alamb commented on code in PR #6940:
URL: https://github.com/apache/arrow-datafusion/pull/6940#discussion_r1264167397
##########
datafusion/physical-expr/src/array_expressions.rs:
##########
@@ -1195,21 +1195,56 @@ pub fn array_to_string(args: &[ArrayRef]) ->
Result<ArrayRef> {
}
let mut arg = String::from("");
- let mut res = compute_array_to_string(
- &mut arg,
- arr.clone(),
- delimeter.clone(),
- null_string,
- with_null_string,
- )?
- .clone();
- match res.as_str() {
- "" => Ok(Arc::new(StringArray::from(vec![Some(res)]))),
+ let mut res: Vec<Option<String>> = Vec::new();
+
+ match arr.data_type() {
+ DataType::List(_) | DataType::LargeList(_) |
DataType::FixedSizeList(_, _) => {
+ let list_array = arr.as_list::<i32>();
+ for (arr, &delimeter) in list_array.iter().zip(delimeters.iter()) {
+ if let (Some(arr), Some(delimeter)) = (arr, delimeter) {
+ arg = String::from("");
+ let s = compute_array_to_string(
+ &mut arg,
+ arr,
+ delimeter.to_string(),
+ null_string.clone(),
+ with_null_string,
+ )?
+ .clone();
+
+ if let Some(s) = s.strip_suffix(delimeter) {
+ res.push(Some(s.to_string()));
+ } else {
+ res.push(Some(s));
+ }
+ } else {
+ res.push(None);
+ }
+ }
+ }
_ => {
- res.truncate(res.len() - delimeter.len());
- Ok(Arc::new(StringArray::from(vec![Some(res)])))
+ // delimeter length is 1
+ assert_eq!(delimeters.len(), 1);
+ let delimeter = delimeters[0].unwrap();
+ let s = compute_array_to_string(
+ &mut arg,
+ arr.clone(),
+ delimeter.to_string(),
+ null_string,
+ with_null_string,
+ )?
+ .clone();
+
+ if !s.is_empty() {
+ let s = s.strip_suffix(delimeter).unwrap().to_string();
+ res.push(Some(s));
Review Comment:
I am not quite sure if it is sure that `delimiter` is always present. If it
is not, this code will panic.
Perhaps it could be something like the following to aavoid the unwrap?
```suggestion
let s = s.strip_suffix(delimeter).unwrap_or(s).to_string();
res.push(Some(s));
```
##########
datafusion/physical-expr/src/array_expressions.rs:
##########
@@ -1195,21 +1195,56 @@ pub fn array_to_string(args: &[ArrayRef]) ->
Result<ArrayRef> {
}
let mut arg = String::from("");
- let mut res = compute_array_to_string(
- &mut arg,
- arr.clone(),
- delimeter.clone(),
- null_string,
- with_null_string,
- )?
- .clone();
- match res.as_str() {
- "" => Ok(Arc::new(StringArray::from(vec![Some(res)]))),
+ let mut res: Vec<Option<String>> = Vec::new();
+
+ match arr.data_type() {
+ DataType::List(_) | DataType::LargeList(_) |
DataType::FixedSizeList(_, _) => {
+ let list_array = arr.as_list::<i32>();
Review Comment:
I don't think this is going to work (will `panic`) for `LargeList` as it
should be `i64` -- we would probably have to make a generic version of this
function.
However, for now I think it looks reasonable and a step forward to me
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
