chore(bigtable): run samples tests pre-submit

[daniel-sanche]

The bigtable library lost the test configs for samples as part of the move to the monorepo, so they are left untested. This PR adds back the configs, and runs tests as part of the pre-submit check

Samples tests are now run from the central bigtable noxfile, instead of using individual noxfiles for each sample directory

[gemini-code-assist]

Code Review

This pull request updates relative imports across several Google Cloud Bigtable samples and adds new configuration files, including noxfile.py and requirements files, for the async data client samples. A critical issue was identified in packages/google-cloud-bigtable/samples/utils.py, where unresolved git merge conflict markers were left in the code, which will lead to a runtime SyntaxError.

[parthea]

parthea force-pushed

To clarify , I clicked the rebase stack button

[parthea]

Please could you add a comment to clarify the reason we don't pin the google-cloud-bigtable library here, and elsewhere?

[daniel-sanche]

We want these sample tests to run against the repo head, so we can catch if any of our changes break anything. Running against an old released version of the library isn't valuable

[parthea]

In that case, perhaps we can remove the requirements files altogether, or only have the dependencies which are not in setup.py

We can then remove session.install("-e", ".", "--no-deps") from the noxfile.py file and have:

    session.install("-e", ".")
    session.install(
        "google-cloud-testutils",
        "mock",
        "pytest",
        "pytest-asyncio",
        *req_args,
    )

[daniel-sanche]

I think the requrements are useful, even if just to help users understand/run the samples. They're referenced in the samples README, and could be seen as of part of the documentation

But if you prefer to get rid of them, I don't feel too strongly about it

[parthea]

Gemini suggested we should have this instead

change

    batcher = table.mutations_batcher(flush_count=2)
    rows = table.read_rows()
    for row in rows:
        row = table.direct_row(row.row_key)
        row.delete_cell(column_family_id="cell_plan", column="data_plan_01gb")

    batcher.mutate_rows(rows)
    batcher.close()

to

    with table.mutations_batcher(flush_count=2) as batcher:
        for row in table.read_rows():
            direct_row = table.direct_row(row.row_key)
            direct_row.delete_cell(column_family_id="cell_plan", column="data_plan_01gb")
            batcher.mutate(direct_row)

This test fails without the fix

def test_streaming_and_batching_actually_deletes(table_id):
    """Verifies that streaming_and_batching actually deletes cells from the table."""
    from google.cloud import bigtable
    from . import deletes_snippets

    client = bigtable.Client(project=PROJECT, admin=True)
    instance = client.instance(BIGTABLE_INSTANCE)
    table = instance.table(table_id)

    # 1. Seed a row in the table with cell_plan:data_plan_01gb
    row_key = b"phone#4c410523#20190501"
    row = table.direct_row(row_key)
    row.set_cell("cell_plan", b"data_plan_01gb", b"true")
    row.commit()

    # 2. Verify row exists before calling the snippet
    seeded_row = table.read_row(row_key)
    assert seeded_row is not None, "Failed to seed row!"
    assert b"data_plan_01gb" in seeded_row.cells["cell_plan"]

    # 3. Run the snippet
    deletes_snippets.streaming_and_batching(
        PROJECT, BIGTABLE_INSTANCE, table_id
    )

    # 4. Check if the cell was deleted
    updated_row = table.read_row(row_key)
    # If the snippet worked, either the row is None (all cells deleted)
    # or the column data_plan_01gb is absent from cell_plan:
    has_cell = (
        updated_row is not None
        and b"data_plan_01gb" in updated_row.cells.get("cell_plan", {})
    )
    assert not has_cell, "Cell was NOT deleted because batcher received an exhausted generator!"

[daniel-sanche]

Good catch, fixed!