WCPS1.5_Wrong operator precedence with coverage subsets

[Piero Campalani]

BEGIN ====================================================
On 18/06/13 16:14, Daems Dirk wrote:

Hi list,

When I execute the following query:

for c in (NDVI_SPOT_VGT)
return encode
   (coverage NDVI_HIST_STDEV
             over $pt t(0:1)
    values ( c[t(28 + 36*$pt),x:"CRS:1"(4001),y:"CRS:1"(4001)])
   ,"csv")

the result is:

{136,189}

When I try to revert to the physical value (= digital value * 0.004 - 0.08):

for c in (NDVI_SPOT_VGT)
return encode
   (coverage NDVI_HIST_STDEV
             over $pt t(0:1)
    values ( c[t(28 + 36*$pt),x:"CRS:1"(4001),y:"CRS:1"(4001)] * 0.004 - 0.08 )
   ,"csv")

the result is:

{-10.336,-14.364}

What actually seems to happen is this: c[t(28 + 36*$pt),x:"CRS:1"(4001),y:"CRS:1"(4001)] * (0.004 - 0.08), which doesn't seem correct according to the operator precedence rules (section 7.2.4 of OGC 08-068r2).
====================================================== END

Example with internal test coverages:

for c in (mr)
    return encode (
        c[x:"CRS:1"(30),y:"CRS:1"(30)] * 1 - 1, 
    "csv")

which gives 0 (<subset> * (1-1)), instead of the correct value.

[Piero Campalani]

Additional related bug (thanks Dirk): '-X-Y' evaluated as '-(X-Y)=-X+Y'

BEGIN ====================================================

  for c in (NDVI_SPOT_VGT)
  return encode
     (coverage NDVI_HIST_STDEV
               over $pt t(0:1)
      values ( (c[t(28 + 36*$pt),x:"CRS:1"(4001),y:"CRS:1"(4001)] * 0.004) - 0.08 - 1 )
     ,"csv")

is actually evaluated as (c[t(28 + 36*$pt),x:"CRS:1"(4001),y:"CRS:1"(4001)] * 0.004) - (0.08 - 1).
====================================================== END

[Dimitar Misev]

There is a workaround for this with parenthesis right? Lowering priority..

[Dimitar Misev]

I spent some time and the conclusion is that this will be pretty complicated to solve because we have mix of CoverageExpr and NumericScalarExpr. Essentially we need to duplicate a bunch of rules in the grammar. For unmixed expressions the precedences are fine. Probably it will be best to postpone for WCPS 2.0 or so.

[Bang Pham Huu]

It is fixed in WCPS1.5, e.g (based on Piero's examples)

c[i(3), j(30)] = 8

Example 1:

for c in (mr)return encode (c[i(30),j(30)] * 1 - 1,"csv")

WCPS1.5 returns {7}, WCPS1.0 returns {0}

Example 2:

for c in (mr)return encode (c[i(30),j(30)]*1  - 0.08 - 1 ,"csv")

WCPS1.5 returns {6.92}, WCPS1.0 returns {-0.64}

then I will close this ticket here as it is already supported in WCPS 1.5.