chrisveness
commented
9 years ago Hoist by my own petard! I emphasised in my explanations that a grid reference is a square, not a point, but then went on to treat a 10-digit grid reference as a point, rather than as a 1-metre square.
I've corrected OsGridRef.osGridToLatLon to convert to the centre of the 1m grid square (and added a round-trip check to the test harness). Round-trip conversion of TQ 44359 80653 now works.
Thanks for pointing that out.
I've not been able to replicate any further problems.
Chris
Hi Chris,
Thanks for this library - it's been very useful.
I've come up with some erroneous behaviour and I wanted to check if it's a bug or my mistake.
The original issue came when I noticed that I was converting a National Grid Reference to a WGS84 and back again and came up with the wrong answer. Here's the data:
I used a third party conversion site to get confirmation of the correct data -- it looks like the NGR -> WGS84 conversion is working fine and the WGS84 -> NGR conversion is in error:
I forked your geodesy codebase to add some tests. The behaviour I ended up with isn't consistent with the above but still deserves attention. See line 169 onwards here: https://github.com/davegurnell/geodesy/blob/master/test.js
These tests appear to reveal two bugs:
Can only convert OSGB36 point to OsGrid. It looks like this is because there's nodatumfield in the locationfollowing conversion from WGS84.Can you provide any hint as to what's going on here? Am I using the library as intended?
It's test case 1 I'm primarily concerned about as I'm not familiar enough with the maths of coordinate conversion to help. I'm happy to provide a bug fix for test case 2 if required.
Best regards,
Dave