There are tools that make it *possible* to productionize this work. But "is possible" isn't the benchmark for "is a best practice."
-
-
Replying to @EmilyGorcenski
I'm really hoping that the next few years of datasci can include an evolution towards better engineering support.
2 replies 2 retweets 4 likes -
Replying to @EmilyGorcenski
"Go read docs to figure out which library this routine is in" is fine for academic stuff but does not fly for engineering work.
4 replies 1 retweet 4 likes -
Replying to @EmilyGorcenski
Imagine if core file IO functions were spread across 5 different libraries. That's the status of datasci right now.
2 replies 3 retweets 8 likes -
Replying to @EmilyGorcenski
There is just no good reason for PCA to be in scikit-learn and SVD to be in numpy. They're the same damn thing modulo a symmetrization op.
3 replies 0 retweets 8 likes -
Replying to @EmilyGorcenski
I've wondered the same thing. Any hints of an effort to rationalize the ecosystem?
2 replies 0 retweets 0 likes -
Replying to @gp_wisconsin @EmilyGorcenski
@NumFOCUS is step in right direction. most Python datasci projects are all volunteer. Some things require FTEs to get done right.1 reply 0 retweets 3 likes -
Replying to @tacaswell @gp_wisconsin and
also consider the cost of "rationalizing". can not break back compatibility on where things are now.
2 replies 0 retweets 0 likes -
why? status quo is everything is ad hoc.
1 reply 0 retweets 0 likes -
because that ad hoc-ness is encoded in millions of lines of user code.
1 reply 0 retweets 0 likes
using tools they don't intend to upgrade anyway
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.