Support for OMega's UFO functionality

[kilian]

The existing API for UFO models handled by O'Mega has to be supported by WHIZARD (and Sindarin). The necessary ingredients appear to be present now.

[kilian]

Related to #56 and #59.

[kilian]

Thorsten's UFO interface is already functional (some minor issues remaining); calling the interface works on the unit-test level. The main issue is proper treatment of the generated model (it appears as if the .mdl file has to be generated before the process is handled, actually already in the compile pass of Sindarin) ... otherwise the model variables cannot be recognized by Sindarin.

[kilian]

Contrary to my previous statement, the syntax model = SM { ?ufo=true } won't work. Models have to be read on the first pass of the script, so they can provide their model variables to the interpreter. Variable values are not available in the first pass.

So, why not consider ufo as a scheme. (We can reasonably expect that the UFO format itself won't support schemes in the way we introduced this feature. Or we can allow for two model arguments.)

The syntax will thus be model = SM (ufo). Simple enough.

I'll have to take some precautions that the UFO-generated SM doesn't clash with the built-in SM. This is a danger for the test suite.

[Juergen Reuter]

I actually like the new syntax <model> (ufo). However, does this mean that there always needs to be a model <model>.mdl be available inside WHIZARD? This would be a killer. If not, then why not use a different name for the tests like SM_UFO (ufo)? Or is it TO's UFO testfile that dictates this?

[kilian]

UFO and non-UFO models should not know about each other, even if the names are identical. We could distinguish UFO models by a suffix in their name, but I think it's more future-proof if we don't do this. We don't know whether UFO will become standard when using Whizard, or not. So keep the user syntax simple.

Models in memory do know whether they are UFO or not.

I'm currently looking at the actual model files. I may have a different suffix .ufo.mdl in the file name, but not in the model name. If the distinction is after the dot, no problem.

[kilian]

UFO handling works in the testsuite. Missing parts:

• some problem with SM parameters in the test model
• make installcheck doesn't work because the test model is not installed.
• The working directory is not yet searched for user models.

[kilian]

The directory issues are solved now. Last issue is numerical agreement.

For searching for the a UFO sources of a model, WHIZARD looks at

1. the working directory. If model = Foo (ufo) is requested, the Python files have to reside in subdirectory Foo.
2. the directory share/models/UFO, again inside subdirectory Foo.

The functional tests use the working directory.

[kilian]

The Whizard side of this is complete now; some validation still under way on the O'Mega side.

[Juergen Reuter]

This is finished now, and the UFO SM as well as SM-like models via UFO are completely supported. Completely general UFO models are deferred to #56 and #142.