Fungsi window dan kueri streaming

Transformasi Data dengan Spark SQL di Databricks

Disha Mukherjee

Lead Data Engineer

Saat groupBy tidak cukup

recraft: half: Sebuah kaca pembesar di atas baris data spreadsheet dengan nilai terhitung yang disorot

groupBy() → satu baris per grup (total, rata-rata)
Fungsi window → menambah kolom, semua baris tetap ada
Use case: total berjalan, pemeringkatan, perbandingan baris

Total berjalan

from pyspark.sql.window import Window

window_spec = (
    Window.partitionBy("Customer_ID")
    .orderBy("Date")
    .rowsBetween(Window.unboundedPreceding, Window.currentRow)
)


df_running = df_valid.withColumn(
    "running_total",
    F.round(F.sum("Transaction_Amount").over(window_spec), 2)
)

Total berjalan

+-----------+-------------------+------------------+-------------+
|Customer_ID|Date               |Transaction_Amount|running_total|
+-----------+-------------------+------------------+-------------+
|CUST003    |2023-01-03 00:00:00|5752.36           |5752.36      |
|CUST003    |2023-02-14 00:00:00|12340.00          |18092.36     |
+-----------+-------------------+------------------+-------------+

Memeringkat pelanggan

customer_totals = (
    df_valid.groupBy("Customer_ID")
    .agg(F.round(F.sum("Transaction_Amount"), 2).alias("total_revenue"))
)


rank_window = Window.orderBy(F.col("total_revenue").desc())

df_ranked = customer_totals.withColumn("revenue_rank", F.rank().over(rank_window))

+-----------+-------------+------------+
|Customer_ID|total_revenue|revenue_rank|
+-----------+-------------+------------+
|CUST1469   |99996.20     |1           |
|CUST19129  |99990.14     |2           |
|CUST39417  |99982.21     |3           |
+-----------+-------------+------------+

Apa itu streaming?

Batch - memproses himpunan data tetap sekaligus
Streaming - memproses data bertahap saat masuk
Ideal untuk aliran transaksi, log, dan data peristiwa

recraft: half: Data mengalir terus-menerus sebagai aliran ke mesin pemroses dengan panah

Streaming berbasis file

Direktori stream:
  day_1.csv  (108 KB)
  day_2.csv  (108 KB)
  day_3.csv  (108 KB)
  day_4.csv  (108 KB)
  day_5.csv  (108 KB)

Direktori file CSV = sumber streaming
Setiap file baru = satu mikro-batch
File baru diambil otomatis
Skema harus didefinisikan eksplisit

Membaca stream

df_stream = (
    spark.readStream.format("csv")
    .option("header", "true")
    .schema(streaming_schema)
    .load(STREAM_DIR)
)

print(df_stream.isStreaming)

True

Apa itu checkpoint?

recraft: half: Penanda bookmark atau titik simpan dalam pipa data mengalir dengan progres tersimpan

Checkpoint = direktori metadata di disk
Mencatat file yang sudah diproses
Saat restart, Spark melanjutkan dari titik terakhir
Mencegah pemrosesan ganda

Menulis stream

query = (
    df_stream.writeStream.format("delta")
    .outputMode("append")
    .option("checkpointLocation", CHECKPOINT_DIR)
    .option("path", DELTA_PATH)
    .trigger(availableNow=True)
    .start()
)
query.awaitTermination()

Status:       Stopped
Baris ditulis: 5,000

Pemantauan: status dan lastProgress

print(query.status)

progress = query.lastProgress
print(f"Rows processed: {progress['numInputRows']}")
print(f"Rows/sec: {progress['processedRowsPerSecond']:.0f}")

{'message': 'Stopped', 'isDataAvailable': False, 'isTriggerActive': False}
Baris diproses: 5,000
Baris/detik:    752

Pemulihan checkpoint

query_restart = (
    df_stream.writeStream.format("delta")
    .option("checkpointLocation", CHECKPOINT_DIR)
    .option("path", DELTA_PATH)
    .trigger(availableNow=True)
    .start()
)
query_restart.awaitTermination()

Baris saat restart: 0

Ayo berlatih!

Transformasi Data dengan Spark SQL di Databricks