feat(eap-items): double-write arrays into typed map columns

[phacops]

Summary

Follow-up to #8056, which added the typed Map(String, Array(T)) columns for array-valued attributes (attributes_array_string, attributes_array_int, attributes_array_float, attributes_array_bool) but left populating them as a follow-up.

This PR populates those columns from EAPItemsProcessor (the Rust eap_items processor), double-writing them alongside the existing attributes_array JSON column. The JSON column keeps being written so the read path is unaffected until it is migrated to the typed columns in a separate change.

What it does

In AttributeMap::insert_array, each array attribute is split by element type into the matching typed bucket and inserted under the same key:

• string elements → attributes_array_string
• double elements → attributes_array_float
• int elements → attributes_array_int and attributes_array_float
• bool elements → attributes_array_bool

Homogeneous arrays — the common case — populate exactly one bucket; mixed arrays are split across buckets by element type.

Integer elements are also written as floats, mirroring the scalar insert_int double-write. The read path resolves numeric attribute types (val_int, val_float, val_double, and their array variants) to the float columns via _VALUE_TYPE_TO_COLUMN, so int arrays must be present in attributes_array_float to be queryable, while attributes_array_int retains exact integer values.

The four new columns are appended to EAPItemRow and to EAPItemRow::COLUMN_NAMES (sent explicitly with INSERT ... FORMAT RowBinary) in matching struct order. The serializer already encodes Vec<(String, Vec<T>)> as Map(String, Array(T)).

Tests

• New test_array_typed_columns_double_write covers homogeneous arrays of each type, the int→float double-write, no cross-type bleed, and that the legacy JSON column is still written.
• test_row_binary_attributes now also asserts the typed attributes_array_string column is populated.
• test_column_names_match_struct_layout updated for the four new columns (96 → 100).

All processors::eap_items unit tests pass; cargo clippy --lib and cargo fmt are clean. (test_row_binary_clickhouse_insert requires a live ClickHouse and only runs in CI.)

🤖 Generated with Claude Code

---

Generated by Claude Code

[linear-code]

EAP-573

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 85f5f3a. Configure here.}

[cursor]

Int arrays skip float column

Medium Severity

Homogeneous integer array attributes are written only to attributes_array_int, not also to attributes_array_float. That breaks parity with scalar insert_int and with _VALUE_TYPE_TO_COLUMN, which routes val_int_array to the float column family, so migrated read paths would not find int array data where queries expect it.

 

^{Reviewed by Cursor Bugbot for commit 85f5f3a. Configure here.}

[sentry]

Bug: Integer arrays are only written to attributes_array_int, not the float column, making them unqueryable after the planned read-path migration, which is expected to read from the float column.
_{Severity: HIGH}

Suggested Fix

Modify the insert_array function for Value::IntValue. When an integer array is processed, it should be written to both attributes_array_int and attributes_array_float to mirror the behavior of scalar integers. This ensures that the data will be queryable when the read path is migrated to use the new typed array columns.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent. Verify if this is a real issue. If it is, propose a fix; if not, explain why it's
not valid.

Location: rust_snuba/src/processors/eap_items.rs#L502-L507

Potential issue: The `insert_array` method for integer values writes data exclusively to
the `attributes_array_int` column. This is inconsistent with the existing behavior for
scalar integers, which are double-written to both integer and float columns
(`attributes_float`). The read path currently resolves numeric types to float columns to
enable querying. When the read path is migrated for arrays in a future change, it will
likely look for integer arrays in `attributes_array_float`. Since the data written by
this PR will only be in `attributes_array_int`, queries for this data will return empty
results, leading to silent data inaccessibility.

Also affects:

• rust_snuba/src/processors/eap_items.rs:1393~1394

_{Did we get this right? 👍 / 👎 to inform future reviews.}