beginner
Query Parquet and CSV files with SQL
Find the busiest taxi pickup zones with SQL on real files.
Open the guideTHE SAIL EXAMPLE COLLECTION
Good questions.
Runnable answers.
Explore a city’s taxi trips. Find a photo with a sentence. Bring a customer table up to date.
Real tasks, explained step by step, with Spark SQL and Python on Sail. Pick something that makes you curious.
On your laptop · No Java · No cluster · No API key
FROM THE PHOTO SEARCH EXAMPLE
a dog playing in the snow
01 / Best match
See how the example works Photo source & license
A FEW GOOD PLACES TO START
See it. Run it. Make it yours.
Each guide shows the result,
the code, and one change to try.
intermediate
Search photos by description
Find a photo by describing what you want to see.
Open the guide
intermediate
Upsert into a Delta Lake table with MERGE
Apply customer updates, inserts and deletes in one SQL statement.
Open the guideFROM CURIOUS TO RUNNING
Your first answer is
a few commands away.
Install uv, then run the taxi example. It handles Python and package setup for you.
After cd sail-examples, run all guide commands from that repository root.
The first run downloads about 1.4 MB of trip data and a small zone lookup, plus Python packages.
TERMINAL
git clone https://github.com/lakehq/sail-examples.git
cd sail-examples
uv run quickstart/query-files-with-sql/main.py
FIRST RESULTEast Harlem North 12,141 trips
KNOW WHAT TO EXPECT
How Sail behaves.
From caching to file formats, these checked notes explain behavior that can change how you write a query.
Browse all 38 notes →- Why did my model run again?Saving and reusing results
- Where do I set engine options?Configuration before startup
- What behaves differently from Spark?Types, timestamps and query results
FOLLOW YOUR CURIOSITY
What would you like to build?
Browse all 35 runnable examples.
35 examples
Hello Sail
Run your first Spark SQL and DataFrame queries on Sail.
Start hereQuery Parquet and CSV files with SQL
Find the busiest taxi pickup zones with SQL on real files.
Start hereExplore a year of taxi trips with SQL
Find seasonal trends and rush hours in a year of NYC taxi trips.
AnalyticsReconcile orders, captures and refunds
Find payment discrepancies without counting repeated deliveries twice.
AnalyticsReproduce a report from saved Iceberg snapshots
Recreate an earlier report after its source tables have changed.
AnalyticsSessions, funnels and cohort retention with SQL
See where users drop out and which cohorts come back.
AnalyticsUse SQL inside a Python app with pandas
Query pandas data with SQL and bring the results back to Python.
AnalyticsTurn spoken requests into intent analytics
Turn spoken support requests into a report of customer needs.
AnalyticsApply replayed changes without resurrecting deleted records
Replay updates safely without bringing deleted records back.
Data pipelinesBuild features as they were known at each decision
Build training features using only what was known at the time.
Data pipelinesUpsert into a Delta Lake table with MERGE
Apply customer updates, inserts and deletes in one SQL statement.
Data pipelinesManage a Delta Lake table through its life
Evolve a Delta table and inspect its earlier versions.
Data pipelinesBuild a versioned search evaluation dataset
Compare search methods against a versioned set of relevance judgments.
Data pipelinesManage an Apache Iceberg table and read old snapshots
Update an Iceberg table and read its earlier snapshots.
Data pipelinesNormalize nested order events without losing rejected records
Turn nested orders into clean tables while keeping rejected records.
Data pipelinesReconcile invoice documents against purchase orders
Read scanned invoices and check them against purchase orders.
Data pipelinesRepair daily reports after corrected order records
Repair past sales totals when orders are corrected or deleted.
Data pipelinesBuild a batch ETL job with data quality checks
Build reporting tables from raw taxi files with data quality checks.
Data pipelinesRead a REST API as a table
Query a paginated REST API as a table.
Python integration and testingChoose a Python UDF style
Choose the right way to run Python logic inside a query.
Python integration and testingEmbed text and search it with vector functions
Find related documents using text embeddings and SQL.
AI data preparationEnrich a table with a model
Run a model once per distinct value, then join its answers back.
AI data preparationRoute support tickets with a model in SQL (Jev)
Classify support tickets with a model called from SQL.
AI data preparationMaintain document embeddings as sources change
Update a search dataset without re-embedding unchanged documents.
AI data preparationAlign robot camera frames with state and actions
Match robot camera frames to nearby state and action records.
AI data preparationClassify images with an ONNX model
Label photos and send uncertain predictions for review.
AI data preparationCurate humanoid robot episodes with every camera frame
Check humanoid robot recordings and align every camera frame.
AI data preparationCurate robot demonstration episodes
Check robot demonstrations before splitting them into training sets.
AI data preparationMake thumbnails from a folder of images
Turn a folder of images into thumbnails and a metadata table.
AI data preparationExtract text from scanned PDFs with OCR
Extract text from old scans and see where OCR makes mistakes.
AI data preparationSearch a folder of PDFs
Turn PDFs into searchable passages.
AI data preparationSearch photos by description
Find a photo by describing what you want to see.
AI data preparationFind scenes in videos with text search
Find a scene in a video by describing it.
AI data preparationTranscribe audio files with Whisper
Turn audio files into transcripts and check their accuracy.
AI data preparationTest Spark code with pytest on Sail
Test Spark DataFrame code locally without Java.
Python integration and testingNo examples match those filters.
Try a broader search, or clear the filters to see the whole collection.
CODE YOU CAN INSPECT
Learn from the result.
Keep the useful parts.
Every guide links to its runnable source. The test suite compares portable examples with Apache Spark and checks selected results against independent answers.
These are examples with bounded inputs, not a promise about every workload. Each guide explains its requirements and limits.
How the examples are checked