[GRASS-dev] pyGRASS and mapcalc's pixel modifiers

[Nikos Alexandris]

Nikos Alexandris wrote:

> >>> Consider a Python class with methods that derive a somewhat complex mapcalc
> >>> expression which contains the following pixel modifiers:
> >>>
> >>> ['A[-1, -1]', 'A[-1, 0]', 'A[-1, 1]', 'A[0, -1]', 'A[0, 0]', 'A[0, 1]',
> >>> 'A[1, -1]', 'A[1, 0]', 'A[1, 1]']

> >>> These modifiers built-up a moving window to compute some values (eg. mean, median and other in-between values) and finally end up in a polynomial like:
> >>>
> >>> -9.674 + 0.653 * SomeValue + 9.087 * SomeValue^2
> >>>
> >>> As far as my understanding goes, I can't check for the validity of the
> >>> in-between computed values with mapcalc's powers.  For example, check if the
> >>> average value as derived from the above defined pixel modifiers, lies within a
> >>> specific range.  I have build helper functions, as part of the Python class,
> >>> which could do this.  Then again, I don't have the powers of mapcalc to process
> >>> the data.
> >>>
> >>> I wonder how I should approach this, excluding rewriting my program in C (which
> >>> I don't speak easily).  If I go step-by-step, that is:
> >>>
> >>> 1) compute a map based with the mean of neighboring pixels (see pix-modifiers above)
> >>> 2) use pyGRASS' RasterRow, for example, to be able to apply some helper function and check or do something
> >>> 3) compute another, required, in-between map
> >>> 4) use pyGRASS again
> >>> 5) etc.
> >>>
> >>> up until to reach the final equation of interest.  This sounds really messy.  How would you approach this (excluding an
> >>> implementation in C)?

Moritz:

> >> Have a look at the eval function of r.mapcalc (check the specific
> >> section of the man page). It allows you to create intermediate results
> >> which you can then use in further processing.

Nikos:

> > Thank you Moritz.  Already in use.   And quite happy about it.  eval,
> > anyhow, > > isn't an answer to my question. A simple example of what would
> > satisfy the current need, would be the > > application of a helper function,
> > like the following, on the fly.  That is > > while reading or computing pixel
> > values.

> > def check_cwv(cwv):
> >      """
> >      Check whether a column water vapor value lies within a "valid" range.
> >      """
> >      if cwv < 0.0 or cwv > 6.3:
> >          raise ValueError('The column water vapor estimation is out of the '
> >                           'expected ranfe [0.0, 6.3]')
> >      else:
> >          return True
 
Moritz Lennert:

> It's clear that you cannot raise an error with such a message in 
> r.mapcalc, but I don't understand why you couldn't do such a test (as 
> long as this test is pixel by pixel) ? Instead of the message you could 
> give those pixels a specific value (e.g. -1 or -9999 or whatever).

Hmmm?  You mean to use if-else conditionals, assign an "out-of-range"
value for pixes which are "faulty"?

That sounds like an idea, but will this make my computations more clean?
Thanks a lot for caring & sharing Moritz. 

> But I guess I just don't really understand what you are after...

I am after adding checks in various steps of the algorithm I am implementing.
It's just a somewhat complex equation [*]. That's all.  Hopefully I'll get
it complete and share results soon.

Nikos

[*] Huazhong Ren, Chen Du, Qiming Qin, Rongyuan Liu,
Jinjie Meng, and Jing Li. "Atmospheric Water Vapor Retrieval
from Landsat 8 and Its Validation." 3045–3048. IEEE, 2014.