Data Wrangling with PySpark for Data Scientists Who Know Pandas - Andrew Ray

"Data scientists spend more time wrangling data than making models. Traditional tools like Pandas provide a very powerful data manipulation toolset. Transitioning to big data tools like PySpark allows one to work with much larger datasets, but can come at the cost of productivity.

In this session, learn about data wrangling in PySpark from the perspective of an experienced Pandas user. Topics will include best practices, common pitfalls, performance consideration and debugging.

Session hashtag: #SFds12

Learn more:
Developing Custom Machine Learning Algorithms in PySpark

Introducing Pandas UDF for PySpark

Best Practices for Running PySpark

Session Overview:
- Why?
- What Do i get with pyspark?
- Primer
- Important Concepts
- Architecture
- Setup
- Run
- Load CSV
- View Dataframe
- Rename Columns
- Drop Column
- Filtering
- Add Column
- Fill Nulls
- Aggregation
- Standard Transformations
- Keep it in the JVM
- Row Conditional Statements
- Python when Required
- merge/join dataframes
- Pivot table
- Summary Statistics
- histogram
- SQL
- Make sure to
- Things not to do
- If things go wrong
- Thank you

About: Databricks provides a unified data analytics platform, powered by Apache Spark™, that accelerates innovation by unifying data science, engineering and business.

Connect with us:

[AlessandroBottoni]

Fantastic introduction to PySpark for beginners. Hope to see Andrew Ray again on the stage for other presentations.

[ratkush]

Must watch Q n A session in the end. I loved it.

[fiddlepants]

Really nice how we see pandas and pyspark functions side-by-side!

[enes-the-cat-father]

Thank you for such a great presentation for beginners!

[kevinlin]

This a great video. Exactly what I'm looking for thanks very much.

[tanishasharma]

he provided with a really good comparison between the two!

[ZenvilleErasmus]

Cool talk and key differences nicely illustrated.

[toygraphers]

Thank you very much for your contribution.

[Arjungtk]

I think I need a soundbox on full volume to hear this.

[abrahamf]

My path to data was a little bit unsual to say the least, started to work in the financial industry using databricks and now on side projects started to work on pandas... funny that I actually used this video backwards hehe

[santil.]

Does it mean that using pyspark sql is the best practice in data wrangling using spark?

[over]

PySpark is great with it's read only. It all goes badly wrong when you try and write anything with a typed schema.

[musasall]

by just downloading and writing this code it will not work. You have to create a session.

[krishnakishorepeddisetti]

Which is better in databricks environment?? Python or R or SQL..reply in comments

[elliottharris]

Would this be a good tool for combining large numbers of csvs into a single dataframe quickly and then performing manipulations on that dataframe before outputting a single csv?

[Tyokok]

great tech video, but volume really ...

[Drivebyeasy]

Hey Andrew could you send me your Github link

[kaixianghuang]

LOL good presentation, but unprepared for the Q &A