Late night coding session. Trying to figure out a workaround for plotting a very large (226 million) array of structure members without running into memory trouble. Somehow discovered a class hiding in the library API for getting the data in sections, and, WTF, worked on the first try. And memory error avoided, woot!
Friday afternoon and late night were a fun experience of updating code to 1) override the code signature of a newly introduced 3rd-party native binary included in a Java library, and 2) extend a JNI class to add a new filechooser method. And I mean "fun" in the most enjoyable nerdy way, because I haven't been there in a while. Just working through both issues seems like work getting done, even if no one else will be able to use the result for (just, I hope) a couple weeks.
And if anyone noticed the #netcdf / CDL syntax highlighting, I have a couple of things to tick off before I do a proper launch in the very near future, but see here: https://github.com/jatkinson1000/tree-sitter-cdl
👀
Yay! Just discovered via Github that Unidata received its latest funding tranche and their developers are back to work!
I am really looking forward to a time when scientific data analysis is less of a constant fuckaround and fight with technical bullshit. I'd *really* like
- #netCDF natively supporting complex numbers
- #Python #xarray and #pandas to natively support physical units (#pint is great on its own but the integrations leave a LOT to be desired)
- #Jupyter notebooks to suck less (crashes, glitches, widget plots not saved statically, an effing BUILTIN formatter, etc.)
- proper data pipeline systems
...