Build on your SQL & BI skills / CORE PATH

Joins, nulls & grain

A join is a cardinality decision, not just a merge call.

Use your star/galaxy modeling knowledge to prevent multiplied facts.

Understand the idea

Declare the expected relationship before merging: one-to-one, many-to-one or many-to-many. Validate reference-key uniqueness before a join; counting rows only after multiplication can hide the reason a KPI changed.

Check unmatched keys in both directions and decide what an unknown dimension means. An inner join may discard legitimate facts, while filling every missing value with zero invents measurements.

What to learn

• Declare one row per entity before joining.
• Use pandas merge(validate=...) and inspect unmatched keys.
• Separate None, NaN, empty text and business-unknown values.

Practice on your own data

Join a fact fixture to a duplicate dimension and make the validation fail.

Practice this lesson in the Python lab →

Prerequisites

DataFrames through a SQL lens

Video and playlist companions

Choose one explanation for the concept, then attempt the practice. These are topic mappings, not a claim that every linked lesson covers every outcome.

Sources and further reading

• pandas · getting started tutorials ↗Official documentation
• كورس Python منحة ITI ↗User-selected ITI playlist
• تحليل البيانات في بايثون (الدورة المصغرة) | Python for Data Analysis (Crash Course) ↗User-selected YouTube playlist
• Python for Data Analysis ↗User-selected YouTube playlist
• Data Exploration and Preparation ↗User-selected YouTube playlist

Track progress on this topic → · Markdown source ↗