New method to add parameters

[dlakaplan]

Normally adding new parameters is a pain. new Parameter instances have to be created with the correct class and arguments, new components may be needed, etc.

When reading a model from a file, though, you can use overrides like:

m=get_model(filename, F2=1)

This implements a similar ability for adding to an existing model. e.g.,

m=get_model(filename)
...
m.add_params(F2=1, JUMP="FREQ  1400.0 2000.0 0.1 0")

will add F2 and JUMP1, even if no JUMP component exists. Note that this will not override existing values since there are already plenty of ways to do that (and will raise an exception in those cases).

[abhisrkckl]

Can you clarify what happens in the following scenarios?

1. A binary parameter is added this way, but there is no binary component in the TimingModel?
2. A binary parameter is added this way, but it is not supported by the binary component in the TimingModel (e.g., EPS1 into DD)?
3. F3 is added, but there is no F2 in the model?
4. "EFAC -f abc " is added, but the same EFAC is already present.
5. Can non-numerical parameters (e.g. EPHEM, CLOCK etc) be added like this?

[abhisrkckl]

Please add a test with FDJUMPDM. It is a weird parameter implementation-wise and may lead to pathological cases.

[abhisrkckl]

Also call TimingModel.validate_toas() after adding the parameters.

[dlakaplan]

Can you clarify what happens in the following scenarios?

1. A binary parameter is added this way, but there is no binary component in the TimingModel?

This fails. I think that makes sense, but let me know if not. I can also make that explicit in the docs.

2. A binary parameter is added this way, but it is not supported by the binary component in the TimingModel (e.g., EPS1 into DD)?

That fails. I think that is sensible.

3. F3 is added, but there is no F2 in the model?
   This fails because of standard PINT checking (need to add prefix parameters sequentially). This is tested in tests/test_modeloverride.py::test_paradd_fails[F3--2e-14]

4. "EFAC -f abc " is added, but the same EFAC is already present.
   It is only added once. Not sure what the expected behavior should be here.

5. Can non-numerical parameters (e.g. EPHEM, CLOCK etc) be added like this?

Yes, although not those examples since they are already present. But for the examples like EFAC and JUMP this works OK.

[dlakaplan]

Please add a test with FDJUMPDM. It is a weird parameter implementation-wise and may lead to pathological cases.

Added

[dlakaplan]

Also call TimingModel.validate_toas() after adding the parameters.

How would that work? Doesn't that need TOAs to be defined? This can be done with just a model.

[abhisrkckl]

I think this is ready to go. Shall I merge this?

[dlakaplan]

I think it's fine, unless you think it would help to explain some of those other scenarios in the docs

[abhisrkckl]

Please add a test to check if adding a binary parameter without a binary component fails as expected.

[abhisrkckl]

Is this required?

[dlakaplan]

no, that was some stupid autocomplete insertion. removed.

[abhisrkckl]

I think it's fine, unless you think it would help to explain some of those other scenarios in the docs

I think it will be helpful to explain what scenarios are invalid in the dosctrings.

[dlakaplan]

Please add a test to check if adding a binary parameter without a binary component fails as expected.

Added.

[abhisrkckl]

I will merge this if it is ready.

[dlakaplan]

I will merge this if it is ready.

Had to push new updates.

[abhisrkckl]

Ok I will merge once the tests pass.

[dlakaplan]

Can you merge?