Closed qwhelan closed 1 year ago
@eprbell I implemented your suggested changes and CI is passing.
I'll probably take a stab at the rp2
version of this change but have some unsubmitted dali
changes I'd like to focus on first.
@eprbell Updated the files you flagged but kept src/stubs/ezodf/document.pyi
's header the same as git is treating it as a rename rather than a new file. Let me know if that one needs fixing.
dali
currently outputs a large fraction of empty rows, whichrp2
then processes individually and potentially outputs a log message per empty line. Assumingn
transactions per asset andk
assets, the currentdali
output isk * n + 40
rows per asset/sheet and a total ofk^2 * n + 40k
rows per document. Thus, only1/k
rows are populated andrp2
runtime explodes while doing useless work.The above actually is the best case scenario, as it is even worse when the distribution of transactions per asset is unevenly distributed.
I have included a test case but it is a bit of a no-op as I did not want to touch the golden datasets and the existing code passes without this change. Please let me know if you have any suggestions on the preferred testing approach in this case.
In order to make
mypy
happy, I also updated theezodf
stubs.