Force use of stated nodata value and float32 dtype in pollination

HISTORY.rst

@@ -37,6 +37,14 @@
 
 Unreleased Changes
 ------------------
+* Pollination
+    * Fixed an issue with nodata handling that was causing some outputs to be
+      filled either with the float32 value for positive infinity, or else with
+      a value very close to it. https://github.com/natcap/invest/issues/1635
+    * While working on https://github.com/natcap/invest/issues/1635, we also
+      updated the stated dtype of most pollination model outputs to be float32
+      instead of the float64 dtype that was being assumed previously.  This
+      will result in smaller output filesizes with minimal loss of precision.
 * Workbench
     * Several small updates to the model input form UI to improve usability
       and visual consistency (https://github.com/natcap/invest/issues/912)

scripts/convert-requirements-to-conda-yml.py

@@ -2,6 +2,7 @@
 """convert-requirements-to-conda-yml.py"""
 
 import argparse
+import os
 import platform
 import sys
 
@@ -14,6 +15,46 @@
 """
 
 
+# Environment marker handling is taken straight from
+# https://peps.python.org/pep-0508/#environment-markers
+def _get_implementation_version():
+    def format_full_version(info):
+        version = '{0.major}.{0.minor}.{0.micro}'.format(info)
+        kind = info.releaselevel
+        if kind != 'final':
+            version += kind[0] + str(info.serial)
+        return version
+
+    if hasattr(sys, 'implementation'):
+        implementation_version = format_full_version(
+            sys.implementation.version)
+    else:
+        implementation_version = "0"
+    return implementation_version
+
+
+# Environment marker handling is taken straight from
+# https://peps.python.org/pep-0508/#environment-markers
+ENV_MARKERS = {
+    "os_name": os.name,
+    "sys_platform": sys.platform,
+    "platform_machine": platform.machine(),
+    "platform_python_impl": platform.python_implementation(),
+    "platform_release": platform.release(),
+    "platform_system": platform.system(),
+    "platform_version": platform.version(),
+
+    # Deliberately not supporting python_version marker in order to avoid
+    # possible issues with the python version not yet being known when this
+    # script is run.
+    # "python_version": '.'.join(platform.python_version_tuple()[:2]),
+
+    "python_full_version": platform.python_version(),
+    "implementation_name": sys.implementation.name,
+    "implementation_version": _get_implementation_version(),
+}
+
+
 def build_environment_from_requirements(cli_args):
     """Build a conda environment.yml from requirements.txt files.
 
@@ -62,7 +103,16 @@ def build_environment_from_requirements(cli_args):
                     # requirement if we're using pip.
                     conda_requirements.add('pip')
 
-                    pip_requirements.add(line)
+                    # Handle environment specifiers and see if the requirement
+                    # should exist in this environment.
+                    # InVEST pretty much just uses this for checking operating
+                    # systems.
+                    if ";" in line:
+                        package, marker = line.split(';')
+                        if eval(marker, ENV_MARKERS):
+                            pip_requirements.add(package)
+                    else:
+                        pip_requirements.add(line)
 
                     # If an scm needs to be installed for pip to clone to a
                     # revision, add it to the conda package list.

src/natcap/invest/pollination.py

@@ -855,6 +855,7 @@ def execute(args):
                             floral_resources_index_path_map[species],
                             convolve_ps_path],
                         target_path=pollinator_abundance_path,
+                        target_dtype=numpy.float32,
                         target_nodata=_INDEX_NODATA),
                     dependent_task_list=[
                         foraged_flowers_index_task_map[(species, season)],
@@ -935,7 +936,9 @@ def execute(args):
                 rasters=[
                     half_saturation_raster_path,
                     total_pollinator_abundance_index_path],
-                target_path=farm_pollinator_season_path),
+                target_path=farm_pollinator_season_path,
+                target_dtype=numpy.float32,
+                target_nodata=_INDEX_NODATA),
             dependent_task_list=[
                 half_saturation_task, total_pollinator_abundance_task[season]],
             target_path_list=[farm_pollinator_season_path]))
@@ -976,20 +979,24 @@ def execute(args):
         kwargs=dict(
             op=pyt_op,
             rasters=[managed_pollinator_path, farm_pollinator_path],
-            target_path=total_pollinator_yield_path),
+            target_path=total_pollinator_yield_path,
+            target_dtype=numpy.float32,
+            target_nodata=_INDEX_NODATA),
         dependent_task_list=[farm_pollinator_task, managed_pollinator_task],
         target_path_list=[total_pollinator_yield_path])
 
     # calculate PYW
     wild_pollinator_yield_path = os.path.join(
         output_dir, _WILD_POLLINATOR_YIELD_FILE_PATTERN % file_suffix)
     wild_pollinator_task = task_graph.add_task(
-        task_name='calcualte_wild_pollinators',
+        task_name='calculate_wild_pollinators',
         func=pygeoprocessing.raster_map,
         kwargs=dict(
             op=pyw_op,
             rasters=[managed_pollinator_path, total_pollinator_yield_path],
-            target_path=wild_pollinator_yield_path),
+            target_path=wild_pollinator_yield_path,
+            target_dtype=numpy.float32,
+            target_nodata=_INDEX_NODATA),
         dependent_task_list=[pyt_task, managed_pollinator_task],
         target_path_list=[wild_pollinator_yield_path])
 
@@ -1392,7 +1399,8 @@ def _sum_arrays(*array_list):
     result = numpy.empty_like(array_list[0])
     result[:] = 0
     for array in array_list:
-        local_valid_mask = ~pygeoprocessing.array_equals_nodata(array, _INDEX_NODATA)
+        local_valid_mask = ~pygeoprocessing.array_equals_nodata(
+            array, _INDEX_NODATA)
         result[local_valid_mask] += array[local_valid_mask]
         valid_mask |= local_valid_mask
     result[~valid_mask] = _INDEX_NODATA
@@ -1423,6 +1431,7 @@ def max_op(*substrate_index_arrays):
     pygeoprocessing.raster_map(
         op=max_op,
         rasters=substrate_path_list,
+        target_dtype=numpy.float32,
         target_path=target_habitat_nesting_index_path)
 
 
@@ -1432,6 +1441,7 @@ def _multiply_by_scalar(raster_path, scalar, target_path):
         op=lambda array: array * scalar,
         rasters=[raster_path],
         target_path=target_path,
+        target_dtype=numpy.float32,
         target_nodata=_INDEX_NODATA,
     )
 
@@ -1455,6 +1465,7 @@ def _calculate_pollinator_supply_index(
         op=lambda f_r, h_n: species_abundance * f_r * h_n,
         rasters=[habitat_nesting_suitability_path, floral_resources_path],
         target_path=target_path,
+        target_dtype=numpy.float32,
         target_nodata=_INDEX_NODATA
     )