Problem description

As observed in #1112 (and likely elsewhere), some of the human-readable output generated by YAML emitters could be improved.

Steps to reproduce

The YAML output below can be produced using the new save method proposed for 1D objects (#1112), but the underlying implementation predates this PR (goes back to #984).

Behavior

There are three formatting issues documented thus far:

(1) Scalars / Rounding

As already pointed out in #984 (where serialization was initially introduced), there are some "floating point rounding issues [that] result in kind of ugly values". For example:

    tolerances:
      steady-abstol: 1.0e-09
      steady-reltol: 1.0e-04
      transient-abstol: 9.999999999999999e-12 # <-- should be 1.0e-11
      transient-reltol: 1.0e-04

I have not looked into details, but at least for scalar values there may be a way to catch these items.

(2) Vectors / Efficient use of space

Number blocks don't use horizontal space efficiently; this may be addressed by changing:

    velocity: [0.7112641405849059, 0.7112641405822596, 0.7112641404451637,
    0.7112641377426332, 0.7112641104415781, 0.7112637339406135,
    0.7112617435453226, 0.7112560064702751, 0.7112322153842345,
    ...
    4.055246182368149, 4.055246182377726]

to

    velocity: [
      0.7112641405849059, 0.7112641405822596, 0.7112641404451637, 0.7112641377426332, 
      0.7112641104415781, 0.7112637339406135, 0.7112617435453226, 0.7112560064702751, 
      0.7112322153842345, 0.7111349509095231, 0.7109781680863129, 0.7106086071388092,
      ...
      4.055246182368149, 4.055246182377726]

which uses space more efficiently and also improves readability from a human perspective. Beyond, there is likely an optimized number of characters per line that prevents ragged output in most cases.

System information

• Cantera version: 2.6
• OS: any
• Python/MATLAB/other software versions: any

Additional context

• PR Enable serialization of Cantera objects #984
• PR Use YAML for saving and loading 1D flame simulations #1112