Thanks @rnd-debug , looks good at first sight 👍 . I hope to review your PR coming weekend.

One test case is failing on Firefox. Still haven't found the time to check what is happening. Hoping to do it this weekend.

@josdejong failing tests fixed: ready to be merged.
[Edit] codecov reports coverage changes on an unchanged file (median.js).

Your PR looks good @rnd-debug , thanks 👍 . Let's just ignore the codecov warnings, I'm not sure what's going on there but the new functionality is well tested.

A few feedback points:

1. Can you add the new syntax math.floor(x, n) to the main comment in floor.js (used for the docs), like you did for math.ceil(x, n)?
2. I didn't think about it before, but maybe we should also implement rounding with precision for the function fix, then all related functions (round, floor, ceil, fix) have the same behavior. What do you think?

@josdejong fixing the 'fix' looks like a reasonable idea: will give it more time this weekend. Keeping in mind docs + trying to hack codecov by providing more tests.

@josdejong I am confused about how to deal with rounding errors. According to the proposal for floor(x,n), we will get the following:

mathjs.floor(7.999999999999999, 2);                             // 8  But should it be 8?
mathjs.floor(-7.999999999999999, 2);                           //  -8 

When the external library gives us :

roundTo.down(7.999999999999999, 2);                        // 7.99   not cool
roundTo.down(-7.999999999999999, 2);                      // -8  

Similar issues can appear with mathjs.ceil. Should we be compliant to the external library?

@josdejong I am confused about how to deal with rounding errors.

...

Similar issues can appear with mathjs.ceil. Should we be compliant to the external library?

If you calculate ceil(0.1 + 0.2, 1), what whould you expect to get? 0.1 + 0.2 returns 0.30000000000000004, and technically ceil(0.30000000000000004, 1) should return 0.4. But that's not what we expect: we expect the result to be 0.3, right?

To work around these issues with equal, larger, smaller, but also ceil, fix, floor in mathjs, we first apply a rounding to EPSILON (1e-12 by default) to get rid of those round off errors. After that, we can do the actualy thing we want to do: checking equality, or applying ceil/fix/floor. Does that make sense?

@josdejong Ok, now I got the big picture. Then the branch is ready to be merged.
Also provided a couple of tests for round that seemed to raise the coverage.

Thanks for implementing this extension for fix too, and for the extensive unit tests 👍 ! All looks good, I'll fix the typo I saw in the comments of round.

Available now in v7.4.0