从头构建 DataFrame

Before loading real files, building a small DataFrame by hand lets you test ideas and learn the structure with zero setup. 🧪

Every project starts the same way: import pandas under the nickname pd. That short alias is a universal convention you will see everywhere.

The most common recipe is a dict where each key is a column name and each value is a list of that column's data.

With the dict approach, the dictionary keys turn into your column headers automatically. Naming columns is as easy as naming keys.

You can also pass a list of rows and supply the column names separately. Handy when your data already arrives row by row.

A list of dicts works too, with one dict per row. Missing keys simply become blanks, which is great for uneven records.

Have a NumPy array of numbers? Pass it straight in and add column names. This bridges fast NumPy math with labeled tables.

Pass an index argument to label your rows the moment you build the frame, instead of relying on default 0, 1, 2 numbers.

Want columns in a specific order? Pass a columns list and pandas arranges them exactly as you list, ignoring dict ordering quirks.

Every column you provide must have the same length. Mismatched list sizes raise an error, so keep your columns even.

After building, glance at dtypes to confirm pandas inferred sensible types like int for ages and object for names.

Pick the building block behind the most common DataFrame recipe.

Dicts, lists, list-of-dicts, or NumPy arrays all become a DataFrame. Pick whatever matches the shape of your data. 🛠️