1.a - Anaconda does indeed have its own set of packages/libraries, so yes, there is a lot of duplication. But this does ensure Anaconda had exactly the libraries it needs. For the Mac, it actually is our recommended method of installing PyFerret.
(again, substituting /my/path/to/miniconda appropriately; remember to make the script executable - "chmod 755 pyferret") which allows the user to run PyFerret without having to worry about being in a bash shell and activating the environment.
1.c - I do not know if there are any conflicts with having both MacPorts and Anaconda. I am not sure if that might have been a source of complications for you.
2.a - It would be nice if there was a proper "configure" for PyFerret (and possibly Ferret), but we just haven't had time available to put that together. With the Anaconda build available, there i
I just succeeded in installing ferret 7.442 on a new Macbook running OS
X 10.14.4, and I have some experiences to report.
(1) Sure I wanted to try out pyferret. My experience with this install
was not good, however. There are confusing circular references all
throughout the website, no possibility of precompiled binaries, and
anaconda is astonishingly complex. It looks like a whole duplicate set
of unix packages, like another MacPorts/Homebrew system. I eventually
got miniconda installed, but could not convince it to download, compile,
install, or run ferret. Happy to rm -rf that whole mess.
(2) I was able to compile from source by cloning the git repo following
https://github.com/NOAA-PMEL/Ferret/blob/master/README_ferret_mac_homebrew.md
and making appropriate changes to platform_specific.mk.intel-mac and and
site_specific.mk. Yay for the good old days. My main challenge is that
I use MacPorts, not Homebrew, for the netCDF, readline, szip, etc.
libraries. I could not get MacPorts to install static libraries in many
cases (this apparently needs to be specifically added by package
developer). Here's what I had to do:
- copy /opt/local/bin/gfortran-mp-8 to /opt/local/bin/gfortran. Could
not convince the Make system to use a different FC. This ought to be fixed.
- needed to install not just the MacPorts netcdf package, but also
netcdf-fortran.
- all LIBDIR variables in
site_specific.mk are set to /opt/local/lib
- the references to .a static libraries in
platform_specific.mk.intel-mac had to be changed to .dylib:
else ifeq ($(strip $(SZ_LIBDIR)),)
# CDFLIB = $(NETCDF_LIBDIR)/libnetcdff.a \
# $(NETCDF_LIBDIR)/libnetcdf.a \
# $(HDF5_LIBDIR)/libhdf5_hl.a \
# $(HDF5_LIBDIR)/libhdf5.a
CDFLIB = $(NETCDF_LIBDIR)/libnetcdff.dylib \
$(NETCDF_LIBDIR)/libnetcdf.dylib \
$(HDF5_LIBDIR)/libhdf5_hl.dylib \
$(HDF5_LIBDIR)/libhdf5.dylib
else
# CDFLIB = $(NETCDF_LIBDIR)/libnetcdff.a \
# $(NETCDF_LIBDIR)/libnetcdf.a \
# $(HDF5_LIBDIR)/libhdf5_hl.a \
# $(HDF5_LIBDIR)/libhdf5.a \
# $(SZ_LIBDIR)/libsz.a
CDFLIB = $(NETCDF_LIBDIR)/libnetcdff.dylib \
$(NETCDF_LIBDIR)/libnetcdf.dylib \
$(HDF5_LIBDIR)/libhdf5_hl.dylib \
$(HDF5_LIBDIR)/libhdf5.dylib \
$(SZ_LIBDIR)/libsz.dylib
Sorry if this is a knock on pyferret, but my summary is that the install
system is not ready for prime time.
--
Andy Jacobson
andy.jacobson@xxxxxxxx
NOAA Earth System Research Lab
Global Monitoring Division
325 Broadway R/GMD1
Boulder, Colorado 80305
303/497-4916