Percentage rounding error? (yet another issue about them)

[maniqui]

I was tinkering with some Bootstrap mixins, in this case, the one for generating a grid system.
I noticed some weird rounding for one value (.col-sm-11-20) on the output when generating a simple 20-columns grid.
I tested both with 1.3.3 and 1.4.0-beta at http://less2css.org/

Is this a bug or anything to be worried about?

Input

//
// Mixins
// --------------------------------------------------

@grid-columns: 20;

// Small grid columns
.generate-grid-columns(@grid-columns) {

  .col-sm-X (@index) when (@index > 0) {
    .col-sm-@{index}-@{grid-columns} { .col-sm-(@index); }
    .col-sm-X((@index - 1));
  }
  .col-sm-X (0) {}

  // Generate columns
  .col-sm-X(@grid-columns);

  // Apply the styles
  .col-sm-(@columns) {
    width: percentage((@columns / @grid-columns));
  }
}

// Generate small grid classes first
.generate-grid-columns(@grid-columns);

Output

.col-sm-20-20 {
  width: 100%;
}
.col-sm-19-20 {
  width: 95%;
}
.col-sm-18-20 {
  width: 90%;
}
.col-sm-17-20 {
  width: 85%;
}
.col-sm-16-20 {
  width: 80%;
}
.col-sm-15-20 {
  width: 75%;
}
.col-sm-14-20 {
  width: 70%;
}
.col-sm-13-20 {
  width: 65%;
}
.col-sm-12-20 {
  width: 60%;
}
.col-sm-11-20 {
  width: 55.00000000000001%;
}
.col-sm-10-20 {
  width: 50%;
}
.col-sm-9-20 {
  width: 45%;
}
.col-sm-8-20 {
  width: 40%;
}
.col-sm-7-20 {
  width: 35%;
}
.col-sm-6-20 {
  width: 30%;
}
.col-sm-5-20 {
  width: 25%;
}
.col-sm-4-20 {
  width: 20%;
}
.col-sm-3-20 {
  width: 15%;
}
.col-sm-2-20 {
  width: 10%;
}
.col-sm-1-20 {
  width: 5%;
}

[lukeapage]

thats strange.. but I assume its just JavaScripts terrible number system. I did think we rounded a certain number of places so we weren't effected by it - that then causes issues for people wanting a million decimal places.

Is this causing an issue or is it just that it looks strange? I wouldn't be too worried by it.

[maniqui]

Haven't yet tested if it causes any issue, particularly when complementing it with one or some columns that add up to the remaining 45%. Will check whenever possible and let you know.

[matthew-dean]

We could do some math to round numbers that are "very close" to their whole number values. That is, cap the number of decimal places.

[matthew-dean]

Stack overflow says on a related issue:

"...due to the way floats are represented, you may not get a clean number that's to two decimal places. For display purposes, always do num.toFixed(2)."

I wonder if that's related. It also provides this link to explain floats, with a deep level of nerdiness: http://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html