02. Top mentions

mention_histogram_alt_python, the mirror of example 01, but for @user mentions instead of #hashtags. Internally it uses advertools.extract_mentions partitioned across workers; advertools normalizes mentions to lowercase, so @Alice and @ALICE collapse onto the same @alice bucket.

What you'll see

Loaded 10 tweets.
Top 5 mentions:
    Mentions  Frequency
     @kaggle          5
     @openai          4
@huggingface          3
       @nasa          2
        @bbc          1

The code

The inline dataset pairs two mentions per row so the frequencies are easy to verify by eye:

_ROWS = [
    ('2022-01-01T00:00:00', 'morning tools roundup @kaggle @openai'),
    ('2022-01-01T01:00:00', 'new model leaderboard @kaggle @huggingface'),
    ('2022-01-01T02:00:00', 'climate dashboard @kaggle @nasa'),
    ('2022-01-01T03:00:00', 'fine-tuning report @openai @huggingface'),
    # ...5 more rows...
    ('2022-01-01T09:00:00', 'product launch'),
]

The work is identical to tutorial 01 except for the analytic call:

from whistlerlib import Context

ctx = Context('processes', 'localhost', 8786)
ds = ctx.load_csv(
    filen=csv_path,
    meta={
        'column_mapping': {'date_column': 'Date', 'text_column': 'text'},
        'file_encoding': 'utf-8',
    },
    num_partitions=2,
)
print(f'Loaded {ds.tweet_count()} tweets.')
histogram = ds.mention_histogram_alt_python(k=5)
print(histogram.to_string(index=False))

mention_histogram_alt_python(k=5) ships a map_partitions closure that calls advertools.extract_mentions per partition; the scheduler merges the partial frequency tables and returns the top-5 as a pandas DataFrame.

The full file (including the tempfile setup and CLI shim) is at examples/02-mention-histogram/example.py.

Differences from example 01

• The result columns are ['Mentions', 'Frequency'] (capital-M, capital-F) rather than ['tag', 'freq']. This mirrors the underlying advertools field names. Phase 2 standardized the alt-python hashtag path to use ['tag', 'freq'] for consistency with the R-bridge; the mention path kept the advertools names because they're already used widely.
• Mentions are returned in lowercase regardless of original casing.

Run it

# From examples/02-mention-histogram/, bring up a local Dask cluster, run the example, tear it down.
docker compose -f ../../docker/docker-compose.yml up -d
python example.py
docker compose -f ../../docker/docker-compose.yml down