Every column in a DataFrame has a dtype — the type of value it holds under the hood. Numbers are usually int64 or float64, and text (including dates that haven't been converted yet!) shows up as object. Check any DataFrame's dtypes with df.dtypes — it's the pandas equivalent of looking at a table's column types in SQL (information_schema or DESCRIBE).

Real-world data rarely arrives in the dtype you want. pandas gives you a few converters:

df['col'].astype(float) (or int, str, etc.) — a direct, general-purpose type conversion.
pd.to_numeric(df['col']) — specifically for turning text-that-looks-like-numbers into real numbers.
pd.to_datetime(df['col']) — specifically for turning text-that-looks-like-dates into real datetimes.

All three return a new Series — you still have to assign it back: orders['order_date'] = pd.to_datetime(orders['order_date']).

Data also gets duplicated by accident — a form submitted twice, a sync job that ran twice. df.duplicated() returns a True/False Series flagging any row that's an exact repeat of an earlier row (the first occurrence stays False). df.drop_duplicates() then removes those repeats, keeping just the first copy of each — pandas' equivalent of SQL's SELECT DISTINCT, but for whole rows.

SQL	pandas
`DESCRIBE table` / `information_schema`	`df.dtypes`
`CAST(col AS FLOAT)`	`df['col'].astype(float)`
`SELECT DISTINCT *`	`df.drop_duplicates()`

A quick rule of thumb: if a column full of numbers or dates shows up as object in df.dtypes, it's stored as text, and you'll need to convert it before .dt, >/<, or math will behave correctly.

Loading SQL engine…

Couldn't load the SQL engine

Database schema

Your SQL level

Metric

Framework

Log in