Conditionally updating values of a DataFrame in Pandas
Start your free 7-days trial now!
Consider the following DataFrame:
df = pd.DataFrame({"A":[3,4],"B":[5,6]})df
A B0 3 51 4 6
Conditionally updating all values
To update values that are larger than 3 in the entire DataFrame:
df[df > 3] = 10df
A B0 3 101 10 10
Explanation
Here, we're first creating a DataFrame of booleans based on our criteria:
df > 3
A B0 False True1 True True
True represents entries that match our criteria. Placing this mask into our df using [~] returns the references to the matched entries:
df[df > 3]
A B0 NaN 51 4.0 6
We can then update the values using = like so:
df[df > 3] = 10df
A B0 3 101 10 10
Conditionally updating values for specific columns
Consider the same DataFrame we had before:
df = pd.DataFrame({"A":[3,4],"B":[5,6]})df
A B0 3 51 4 6
Instead of updating the values of the entire DataFrame, we can select the columns to conditionally update using the loc property:
df.loc[df["A"] > 3, "A"] = 10df
A B0 3 51 10 6
Here, we are updating values that are greater than 3 in column A.
Explanation
To break down the components of loc, here's the boolean mask we are passing in:
df["A"] > 3
0 False1 TrueName: A, dtype: bool
This is a Series, where True indicates the entry that satisfied the criteria.
The trap here is that, if we just pass this mask directly into loc, we end up with the second row being updated:
df.loc[df["A"] > 3] = 10df
A B0 3 51 10 10
This is not what we want since we want to perform updates on column A only. To this end, we need to specify the columns like so:
df.loc[df["A"] > 3, "A"] = 10df
A B0 3 51 10 6
Conditionally updating values based on their value
Consider the following DataFrame:
df = pd.DataFrame({"A":[3,4],"B":[5,6]})df
A B0 3 51 4 6
All values in the DataFrame
To update values based on their value, use the applymap(~) method like so:
df = df.applymap(lambda val: 2*val if val > 3 else val)df
A B0 3 101 8 12
Here, we're doubling values that are greater than 3. This approach gives you the flexibility of setting a new value that is based on the value to be updated, which isn't possible by using loc alone.
Values of specific columns
To update values of specific columns based on their value:
df["A"] = df["A"].apply(lambda val: 2*val if val > 3 else val)df
A B0 3 51 8 6
Here, note the following:
we're doubling values in column
Athat are greater than3.since
Seriesdoes not haveapplymap(~), we usedapply(~)instead.
