On Mar 14, 2011, at 4:04 PM, Jed Brown wrote: > I want to reiterate a statement I made a while back that any alternative > input format really needs to support references. This is a good reason to > prefer Yaml over JSON. Yaml is a superset of JSON, but offers a more readily > human-readable layout, example: http://www.yaml.org/start.html > > The important feature here is that references can be reused so it becomes > possible to define multiple solver configurations, then use each one (perhaps > with local modifications) in multiple places or not at all. If you just do > plain JSON, it's essentially the same expressiveness as current options > files, albeit easier to manipulate from other languages.
I fully agree on the principle, but the scope is beyond what I can reasonably get done in a limited time. To me, "easier to manipulate from other languages" is already a huge improvement. I have folders with 100's of computations corresponding to different sets of parameters. I need an easier way to import these simulation parameters in scripts. And I need it yesterday, not next year. To be honest windows style .ini files would be fine for what I need to do. JSON is just a bit more hip. Blaise -- Department of Mathematics and Center for Computation & Technology Louisiana State University, Baton Rouge, LA 70803, USA Tel. +1 (225) 578 1612, Fax +1 (225) 578 4276 http://www.math.lsu.edu/~bourdin -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20110314/0748616a/attachment.html>
