LoginSign Up
← Back to all sparks
D

dplyr

ANALYTICS
Velocity0.0

Grammar of data manipulation for R

github.com/tidyverse/dplyr ↗

After two quiet years dplyr widened its verb vocabulary in one release

rdata-manipulationtidyverseapi-expansionperformance
◆Current state
dplyr sat on patch releases from late 2023 until 1.2.0 landed in February 2026, and that release did a lot at once: a filter_out() counterpart to filter(), elementwise when_any() and when_all(), and three new recoding verbs alongside case_when(). It also rewrote if_else(), case_when() and coalesce() in C via vctrs, and promoted .by and reframe() from experimental to stable. The follow-up 1.2.1 is a compliance patch.
◆Where it's heading
The package is expanding its verb set deliberately, through published Tidyup design proposals rather than ad-hoc additions, and each new verb targets a case where the old idiom was error-prone - most obviously NA handling in negated filters. Underneath, hot paths keep moving from R into C, so the API grows while the runtime cost falls.
◆Prediction
Expect the remaining experimental surface to follow .by and reframe() toward stable, and further hot paths to be rewritten in C via vctrs. The two Tidyup proposals referenced here suggest more of the filter and recode families is still being designed.

◆Recent moves

  1. 4mo ago

    Full compliance with the R C API

    A one-line patch bringing dplyr into full compliance with the R C API. Invisible to users writing dplyr code, but it keeps the package clean against R's own interface rules.

    View source ↗
  2. 6mo ago

    filter_out(), when_any() and three recoding verbs land in 1.2.0

    ⚡ SPARK

    The release that ends the long 1.1.x stretch and widens dplyr's core vocabulary: filter_out(), when_any()/when_all(), and three new recoding verbs, plus a C rewrite of if_else(), case_when() and coalesce() and stable status for .by and reframe().

    View source ↗
  3. 2y ago

    Namespaced join_by() helpers and refreshed bundled datasets

    A patch release mixing real usability work - namespaced join_by() helpers, a specific error when a join would overflow - with refreshed storms and starwars datasets. Incremental polish on the 1.1 line.

    View source ↗
  4. 2y ago

    Deprecation message and setequal() consistency fixes

    Corrects deprecation messages for mutate_each() and summarise_each() and tightens setequal() to require compatible inputs like the other set operations. Consistency fixes only.

    View source ↗
  5. 3y ago

    All-NA join key fix and count() documentation

    Documentation for count()'s .drop argument, a join fix for all-NA key columns, and test compatibility with an upcoming waldo release. Routine patch work.

    View source ↗
  6. 3y ago

    Joins gain a relationship argument and warn far less often

    Mutating joins stop warning on one-to-many relationships and warn only on the genuinely dangerous many-to-many case, with a new relationship argument to state the expected cardinality. A meaningful correction to behaviour shipped only weeks earlier.

    View source ↗