[BEAM-9331] Add better Row builders #10883
Conversation
| // passed in, it could result in strange errors later in the pipeline. This method is largely | ||
| // used internal | ||
| // to Beam. | ||
| @Internal | ||
| public Builder attachValues(List<Object> values) { |
There was a problem hiding this comment.
Maybe this is an opportunity to change the attachValues. No values should be set before or after the attach. I see 2 options to improve this:
- In the attach first see if values are already set. Let the attachValues return the new Row directly. This is maybe a bit strange as it violates a builder pattern.
- Have 4 build in builders. The starting one (that includes an
attachValues,addandwithFieldValue), all of them return a specific builder: the newModifyingBuilderand a newAddValuesBuilderthat only has the add methods and anAttachBuilderthat only has build. This also eliminates some elaborate if/then/else's in the builder().
There was a problem hiding this comment.
I like making attachValues return a Row. I thin the same for withFieldValueBuilders. this simplifies the builder code
| } | ||
|
|
||
| /** Builder for {@link Row} that bases a row on another row. */ | ||
| public static class ModifyingBuilder { |
There was a problem hiding this comment.
Can't this be tweaked a bit, that this is the builder specifically for use with withFieldValue. Meaning that if when nit doesn't have a source row it just assumes null values for the fields not set. See remark on withFieldValue on initial builder as well.
There was a problem hiding this comment.
Yes, good suggestion. This should simplify the code
| * Set a field value using the field name. Nested values can be set using the field selection | ||
| * syntax. | ||
| */ | ||
| public Builder withFieldValue(String fieldName, Object value) { |
There was a problem hiding this comment.
Maybe this one can return the ModifyingBuilder so that no other methods can used (no attachValues, no add's).
| return sourceRow.getSchema(); | ||
| } | ||
|
|
||
| public ModifyingBuilder withFieldValue(String fieldName, Object value) { |
There was a problem hiding this comment.
Wouldn't it be useful to have a withFieldValue with an index?
|
Maybe rebase, then we could get this in. This is a useful addition. |
|
@alexvanboxel there are still some bugs in this PR related to logical types, which is why I haven't pushed it in yet. |
|
Run Java PreCommit |
reuvenlax
force-pushed
the
better_row_builder
branch
from
778bf70 to
52695aa
Compare
|
@alexvanboxel rebased and fixed bugs. Previously I was blocked on getting logical types to work, but now that we natively store logical types in Row, it's become much easier. |
|
run sql postcommit |
|
@alexvanboxel I would rather not in this PR, because the RowUtils API wasn't designed for public usage. If we were to make it public, I would prefer to spend a lot more time on the API design, and I would also want to understand the use cases a bit better. |
Understand LGTM |
|
run sql postcommit |
1 similar comment
|
run sql postcommit |
This PR adds two builders to the Row object. The first allows building a Row by specifying fields by name:
Row row = Row.withSchema(schema)
.withFieldValue("userId", "user1)
.withFieldValue("location.city", "seattle")
.withFieldValue("location.state", "wa")
.build();
The second allows building. a Row based on a previous row by specifying only the fields to change:
Row modifiedRow =
Row.fromRow(row)
.withFieldValue("location.city", "tacoma")
.build();
R: @rezarokni