Pandas

Pandas is the in-memory table tool to reach for when labeled rows and columns, heterogeneous dtypes, and data-cleaning ergonomics matter more than distributed execution. These notes are a decision map, not a local copy of the API reference.

Working path

Need	Start with
Understand the objects	Series, then DataFrame
Load and inspect data	DataFrame indexing and loading
Select rows or columns	querying and indexing pages
Repair data quality	missing values, then data cleaning
Combine tables	merging DataFrames
Summarize	grouping data and pivot tables
Reshape or encode	scales and pivot tables
Work with time	date functionality

House rules

1. Make labels, dtypes, units, time zones, and missing-value policy explicit.
2. Use .loc for labels and .iloc for positions; do not rely on ambiguous [] behavior.
3. Prefer vectorized operations, agg, and transform before row-wise apply.
4. Validate merge cardinality and inspect unmatched keys.
5. Treat a DataFrame as an intermediate analytical object, not a database or a distributed compute engine.

Source of truth

• Pandas user guide
• 10 minutes to pandas
• API reference