Closed mdsumner closed 1 year ago
Thanks for this suggestion. If you'd like to submit this via a pull request, making sure that RTriangle's tests are passed, I'll take a look at it. (I might take a look at it eventually, but the pull request would speed up matters.)
I've now tested the patch, also on my Retistruct revdep. Are there any other speed improvements you can think of before I submit to CRAN?
Possibly the use of anyNA for the missing value checks in triangulate(), but I haven't checked - nothing else obviously stands out .
Thanks!
This was submitted to CRAN on 2018-01-30, so closing the issue.
anyDuplicated.matrix
is quite slow, as it usesapply()
, which meanspslg
is a fair bit slower than the actualtriangulate
call. There's a substantial speed up in using the.data.frame
method rather than.matrix
, as inpslg2
below.I think there are faster ways still to speed this up, and while it's easy enough to construct the pslg manually with my own validations, this might be a useful modification - thanks!