EDR report no longer shows sample rows for failures and warnings

[orenniapete]

Describe the bug
The edr report used to show table rows of failed data when sample rows are enabled, this is no longer the case. Failed tests that show counts do not show the sample rows.

To Reproduce
Steps to reproduce the behavior:

1. Run a test where a failure or warning is triggered with one or more counts, ensure the project config has sample row enabled
2. Run the edr report and look at the failure/warning test results, expand each result row to show details, counts, and sample rows

Expected behavior
Depending on the type of test, the column, table, or singular test details are shows with a description, count of failure rows, and example rows limited to a number set in the profile/project config.

Environment (please complete the following information):
dbt-core==1.11.5
dbt-snowflake==1.11.0
elementary-data==0.24.0

Additional context
Used to work fine in version 0.22.0

Would you be willing to contribute a fix for this issue?
I'm a user not a developer so most likely not very helpful in terms of a fix.

[orenniapete]

With some LLM help I managed to fix the sample rows locally by updating the results after report generation:

I investigated this in detail and can confirm the missing rows are not a UI-only problem:

Failing tests have failure counts and corresponding stored sample rows in warehouse tables.
The generated report payload has empty results_sample for those tests.
The HTML report is generated from embedded window.elementaryData, so if results_sample is empty during report generation, the UI cannot show rows.
Proposed fix:

In report data assembly, add a fallback hydration step:
If test_results.results_sample is empty and failed_rows_count > 0
Query existing stored test result rows (same invocation/test unique id)
Populate results_sample before serialization
Keep this behind existing sample-row controls (disable-samples, row limits).

What to change

Primary fix location:
elementary/monitor/api/tests/tests.py
Method: _get_test_result_from_test_result_db_row
Optional normalization location:
elementary/monitor/fetchers/tests/schema.py
Field: sample_data on TestResultDBRowSchema
Tests to add:
tests/unit/monitor/api/tests/test_tests_api.py
tests/mocks/fetchers/tests_fetcher_mock.py
tests/e2e/report/test_report.py (or the fixture used by it)
Example fix (minimal, safe)

File: elementary/monitor/api/tests/tests.py

Add a helper to normalize sample payload shape before filling results_sample:

@staticmethod
def _normalize_results_sample(sample_data):
    if isinstance(sample_data, list):
        return sample_data

    if isinstance(sample_data, dict):
        # Support common payload shapes from dbt macro output
        for key in ("rows", "sample_rows", "results_sample", "data"):
            value = sample_data.get(key)
            if isinstance(value, list):
                return value

    return None

Then update dbt_test mapping:

if test_result_db_row.test_type == "dbt_test":
    sample_data = TestsAPI._normalize_results_sample(test_result_db_row.sample_data)
    if disable_samples:
        sample_data = None

    test_results = DbtTestResultSchema(
        display_name=test_result_db_row.test_name,
        results_sample=sample_data,
        error_message=test_result_db_row.test_results_description,
        failed_rows_count=TestsAPI._get_failed_rows_count(test_result_db_row),
    )

Then update dbt_test mapping:

if test_result_db_row.test_type == "dbt_test":
    sample_data = TestsAPI._normalize_results_sample(test_result_db_row.sample_data)
    if disable_samples:
        sample_data = None

    test_results = DbtTestResultSchema(
        display_name=test_result_db_row.test_name,
        results_sample=sample_data,
        error_message=test_result_db_row.test_results_description,
        failed_rows_count=TestsAPI._get_failed_rows_count(test_result_db_row),
    )

Why this is a good first fix:

It is backward compatible with list payloads.
It handles dict payloads without changing report schema.
It targets the exact serialization path that feeds report output.
Optional hardening

File: elementary/monitor/fetchers/tests/schema.py

Add a validator for sample_data so downstream code always sees a normalized shape:

@validator("sample_data", pre=True)
def normalize_sample_data(cls, sample_data):
    if isinstance(sample_data, dict):
        for key in ("rows", "sample_rows", "results_sample", "data"):
            value = sample_data.get(key)
            if isinstance(value, list):
                return value
    return sample_data