Getting cmake working.

Author: samharrison7

.gitignore

@@ -11,6 +11,8 @@ doc/graph/
 doxygen/
 bin/
 build/
+release/
+debug/
 obj/
 log/
 cache/
@@ -58,6 +60,10 @@ scratch/
 *.out
 *.app
 
+# Debugging
+massif*
+gmon.out
+
 # IDEs
 *.cbp
 *.depend
@@ -68,6 +74,7 @@ scratch/
 .vscode/
 .project
 Debug/
+Release/
 *.sln
 *.u2d
 *.vfproj*

.gitmodules

@@ -7,10 +7,6 @@
     path = vendor/mo_netcdf
     url = https://github.com/schaefed/mo_netcdf
 
-[submodule "json2netcdf"]
-    path = vendor/json2netcdf
-    url = https://github.com/samharrison7/json2netcdf
-
 [submodule "vendor/datetime-fortran"]
 	path = vendor/datetime-fortran
 	url = https://github.com/samharrison7/datetime-fortran

CHANGELOG.md

@@ -1,9 +1,20 @@
 # Changelog
 
-All notable changes to the model will be documented in this file.
+All notable changes to the model will be documented in this file. Breaking changes (i.e. those that cause changes to the model's interface) are denoted by ⚠️
 
 ## [Unreleased]
 
+### Added
+
+- ⚠️ Added option to write to NetCDF file. Specify `&output > write_netcdf = .true.` to use. This requires the compilation of a new Fortran source file, `src/Data/NetCDFOutputModule.f90`, and your compilation process will need ammending. The [example Makefile](./Makefile.example) has been updated accordingly. NetCDF output variables are stored in memory for the entire model run (as opposed to CSV files, which are written iteratively on each time step), which means choosing to write to NetCDF may use significantly more memory. 
+- Along with this, added the option for NetCDF output to be chunked when the model is run in batch mode. If `&output > chunk_netcdf = .true.`, then a NetCDF file is written at the end of every chunk in the batch run. If `.false.`, the NetCDF is written at the end of the entire batch. As output variables are stored in memory, for particularly long batch runs, memory usage may be very high for non-chunked NetCDF output. The `nco` command line tools can be used to easily merge chunked NetCDF files after the run, e.g. using [`ncrcat`](http://nco.sourceforge.net/nco.html#ncrcat).
+
+### Changed
+
+- ⚠️ `src/CheckpointModule1.f90` renamed to `src/CheckpointModule.f90`. Your compilation process may need ammending. The [example Makefile](./Makefile.example) has been updated.
+- Submodule `vendor/mo_netcdf` updated to latest version to allow the creation of scalar variables.
+- Submodule `vendor/json2netcdf` removed.
+
 ## [0.0.1] - 2021-03-20
 
 ### Added

CMakeLists.txt

@@ -3,30 +3,35 @@ cmake_minimum_required(VERSION 3.13)
 project(nanofase)
 enable_language(Fortran)
 
-if (CMAKE_Fortran_COMPILER_ID MATCHES "GNU")
-    set(debug_flags "-Wall -O0 -fcheck=all -fbackslash -g -ffpe-trap=zero,invalid,overflow,underflow -pg")
+# GFortran compiler
+if (CMAKE_Fortran_COMPILER_ID MATCHES GNU)
+    set(debug_flags "-Wall -O1 -fcheck=all -fbackslash -g -ffpe-trap=zero,invalid,overflow,underflow -pg")
     set(release_flags "-O3 -fbackslash")
 endif()
 
-if (CMAKE_Fortran_COMPILER_ID MATCHES "Intel")
+# Intel Fortran compiler
+if (CMAKE_Fortran_COMPILER_ID MATCHES Intel)
     set(debug_flags "/assume:bscc /fpe:0 /check:bounds /traceback /check:stack /debug:full /Od")
     set(release_flags "/O3 /assume:bscc")
 endif()
 
-# Get the NetCDF flags by running nf-config command
-execute_process(COMMAND nf-config --fflags --flibs OUTPUT_VARIABLE netcdf_flags)
-# For some reason nf-config returns string with newline characters, so we need to remove these
-string(REPLACE "\n" " " netcdf_flags ${netcdf_flags})
-
-set(CMAKE_Fortran_FLAGS_DEBUG "${CMAKE_Fortran_FLAGS_DEBUG} ${debug_flags} ${netcdf_flags}")
-set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} ${release_flags} ${netcdf_flags}")
-# set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} ${netcdf_flags}")
+set(CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake" CACHE INTERNAL "Location of CMake modules.")
+set(NETCDF_F90 "YES")
+find_package(NetCDF REQUIRED)
 
+set(CMAKE_Fortran_FLAGS_DEBUG "${CMAKE_Fortran_FLAGS_DEBUG} ${debug_flags}")
+set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} ${release_flags}")
 
+# Sources to compile
 file(GLOB_RECURSE sources src/*.f90 vendor/*.f90)
 list(FILTER sources EXCLUDE REGEX ".*vendor/feh/example/.*f90$")
-list(FILTER sources EXCLUDE REGEX ".*vendor/datetime-fortran/src/tests/datetime_tests.f90$")
+list(FILTER sources EXCLUDE REGEX ".*vendor/feh/tests/.*f90$")
+list(FILTER sources EXCLUDE REGEX ".*vendor/datetime-fortran/src/tests/.*f90$")
 list(FILTER sources EXCLUDE REGEX ".*vendor/mo_netcdf/examples/.*f90$")
 list(FILTER sources EXCLUDE REGEX ".*vendor/mo_netcdf/tests/.*f90$")
-add_executable(prog ${sources})
-target_link_options(prog INTERFACE "${netcdf_flags}")
+list(FILTER sources EXCLUDE REGEX "src/Data/DataOutputModule.f90")
+
+# Compile to executable and link to NetCDF
+add_executable(nanofase ${sources})
+target_include_directories(nanofase PUBLIC "${NETCDF_INCLUDES}")
+target_link_libraries(nanofase PRIVATE "${NETCDF_LIBRARIES_F90}")

Makefile.example

@@ -80,8 +80,9 @@ OBJ = vendor/feh/src/ErrorInstance.o \
 	src/GridCell/classGridCell2.o \
 	src/Environment/spcEnvironment.o \
 	src/Environment/classEnvironment1.o \
+	src/Data/NetCDFOutputModule.o \
 	src/Data/DataOutputModule1.o \
-	src/CheckpointModule1.o \
+	src/CheckpointModule.o \
 	src/main.o 
 
 OBJECTS = $(notdir $(OBJ))

cmake/FindNetCDF.cmake

@@ -0,0 +1,71 @@
+# - Find NetCDF
+# Find the native NetCDF includes and library
+#
+#  NETCDF_INCLUDES    - where to find netcdf.h, etc
+#  NETCDF_LIBRARIES   - Link these libraries when using NetCDF
+#  NETCDF_FOUND       - True if NetCDF found including required interfaces (see below)
+#
+# Your package can require certain interfaces to be FOUND by setting these
+#
+#  NETCDF_CXX         - require the C++ interface and link the C++ library
+#  NETCDF_F77         - require the F77 interface and link the fortran library
+#  NETCDF_F90         - require the F90 interface and link the fortran library
+#
+# The following are not for general use and are included in
+# NETCDF_LIBRARIES if the corresponding option above is set.
+#
+#  NETCDF_LIBRARIES_C    - Just the C interface
+#  NETCDF_LIBRARIES_CXX  - C++ interface, if available
+#  NETCDF_LIBRARIES_F77  - Fortran 77 interface, if available
+#  NETCDF_LIBRARIES_F90  - Fortran 90 interface, if available
+#
+# Normal usage would be:
+#  set (NETCDF_F90 "YES")
+#  find_package (NetCDF REQUIRED)
+#  target_link_libraries (uses_f90_interface ${NETCDF_LIBRARIES})
+#  target_link_libraries (only_uses_c_interface ${NETCDF_LIBRARIES_C})
+
+if (NETCDF_INCLUDES AND NETCDF_LIBRARIES)
+  # Already in cache, be silent
+  set (NETCDF_FIND_QUIETLY TRUE)
+endif (NETCDF_INCLUDES AND NETCDF_LIBRARIES)
+
+find_path (NETCDF_INCLUDES netcdf.h
+  HINTS NETCDF_DIR ENV NETCDF_DIR)
+
+find_library (NETCDF_LIBRARIES_C       NAMES netcdf)
+mark_as_advanced(NETCDF_LIBRARIES_C)
+
+set (NetCDF_has_interfaces "YES") # will be set to NO if we're missing any interfaces
+set (NetCDF_libs "${NETCDF_LIBRARIES_C}")
+
+get_filename_component (NetCDF_lib_dirs "${NETCDF_LIBRARIES_C}" PATH)
+
+macro (NetCDF_check_interface lang header libs)
+  if (NETCDF_${lang})
+    find_path (NETCDF_INCLUDES_${lang} NAMES ${header}
+      HINTS "${NETCDF_INCLUDES}" NO_DEFAULT_PATH)
+    find_library (NETCDF_LIBRARIES_${lang} NAMES ${libs}
+      HINTS "${NetCDF_lib_dirs}" NO_DEFAULT_PATH)
+    mark_as_advanced (NETCDF_INCLUDES_${lang} NETCDF_LIBRARIES_${lang})
+    if (NETCDF_INCLUDES_${lang} AND NETCDF_LIBRARIES_${lang})
+      list (INSERT NetCDF_libs 0 ${NETCDF_LIBRARIES_${lang}}) # prepend so that -lnetcdf is last
+    else (NETCDF_INCLUDES_${lang} AND NETCDF_LIBRARIES_${lang})
+      set (NetCDF_has_interfaces "NO")
+      message (STATUS "Failed to find NetCDF interface for ${lang}")
+    endif (NETCDF_INCLUDES_${lang} AND NETCDF_LIBRARIES_${lang})
+  endif (NETCDF_${lang})
+endmacro (NetCDF_check_interface)
+
+NetCDF_check_interface (CXX netcdfcpp.h netcdf_c++)
+NetCDF_check_interface (F77 netcdf.inc  netcdff)
+NetCDF_check_interface (F90 netcdf.mod  netcdff)
+
+set (NETCDF_LIBRARIES "${NetCDF_libs}" CACHE STRING "All NetCDF libraries required for interface level")
+
+# handle the QUIETLY and REQUIRED arguments and set NETCDF_FOUND to TRUE if
+# all listed variables are TRUE
+include (FindPackageHandleStandardArgs)
+find_package_handle_standard_args (NetCDF DEFAULT_MSG NETCDF_LIBRARIES NETCDF_INCLUDES NetCDF_has_interfaces)
+
+mark_as_advanced (NETCDF_LIBRARIES NETCDF_INCLUDES)

config.example/batch_config_test-scenario.example.nml

@@ -0,0 +1,10 @@
+&batch_config
+n_chunks = 3
+/
+
+&chunks
+input_files = "data.example/test-scenario.nc", "data.example/test-scenario-t11.nc", "data.example/test-scenario.nc"
+constants_files = "data.example/constants_test-scenario.nml", "data.example/constants_test-scenario.nml", "data.example/constants_test-scenario.nml"
+start_dates = "2015-01-01", "2015-01-11", "2015-01-23" 
+n_timesteps_per_chunk = 10, 11, 10
+/

config.example/config.example.nml

@@ -32,6 +32,8 @@ output_path = "data/output/"                            ! Path to store the outp
 ! the more data that is output and the larger the output data file sizes
 &output
 write_csv = .true.                                      ! Should we write output data to CSV files (n.b. turning this off currently suppresses all output, except for runs to steady state)
+write_netcdf = .true.                                   ! Should we write output data to a NetCDF file?
+netcdf_write_mode = 'itr'                           ! When should we write to the NetCDF file? Every time step ('itr') or at the end of the run ('end'). Every time step is slower, end uses much more memory
 write_metadata_as_comment = .true.                      ! Should output data metadata (e.g. column descriptions) be included as comments at top of CSV files?
 write_compartment_stats = .true.                        ! Should a file with summary stats for each compartment (soil, water, sediment) be written?
 include_waterbody_breakdown = .true.                    ! For surface water output, include breakdown over waterbodies or aggregate at grid cell level?

config.example/test-scenario.example.nml

@@ -8,7 +8,6 @@ n_sediment_layers = 4                                   ! Number of sediment lay
 n_nm_size_classes = 5                                   ! Number of NM size classes
 n_spm_size_classes = 5                                  ! Number of SPM size classes
 n_fractional_compositions = 4                           ! Number of fractional compositions for sediment
-n_other_sites = 1                                       ! Number of other sampling sites
 /
 
 &nanomaterial
@@ -17,15 +16,6 @@ n_nm_forms = 4                                          ! Number of NM forms (fr
 n_nm_extra_states = 2                                   ! Number of extra NM states, other than heteroaggregated to SPM
 / 
 
-! Run the model in calibration mode? Documentation here: [doc/calibration.md](../doc/calibration.md)
-&calibrate
-calibration_run = .false.                               ! Run in calibration mode or not? Following variables only required if this is true
-site_data = ""                                          ! Path to CSV of sampling sites
-start_site = ""                                         ! Code of start site (most upstream) in sampling sites CSV
-end_site = ""                                           ! Code of end site (most downstream) in sampling sites CSV
-other_sites = ""                                        ! Which other sites should be used from sampling sites CSV? Array of length &allocatable_array_sizes > n_other_sites
-/
-
 ! Paths to data. For info on compiling data for the NanoFASE model, see the nanofase-data repo: https://github.com/NERC-CEH/nanofase-data
 &data
 input_file = "data.example/test-scenario.nc"            ! Path to NetCDF input data, which includes most of the spatial and temporally resolved data
@@ -34,14 +24,16 @@ output_path = "data/output/"                            ! Path to store the outp
 /
 
 &output
-write_csv = .true.                                      ! Should we write output data to CSV files (n.b. turning this off currently suppresses all output, except for runs to steady state)
-write_metadata_as_comment = .true.                      ! Should output data metadata (e.g. column descriptions) be included as comment at top of CSV file?
+write_csv = .true.                                      ! Should we write output data to CSV files?
+write_netcdf = .true.                                   ! Should we write output data to a NetCDF file?
+netcdf_write_mode = 'end'
+write_metadata_as_comment = .false.                      ! Should output data metadata (e.g. column descriptions) be included as comment at top of CSV file?
+include_waterbody_breakdown = .true.
 include_sediment_layer_breakdown = .true.               ! Include breakdown of data over sediment layers?
 include_soil_layer_breakdown = .false.                  ! Include breaedown of data over soil layers?
 soil_pec_units = 'kg/kg'                                ! What units to use for soil PEC - kg/kg or kg/m3?
 sediment_pec_units = 'kg/kg'                            ! What units to use for sediment PEC - kg/kg or kg/m3?
 include_soil_state_breakdown = .false.                  ! Include breakdown of NM state - free vs attached to soil matrix
-! output_rate_constant_summary = .true.
 /
 
 &run
@@ -56,19 +48,6 @@ log_file_path = "log/"                                  ! Where to place model l
 warm_up_period = 0                                      ! Warm up period before main simulation begin. *Not yet implemented.*
 /
 
-&checkpoint
-checkpoint_file = "./checkpoint.dat"                    ! Location of checkpoint file to read from and/or save to
-save_checkpoint = .false.                               ! Save a checkpoint file when the run is finished? Defaults to false
-reinstate_checkpoint = .false.                          ! Reinstate a checkpoint from checkpoint_file? Defaults to false
-preserve_timestep = .false.                             ! Should the timestep from the checkpoint be used as a starting timestep in a reinstated run?
-/
-
-&steady_state
-run_to_steady_state = .false.
-mode = 'sediment_size_distribution'
-delta = 1e-4                                            ! Delta value to test whether at steady state
-/
-
 &soil
 soil_layer_depth = 0.05, 0.15, 0.2                      ! Depth of each soil layer. Array of length &allocatable_array_sizes > n_soil_layers
 include_bioturbation = .true.                           ! Should bioturbation be modelled?
@@ -77,7 +56,7 @@ include_attachment = .true.                             ! Should attachment be m
 
 &sediment
 sediment_layer_depth = 0.01, 0.01, 0.01, 0.01           ! Depth of each sediment layer. Array of length &allocatable_array_sizes > n_sediment_layers
-spm_size_classes = 1e-6, 3e-6, 10e-6, 30e-6, 100e-6     ! Diameter of SPM in each binned size class [m]
+spm_size_classes = 0.002e-3, 0.06e-3, 0.2e-3, 0.6e-3, 2.0e-3    ! Diameter of SPM in each binned size class [m]
 include_bed_sediment = .true.                           ! Should bed sediment be modelled?
 sediment_particle_densities = 1500, 2600, 2600, 2600    ! Density of sediment particles in each fractional composition class [kg/m3]
 /