lat/lon to utm to lat/lon is extremely flawed, how come?

Question

I've tried the following, input: lat/lon data then I'll calculate a box around it by, let's say 50 m, so +/- 50 m on easting/northing value.

Now I reconvert it to lat/lon and with a script:

http://robotics.ai.uiuc.edu/~hyoon24/LatLongUTMconversion.py I get a result that just can't be, lon before is around 7, afterwards around 2.

zone, easting, northing = LLtoUTM(23, location.get_lat(), location.get_lon()) 

topUTM = northing + error
bottomUTM = northing - error
leftUTM = easting - error
rightUTM = easting + error
left, top = UTMtoLL(23, leftUTM, topUTM, zone)

Is the error in my code, or might be the script flawed?

So I've tried to use pyproj, just lat/lon to utm to lat/lon to see what happens

>>> p = pyproj.Proj(proj='utm', zone=32, ellps='WGS84')
>>> p
<pyproj.Proj object at 0x7ff9b8487dd0>
>>> x,y = p(47.9941214, 7.8509671)
>>> print x,y
5159550.36822 1114087.43925
>>> print p(x,y,inverse=True)
(47.971558538495991, 7.8546573140162605)

And here it's not as extremely far off as with the script from above, but it still seems strongly enough incorrect as to not be able to use it. How come? What can I do to get more exact results?

EDIT:

I ran test() and it all tests passed.

in epsg file there is no such thing. The closest I've found was this:

<32632> +proj=utm +zone=32 +ellps=WGS84 +datum=WGS84 +units=m +no_defs <>

no tmerc. Also What would I need to pass the towgs84 as parameters? The ones above?

Answer 1

I've created a small UTM conversion library for Python last week and uploaded it to the Python Package Index: http://pypi.python.org/pypi/utm

I have compared it to using pyproj and it is faster and more accurate. Given your sample data, this is the result:

>>> import utm

>>> u = utm.from_latlon(47.9941214, 7.8509671)
>>> print u
(414278, 5316285, 32, 'T')

>>> print utm.to_latlon(*u)
(47.994157948891505, 7.850963967574302)

UPDATE: Richards answer below describes the real solution for this issue.