Skip to content

Commit a545faa

Browse files
committed
Document that strided buffers do not need to be supported
1 parent f9f259c commit a545faa

File tree

1 file changed

+7
-1
lines changed

1 file changed

+7
-1
lines changed

protocol/dataframe_protocol_summary.md

+7-1
Original file line numberDiff line numberDiff line change
@@ -172,7 +172,13 @@ We'll also list some things that were discussed but are not requirements:
172172
3. Extension dtypes, i.e. a way to extend the set of dtypes that is
173173
explicitly support, are out of scope.
174174
_Rationale: complex to support, not used enough to justify that complexity._
175-
4. "virtual columns", i.e. columns for which the data is not yet in memory
175+
4. Support for strided storage in buffers.
176+
_Rationale: this is supported by a subset of dataframes only, mainly those
177+
that use NumPy arrays. In many real-world use cases, strided arrays will
178+
force a copy at some point, so requiring contiguous memory layout (and hence
179+
an extra copy at the moment `__dataframe__` is used) is considered a good
180+
trade-off for reduced implementation complexity._
181+
5. "virtual columns", i.e. columns for which the data is not yet in memory
176182
because it uses lazy evaluation, are not supported other than through
177183
letting the producer materialize the data in memory when the consumer
178184
calls `__dataframe__`.

0 commit comments

Comments
 (0)