Allow dynamic raster colorisation based on index

[apercas]

Add a legend with the color steps in it when a raster layer is displayed. Coordinate with @agrosmartrural for data validation.

[apercas]

• First: create a table where we'll store the type of each dataset:

raster name (varchar)	color type (varchar)
my_awesome_raster_field5_raster	NDVI
second_raster_temperature	Temperature

• Precreate raster legends for those type of colorisation.

Current plan for once a user drags a raster dataset:

• Check raster and look for colours on tiff_loader.rb and raster2pgsql.rb
• Try guessing type (NDVI, NDRE etc)
• Add a row to our table with the raster name and the type. We'll have to bypass the apikey here somehow.
• On the raster_layer.js class from our fork of cartodb.js query over the SQL API and look for the type.
• Display a legend from the array we created before making use of Leaflet and maybe in the future use Builder's legends streaming from here the presets.

Ideas @jjmata @ivanprado?

[ivanprado]

Some thoughts:

• Probably legends will be dynamic. I mean, there are absolute values, at least for maps like temperature. Which means that what was colored as red in one map could means 30 degrees, but for other red means 20 degrees. I'm not 100% sure of that, I think it should be checked with Diego. Must colorization be fixed, or it is relative/normalized for each map? Do legend should show relative values or absolute values?
• I think platform should take care of colorization when loading (if possible and enough simple). The way of doing it is using Gdal, see here, but it seems that previously it is needed to perform a color classification to assign colors to ranges. I have found that and that but maybe there are better/simpler ways for this calculation.
• Legend could be based on QGis format. I would have one legend/color_map associated to each layer as it allows to dynamic legends and it would be pretty flexible:

  # Archivo de exportación de mapa de colores generado por QGIS
  INTERPOLATION:INTERPOLATED
  -0.012712,26,150,65,255,mu verde
  0.066397,166,217,106,255,menos verde
  0.145507,255,255,192,255,amarillea
  0.224616,253,174,97,255,esta reguleras
  0.303725,215,25,28,255,esta fatal

[apercas]

Just as a reference, QGIS file for styling example (please note that we also have the opacity value in case we don't want to display them with the same transparency):

<!DOCTYPE qgis PUBLIC 'http://mrcc.com/qgis.dtd' 'SYSTEM'>
<qgis version="2.18.2" minimumScale="inf" maximumScale="1e+08" hasScaleBasedVisibilityFlag="0">
  <pipe>
    <rasterrenderer opacity="1" alphaBand="0" classificationMax="50.4219" classificationMinMaxOrigin="CumulativeCutSubExtentEstimated" band="1" classificationMin="29.2193" type="singlebandpseudocolor">
      <rasterTransparency/>
      <rastershader>
        <colorrampshader colorRampType="INTERPOLATED" clip="0">
          <item alpha="255" value="29.2" label="29.2" color="#d7191c"/>
          <item alpha="255" value="34.5" label="34.5" color="#fdae61"/>
          <item alpha="255" value="39.8" label="39.8" color="#ffffbf"/>
          <item alpha="255" value="45.1" label="45.1" color="#abd9e9"/>
          <item alpha="255" value="50.4" label="50.4" color="#2c7bb6"/>
        </colorrampshader>
      </rastershader>
    </rasterrenderer>
    <brightnesscontrast brightness="0" contrast="0"/>
    <huesaturation colorizeGreen="128" colorizeOn="0" colorizeRed="255" colorizeBlue="128" grayscaleMode="0" saturation="0" colorizeStrength="100"/>
    <rasterresampler maxOversampling="2"/>
  </pipe>
  <blendMode>0</blendMode>
</qgis>