# Rounding up and Down

My arduino is meisuring the tempature and gives values like: 28.58 Degrees Celcius. I would like to round this up to one decimal. I could make the variable one shorter by cutting the "8" off which would desplay 28.5. Yet this feels a bit like cheating as it mathimatical not correct. Does anyone know how to round up/down. Which would show my tempature as 28.6.

Extra advice needed. I am sending the tempature by Serial to my RPi which has a python code running. Would you suggest doing the rounding in the Arduino code or in The Python code.

• Wouldn't it be better to round to nearest rather than systematically rounding up. – Edgar Bonet Jun 20 '17 at 9:29
• Yea I would love to able to round it the mathematical way. 28.65 <, will become 28.6 and 28.65 ≥, will become 28.7 – Anton van der Wel Jun 20 '17 at 10:36
• Then let `Serial.print()` do the rounding, as per @Jot's answer. – Edgar Bonet Jun 20 '17 at 11:22
• Personally, I would not round until the data is on the Pi. Why lose precision when you don't need to? What if, at some later point, you decide you need to do more calculations on the data on the Pi? Python has a round() function built in. – Johnny Mopp Jun 20 '17 at 12:06

As far as I know, the Serial.print (or println) does rounding up and down. All you have to do is take a float variable and do `Serial.print( value, 1);`

Without extra parameter, the default is choosen, which is 2 decimal digits: `Serial.print( value, 2);`

Why do you want to shorten the bytes over the Serial ? You can just as well send two decimal digits and process the value in Python.

• Thats what the last part of my question is about ;) wondering where to process, in python on my Rpi or before sending. – Anton van der Wel Jun 20 '17 at 10:20
• Any rounding should be done as close to the final output as possible. If your serial port is fast enough then send 2 or 3 digits and round when the Pi is outputting the data. You don't want to base calculations on rounded values if you can help it, that introduces the possibility of errors. That said it's a complete waste of time sending random numbers, if the temperature sensor has an error of +/- 1 degree then the first decimal place may be useful, anything after that is meaningless. – Andrew Jun 20 '17 at 14:15
• Okay agreed! It more precise to process at the final stage, yet the accuracy of the sensor is not there, so I will save bandwidth by using Serial.print(value, 1); – Anton van der Wel Jun 20 '17 at 15:00

You can round in C by multiplying for the significance, adding +0.5, round down (equals as casting to an integer) and divide.

``````float f_rounded = ((int) (f * 10.0 + 0.5) / 10.0);
``````

28.6 will be:

``````float f_rounded = ((int) (28.6 * 10.0 + 0.5) / 10.0)
= ((int) 286.5) / 10.0 = 286 / 10.0 = 28.6
``````

In Python it is equal (except the cast operator is int(...)

Update: for negative numbers +0.5 should be -0.5

Credits for Edgar Bonet: using round works for both positive and negative numbers:

``````float f_rounded = round(f * 10) / 10.0;
``````
• This would round `-28.58` to `-28.5`. – Johnny Mopp Jun 20 '17 at 12:04
• @JohnnyMopp Updated my answers, thanks for the comment – Michel Keijzers Jun 20 '17 at 12:29
• Why not just `float f_rounded = round(f * 10) / 10.0;`? This handles both the positive and negative cases. – Edgar Bonet Jun 20 '17 at 12:40