Skip to content

Make flattened synthetic source concatenate object keys on scalar/object mismatch #129600

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 12 commits into
base: main
Choose a base branch
from

Conversation

parkertimmins
Copy link
Contributor

@parkertimmins parkertimmins commented Jun 18, 2025

There is an issue where for Flattened fields with synthetic source, if there is a key with a scalar value, and a duplicate key with an object value, one of the values will be left out of the produced synthetic source.

This fixes the issue by replacing the problematic object with paths to each of its keys. These paths consist of the concatenation of all keys going down to a given scalar, joined by .. For example, they are of the form foo.bar.baz. This applies recursively, so that every value within the object, no matter how nested, will be accessible through a full specified path.

For example if the following flattened field values is indexed:

{
   "a": {
       "b": 5
       "b": {
          "c": 10,
          "d": { "e": 15 }
       }
   }
}

The following synthetic source will be produced:

{
   "a": {
      "b": 5
      "b.c": 10,
      "b.d.e": 15
   }
}

Fixes #122936

next = nextValue == null ? KeyValue.EMPTY : new KeyValue(nextValue);

var startPrefix = curr.prefix.diff(openObjects);
if (startPrefix.prefix.isEmpty() == false && startPrefix.prefix.getFirst().equals(lastScalarSingleLeaf)) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What if the conflict doesn't happen on the first part of the prefix, e.g.

field {
  path {
    to: 10
    to {
      foo: bar
    }
  }
}

Would this be caught here?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yep, so this will become:

field {
  path {
    to: 10
    to.foo: bar
 }
}

When it get to the first key/value field.path.to|10 it will take the else block and traverse down into the object, adding field and path to the openObject context. When it reaches the key value field.path.to.foo|bar that object will still be open, and seeing that lastScalarSingleLeaf has a value of to, and that to is the first token in the startPrexix (to.foo), it will make a concatenated path.

(Updated a test to this situation to verify it)

@@ -103,7 +103,7 @@ public void testSingleObject() throws IOException {

// THEN
assertEquals(
"{\"a\":\"value_a\",\"a\":{\"b\":\"value_b\",\"b\":{\"c\":\"value_c\"},\"d\":\"value_d\"}}",
"{\"a\":\"value_a\",\"a.b\":\"value_b\",\"a.b.c\":\"value_c\",\"a.d\":\"value_d\"}",
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Interestingly, there was a test which had a scalar/object mismatch. But it produced duplicate keys. When the xcontent was converted to jsont these duplicate keys originally threw an error, but now just drop the duplicates. (Something must have changed in xcontent stuff since the issue was opened to cause this change from an error to deduplication)

@parkertimmins parkertimmins marked this pull request as ready for review June 18, 2025 19:32
@parkertimmins parkertimmins requested review from kkrik-es and lkts June 18, 2025 19:33
@elasticsearchmachine elasticsearchmachine added the needs:triage Requires assignment of a team area label label Jun 18, 2025
@parkertimmins parkertimmins added auto-backport Automatically create backport pull requests when merged :StorageEngine/Mapping The storage related side of mappings labels Jun 18, 2025
@elasticsearchmachine
Copy link
Collaborator

Pinging @elastic/es-storage-engine (Team:StorageEngine)

@elasticsearchmachine elasticsearchmachine removed the needs:triage Requires assignment of a team area label label Jun 18, 2025
@parkertimmins parkertimmins added needs:triage Requires assignment of a team area label v8.19.0 v9.0.4 v8.18.4 and removed Team:StorageEngine needs:triage Requires assignment of a team area label labels Jun 18, 2025
@elasticsearchmachine
Copy link
Collaborator

Hi @parkertimmins, I've created a changelog YAML for you.

Copy link
Contributor

@lkts lkts left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we document this? Otherwise LGTM

// In the open object, there is a leaf with a scalar value, which is also the first
// part of the current path. Instead of traversing down into the path and building objects,
// combine the path into a single leaf and add it as a field.
if (curr.pathEquals(next) == false) {
Copy link
Contributor

@lkts lkts Jun 18, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you explain what this if does to me?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This if is where we add fields to the output. Specifically, fields that need to be specified with the concatenated path due to a scalar/object mismatch. The similar if in the else block below serves the same purpose. The if is necessary to handle multiple values for the same field. For example if we have the following key/value pairs (where I'm using | instead of \0 to separate key and value):

foo|5
foo.bar|10
foo.bar|20
baz|99

Assume we've already written out foo|5. When we go through the loop for foo.bar|10, we'll add 10 to the values array (which will be empty) at the beginning of the loop. But when we get to this if statement, we'll see that the next pair (foo.bar|20) has the same path foo.bar. Because of this we won't enter the if and won't write out the field yet. In the next loop iteration, curr is now foo.bar|20. We'll add 20 to values, making it [10, 20]. When we get to this if statement, we see that next path is baz. Since this not equal to foo.bar we know that we have seen all values for foo.bar. So we enter the if statement and write out all values for the field.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a good example, consider adding it to the comment above, or splitting some of that before the first branch.

Copy link
Contributor

@lkts lkts Jun 19, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you @parkertimmins, i think it's worth a comment too. Alternatively it sounds like we could restructure this to first read all the values for a field (move the if into a separate loop above this one) and then write them but this is not critical of course.

Copy link
Contributor

🔍 Preview links for changed docs:

🔔 The preview site may take up to 3 minutes to finish building. These links will become live once it completes.

`"bar"` has both a scalar value `"10"`, and an object value of `{ "baz": "20" }`.

With synthetic source, objects with such duplicate fields will appear
differently in `_source`. For example, if the above field has synthetic
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

More accurately: if the field is defined in an index configured with synthetic source.

This is because, when constructing the source, the key `"foo.bar"` is eagerly expanding into the key `"foo"` with an object value.
Then `"bar"` is added to the object with the scalar value of `"10"`. Then, because another
`"bar"` cannot be added with an object value, `"bar"` and `"baz"` are collapsed
in the flat key `"bar.baz"` with a scalar value of `"20"`
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd probably skip this paragraph, and maybe mention above:

... to produce a valid JSON output.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1, let's not go into the details

var syntheticSource = syntheticSource(mapper, b -> {
b.startObject("field");
{
b.field("key1.key2", "foo");
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe add also: `b.field("key1", "baz");

final FlattenedFieldSyntheticWriterHelper writer = new FlattenedFieldSyntheticWriterHelper(new SortedSetSortedKeyedValues(dv));
final ByteArrayOutputStream baos = new ByteArrayOutputStream();
final XContentBuilder builder = new XContentBuilder(XContentType.JSON.xContent(), baos);
final List<byte[]> bytes = List.of("a.b.c" + '\0' + "10", "a.b.c.d" + '\0' + "20")
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's also add something for a and a.b.

});
assertThat(syntheticSource, equalTo("""
{"field":{"key1":{"key2":"foo","key2.key3":"bar"}}}"""));
}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Shall we also add an example with arrays?

Copy link
Contributor

@kkrik-es kkrik-es left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nice, some minor comments to address but otherwise it looks good.

@@ -364,3 +364,42 @@ Will become (note the nested objects instead of the "flattened" array):
}
}
```

Flattened fields allow for a duplicate key to contain both an object and a scalar value.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
Flattened fields allow for a duplicate key to contain both an object and a scalar value.
Flattened fields allow for a key to contain both an object and a scalar value.

This is because, when constructing the source, the key `"foo.bar"` is eagerly expanding into the key `"foo"` with an object value.
Then `"bar"` is added to the object with the scalar value of `"10"`. Then, because another
`"bar"` cannot be added with an object value, `"bar"` and `"baz"` are collapsed
in the flat key `"bar.baz"` with a scalar value of `"20"`
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1, let's not go into the details

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
auto-backport Automatically create backport pull requests when merged >bug :StorageEngine/Mapping The storage related side of mappings Team:StorageEngine v8.18.4 v8.19.0 v9.0.4 v9.1.0
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Synthetic _source can't be retrieved with overlapping keys in flattened field
4 participants