[ET Trac] [Einstein Toolkit] #469: Improve default directory locations

[Einstein Toolkit]

#469: Improve default directory locations
-------------------------+--------------------------------------------------
 Reporter:  hinder       |       Owner:  eschnett
     Type:  enhancement  |      Status:  new     
 Priority:  minor        |   Milestone:          
Component:  SimFactory   |     Version:          
 Keywords:               |  
-------------------------+--------------------------------------------------
 It is usually desirable to set the user's source base directory to be in
 their home directory.  On many systems, this can be inferred from the user
 name; e.g. /home/@USER at .  However, on systems with very large numbers of
 users, the home directory is sometimes not predictable in this way.  For
 example, /home1/00915/@USER@ (this is on LoneStar), where the numbers are
 different for different users.  This also happens on Kraken (see ticket
 #382).  Having to specify these details manually for each machine that you
 want to use is an extra step in configuring SimFactory which we would like
 to avoid.

 Option 1: Provide a @HOME@ definition which is determined automatically
 from the user's home directory.

 This would be possible, but it only solves the problem for the home
 directory.  On some systems (again, LoneStar and Kraken), the home
 directory might not be large enough to use as a source base directory, and
 instead you might want to use the "work" directory instead (this is the
 current simfactory setting for LoneStar).  This directory also suffers
 from the same unpredictable numbering problems.  As the home directory is
 available in the HOME environment variable, the work directory is
 available in the WORK environment variable.

 Option 2: Provide a general method for accessing environment variables and
 storing them in definitions.  For example, you could write @ENV(HOME)@ or
 @ENV(WORK)@.  This requires the environment to have been correctly set up
 (for remote operation, this requires a login shell, but I think we have
 that already).  In fact it looks like there is preliminary code in
 simsubs.py, but it is incomplete and is not called from anywhere.

 I would favour option 2.  Thoughts?

-- 
Ticket URL: <https://trac.einsteintoolkit.org/ticket/469>
Einstein Toolkit <http://einsteintoolkit.org>
The Einstein Toolkit