JSON Input: Constraints about geometric HOT 8 CLOSED

from geometric.

Can you use:

if isinstance(constraint, str):
    ...
elif isinstance(constraint, (tuple, list)):
    ...
else:
   raise TypeError(...)

from geometric.

I think that could work - @yudongqiu , any thoughts?

Thanks,

• Lee-Ping

from geometric.

I think list of tuples can work, dictionary may also work, the main thing is we need to be careful that it fully reproduces the original functionality.

from geometric.

Currently, an example of the constraints input string looks like this:

$freeze
xyz 1-5

$set
angle 2 1 5 110.0

# A comment
$scan
distance 2 1 1.0 1.2 3 # Another comment
dihedral 1 5 6 7 110.0 150.0 3

This corresponds to a dictionary that looks like this:

{ 
  'freeze' : [('xyz', '1-5')], 
  'set'    : [('angle', '2', '1', '5', '110.0')], 
  'scan'   : [('distance', '2', '1', '1.0', '1.2', '3'), 
              ('dihedral', '1', '5', '6', '7', '110.0', '150.0', '3')] 
}

I think we can easily have the JSON constraint input be a dictionary like the one above. It's much easier if we require all of the elements in each tuple to be "words" (i.e. strings without spaces). If so, all we need to do is write an extra function to convert the dictionary to the constraints input string and pass that to optimize.py::ParseConstraints() without modifying the latter.

from geometric.

I understand the alternative is to rewrite optimize.py::ParseConstraints() so that it supports multiple input types, but I don't see how that is possible without substantially restructuring it. That function is quite complicated because it supports simultaneous freezing / setting / scanning of various combinations of constraints.

from geometric.

@leeping Thanks a lot for the clear specs of the new constraints input format. I will implement that and make a PR soon.

from geometric.

Addressed in PR #27 .

from geometric.