Loading arxml with Ethernet data Performance is extremely bad on development branch(2023-12-15) about canmatrix

[ebroecker]

Hi @xRowe

Thanks for the analysis.

I had already a look but did not yet find the patch witch made it worse. I'll keep on searching

[xRowe]

Hello @ebroecker

From my test, I found the below code takes much of time.
https://github.com/ebroecker/canmatrix/blob/d9902ace093fc1bb31af02599243847b529b0b44/src/canmatrix/formats/arxml.py#L1135C1-L1141C21

If I comment the code, the perfomance will be good.

Sorry, I am new on arxml, could not give some further investigate now

Thanks,
Rowe

[ebroecker]

Hi @xRowe
Thank you for this analysis.
So I know where to look at...

[ebroecker]

Hi @xRowe,

could you try branch https://github.com/ebroecker/canmatrix/tree/issue_750

pip install git+https://github.com/ebroecker/canmatrix/tree/issue_750

Maybe this little fix speeds up the parsing again

[xRowe]

Hello @ebroecker
Thanks for your work,
As I check, Branch issue_750 created from official v1.0, with which cannot parse my arxml, but the latest development can do.
Traceback as below

Traceback (most recent call last):
  File "d:\Users\aaaa\Desktop\arxml\aaaa.py", line 45, in <module>
    dbc = load(arxml, decode_ethernet=True)
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "d:\Users\aaaa\Desktop\arxml\canmatrix\formats\arxml.py", line 2094, in load
    source_signal = ea.selector(source_triggering[0], "/I-SIGNAL-REF")
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "d:\Users\aaaa\Desktop\arxml\canmatrix\formats\arxml.py", line 306, in selector
    return sorted(result_list, key=lambda element: element.sourceline)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "d:\Users\aaaa\Desktop\arxml\canmatrix\formats\arxml.py", line 306, in <lambda>
    return sorted(result_list, key=lambda element: element.sourceline)
                                                   ^^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'sourceline'

So I update the commited code to the latest development branch and have a try, below are the test result:
Time consumption: 2 ~ 2.1s
vs
https://github.com/ebroecker/canmatrix/blob/d9902ace093fc1bb31af02599243847b529b0b44/src/canmatrix/formats/arxml.py#L1135C1-L1141C21 is commented
Time Consumption: 1.7 ~ 1.8s

it is getting the signal.receivers

[ebroecker]

Hi @xRowe

thank you for testing!
issue_750 should be created from development, but I just now did an update of the branch. It should now depend on current development branch.

But anyway, you tried it, and it seems to fix the time-issue (more or less).
I'll do some cleanup and then merge it into development.