alamb commented on a change in pull request #812:
URL: https://github.com/apache/arrow-datafusion/pull/812#discussion_r682777371
##########
File path: datafusion/src/physical_plan/hash_utils.rs
##########
@@ -245,9 +249,60 @@ macro_rules! hash_array_float {
};
}
-/// Creates hash values for every row, based on the values in the columns
+/// Hash the values in a dictionary array
+fn create_hashes_dictionary<K: ArrowDictionaryKeyType>(
+ array: &ArrayRef,
+ random_state: &RandomState,
+ hashes_buffer: &mut Vec<u64>,
+ multi_col: bool,
+) -> Result<()> {
+ let dict_array =
array.as_any().downcast_ref::<DictionaryArray<K>>().unwrap();
+
+ // Hash each dictionary value once, and then use that computed
+ // hash for each key value to avoid a potentially expensive
+ // redundant hashing for large dictionary elements (e.g. strings)
+ let dict_values = Arc::clone(dict_array.values());
+ let mut dict_hashes = vec![0; dict_values.len()];
+ create_hashes(&[dict_values], random_state, &mut dict_hashes)?;
+
+ // combine hash for each index in values
+ if multi_col {
+ for (hash, key) in
hashes_buffer.iter_mut().zip(dict_array.keys().iter()) {
+ if let Some(key) = key {
+ let idx = key
+ .to_usize()
+ .ok_or_else(|| {
+ DataFusionError::Internal(format!(
+ "Can not convert key value {:?} to usize in
dictionary of type {:?}",
+ key, dict_array.data_type()
+ ))
+ })?;
+ *hash = combine_hashes(dict_hashes[idx], *hash)
+ } // no update for Null, consistent with other hashes
+ }
+ } else {
+ for (hash, key) in
hashes_buffer.iter_mut().zip(dict_array.keys().iter()) {
+ if let Some(key) = key {
+ let idx = key
+ .to_usize()
+ .ok_or_else(|| {
+ DataFusionError::Internal(format!(
+ "Can not convert key value {:?} to usize in
dictionary of type {:?}",
+ key, dict_array.data_type()
+ ))
+ })?;
+ *hash = dict_hashes[idx]
+ } // no update for Null, consistent with other hashes
Review comment:
Yes, I agree in general the hashing of nulls needs some more thought. I
think it is best to have dictionary hashing be consistent with other types but
I think the hashing of nulls for everything should be considered.
I have filed https://github.com/apache/arrow-datafusion/issues/822 to track
the issue
--
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]
