Closed ue71603 closed 2 months ago
Your approach loading in a single tree, in order to have it available, this obviously makes sence for smarter access and limited code changes, but how the code now works is trying to minimize the in memory usage therefore it uses xpath to parse a file. I think it is a better approach to have the code in some way 'aware' what tree to use for which datatype. So I think the idea to make a network-delivery from a line-delivery is a sound but bloated way.
I am not sure, that this will help in some cases (e.g. for the Interchanges, when they are only in one tree). Also some transforms need to have a lot of the things in memory otherwise they can't optimise the output (e.g. reuse of AvailabilityConditions).
What could work would be a database / tree that stores: (class,id,version,file) But then one would need to optimise access by sorting them by file for each needed part) Algorithm:
netex-to-db.py has the ability to transparently process zip files, xml files, and gzipped xml files.
I think I need some progress for the Swiss files.
When do you think that you could propose a db based solution.
For the current Swiss files there could be an intermediate step:
Currently we have
I suggest to rewrite any-toepip.py for the time being to:
l``` load common frame and add it to aux_tre load resource frame and add it to aux_tre load site frame and add it to aux_tre load service frame and add it to aux_tre load service calendar frame and add it to aux_tre for each timetable file load time table file convert(aux_tree, timetable file)