GDAL, Nan Calculations, and HDFS for loading images

[SteveAIS]

Adding changes to use GDAL for loading TIF images.

Adding changes to allow nan calculations (i.e. nanmean, nanmin, nanmax)

Adding changes to allow reading from HDFS when loading images.

[freeman-lab]

@stevevarner thanks for putting this together! A few top-level comments that will hopefully make it easier to incorporate these changes

• it'd be ideal if you could split this into three separate PRs for the three separate changes (GDAL, HDFS, nan)
• it'd be great to write new tests for each new piece of functionality, you can look at the existing tests as a guide
• there's code in the commit that adds the crop method back in, but it's not mentioned in the PR or used in any of the other changes, maybe this was included by accident?
• the gdal related code seem to have some hardcoded paths, maybe that was included by accident too?

thanks!

[techchrj]

@freeman-lab

In reference to gdal:

you use FileFromMemBuffer() to create a virtual file and open it with the buffer

Here's the C API for it :
http://gdal.org/cpl__vsi_8h.html#1ecb3a78729ecea4d2ce22065a605244

And here's an extract of a test in the Python autotest suite.

content = open('data/byte.tif', mode='rb').read()

# Create in-memory file and initialize it with the content 
gdal.FileFromMemBuffer('/vsimem/tiffinmem', content)

# Open the in-memory file
ds = gdal.Open('/vsimem/tiffinmem')

# do something with the dataset

# Close the dataset
ds = None

# Release memory associated to the in-memory file
gdal.Unlink('/vsimem/tiffinmem')