Merge GDAL interface into ESMF develop. #474
Conversation
…n into individual file format methods and reaarange PIO ifdefs.
…ssue with file format flag Fortran-C correspondance.
and working in serial (local_pet == 0). MSL
so it's being committed. Still needs some clean-up but that can wait MSL
read now. (ESMCI_IO_GDAL* files already created) - MSL
data/field read from shapefile MSL
GDALClose still segfaults, because it's still done on all nodes. Close needs to happen only once. At this point, freezing this branch, and cloning into another for a restart. It looks like we need the PIO CPU topo and file handling types to manage the interface. So PIO will be cloned and GDAL built in in place of NETCDF. Lots to do! MSL
Going to clone PIO and convert from NETCDF to GDAL. This reverts commit 83dccf0.
which will hold the GDAL-specific routines -- likely all routines in one module. Maybe they'll split apart in the future. MSL
the data directly. -- Need to map from Shapefile features to Mesh elements when features < elements (as in the case with Multipolygons) -- What is going on with start/count in PIOc_InitDecomp()?
The process of reading a SHP file in ESMF broke multipolygons apart into individual elements. This revision preserves the multipolygons as a whole element. This facilitates data reading/writing on a Mesh. -- rewrote the GDAL Util routines. -- Made extensive use of std::vectors -- Now, no longer strictly a C project. Now C++ too MSL
a mesh created from that shapefile. -- serial read, mpi broadcast. -- Only OFTReal is possible at the moment. It seems that OFTString is interpreted as a real by GDAL/ESMF, so that a string field with an integer in it is read as a double. Weird MSL
writing data yet. -- MSL
It has been cleaned up for merging with the main repository I expect a lot of feedback and edits. There is much more work to be done. -- MSL
in prep for a PR to diagnose the GDAL read issues. MSL
garbled after running ESMF_FieldRegridStore() MSL
by a DE from a shapefile. And a makefile fix MSL
Make sure that everything is created > 0. MSL
…ELIBS that need looking out for. MSL
Not sure the parallel read works, but it at least works on 1 proc MSL
be better if we set the FIDs at file open and got them here as i. MSL
|
core team: currently on a branch but will take non-trivial work on PIO and get this merged into ESMF. Erik at NCAR had also asked about shapefile. Because it is related to PIO, we need to talk to Jim E. before he retires. (and Brian and Bill) |
|
@anntsay |
|
Hi @msl3v, We prefer to include third party libraries externally but PIO has been an exception. If we're ever bringing in a new library then we need to separate the pull request into 1. additional embedded external code 2. modification of existing code. And if there is an upstream then we have to discuss code synchronization practices. |
4 participants
This PR is meant to pull in the changes made to include GDAL and functionality for GIS files in ESMF. Right now it's limited to shapefiles. It can open files to create meshes or LocStreams, read fields onto a mesh or LocStream. It can write mesh fields onto existing shapefiles, but cannot create the shapefile itself.
Changes made to IO, Mesh, Field, and LocStream. And do build & cmake files.