Skip to main content
The datatable.str module provides functions that operate on string (str32 / str64) columns inside datatable expressions. Functions accept FExpr arguments and return FExpr results, composing naturally with f selectors and other datatable operations. 
import datatable as dt
from datatable import f

DT = dt.Frame(name=["Alice", "Bob", "Charlie"])

# Compute string lengths
DT[:, dt.str.len(f.name)]
Functions in dt.str operate lazily. They return an FExpr that is only evaluated when used inside a DT[rows, cols] expression.

​
Functions

​
dt.str.len(column)

Compute the length (number of characters) of each string value in column. Parameters
​
column
FExpr[str]
A string column expression.
Returns FExpr[int64] — an integer column containing the character count for each row. NA strings produce NA output.

​
dt.str.slice(col, start, stop, step=1)

Apply the slice [start:stop:step] to each string value in col. You can also write f.col[start:stop:step] as a shorthand. Parameters
​
col
FExpr[str]
The string column to slice.
​
start
int | None
Start index of the slice (inclusive). Negative indices count from the end. None means the beginning of the string.
​
stop
int | None
Stop index of the slice (exclusive). Negative indices count from the end. None means the end of the string.
​
step
int
default:"1"
Step size. Defaults to 1.
Returns FExpr[str] — a string column with the sliced values.

​
dt.str.split_into_nhot(frame, sep=",", sort=False)

Split and n-hot encode a single-column string frame. Each value is split on sep, whitespace is trimmed, and the resulting labels become boolean columns in the output frame. Parameters
​
frame
Frame
A single-column frame with str32 or str64 stype.
​
sep
str
default:"\",\""
Single-character delimiter to split on.
​
sort
bool
default:"false"
If True, output columns are sorted alphabetically by label. Due to parallelization, column order is otherwise not guaranteed.
Returns Frame — one boolean column per unique label, one row per input row.
split_into_nhot operates on a full Frame (not a lazy FExpr). Pass the single-column frame directly to this function, not inside a DT[rows, cols] expression.

​
Examples

​
Compute string lengths and filter

import datatable as dt
from datatable import f

DT = dt.Frame(word=["cat", "elephant", "ox", None, "rhinoceros"])

# Add a length column, then keep only words longer than 3 characters
result = DT[:, {"word": f.word, "length": dt.str.len(f.word)}]
result = result[f.length > 3, :]

​
Slice the first five characters of each string

import datatable as dt
from datatable import f

DT = dt.Frame(A=["apples", "bananas", "cherries", "dates"])

# Using dt.str.slice
DT[:, dt.str.slice(f.A, None, 5)]

# Equivalent shorthand via f-expression subscript
DT[:, f.A[:5]]

​
N-hot encode a multi-label column

import datatable as dt

DT = dt.Frame(tags=["python,data", "data,ml", "python,ml,deep-learning"])

# Produce one boolean column per unique tag
encoded = dt.str.split_into_nhot(DT[:, "tags"], sep=",", sort=True)

Build docs developers (and LLMs) love

Get started for freeTalk to us