dplyr cur_group_id() in R: Stable Integer ID Per Group

The cur_group_id() function in dplyr returns a unique integer 1..n_groups identifying the current group inside summarise() or mutate() on a grouped tibble.

df |> group_by(g) |> mutate(grp_id = cur_group_id())
df |> group_by(g) |> summarise(grp_id = cur_group_id(), n = n())
cur_group()$g                            # alternative: actual group values
match(g, unique(g))                       # base R equivalent (less robust)

What cur_group_id() does in one sentence

cur_group_id() returns a single integer between 1 and n_groups, unique per group, when called inside a dplyr verb on a grouped tibble. Useful for adding integer group IDs without manually computing them.

Syntax

No arguments. Must be inside a dplyr verb on a grouped tibble.

Five common patterns

1. Add stable group ID

2. Use ID as factor levels

3. Group-specific seed for randomness

4. Track progress in long pipelines

5. Combine with cur_group for full context

cur_group_id() vs cur_group() vs match

Three ways to identify groups in dplyr.

When to use which:

• cur_group_id for stable ID that doesn't change with row order.
• cur_group for the actual grouping column values.
• cur_group_rows for advanced per-group operations.

A practical workflow

Use cur_group_id when you need a stable integer key per group for downstream use.

The integer is consistent across reruns of the same data.

Common pitfalls

Pitfall 1: ID order is dplyr-defined, not value-sorted. Don't assume cur_group_id == 1 means the smallest grouping value. dplyr's internal ordering may differ.

Pitfall 2: errors outside dplyr verbs. mtcars |> cur_group_id() errors. Must be inside summarise / mutate / filter.

Try it yourself

Related dplyr functions

After mastering cur_group_id, look at:

• cur_group(): actual grouping values
• cur_group_rows(): row indices
• n(): row count in current group
• group_by(): parent context for all cur_group_*
• dense_rank(): similar but operates on a column directly

For a column-based equivalent, dense_rank(category) produces stable integers from a categorical column without group_by.

Why cur_group_id beats hand-computed IDs

Computing group IDs manually with match or factor() works but breaks down with grouped pipelines. cur_group_id is dplyr-aware and integrates naturally. It also produces consistent results across versions of dplyr, whereas custom match-based logic depends on data ordering. For reproducibility, prefer cur_group_id.

FAQ

What does cur_group_id return?

A single integer between 1 and n_groups, unique to the current group inside summarise / mutate / filter.

What is the difference between cur_group_id and cur_group?

cur_group_id returns a single integer; cur_group returns a 1-row tibble of the actual grouping column values. Use ID for stable keys; cur_group for actual values.

Are cur_group_id values stable across runs?

Yes, for the same input data and the same group_by() call. Different group_by calls or different data may produce different IDs.

Can I use cur_group_id outside dplyr?

No. It must be inside summarise, mutate, or filter on a grouped tibble.

How do I get a stable group ID without group_by?

Use dense_rank(col) for a column-based stable integer. cur_group_id is for grouped contexts; dense_rank works at the top level.