k++,t+=step){varv=kde(t);cdi.density[k]={v:v,t:t};maxKDE=Math.max(maxKDE,v);}

When step is extremely small (≈1e-14), repeated floating-point addition in t += step accumulates error:

• In some spans, t exceeds span[1] + step/2 early → underrun (holes or shorter array).
• In others, drift keeps t below the limit too long → overrun (extra sample).

Suggested fix

Instead, compute t fresh each iteration from the integer index k:

- for (var k = 0, t = span[0]; t < (span[1] + step / 2); k++, t += step) {+ for (var k = 0; k < n; k++) {+     var t = span[0] + k * step;      var v = kde(t);      cdi.density[k] = { v: v, t: t };      maxKDE = Math.max(maxKDE, v);}

Notes

Related: supercedesplotly/plotly.js#6490

That issue was closed as stale, but a maintainer noted that a PR would be welcome.

This report includes a confirmed repro and working fix.

I will be making a Pull Request next with the working fix referencing this issue.