OMParser update

[syntron]

update OMParser; mainly just change code updates to get it safe for reuse in ModelicaDoE usecases

• please check as I do not have all details of the code workflow
• should be after PR F001 restructure cleanup #444 but it is save anytime as there is no other PR touching OMParser.py

[syntron]

@adeas31 The current code still uses Python2 definitions - thus it fails:

    def formatRecords(strings):
        result['RecordResults'] = {}
        recordName = strings[strings.find("record ") + 1:strings.find("\n")]
        recordName = recordName.replace("ecord ", '').strip()
        strings = strings.replace(("end " + recordName + ";"), '').strip()
        recordItems = strings[strings.find("\n") + 1: len(strings)]
>       recordItems = recordItems.translate(None, "\\")
E       TypeError: str.translate() takes exactly one argument (2 given)

Should it be converted to Python3 at all or removed? This PR tries to update the code to get it running (see: 1b592c6 for Python2 => Python3 conversion) and thread save but is it still be required or should it be completly removed? Overall it was not working for some time / any Python3 usage

[adeas31]

As far as I can see the code is used to format the records, the record returned by simulate API is handled via formatSimRes. The other APIs that return a record are getMessagesStringInternal and checkSettings or maybe more. You can test getMessagesStringInternal and see if its result record is formatted correctly.

I recommend converting the code to python 3 and adding a test for it.

[syntron]

@adeas31 I'm not that familar with the OM return values to check this in detail; regarding OMParser, I noted that it it does not return the expected results:

running test_everything from test_OMParser (currently commented out; content of results)

{'RecordResults': {',5.9,6,NONE ( )},record AB': '"', 'startTime': 'ErrorLevel.warning', "'stop*Time'": 'SOME(1.0', 'nd ABC;}': 'end ABC;}', 'RecordName': ''}}

the same test in test_OMTypedParser:

(1.0, ((1, True, 3), ('4"\n', 5.9, 6, None), {'startTime': 'ErrorLevel.warning', "'stop*Time'": 1.0}))

It is the same input data to parse! The changes in this PR have no effect on results of test_OMParser.test_everything() - it is the same results data before / after applying it; just 675997b is always needed to fix Python 2 code.

Input data as reference:

    testdata = """
    (1.0,{{1,true,3},{"4\\"
",5.9,6,NONE ( )},record ABC
    startTime = ErrorLevel.warning,
    'stop*Time' = SOME(1.0)
end ABC;})
    """

[adeas31]

Can't we simply say Python >= 3.10 or Python >= 3.12 supported. Do we need to be verbose here?

[syntron]

we can go with Python >= 3.10; however, the OMPath functionality is only supported for Python >= 3.12 - Python 3.10 has only a compatibility layer to get something working

PS: this should be in the discussion for PR #444

[adeas31]

Can this be merged now or does it depend on any other PR?

[syntron]

@adeas31 it can be merged as I hope to not change any functionality; however, I did only some basic tests as I have no real use cases for this part of the code ...

PR #444 should be run before - it is just some simple / small cleanup ...; else it is included in this PR ...

[syntron]

I just run some checks; the return values from OMParser would crash OMPython ...

Example from run of test_ModelicaSystemOMC.test_getSolutions():

print(expr) => buildModel(M, variableFilter=".*")
print(result) => {"/tmp/88ab0555-6bbe-4470-b43e-2d59df7e50d4/M", "M_init.xml"}
print(om_parser_typed) => ('/tmp/88ab0555-6bbe-4470-b43e-2d59df7e50d4/M', 'M_init.xml')
print(om_parser_basic) => {'SET1': {'Values': ['"/tmp/88ab0555-6bbe-4470-b43e-2d59df7e50d4/M", "M_init.xml"']}}

The data expected by the code is given by om_parser_typed; om_parser_basic would crash the code if returned; nevertheless, it contains the same / the expected information ...

Did I missed some additional conversion, i.e. SET1 => set()?

(test run with both versions of the code - same results!)

[sjoelund]

If we do remove the untyped API from OM finally, the parsers aren't even needed. We could switch to --interactiveDumpFormat=json. Older OMC versions still need a parser though :(