Java double vs BigDecimal for latitude/longitude

Question

When storing latitudes/longitudes which are typically of the format: 44.087585 (i.e. max 2 numbers before the dot and 6dp) do I need to bother with bigdecimals?

Irrelevant for the question, but longitude values range from -180 to 180, so that's 3 numbers before the dot. And a potential minus sign :P — Tom Ladek, Jun 19 '17 at 10:27

score 199 · Accepted Answer · edited Apr 07 '21 at 22:31

Using double has enough precision for accurate lat/lon down to inches for 6-7 decimal places. In aviation, if decimal degrees are used, they typically go to at least 7 decimal places. In our NASA simulations, lat/lon data are doubles while all other attitude and altitude are floats. In other words, the 6th decimal place for altitude isn't significant while the 6th decimal place for lat/lon is for sub-foot accuracy. You should not use float for lat/lon if you need precision down to sub-feet.

You can play around in Google Earth to see what accuracy you get by placing markers and manipulating the last decimal digit.

+1 for the good explanation of the decimal point accuracy with respect to logitudes/latitudes. — CoolBeans, Jul 20 '11 at 18:11

score 41 · Answer 2 · edited Dec 30 '14 at 16:39

41

double should be fine for latitudes/longitudes. BigDecimal becomes more important once you start manipulating currency or need to have more control over precision.

edited Dec 30 '14 at 16:39

bluish

26,356
27
122
180

answered Jul 19 '11 at 22:31

CoolBeans

20,654
10
86
101

12

Downvoter: Why the down vote? Feel free to leave a comment - it will make you look reasonable. – CoolBeans Jul 20 '11 at 03:36

score 7 · Answer 3 · answered Jul 19 '11 at 22:32

7

A double has enough precision for storing that. However, when displaying to users, use a format string like %.6f so that you don't print garbage digits out.

answered Jul 19 '11 at 22:32

C. K. Young

219,335
46
382
435

score 6 · Answer 4 · answered Jan 15 '14 at 19:55

6

The worst accuracy of storing as double is about 3 nm.

If you'd like to calculate it in Java:

Math.ulp((double) 180) * 60 * 1852

3.1582203519064933E-9

As a bonus; the worst accuracy you get from storing in floats is 1.7 meters.

answered Jan 15 '14 at 19:55

Gussoh

799
1
6
13

1

I just realized that I also need to make pythagoras with the greatest error in the other direction too. So a little bit worse. :) – Gussoh Mar 19 '14 at 14:52

score 6 · Answer 5 · edited May 23 '17 at 12:26

If you're concerned about precision, you should probably go with BigDecimal for lots of reasons covered in this post. That said, double or float should be just fine to store what you need in terms of being able to satisfy the range of values that you need.

At just six decimal places of precision I don't think double or float will trip you up, but if you start aggregating the values in any way, the minor imprecisions can start to add up.

score 3 · Answer 6 · answered Jul 19 '11 at 22:35

3

if you want to control nano-robots in their nano-travels using nano-GPS - go for BigDecimal. otherwise - float/double is enough.

answered Jul 19 '11 at 22:35

keymone

8,006
1
28
33

Java double vs BigDecimal for latitude/longitude

6 Answers6

Linked