Some things sitting around uncommitted for a while
Some additional type hint cleanups/fixes
One (probably never-hit) bug fixed: "lines_between" had a logic error that would not search for the start marker if the end marker was found before it
Reduced memory consumption slightly by not caching the xml element on every TwincatItem
Add EnumerationTextList with get_source_code support; as previously these translatable types were missing
Parser classes background
So there are some unfortunate things in the way that the parser breaks out TwincatItem classes. Here's the background on why the parser items are as they are:
I wanted easy mappings of TwinCAT <xml-tag> to Python class, where methods and properties could then be context-specific (e.g., give me the source code of the GVL)
I wanted things like gvl.Declaration or plc.TcSmItem.PlcProject.Mappings to be a thing you could do, since I thought __getitem__-based plc["TcSmItem"]["PlcProject"]["Mappings"] was verbose and ugly
I was unaware of which XML items would be able to have more than one entry in a given scope. (In retrospect, TwinCAT has xsds that provide this information)
This led to the current (ugly) structure of plc.TcSmItem[0].PlcProject[0].Mappings[...], along with all of the code that grabs the only element from an assumed single-element list
What about better type hints?
Now, type hints weren't really a thing when pytmc was getting on its feet, at least with the versions we used and whatnot.
So type hints in TwincatItem subclasses are unfortunately really bad, given how the classes are structured:
class EnumInfo(_TmcItem):
# Some have type hints like this now
Comment: list
# What we really want to say to describe the structure as-is
Comment: Optional[typing.List[Comment]]
# With a refactor to de-listify single items, this would be ideal
Comment: Optional[Comment]
The latter two appear to be circular references rather than one to a corresponding Comment class.
Where do we go from here? Though I'd really like to see this improved, I guess we just leave it as-is until a real refactor could be justified - unless anyone has a better idea.
Changes
EnumerationTextListwithget_source_codesupport; as previously these translatable types were missingParser classes background
So there are some unfortunate things in the way that the parser breaks out
TwincatItemclasses. Here's the background on why the parser items are as they are:<xml-tag>to Python class, where methods and properties could then be context-specific (e.g., give me the source code of the GVL)gvl.Declarationorplc.TcSmItem.PlcProject.Mappingsto be a thing you could do, since I thought__getitem__-basedplc["TcSmItem"]["PlcProject"]["Mappings"]was verbose and uglyplc.TcSmItem[0].PlcProject[0].Mappings[...], along with all of the code that grabs the only element from an assumed single-element listWhat about better type hints?
Now, type hints weren't really a thing when pytmc was getting on its feet, at least with the versions we used and whatnot. So type hints in
TwincatItemsubclasses are unfortunately really bad, given how the classes are structured:The latter two appear to be circular references rather than one to a corresponding
Commentclass.Where do we go from here? Though I'd really like to see this improved, I guess we just leave it as-is until a real refactor could be justified - unless anyone has a better idea.