Greg Landrum <greg.land...@gmail.com> writes:
> Hi Loris,
> On Thu, Sep 14, 2017 at 2:25 PM, Loris Bennett <loris.benn...@fu-berlin.de>
> I am trying to install RDKit on a university cluster running Linux from
> source. The build seem to go OK and 'make install' copied the
> to the NFS share where the software should reside. I then do
> export RDBASE=/cm/shared/apps/rdkit/rdkit_2017_03_3
> export PYTHONPATH=$PYTHONPATH:$RDBASE
> export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$RDBASE/lib
> However when I then run Python (2.6.6) and try
> Just to do some expectation management: python 2.6 is pretty ancient
> and there's no guarantee that all of the RDKit code will work with
> it. Python 2.7 is the minimum version that we "officially"
> support. It's a very good idea to update.
OK. I didn't notice that 2.6 was deprecated - maybe this could be
explicitly mentioned in the install instructions. I'm running the
RedHat clone Scientific Linux 6, so everything in this thread on
RH/Python applies. So I can use either Python 2.7 or Python 3.5. I can
ask the users what they prefer - although, as you seem know my users
here in Berlin, maybe you know too ;-)
> import rdkit
> I get
> ImportError: No module named rdkit
> I am not a Python person and my naive expectation was that there should
> be a file called
> Based on the info provided so far, there should be a directory called
> rdkit in the directory: /cm/shared/apps/rdkit/rdkit_2017_03_3
This directory exists.
> That directory should contain a number of sub dirs, other files, and a
> file called __init__.py (this is the one that tells Python that it can
> import the directory as a package). What do you see there?
The directory just contains
an nothing else, in particular, no __init__.py. I have plenty of
__init__.pys in the build directory, so I assume I must have done some
thing wrong when running cmake and/or make install.
I must admit that I found the installation instructions somewhat unclear
on that point. I would find it clearer if things were couched in terms
of 'source' and 'destination'. For me, as a make-guy rather than a
cmake-guy, it would also be helpful if it were made clearer at which
point the destination directory should be specified. I ended up with
RDKit being installed under a very long path with included both my
intended path and the original build path, so I had to move things
around and may have goofed up at that point.
> which has to be on my PYTHONPATH. However, since the unpacked sources
> together with the build don't seem to contain such a file, either
> something is broken or the rdkit module should be found by some other
> Again, based on the info above, I would expect that you want "make
> install" to copy the "rdkit" and "lib" directories (as well as a
> couple others) to /cm/shared/apps/rdkit/rdkit_2017_03_3. Once we
> figure out what actually happened I can maybe help you figure out how
> to fix it.
This is what I did:
module add boost # this just sets the boost stuff up
and then probably
so I may have over-egged my install-path-cake. I started all the
fiddling with DESTDIR and CMAKE_INSTALL_PREFIX, because my initial
attempt resulted in the destination directory being the same as the
build directory, which didn't work so well.
Thanks for the help - I'll have another go Python 3.5 and try to keep my
eye on __init__.py.
Dr. Loris Bennett (Mr.)
ZEDAT, Freie Universität Berlin Email loris.benn...@fu-berlin.de
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
Rdkit-discuss mailing list