Michael Behrisch
Michael Behrisch
Now the new numbers after some optimizations and the fix for csv.gz: format|write|size | pandas.read | parse_fast_nested ------|-----:|--------: | ---: | ---: fcd.xml|12s|357MB | 11s | 8s fcd.csv|12s|194MB | 2s...
Things to discuss - do we need (configurable) compression? - what about duplicate attributes (column names) - always prefix with element name (like xml2csv.py does it) - this would not...
> Now I can run the simulation, but the fcd output is empty. I think there is conflict with my other sumo build, because the columns are not all str....
A proposal to solve open things after discussion with @namdre - handling duplicate attributes should be configurable (default always prefix) - float32 is probably enough but maybe x, y should...
Does a simple `sumo` call report something like: ``` Eclipse SUMO sumo Version v1_23_1+0790-d6e148261c2 Build features: Linux-6.8.0-53-generic x86_64 GNU 13.3.0 Release FMI Proj GUI FMT Intl SWIG tcmalloc Parquet Eigen...
OK it seems to be the correct version. Very strange. And the same call with generating XML or CSV works fine?
OK, I will try to get my hands on a mac. Just to make sure, you are installing arrow via homebrew and it is version 20.0.0?
I don't think so, parquet does not use fmt, it is only used by XML and CSV. You could try whether CSV output works just to have one more datapoint.
I added a small test into the sumo test suite which at least runs successfully on the last github CI. You can see the results and download the outputs here:...
Please retry whether it works if you install libsumo via pip