implement searchsorted for multindex (bug #14833) #23210
Conversation
|
Hello @btel! Thanks for updating the PR.
Comment last updated on October 17, 2018 at 19:42 Hours UTC |
Codecov Report
@@ Coverage Diff @@
## master #23210 +/- ##
==========================================
+ Coverage 92.19% 92.19% +<.01%
==========================================
Files 169 169
Lines 50954 50961 +7
==========================================
+ Hits 46975 46982 +7
Misses 3979 3979
Continue to review full report at Codecov.
|
jreback
left a comment
jreback
left a comment
There was a problem hiding this comment.
I am not convinced of the utility of using searchsorted here, what are you going to do with this?
| return np.lib.arraysetops.in1d(labs, sought_labels) | ||
|
|
||
| def searchsorted(self, arr): | ||
| dtype = [l.dtype.descr for l in self.levels] |
There was a problem hiding this comment.
this will be much more performant to do this level by level (as you iterate over the key). which btw must be fully specified. needs a doc-string and error checking.
| result = midx.get_indexer(midx) | ||
| tm.assert_numpy_array_equal(result, np.arange(9, dtype=np.intp)) | ||
|
|
||
|
|
There was a problem hiding this comment.
needs error checking, and a full-on test where the element is there (to be honest you should simply do a .get_loc()) first to find it then if not found do the searchsorted. You must also assert that the index is lexsorted as well.
jreback
added
MultiIndex
Indexing
|
cc @toobaz |
|
Can you merge master and address comments? |
|
closing as stale, but ping to reopen if can continue |
4 participants
git diff upstream/master -u -- "*.py" | flake8 --diffProof-of-concept to fix #14833.
This works: