Numpy interface gives misleading results when data spans multiple zones

[jgodwin]

It appears that the numpy interface returns all the data in the lat/long arrays in the same zone regardless of how many zones the data actually spans. For example, here are some lat/longs in Texas that span 6 zones. When you run each data point through utm.from_latlong individually you get the right UTM zones (6) with output as shown below:

  lat        long       northing                        easting          utmzone

31.536598 -103.669795 626277.957127 3.489840e+06 13R
33.044924 -102.473123 721477.858368 3.601393e+06 13S
30.427469 -96.538430 736424.325834 3.368727e+06 14R
33.349272 -100.655458 344773.270375 3.694674e+06 14S
31.276065 -95.483046 263604.667649 3.462858e+06 15R
32.061439 -95.761137 239325.564516 3.550582e+06 15S

To be clear, this result is from doing:

result = []
for lat,long in zip(df['lat'],df['long']):
   result.append(utm.from_latlong(lat,long))

If you instead put those lat/long pairs into numpy arrays and then feed them through utm.from_latlong all at once, they all get projected into a single utm zone, which appears to be the first one in the dataset:

(array([ 626277.95712696, 735961.8832752 , 1314026.08573693,904396.7559105 , 1407848.61383494, 1373735.1761227 ]), array([3489839.70071501, 3659105.6188487 , 3396735.53928308, 3698446.73936288, 3499566.456626 , 3584840.23674779]), 13, 'R')

result = utm.from_latlong(df['lat'].values,df['long'].values)

[sdhiscocks]

I think the current behaviour is correct, as a typical use case would require all points in the same projection.