[raster] Add st_tpi

[nclay]

Add the ability to calculate the Topographic Position Index for a DEM raster.

[nclay]

st_tpi

[Bborie Park]

Can I get some reference material describing TPI and TRI? That way appropriate regression tests can be created. I can't really consider committing this without a good understanding of the operations and verifying that the code is doing what it is saying it is doing.

[nclay]

Definition of TPI via BPI pg 10

[Bborie Park]

No regression tests or docs… punting to the future. Should also add roughness…

[nclay]

Added docs, is there anything other than regression test that I need to do?

Thanks,

Nathan

[robe]

nclay — roughness?(though guess that can wait if it's a feature) Also the id of your function in doc has to be <refentry id="RT_ST_TPI"> to agree with other documented raster functions.

I'm willing to accept this for 2.1 (since it was originally slated for 2.1) if you can get the regress in by next week and dustymugs is fine with it.

dustymugs — what d'ya think?

[Bborie Park]

Assuming regression tests are done and code review, I'm fine with 2.1. roughness would be a separate function and the last one needed to complete the "basic" set of DEM processing functions.

[nclay]

Regression tests, how should or should I, I add rasters to the testing environment? What rasters am I to expect to be available, for my regression scripts? I know that these are basic questions but I want to get things right and up to snuff.

Thanks,

Nathan

[Bborie Park]

Generally, you want to create a raster within the database using ST_MakeEmptyRaster(), ST_AddBand() and ST_SetValues(). That way, everything is portable regardless of filesystem.

[nclay]

Just to be clear, I should build up and tear down any needed raster within the regression script.

[Bborie Park]

Yes. You may want to consider regression tests that focus only upon the callback function. Take a look at rt_4ma.sql and rt_invdistweight4ma.sql in raster/test/regress. The idea is to test just the code that needs testing rather than the whole map algebra framework, which is tested elsewhere. This way any failures can only originate in the callback function.

[nclay]

I've added the regression tests.

[Bborie Park]

Added in r11378