pandas is a Python library for working with tabular data. Its central structure, the data frame, holds a two-dimensional grid with row and column labels. It makes repeated tasks—loading, cleaning, selecting, combining, and summarizing data—less cumbersome than writing loops and list comprehensions for each task. section 24.2.1 section 24.2.3
Start with a data frame
Given `grid = [[1, 2, 3], [4, 5, 6], [7, 8, 9]]`, create `df = pd.DataFrame(grid, columns=["one", "two", "three"])` after `import pandas as pd`. Then `df["two"]` selects the values 2, 5, and 8 by column name. `df[["one", "three"]]` returns another data frame containing those two columns; adding `.add(2)` adds 2 to every selected value. The double brackets matter: they pass a list of column names. section 24.2.3
Common operations fit into a short workflow:
- Load a file with `pd.read_csv(...)` or `pd.read_json(...)`. For CSV data, `na_values=['Missing']` converts that marker to `NaN` during loading; `usecols` can omit unwanted columns. Save a frame with `to_csv(...)` or `to_json(...)`. section 24.3.1 section 24.3.2
- Clean a column with column-wide operations. For percentage strings, `temp[17] = pd.to_numeric(temp[17].str.strip('%')).div(100)` removes the sign, converts the values to numbers, and expresses them as fractions. section 24.3.2
- Combine related frames with `pd.merge(calls, revenue, on=['Territory', 'Month'])`. A shared pair of columns links sales calls to revenue, much like a relational database join. section 24.4.1
- Filter rows with a condition, such as `calls_revenue[calls_revenue.Territory == 3]`. Calculate a new column with `calls_revenue['Call_Amount'] = calls_revenue.Amount / calls_revenue.Calls`. section 24.4.2
- Summarize a column with methods such as `sum()`, `mean()`, or `median()`, and use `groupby(...)` to calculate summaries per month or territory. section 24.4.3
- Plot a grouped frame in a Jupyter notebook with `.plot.bar()`; pandas uses Matplotlib for the plot. `plot.line()` and `plot.pie()` offer other simple views. section 24.5
print(calls_revenue[['Month', 'Calls', 'Amount']].groupby(['Month']).sum())
Calls Amount
Month
1 318 141165
2 302 164279
3 267 132238
4 328 135201
print(calls_revenue[['Territory', 'Calls', 'Amount']].groupby(['Territory']).sum())
Calls Amount
Territory
1 444 211532
2 379 149931
3 392 211420
Compare the two grouped outputs: the same calls and revenue columns yield totals by month or by territory simply by changing the grouping column. section 24.4.3
Install pandas with `pip install pandas`; a common import is `import pandas as pd`. In a Jupyter notebook, `%matplotlib inline` enables inline Matplotlib plots. Some file and database operations rely on additional libraries, so an import problem may lie outside pandas itself. section 24.2.2 section 24.3.1
Choose pandas for data that fits its tabular approach and benefits from these operations—not automatically for every dataset. Learning its conventions takes time, and plain Python may suit large datasets needing little mathematical manipulation, awkwardly structured data, or basic transaction-stream processing better. section 24.6
