/bikeshareapi

bikeshare api (Capstone Project)

Primary LanguageJupyter Notebook

Bike Sharing API

For original README (Introduction of capstone project) visit OLD README

SETUP

Pengaturan environment dapat menggunakan environment.yml (mengutamakan channel conda-forge). Untuk pip dapat menggunakan requirements.txt. Environment ini digunakan saat menggunakan VSCODE (ipykernel) dan test API (requests).

  • Install langsung menggunakan conda
conda env create -f "https://raw.githubusercontent.com/taruma/bikeshareapi/main/environment.yml"

TEST FOR RUBRIC

1 [1] Created Flask App

  • Menjalankan aplikasi flask app.py.

2 [2] Created functionality to read to get specific data from the database

  • Mengambil data berdasarkan id dari database trips.

3 [4] Created functionality to input new data into each table for the databases

  • Memasukkan data austin_bikeshare_trips_2021.csv ke database trips.

image

4 [3] Created static endpoints which return analytical result (must be different from point 2,3)

  • Memperoleh daftar tipe subscriber yang diurutkan dari terbanyak hingga tersedikit. Return: JSON (dari DataFrame).
def top_subscribers(conn):
    query = """
    SELECT 
        trips.subscriber_type,
        COUNT(subscriber_type) AS total_subscriber
    FROM trips
    GROUP BY trips.subscriber_type
    ORDER BY total_subscriber DESC
    """
    return pd.read_sql_query(query, conn)

# ...

### Static Endpoint
@app.route('/trips/top_subscribers/')
def route_trips_top_subscriber():
    conn = make_connection()
    top_list = top_subscribers(conn)
    return top_list.to_json()

5 [3] Created dynamic endpoints which return analytical result (must be different from point 2,3,4)

  • Menampilkan jumlah penggunaan bike pada hari Senin-Minggu pada stasiun tertentu (start_station_id), diurutkan dari yang tertinggi hingga terendah. RETURN: JSON (from DataFrame) atau Text (jika station tidak ditemukan).
def top_days_start_in_station(start_station_id, conn):
    try:    
        return (
            pd.read_sql_query(
                f"""
                SELECT bikeid, start_time, start_station_id
                FROM trips
                WHERE trips.start_station_id = {start_station_id}
                """,
                conn,
                parse_dates='start_time'
            ).assign(
                start_day_of_week=lambda x: x.start_time.dt.day_name()
            ).pivot_table(
                index='start_day_of_week',
                values='start_time',
                aggfunc='count'
            ).sort_values('start_time', ascending=False)
        )
    except:
        return f"STATION ID {start_station_id} NOT FOUND!"

### ...

### dynamic endpoints
@app.route('/trips/top_days_start_in_station/<station_id>')
def route_top_days_start_in_station(station_id):
    conn = make_connection()
    top_days = top_days_start_in_station(station_id, conn)
    return top_days.to_json() if isinstance(top_days, pd.DataFrame) else top_days

6 [3] Created POST endpoint which receive input data, then utilize it to get analytical result (must be different from point 2,3,4,5)

  • Mengirimkan informasi subscriber_type dan bulan (dalam bentuk angka 00) dan memperoleh informasi daftar informasi jumlah penggunaan bike, rata-rata + median + maksimum durasi penggunaan bike yang diagregasi setiap tahun dan hari Senin-Minggu. RETURN: JSON (dari DataFrame).
def top_duration_mean(subscriber_type, month_number, conn):
    return (
        pd.read_sql_query(
            f"""
            SELECT *
            FROM trips
            WHERE 
                trips.subscriber_type LIKE '{subscriber_type}'
                AND trips.start_time LIKE '%-{month_number}-%'
            """,
            conn,
            parse_dates='start_time'
        )
        .assign(
            start_year=lambda df: df.start_time.dt.year,
            start_dow=lambda df: df.start_time.dt.day_name()
        )
        .groupby(['start_year', 'start_dow']).agg({
            'bikeid': 'count',
            'duration_minutes': ['median', 'mean', 'max'],
        })
        .droplevel(0, axis='columns')
        .sort_values(['start_year', 'mean'], ascending=[False, False], axis='index')
        .rename({'median': 'duration_median', 'mean':'duration_mean', 'max':'duration_max'}, axis='columns')
    )

# ...

### post endpoint
@app.route('/trips/top_duration_mean', methods=['POST'])
def route_top_duration_mean():
    data_input = request.get_json(force=True)
    conn = make_connection()
    return top_duration_mean(conn=conn, **data_input).to_json()

Kekurangan / TO-DO LIST

  • penggunaan nama fungsi yang masih dibilang kurang singkat dan jelas. Seperti fungsi top_subscriber() yang sebaiknya menjadi top_subscribers_type().
  • endpoint harusnya menggunakan dash - dibandingkan _. Info: Hyphen, underscore, or camelCase as word delimiter in URIs?
  • pada endpoint /trips/top_duration_mean, seharusnya jangan menggunakan unpacking dictionary pada fungsinya, untuk menghindari error ketika menerima keyword yang tidak ada.
  • memisahkan logic function ke file yang berbeda.
  • validasi data sebelum diproses. (data yang hilang, dlsbnya).
  • untuk operasi count seharusnya mengacu pada bikeid.
  • single output endpoint, jadi endpointnya mengeluarkan nilai, bukan dataframe.to_json().
  • refactor kode. šŸ¤£