TopolearnNearly all pandas work means building, inspecting and transforming DataFrames.
A DataFrame is a two-dimensional table: a collection of columns, each a Series with its own dtype, all sharing the same row index. Column labels live in df.columns.
The most common way to build one by hand is from a dict of lists (one key per column) or a list of dicts (one per row):
df = pd.DataFrame({"city": ["Oslo", "Lima"], "temp": [3.5, 19.0]})First things to do with any DataFrame:
df.shapegives(rows, columns);len(df)gives the row countdf.head()anddf.tail()show the first or last rows;df.sample(5)a random fewdf.dtypeslists each column's type;df.info()adds non-null counts and memory usedf.columnsanddf.indexshow the labels
Selecting one column with df["temp"] returns a Series; a list of columns, df[["city", "temp"]], returns a DataFrame. Attribute access (df.temp) also works for simple names, but breaks for names with spaces or that collide with methods (df.count is a method), so square brackets are the safer habit.
Because columns are stored separately and can have different types, think of a DataFrame as column-oriented. Operations on whole columns are fast; looping over rows is slow and rarely necessary.
The row index defaults to 0..n-1 (a memory-free RangeIndex). You can promote a column to be the index with df.set_index("city"), which makes label lookups by city possible, and undo it with reset_index(). Most methods return a new DataFrame rather than modifying the original, so assign the result: df = df.set_index("city").
Example
import pandas as pd
df = pd.DataFrame({
"city": ["Oslo", "Lima", "Cairo", "Perth"],
"temp": [3.5, 19.0, 28.2, 22.4],
"rain_mm": [49, 1, 0, 14],
})
print(df.shape) # (4, 3)
print(df.head(2))
# city temp rain_mm
# 0 Oslo 3.5 49
# 1 Lima 19.0 1
print(df["temp"].max()) # 28.2 a column is a Series
print(df[["city", "rain_mm"]].columns.tolist()) # ['city', 'rain_mm']
print(df.dtypes["rain_mm"]) # int64
rows = [{"city": "Rome", "temp": 15.1, "rain_mm": 70}]
print(pd.DataFrame(rows))
by_city = df.set_index("city")
print(by_city.loc["Lima", "temp"]) # 19.0Remember
- A DataFrame is a set of Series columns sharing one row index
- Start with
shape,head,dtypesandinfo df['col']gives a Series,df[['a', 'b']]a DataFrame- Most methods return a new object; assign the result
Watch out for
- Calling
df.set_index('id')and expectingdfto change: it returns a new DataFrame.