stringr str_split_fixed() in R: Split Strings to Matrix

The stringr str_split_fixed() function splits each string in a character vector by a pattern and returns a character matrix with exactly n columns. Short inputs are padded with empty strings; long inputs collapse extra pieces into the final column.

str_split_fixed(x, "/", n = 3)              # 3-column matrix
str_split_fixed(x, ",", n = 2)              # split once into key, value
str_split_fixed(x, "-", n = 4)[, 2]         # take 2nd field as vector
str_split_fixed(x, regex("[,;]"), n = 2)    # regex delimiter
str_split_fixed(x, fixed("."), n = 3)       # literal dot, not regex
data.frame(str_split_fixed(x, "/", n = 3))  # straight to data frame
str_split_fixed(x, "\\s+", n = 2)           # split on any whitespace run

What str_split_fixed() does in one sentence

str_split_fixed(string, pattern, n) returns a character matrix with length(string) rows and exactly n columns. Each row holds the pieces from one input string. The fixed shape is what makes it useful for tabular conversion, where str_split() returns a ragged list that you would have to coerce.

The trade off is data loss when an input has more than n pieces: the final column absorbs the leftover pattern and pieces verbatim. Choose n to match your maximum expected piece count.

Syntax

str_split_fixed(string, pattern, n) requires all three arguments. Unlike str_split(), where n defaults to Inf, str_split_fixed() has no default for n because the matrix shape must be known in advance.

The result is a 3 by 3 character matrix. Every cell is a string, including the price column. Convert numeric columns explicitly with as.numeric() before doing math on them.

Five common patterns

1. Split into named data frame columns

The most common downstream use is a data frame with one column per piece. Wrap the matrix in data.frame() and rename the columns.

2. Pad short rows with empty strings

When inputs have fewer than n pieces, the missing slots become "" (empty string), not NA.

3. Cap pieces, collapse the rest

When inputs have MORE than n pieces, the final column receives the rest of the string with the delimiter intact. This is the most common surprise for new users.

4. Take a single field as a vector

Indexing the matrix with [, k] returns column k as a character vector. This is the cleanest way to pull "second field" from a structured string.

5. Split on a regex delimiter

Pass a regex when the delimiter varies. A character class like [,;] matches either a comma or a semicolon.

str_split_fixed vs alternatives

Use this table to pick between str_split_fixed and the four most-confused alternatives. All four solve overlapping problems but optimize for different output shapes.

str_split_fixed() wins when your data has a known schema (CSV-like rows of three fields, hyphen-separated codes with four parts) and you want to feed the result to a model, table, or plot.

Common pitfalls

Three mistakes recur with str_split_fixed(): forgetting that empty cells are not NA, treating numeric columns as numbers without conversion, and choosing n smaller than the actual maximum.

Convert with dplyr::na_if(mat[, 3], "") or ifelse(mat == "", NA, mat) if downstream code expects NA.

Without as.numeric(), sum() raises invalid 'type' (character) of argument.

Try it yourself

Related stringr functions

For other split and extract patterns, see:

• str_split(): list output for variable piece counts
• str_split_1(): single string into a flat vector
• str_split_i(): pull the i-th piece directly
• str_extract(): capture matching pieces instead of splitting
• str_match(): regex groups returned as a matrix

The official reference is the stringr str_split documentation, which covers every variant in the family.

FAQ

What is the difference between str_split() and str_split_fixed()?

str_split() returns a list because each input string can produce a different number of pieces. str_split_fixed() returns a character matrix with a fixed column count, padding short rows with "" and collapsing long ones into the final column. Pick the matrix when downstream code needs uniform shape (a data frame, a model matrix); pick the list when piece counts vary and you want every piece.

Why does str_split_fixed() return empty strings instead of NA for missing pieces?

The output type is character, and empty string is the zero value for character vectors. NA_character_ would also work, but the design choice in stringr matches strsplit() behavior. Convert empty cells with dplyr::na_if(x, "") or ifelse(x == "", NA, x) when you want explicit missingness.

How do I split on a literal dot or special regex character?

The pattern argument is regex by default, so "." matches any character. Wrap the delimiter in stringr::fixed() (fixed(".")) for literal matching, or escape it with backslashes ("\\."). Use fixed() when readability matters and you have no regex needs.

Can str_split_fixed() handle a vector of patterns?

No, the pattern argument is a single regex or fixed()/coll() modifier applied to every input string. To split different rows on different patterns, loop with mapply() over strsplit() or use purrr::map2() with str_split().

How do I convert the result of str_split_fixed() to a data frame with named columns?

Wrap the matrix in data.frame() and assign column names: setNames(data.frame(str_split_fixed(x, "/", n = 3)), c("a", "b", "c")). Or use as_tibble(.name_repair = "minimal") from tibble if you prefer tidyverse output. Both preserve the character type, so cast numeric columns with as.numeric() afterwards.