Navigate
Home Compendium
Exercises
Tidyverse packages
dplyr ggplot2 tidyr stringr lubridate purrr forcats readr broom Tidyverse R for Data Science
Deep dives
dplyr Joins dplyr Window Functions dplyr group_by ggplot2 Themes ggplot2 Facets ggplot2 Color Scales ggplot2 Bar Charts ggplot2 Heatmaps tidyr Pivot tidyr Nest/Unnest Regex
Wrangling & EDA
Data Wrangling Data Cleaning EDA Data Visualization data.table dbplyr / SQL Web Scraping API Calls
Statistics
Hypothesis Testing Linear Regression Logistic Regression Correlation Probability Distributions
Machine Learning
Machine Learning tidymodels Random Forest XGBoost Clustering PCA Cross-Validation
Time Series
Time Series ARIMA
By Industry
Finance Marketing Analytics Healthcare Biostatistics Genomics Sports Analytics A/B Testing
Reporting & Apps
Shiny R Markdown
Levels
R for Beginners R Interview Questions
R r‑statistics.co

dplyr n_distinct() in R: Count Unique Values Fast

The n_distinct() function in dplyr counts the number of unique values in one or more vectors. It is the fast, dplyr-native equivalent of length(unique(x)).

By Selva Prabhakaran  ·  Published May 12, 2026  ·  Last updated May 12, 2026
⚡ Quick Answer
n_distinct(x)                          # unique values in x
n_distinct(x, na.rm = TRUE)            # exclude NAs
n_distinct(x, y)                       # unique combinations of x and y
df |> summarise(n_unique = n_distinct(col))
df |> group_by(g) |> summarise(n_unique = n_distinct(col))
length(unique(x))                       # base R equivalent (slower)

Need explanation? Read on for examples and pitfalls.

📊 Is n_distinct() the right tool?
STARTcount unique values in a vectorn_distinct(x)count unique combinations of multiple colsn_distinct(x, y, z)exclude NA from the countn_distinct(x, na.rm = TRUE)count rowsn() or nrow()show the distinct valuesdistinct(df, col) or unique(x)per-group unique countgroup_by + summarise(n_distinct(col))

What n_distinct() does in one sentence

n_distinct(x, ..., na.rm = FALSE) returns the integer count of unique values in x (or unique combinations across x, ...). It is faster than length(unique(x)) because it avoids materializing the full unique vector.

The standard "how many unique customers / products / categories?" function in dplyr.

Syntax

n_distinct(..., na.rm = FALSE). Pass one or more vectors; counts unique combinations across them.

Run live
Run live, no install needed. Every R block on this page runs in your browser. Click Run, edit the code, re-run instantly. No setup.
RUnique values in a vector
library(dplyr) x <- c(1, 2, 2, 3, 3, 3, NA) n_distinct(x) #> [1] 4 (1, 2, 3, NA all counted as distinct) n_distinct(x, na.rm = TRUE) #> [1] 3

  

  





Tip
n_distinct(x) is faster than length(unique(x)). dplyr uses a hash-based approach that avoids materializing the unique vector. On a million-element vector the difference is significant.



Five common patterns



1. Unique values in one column




  

    
RHow many cylinder counts in mtcars?

    
n_distinct(mtcars$cyl)
#> [1] 3

    

      
      
    

    

  

  





2. Inside summarise




  

    
RPer-group unique count

    
mtcars |>
  group_by(cyl) |>
  summarise(n_gears = n_distinct(gear))
#>     cyl n_gears
#>     4       3
#>     6       3
#>     8       2

    

      
      
    

    

  

  





3. Multi-column combinations




  

    
RUnique (cyl, gear) pairs

    
n_distinct(mtcars$cyl, mtcars$gear)
#> [1] 8

    

      
      
    

    

  

  





Pass multiple vectors to count unique combinations.



4. Excluding NAs




  

    
RSkip NAs in the count

    
x <- c("a", "b", NA, "a", NA)
n_distinct(x, na.rm = TRUE)
#> [1] 2

n_distinct(x, na.rm = FALSE)
#> [1] 3   (NA counted as one of the distinct values)

    

      
      
    

    

  

  





5. Inside mutate (per-group)




  

    
RAdd unique count column

    
mtcars |>
  group_by(cyl) |>
  mutate(n_gear = n_distinct(gear))
#> Each row gets the count of unique gear in its cyl group

    

      
      
    

    

  

  





Key Insight
n_distinct() differs from n(): it counts UNIQUE values, not ROWS. n() returns the group size; n_distinct(col) returns how many distinct values appear in that column. Easy to confuse but very different semantics.



n_distinct() vs length(unique()) vs distinct() vs n()



Four ways to handle "uniqueness" questions in dplyr / R.







Function
Returns
Best for






n_distinct(x)
Integer count
Quick count, dplyr summarise




length(unique(x))
Integer count
Base R, equivalent but slower




dplyr::distinct(df, col)
Filtered tibble
"Show me the unique rows"




unique(x)
Vector of unique values
Inspect what those values are




n()
Group size (row count)
Different question







When to use which:



    


  • n_distinct(x) inside summarise/mutate.
    • 


  • length(unique(x)) for base R; same result, slightly slower.
    • 


  • distinct(df, col) to keep one row per unique value.
    • 


  • unique(x) to see the actual unique values.
    • 





A practical workflow



The "audit" pattern is the most common n_distinct use case in summary tables.




  

    
RDataset audit

    
df |>
  summarise(
    rows         = n(),
    unique_users = n_distinct(user_id),
    unique_items = n_distinct(item_id),
    avg_qty      = mean(qty),
    .groups      = "drop"
  )

    

      
      
    

    

  

  





A first-pass dataset audit: how many rows, how many unique users, how many unique items. Tells you the table's shape at a glance.



For per-group audits:




  

    
RDaily session and unique-user counts

    
df |>
  group_by(date) |>
  summarise(
    sessions       = n(),
    unique_users   = n_distinct(user_id),
    .groups        = "drop"
  )

    

      
      
    

    

  

  





Daily session and unique-user counts.



Common pitfalls



Pitfall 1: NA counted as distinct. Default na.rm = FALSE includes NA in the count. n_distinct(c(1, NA, 1)) returns 2 (1 and NA). Add na.rm = TRUE to exclude.



Pitfall 2: passing multiple cols treats them as combinations. n_distinct(x, y) counts unique (x, y) PAIRS, not unique values in either. To count separately, call twice.



Warning
n_distinct() is faster than length(unique(x)) but they CAN differ on edge cases. With factors and NA handling, results may vary. Pick one in a project and stick with it for consistency.



Performance note



For very large data, n_distinct() is faster than length(unique(x)) thanks to a hash-based implementation. On vectors with millions of elements the difference can be 2-10x. For everyday data sizes (thousands to hundreds of thousands of rows) both functions feel instant, so pick by readability and consistency. Inside dplyr pipelines, n_distinct is the idiomatic choice. For data.table users, uniqueN() plays the same role with similar performance. The underlying algorithm uses a hash set internally rather than building the full unique vector, which is what saves memory and time on big inputs.



Try it yourself





Try it: For each cyl group in mtcars, count the unique number of gears AND the unique number of carburetors. Save to ex_uniq.



  

    
RYour turn: 2 unique-counts per group

    
ex_uniq <- mtcars |>
  # your code here

ex_uniq
#> Expected: 3 rows (one per cyl) with n_gear and n_carb columns

    

      
      
    

    

  

  





Click to reveal solution



  

    
RSolution

    
ex_uniq <- mtcars |>
  group_by(cyl) |>
  summarise(
    n_gear = n_distinct(gear),
    n_carb = n_distinct(carb),
    .groups = "drop"
  )

ex_uniq
#> # A tibble: 3 x 3
#>     cyl n_gear n_carb
#>     4      3      2
#>     6      3      4
#>     8      2      4

    

      
      
    

    

  

  





Explanation: n_distinct(gear) counts unique gear values per cyl. Same for carb. Two summary stats per group.










After mastering n_distinct, look at:



    


  • n(): row count of current group
    • 


  • count(df, g): count rows per group
    • 


  • distinct(df, col): keep one row per unique value
    • 


  • unique(x): base R; show the unique values
    • 


  • summarise(): standard aggregation context
    • 


  • group_by(): per-group n_distinct
    • 





For "show me the actual unique rows", distinct(df, col, .keep_all = TRUE) is the cleaner tool.



FAQ



What does n_distinct do in dplyr?



n_distinct(x) returns the count of unique values in x as an integer. Faster than length(unique(x)) and integrates with summarise / mutate.



What is the difference between n_distinct and length(unique())?



Both return the same count. n_distinct is faster (hash-based) and is the dplyr-native idiom. length(unique()) is base R; works anywhere.



How do I exclude NAs from n_distinct?



Pass na.rm = TRUE: n_distinct(x, na.rm = TRUE). Default counts NA as one of the distinct values.



How do I count unique combinations of multiple columns?



n_distinct(x, y, z) counts unique tuples across the three vectors. Pass each column as a separate argument.



Is n_distinct different from n() in dplyr?



Yes. n() counts ROWS in the current group. n_distinct(col) counts UNIQUE values in a column. Different questions.



          

        


        
      


      

        

        
© 2016-2026 Selva Prabhakaran.
          This work is licensed under the Creative Commons License.
        

        

          About ·
          RSS