Fixes #7228: Round °F temperature to one decimal place

[royreznik]

Fixes: #7228

Rounding the °F temperature to the first decimal place

before:

after:

[candlerb]

Does this actually work? Certain numbers cannot be represented exactly using floating point (0.1 is one example); rounding doesn't change this.

I would expect the problem is at the point of rendering it back to decimal.

[royreznik]

Can you please provide an example for this?

I checked it with 0.1 and 0.3 and still got the expected value

Math.round((0.1 + Number.EPSILON) * 10) / 10; # 0.1
Math.round((0.3 + Number.EPSILON) * 10) / 10; # 0.3

[candlerb]

OK, I think I can replicate via Javascript console in Chrome:

The problem is indeed fixed by your method, but actually it would be simpler to multiply by 9 and then divide by 5, rather than multiply by (9/5) which is an inexact floating point value; it appears that the error in (9/5) is magnified by multiplying by the degrees C value.

[candlerb]

There's no guarantee that doing *9/5 will give the right answer though. I believe the correct solution is to convert the number to a string with the correct number of decimal places, using toFixed:

> (42 * 1.8)
< 75.60000000000001
> (42 * 1.8).toFixed(2)
< '75.60'
> (42 * 1.8).toFixed(1)
< '75.6'

[royreznik]

As you said, There's no guarantee that doing *9/5 will give the right answer though, i wont going to change this

using only toFixed can result a different inacurate like this:

(1.005).toFixed(2) # 1.00

[candlerb]

Yes you're right:

> (1.006).toFixed(2)
< '1.01'
> (1.005).toFixed(2)
< '1.00'
> (1.00501).toFixed(2)
< '1.01'

and yet:

> (1.05).toFixed(1)
< '1.1'

Floating point is broken, and Javascript especially so :-(

However I think toFixed is OK here. I just want never to see a number with spurious 13 places of decimals. I don't mind if it's off by 0.1 degrees F, which it almost never will be anyway.

[royreznik]

I see what you meant. Floating point is evil :-(

also toFixed will do this:

(1).toFixed(1) // "1.0 "
Math.round((1 + Number.EPSILON) * 10) / 10; //  1

this will add .0 for integers.

Anyway, I think my solution fix the bug we had and maybe different approch can be achived in other PR :)