After perusing the archives for explanations of Modeller and Python interactions, it isn't apparent to me that Modeller couldn't use a later interpreter and modules if the paths are set correctly. I understand that Modeller can be made aware of standard python 2.3 modules with the PYTHONHOME variable. However, I haven't found a clear yes or no on using a Python 2.4 standard installation with Modeller. Is it possible to disable the internal Python 2.3 interpreter linked into Modeller and force Modeller to use a system's Python 2.4 installation? If not, then I could just compile/install Python 2.3. I use SUSE 9.3 and 10.0 mainly at the moment in our lab at UNCG on all the machines so I don't have a native install of the older Python. Uninstalling Python 2.4 breaks several packages so a side by side compile/install is probably easiest. Thanks for your help, Dow
No sig.
Dow_Hurst wrote: > After perusing the archives for explanations of Modeller and Python > interactions, it isn't apparent to me that Modeller couldn't use a > later interpreter and modules if the paths are set correctly.
The Python interpreter is built into the executable, so you cannot use a different one. You can, however, use other *modules*, as long as they are compatible with the Python 2.3 interpreter. This generally means that earlier pure Python (.py) modules will work, or later .py modules that do not require any 2.4 features will also work. What will generally _not_ work, however, are any modules which have a binary component (i.e. .pyd, .dll, or .so files) as these generally require a specific Python interpreter version.
> I understand that Modeller can be made aware of standard python 2.3 > modules with the PYTHONHOME variable. However, I haven't found a > clear yes or no on using a Python 2.4 standard installation with > Modeller. Is it possible to disable the internal Python 2.3 > interpreter linked into Modeller and force Modeller to use a system's > Python 2.4 installation?
No.
> If not, then I could just compile/install > Python 2.3. I use SUSE 9.3 and 10.0 mainly at the moment in our lab > at UNCG on all the machines so I don't have a native install of the > older Python. Uninstalling Python 2.4 breaks several packages so a > side by side compile/install is probably easiest.
I certainly recommend this approach: http://salilab.org/modeller/8v2/manual/node10.html Compiling Python 2.3 from source is fairly straightforward on a modern Linux system, and you can install in /usr/local or some other location which won't conflict with the existing 2.4 install.
The next release of Modeller will be available as a 'true' Python module, so will be interoperable with any sufficiently recent Python interpreter.
Ben Webb, Modeller Caretaker
I installed the tar bzip2 source of python 2.3.5 from the main python ftp site. The install for SUSE 9.3 worked great like this:
cd /tmp bunzip2 -c python2.3.5.tar.bz2 | tar xvf - cd python2.3.5 ./configure --prefix=/usr make make install
then I did what the Modeller docs suggest and in the bash shell I was going to run Modeller from I defined:
PYTHONHOME=/usr/lib/python2.3.5
then ran Modeller 8v2 from that shell. It picked up the 2.3 based modules not built into Modeller from the external 2.3 install.
I did try using checkinstall instead of "make install" at first but it failed with errors near the end so I couldn't get an rpm of python. I'm sure an rpm could be built for someone who wanted to put in the time, but this standard method worked for me. I just wanted to post this for those who need to get this functionality working. Dow
Modeller Caretaker wrote: > Dow_Hurst wrote: >> After perusing the archives for explanations of Modeller and Python >> interactions, it isn't apparent to me that Modeller couldn't use a >> later interpreter and modules if the paths are set correctly. > > The Python interpreter is built into the executable, so you cannot use > a different one. You can, however, use other *modules*, as long as > they are compatible with the Python 2.3 interpreter. This generally > means that earlier pure Python (.py) modules will work, or later .py > modules that do not require any 2.4 features will also work. What will > generally _not_ work, however, are any modules which have a binary > component (i.e. .pyd, .dll, or .so files) as these generally require a > specific Python interpreter version. > >> I understand that Modeller can be made aware of standard python 2.3 >> modules with the PYTHONHOME variable. However, I haven't found a >> clear yes or no on using a Python 2.4 standard installation with >> Modeller. Is it possible to disable the internal Python 2.3 >> interpreter linked into Modeller and force Modeller to use a system's >> Python 2.4 installation? > > No. > >> If not, then I could just compile/install >> Python 2.3. I use SUSE 9.3 and 10.0 mainly at the moment in our lab >> at UNCG on all the machines so I don't have a native install of the >> older Python. Uninstalling Python 2.4 breaks several packages so a >> side by side compile/install is probably easiest. > > I certainly recommend this approach: > http://salilab.org/modeller/8v2/manual/node10.html > Compiling Python 2.3 from source is fairly straightforward on a modern > Linux system, and you can install in /usr/local or some other location > which won't conflict with the existing 2.4 install. > > The next release of Modeller will be available as a 'true' Python > module, so will be interoperable with any sufficiently recent Python > interpreter. > > Ben Webb, Modeller Caretaker