milesgranger commented on code in PR #14395:
URL: https://github.com/apache/arrow/pull/14395#discussion_r1007723825
##########
cpp/src/arrow/compute/kernels/scalar_nested.cc:
##########
@@ -87,6 +89,206 @@ Status GetListElementIndex(const ExecValue& value, T* out) {
return Status::OK();
}
+template <typename Type, typename IndexType>
+struct ListSlice {
+ using offset_type = typename Type::offset_type;
+
+ static Status Exec(KernelContext* ctx, const ExecSpan& batch, ExecResult*
out) {
+ const auto opts = OptionsWrapper<ListSliceOptions>::Get(ctx);
+
+ // Invariants
+ if (opts.start < 0 || (opts.start >= opts.stop && opts.stop != -1)) {
+ // TODO: support start == stop which should give empty lists
+ return Status::Invalid("`start`(", opts.start,
+ ") should be greater than 0 and smaller than
`stop`(",
+ opts.stop, ")");
+ }
+ if (opts.step != 1) {
+ // TODO: support step in slicing
+ return Status::NotImplemented(
+ "Setting `step` to anything other than 1 is not supported; got
step=",
+ opts.step);
+ }
+ if (opts.stop == -1) {
+ // TODO: Support slicing to arbitrary end
Review Comment:
I did see the default `None` of SliceOptions then is represented as the max
int64. I think this doesn't match for this case for at least two reasons:
- If user wanted fixed size output, then it's ambiguous if they literally
put in the max int value, or if this is slice to end of the largest list item;
`-1` makes that distinction.
- `-1` is a fairly common flag to say "use everything", from how many cpu
cores to use or similar.
> ... if you slice until after the end, does it also get filled with nulls,
or do we automatically slice until the end (like how Python slicing works)?
I think this would depend on the value of `return_fixed_size_list`
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscr...@arrow.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
