Strange ds/head and ds/tail behaviour in v7 about tech.ml.dataset HOT 6 CLOSED

The fix is really two-fold. First, select-rows should clear specifically print-index-range. Second, printing itself should never cause that exception so it needs to always filter print-index-range to be in the range of the current dataset row indexes.

One thing you may really like - select-rows now accepts negative indexes which will index from the end python-style.

from tech.ml.dataset.

And maybe a better example of the issue:

(def x (ds/->dataset {:a (range 1000) :b (range 1000)}))
=> #'jasmine.var/x
(ds/select-rows (ds/tail x 5) [0 1])
Error printing return value (IndexOutOfBoundsException) at ham_fisted.ChunkedList/sublistCheck (ChunkedList.java:216).
End index out of range: end-index(5), n-elems(2)

from tech.ml.dataset.

That is the error I get - index out of bounds - it was because the head/tail operations put print-index-range metadata on the returned dataset and the select-rows operation didn't clear the metadata (which in general it should not). So when you go to print the third, 2 row dataset it has print-index-range of (range 5) from the head operation :-).

from tech.ml.dataset.

New release is up :-)

from tech.ml.dataset.

thanks!!

from tech.ml.dataset.

Wow! Nice!

from tech.ml.dataset.