Directory of Python Binaries

I came across this page today while installing SciPy, which I am using to plot 3D shapes in matplotlib. It is a collection of windows binaries for many useful python modules, and has proven very useful:

Unofficial Windows Binaries for Python Extension Packages

Random Shapefile Data

I often need to quickly generate some random attribute values in a shapefile, either to test code, write an answer on GIS.SE or to mock up some data to test a workflow. There are a few ways to do this simply in pure python using the various random functions, but if all I need to do is populate an attribute table the ArcGIS field calculator will do the job.

The syntax of the command is very un-pythonic, but does allow for a range of different random datasets to be generated. I recently used this command to create some percentage data at work:


Which clearly creates random integers between 0 and 100. After reading through the helpfiles I found a list of all the different random distributions that can be used, which I have reproduced below for my own convenience:

• UNIFORM {Minimum}, {Maximum}—A uniform distribution with a range defined by the {Minimum} and {Maximum}. Both the {Minimum} and {Maximum} are of type double. The default values are 0.0 for {Minimum} and 1.0 for {Maximum}.

• INTEGER {Minimum}, {Maximum}—An integer distribution with a range defined by the {Minimum} and {Maximum}. Both the {Minimum} and {Maximum} are of type long. The default values are 1 for {Minimum} and 10 for {Maximum}.

• NORMAL {Mean}, {Standard Deviation}—A Normal distribution with a defined {Mean} and {Standard Deviation}. Both the {Mean} and {Standard Deviation} are of type double. The default values are 0.0 for {Mean} and 1.0 for {Standard Deviation}.

• EXPONENTIAL {Mean}—An Exponential distribution with a defined {Mean}. The {Mean} is of type double. The default value for {Mean} is 1.0.

• POISSON {Mean}—A Poisson distribution with a defined {Mean}. The {Mean} is of type double. The default value for {Mean} is 1.0.

• GAMMA {Alpha}, {Beta}—A Gamma distribution with a defined {Alpha} and {Beta}. Both the {Alpha} and {Beta} are of type double. The default values are 1.0 for {Alpha} and 1.0 for {Beta}.

• BINOMIAL {N}, {Probability}—A Binomial distribution with a defined {N} and {Probability}. The {N} is of type long, and {Probability} is of type double. The default values are 10 for {N} and 0.5 for {Probability}.

• GEOMETRIC {Probability}—A geometric distribution with a defined {Probability}. The {Probability} is of type double. The default value for {Probability} is 0.5.

• NEGATIVE BINOMIAL {N}, {Probability}—A negative binomial distribution with a defined {N} and {Probability}. The {N} is of type long, and {Probability} is of type double. The default values are 10 for {N} and 0.5 for {Probability}.

There are, of course, many other ways to do this but I find that this weird little function does a pretty good job, once you get to grips with the odd syntax.

GDAL Contour

I wanted to create some contours from a DEM file today and decided to use GDAL instead of Arc or FME and I was again pleasantly surprised by the simplicity and power of the GDAL environment. My first port of call was to read the documentation, which showed that this command's syntax does not differ greatly from
the rasterize command discussed in the last post.

I decided to generate a set of 10m contours as a shapefile using the following command:

The switches do the following:

•  -b 1 selects the band of the image to process, which defaults to 1
•  -a elevation is the name of the contour elevation attribute which will be created
• -snodata -9999 tells GDAL the value of nodata cells in the input raster, so they can be ignored
• ns67ne.tif contour.shp are the input and output files, respectively
• -i 10 is the spacing between each contour

GDAL Rasterize

I wanted to convert a vector outline of the GB coastline into a raster in order to use it as a background layer in MapGuide  Maestro. I would usually do such a process in ArcMap, but I am trying to learn open source alternatives to these types of functions. As a result I have been playing around with GDAL and OGR for a week or so now, and have been very impressed with their power and versatility, I only wish I could run them at a UNIX shell at work instead of at the windows command line. With these two packages, their python bindings, numpy, pyshp and matplotlib I think I could begin to move away from ArcGIS without too much pain.

Version 1.8 and up of GDAL support the creation of output raster files when running the GDAL_RASTERIZE command, which makes this operation a single line process, there are a number of switches and options which can be used but here I will only go through the switches used for this process. The command I used to convert the vector outline to a raster file was:


The individual switches operate as follows:

• -a ID Indicates the attribute to be used as the key when converting the vector file. This can be extended using an SQL statement and the -SQL switch to allow for selections of only parts of a file to be converted.
• -l GB_Fix Is the name of the layer to be converted
• -a_nodata -9999  Assigns a raster nodata value of -9999
• -a_srs EPSG:3857 Sets the coordinate system using an EPSG code
• -tr 1000 1000 Sets the raster cell size, the smaller these values are the bigger the output file will be. I initially ran the command with 10 10 and got a file of over 15Gb
• -ot Float32 The datatype of the output file, either Byte, Float32 or Float64 depending on the file format
• GB_Fix.shp dest.tif The input and output filenames, including extensions

This is just scraping the surface of this one command but serves as a starting point for any future conversions I have to perform using GDAL.