Table.group_by #3305
Table.group_by #3305
Conversation
There was a problem hiding this comment.
Looks good to me.
Some style comments and questions but given it's a prototype, most are just suggestions for the future.
The important thing is the Group_By_Key - IMO it should not repeat the implementations of existing functions if it can delegate to them unless it is really absolutely necessary.
distribution/lib/Standard/Table/0.0.0-dev/src/Data/Aggregate_Column.enso
Outdated
Show resolved
Hide resolved
distribution/lib/Standard/Table/0.0.0-dev/src/Data/Aggregate_Column.enso
Outdated
Show resolved
Hide resolved
| - population argument specifies if group is a sample or the population | ||
| type Standard_Deviation (column:Column|Text|Integer) (name:Text|Nothing=Nothing) (population:Boolean=False) | ||
|
|
||
| ## Creates a new column with the values concatenated together. NULL values will become an empty string. |
There was a problem hiding this comment.
Btw. why is NULL collated with "" here instead of being ignored like in other cases ?
distribution/lib/Standard/Table/0.0.0-dev/src/Data/Aggregate_Column.enso
Outdated
Show resolved
Hide resolved
distribution/lib/Standard/Table/0.0.0-dev/src/Data/Aggregate_Column.enso
Show resolved
Hide resolved
distribution/lib/Standard/Table/0.0.0-dev/src/Data/Group_By.enso
Outdated
Show resolved
Hide resolved
distribution/lib/Standard/Table/0.0.0-dev/src/Data/Group_By.enso
Outdated
Show resolved
Hide resolved
distribution/lib/Standard/Table/0.0.0-dev/src/Data/Group_By.enso
Outdated
Show resolved
Hide resolved
distribution/lib/Standard/Table/0.0.0-dev/src/Data/Group_By.enso
Outdated
Show resolved
Hide resolved
| array = new_table.at (j + key_length) . at 1 . to_array | ||
| current = array . at row_index | ||
| new = aggregator current i | ||
| array . set_at row_index new |
There was a problem hiding this comment.
This may stop working when we fix the "to_array is leaking mutability" issue.
As this prototype may live shorter than the above issue - probably ignore this comment.
But just for future - maybe we could just create an Expandable_Array which wraps a regular, explicitly-mutable array but also has the append operation and does the allocations that give us O(1) amortized complexity of append. Because I don't like that we are abusing the Vector_Builder a bit when we could just create a data structure meant for such use-cases. But since this is a prototype - just commenting a suggestion for future.
jdunkerley
force-pushed
the
wip/jd/group-by-181268783
branch
from
81746c2 to
2f74a42
Compare
Handle no key
Use first/second Expanding tests
Sum of all Nothing == Nothing StDev calculation fixes
Support for Booleans and Nothing in the keys
jdunkerley
force-pushed
the
wip/jd/group-by-181268783
branch
from
434a021 to
9f99739
Compare
distribution/lib/Standard/Table/0.0.0-dev/src/Data/Aggregate_Column.enso
Outdated
Show resolved
Hide resolved
distribution/lib/Standard/Table/0.0.0-dev/src/Data/Group_By.enso
Outdated
Show resolved
Hide resolved
Pull Request Description
Functioning group_by based of Enso Map.
Important Notes
This is an initial version which will be used to establish the API.
The grouping map will need to be moved to Java code for performance.
Checklist
Please include the following checklist in your PR:
./run distand./run watch.