Closed cgiachalis closed 8 months ago
Another excellent catch.
Let me check with the documentation to see what C++, Python, ... do so that I can align accordingly.
Not entirely clear from looking at https://docs.tiledb.com/main/how-to/object-management so I think i wil just do a quick clean-up along the lines you suggested.
It'll also look nicer once TYPE is short. Quick before with a standard single-cell data set I have here.
Excellent! Thanks for your prompt response.
After the fix:
Need to now ponder if this warrants an 'interface change' (I lean towards no) and what it may do to tests. Apparently nothing (good !) but I seem to have left the last PR with a missing re-roxygenize run I may wrap into this.
I will make this two PRs just to be cleaner.
Nice. Is it good time to ask if labels
are coming to the R client? :)
Oh..you might want to update the LICENSE year https://github.com/TileDB-Inc/TileDB-R/blob/d796865475ae79c8d63bb4355d39c65bdcec9604/LICENSE#L1
Wow. Another one.
Nice. Is it good time to ask if
labels
are coming to the R client? :)
Err, what labels
? Am I behind some API offerings not covered?
"Interesting"
I don't see it at https://docs.tiledb.com/main/how-to/arrays/creating-arrays/creating-the-array-schema but will poke.
Might be still experimental?
Apparently, there is a header dimension_label_experimental.h
. Will peruse / discuss. Thanks for the heads-up!
In short, in the output dataframe
TYPE
column holds the uris andURI
column the types. Docs states a dataframe with object type, object uri string columns, so the contents has to be swapped.The root cause is in
libtiledb_object_walk
: https://github.com/TileDB-Inc/TileDB-R/blob/d796865475ae79c8d63bb4355d39c65bdcec9604/src/libtiledb.cpp#L4327-L4328If we swap uris with types we get the desired result:
Let me know if you wish to see a PR.