Programmable filter can add dimensions to context

[gadomski]

Using one or more add_dimension options on a programmable filter will instruct the filter to add those dimensions to the primary PointContext. This way, the programmable filter can set dimensions that might not be present on the PointContext.

My use case for this is manually scaling and "renaming" a dimension coming in from a reader, e.g. this pipeline:

<?xml version="1.0" encoding="utf-8"?>
<Pipeline version="1.0">
    <Writer type="drivers.las.writer">
        <Option name="filename">output.las</Option>
        <Option name="discard_high_return_numbers">true</Option>
        <Filter type="filters.programmable">
            <Option name="source">
import numpy
def reflectance_to_intensity(ins, outs):
    ref = ins["Reflectance"]
    min = numpy.amin(ref)
    max = numpy.amax(ref)
    outs["Intensity"] = (65536 * (ref - min) / (max - min)).astype(numpy.uint16)
    return True
            </Option>
            <Option name="function">reflectance_to_intensity</Option>
            <Option name="module">pyrxp</Option>
            <Option name="add_dimension">Intensity</Option>
            <Reader type="drivers.rxp.reader">
                <Option name="filename">/home/vagrant/PDAL/test/data/rxp/130501_232206_cut.rxp</Option>
            </Reader>
        </Filter>
    </Writer>
</Pipeline>

The rxp reader, which produces the "Reflectance" value, comes from this branch.

[hobu]

So filters.programmable is going to be the way to add dimensions "slots" to a pipeline? Works for me.

[gadomski]

I thought about adding another filter, but since my use case is specific I thought I'd keep it simple for now. I'm open to feedback on that idea, though.